HISTORICAL RESEARCH REPORTResearch Report TM/92/06

1992

Coalminers’ mortality in relation to low-level exposures to radon and thoron daughters. Final report on NRPB Contract(EMR 7/87 90214000 and CEC Contract No.7270-ZZ-003 Maclaren WM

Copyright © 2006 Institute of Occupational Medicine. INSTITUTE OF OCCUPATIONAL MEDICINE No part of this publication may be reproduced, stored Research Avenue North, Riccarton, Edinburgh, EH14 4AP or transmitted in any form or by any means without Tel: +44 (0)870 850 5131 Fax: +44 (0)870 850 5132 written permission from the IOM e-mail publications@iomhq.org.uk

Coalminers’ mortality in relation to low-level exposures to radon and thoron daughters. Final report

on NRPB Contract(EMR 7/87 90214000 and CEC Contract No.7270-ZZ-003

Maclaren WM This document is a facsimile of an original copy of the report, which has been scanned as an image, with searchable text. Because the quality of this scanned image is determined by the clarity of the original text pages, there may be variations in the overall appearance of pages within the report. The scanning of this and the other historical reports in the Research Reports series was funded by a grant from the Wellcome Trust. The IOM’s research reports are freely available for download as PDF files from our web site: http://www.iom-world.org/research/libraryentry.php

HISTORICAL RESEARCH REPORTResearch Report TM/92/06

1992

ii Research Report TM/92/06

COALMINERS' MORTALITY IN RELATION TO

LOW-LEVEL EXPOSURES TO

RADON AND THORON DAUGHTERS

WM Maclaren

June 1992

IOM Report TM/92/06

Report No: TM/92/06

INSTITUTE OF OCCUPATIONAL MEDICINE

COALMINERS1 MORTALITY IN RELATION TO LOW-LEVEL EXPOSURESTO RADON AND THORON DAUGHTERS

by

WM Maclaren

Final Report on NRPB Contract (EMR 7/87) 90214000and CEC Contract No. 7270-ZZ-003

Duration of project: 1 November 1988 - 31 December 1991

Institute of Occupational Medicine8 Roxburgh PlaceEDINBURGHEH8 9SU

Tel: 031 667 5131Telex: 9312100237=10 GFax: 031 667 0136

Price: £40.00 (UK)£45.00 (Overseas)

This report is one of a series of Technical Memoranda (TM) distributed by theInstitute of Occupational Medicine. Current and earlier lists of these reports andof other Institute publications, are available from the Librarian/Information Officer.

(i)

CONTENTS

Page No.

SUMMARY v

1. INTRODUCTION 1

2. MEN AND CHARACTERISTICS STUDIED 5

2.1 The Pneumoconiosis Field Research 5

2.2 Men Participating in the Present Study 6

2.3 Mortality Data 7

2.4 Radon and Thoron Daughter Exposures 7

2.5 Smoking Habit 8

3. METHODS 11

3.1 Establishing Vital Status 11

3.2 The Calculation of Radon and ThoronDaughter Exposures 12

3.2.1 Outline of the method of calculation 123.2.2 Definition of risk categories 123.2.3 Assignment of radioactivity levels

to risk categories 123.2.4 Partitioning of Attendance Records time 133.2.5 Partitioning of estimated working times

derived from Occupational Histories 133.2.6 Calculation of exposures - an example 13

3.3 Smoking Habit 14

3.4 Definition of Risk Period 14

3.5 Statistical Methods 15

3.5.1 Person-years at risk 153.5.2 Case-referent studies 18

4. RESULTS 29

4.1 Vital Status 29

4.2 Exclusions from Statistical Analysis onGrounds of Missing or Unreliable Data 29

4.3 Summary Statistics for the Study Group 30

4.4 External Comparisons of Mortality - AllCauses of Death, and Lung Cancer 31

4.5 The Relationship between Radon and ThoronDaughter Exposure and Cause-Specific DeathRates 31

4.5.1 Lung cancer 314.5.2 Stomach cancer 334.5.3 Other causes of death 34

5. DISCUSSION 79

5.1 Lung Cancer 79

5.2 Other Causes of Death 82

5.3 Quality of Exposure Data 82

5.4 Completeness of Follow-up; Exclusions fromAnalysis on Grounds of Missing orUnreliable Data 85

5.5 Statistical Methods 86

5.5.1 Treatment of exposure variables,and smoking 86

5.5.2 Person-years analyses 875.5.3 Case-referent studies 87

5.6 Conclusions and Recommendations forFurther Analysis 87

5.6.1 Conclusions 875.6.2 Recommendations for further analysis 88

ACKNOWLEDGEMENTS 91

REFERENCES 93

APPENDIX 1: Quantities of radioactivity and theirunits of measurement. 97

(Hi)

APPENDIX 2:

APPENDIX 3:

APPENDIX 4:

APPENDIX 5:

APPENDIX 6:

APPENDIX 7:

APPENDIX 8:

Results of a survey of radon and thorondaughter levels carried out at 11 Britishcollieries between 1972 and 1980. 99

PANDA questionnaire on respiratory symptomsand smoking used at routine PFR surveys. 105

The seam-mean method of exposure estimation. 109

Smoking habit - derivation of codes.

Results bearing on the reliability ofexposures.

Tables of person-years and numbers ofdeaths.

Indoor exposures for the study group.

125

129

135

137

(iv)

(v)

Report No: TM/92/06

INSTITUTE OF OCCUPATIONAL MEDICINE

COALMINERS' MORTALITY IN RELATION TO LOW-LEVEL EXPOSURESTO RADON AND THORON DAUGHTERS

by

WM Maclaren

SUMMARY

Epidemiological studies of many groups of miners, of various ores, have shown thatexposure to radon daughters is associated with increased lung cancer risk. Theseresults, taken in conjunction with surveys of domestic radon levels, suggested thatindoor exposure might also present a hazard for some segments of the generalpopulation. However, the magnitude of risk at the comparatively low levels foundindoors was uncertain, since the exposure of mining groups from whom riskestimates were derived, was in general higher. Surveys of radon gas in Britishcoalmines had shown levels of a similar order of magnitude to those of indoors,and it therefore appeared that coalminers might be a suitable occupational group inwhich to study the effects of low-level exposure to radon daughters. Theresources of the Pneumoconiosis Field Research (PFR) database presented theopportunity of carrying out such a study, which was begun in November 1988, withsupport from the National Radiological Protection Board and also, subsequently,from the Commission of the European Communities. The objective was to studyrelationships between exposure to low levels of radon and thoron daughters at 10British collieries, and subsequent mortality, in particular from lung cancer.

The study group consisted of 19418 male industrial coalmine workers who attendedeither or both of two medical surveys carried out at 10 British collieries during the1950s and early 1960s as part of the PFR. Deaths in this group (up to 31December 1989) were notified to the Institute of Occupational Medicine (IOM) bythe Office of Population Censuses and Surveys. Cumulative exposures to radonand thoron daughters acquired during working time, were calculated for a subgroupof 14956 men using data from two sources: a set of 146 measurements of radonand thoron daughter levels made at the 10 collieries during the 1970s; andextensive information held in the PFR database on time worked underground andon the surface. Indoor exposure acquired outside working hours was notestimated. An overlapping subgroup of 14145 men were categorized by smokinghabit, using questionnaire information obtained at a sequence of medical surveyscarried out during the 1950s, 60s and 70s.

(vi)

Analyses of mortality from two causes (lung and stomach cancer) used theperson-years-at-risk method, and also matched case-referent comparisons. Afurther eight causes of death (malignant neoplasm of: oral cavity, oesophagus,larynx, bone, prostate, kidney; malignant melanoma, leukaemia - excludingchronic lymphoid) were analysed using case-referent methods only. All analysesof exposure-response were carried out entirely within the study group. Externalcomparisons of mortality were carried out for lung cancer and all-causes only;exposure-response was not examined.

Only men with reliable data on vital status, exposure and smoking were included instatistical analyses - a subgroup of 12361 men. Of the complete study group(19418 men), 3.3% were excluded because of unknown vital status. The mediandate of entry to the study was 1960, the median age at entry was 44; theaverage complete cumulative working time exposure to radon daughters was 250WL hour (sd 291), and to thoron daughters, 162 WL hour (sd 121); theproportion of 'never-smokers' among men examined at PFR surveys varied from12% to 14%.

The SMR for all causes was 96, based on 5822 deaths in the study group of 12361men. For lung cancer, the SMR was 87, based on 521 deaths. Person-yearsanalysis did not show any relationship between lagged cumulative exposure to radondaughters and lung cancer death rate. Case-referent analysis showed a jointeffect of smoking and radon daughter exposure upon mortality: relative risks wereelevated for non-smokers and smokers of up to 10 cigarettes per day, but declinedto less than unity for heavier smokers. No relationship between thoron daughterexposure and mortality was identified in either statistical analysis. However thesame pattern of interaction between smoking and exposure was evident, but notstatistically significant, in the case-referent study. Case-referent analysis of acombined dose measure showed a joint effect of smoking and dose upon mortality.The pattern of interaction was similar to that observed for radon daughterexposure. Neither the person-years analysis, nor the case-referent study, showedany relationship between stomach cancer and radon daughter, or thoron daughter,exposure. For oesophageal cancer, an apparently negative association betweenexposure (both radon and thoron daughter) and mortality was observed, whichvaried with age. No associations were found for any other causes of deathanalysed.

The pattern of the joint effect of radon daughter exposure and smoking upon lungcancer mortality presents difficulties of interpretation; the negative associationamong heavy smokers is biologically implausible, and may be due to the operationof a bias, as yet unidentified. However, it would be premature to dismiss thepositive association between lung cancer mortality and radon daughter exposure inlight smokers because of the problematical negative association in heavy smokers.Finally, the joint influence of age and both exposures upon oesophageal cancermortality may be due to health-related movements between jobs.

1

1. INTRODUCTION

The fact that the radioactive decay products of radon gas (known collectively asradon daughters) present a lung cancer hazard is by now well documented.Epidemiological studies of uranium miners in the USA (Lundin et al, 1969),Czechoslovakia (Kunz et al, 1979) and Canada (Muller et al, 1987) have shownexcess lung cancer mortality, compared to general population death rates, whichincreased with increasing cumulative exposure to radon daughters. Miners of oresother than uranium are also at increased risk. Excess lung cancer rates related toradon daughter exposure,, have been reported in Swedish iron-ore miners (Radfordand St Clair Renard, 1984), calcium fluoride (fluorspar) miners in Newfoundland,Canada (Morrison et al, 1985) and tin miners in China (Qiao et al, 1989). InBritain, a study of iron-ore miners (Boyd et al, 1970) showed excess lung cancerin men employed underground, which the authors suggested might be attributable toradioactivity or iron oxide; and a recent study of Cornish tin miners (Hodgsonand Jones, 1990) showed significantly elevated standardized mortality ratios (SMRs)for lung cancer, which increased with length of time worked underground, andwhich the authors attributed to the effect of radon daughter exposure.

The presence of radon gas at potentially hazardous concentrations in the indoor airof homes was reported by Swedjemark (1979), who presented the results of aSwedish country-wide investigation of radon and radon daughters in indoor air.The highest concentrations of radon were found in dwellings built of concrete basedon alum shale. Houses built on waste material from alum shale mining were alsofound to have high levels of radon and radon daughters. Subsequent surveys inNorth America and Europe showed that high indoor concentrations could occur indwellings built on foundations free from any waste material from mines; Nero(1988) gives a table of mean radon levels obtained from surveys in variouscountries. In the UK, results of a survey of radon levels in over 2000 houses(Wrixon et al, 1988) showed wide variations in concentration, and led the NationalRadiological Protection Board (NRPB) to recommend that action be taken to reducehigh indoor radon concentrations.

Several case-referent studies of indoor exposure have been carried out in Sweden;an example is the study of Axelson et al (1979). Radon exposure wascharacterized by type of dwelling: (i) wooden, without a basement, (ii) stone, witha basement, (iii) all other dwellings. A relative risk of lung cancer mortality of1.8 was found for categories (ii) and (iii) versus (i). A later study (Edling etal, 1984) established a gradient in relative risks, using the same characterization ofexposure. In two more recent studies, lung cancer risks were examined inrelation to quantitative estimates of exposure obtained from measurements made inthe dwellings of subjects. Schoenberg et al (1990) compared exposures of 433New Jersey female lung cancer cases with those of 402 controls. Estimates ofindoor radon were obtained from year-long measurements made in dwellings whichhad been occupied for at least 10 years by study subjects. An increasing trendin relative risk, allowing for smoking habit, was significant at the 4% level.Ruosteenoja (1991) studied 318 male lung cancer cases, incident during 1980-85 in19 municipalities in southern Finland. Comparisons were with a frequencymatched random sample of 495 non-lung cancer cases. Exposure estimates werebased partly on two-month-long samples taken in dwellings occupied by cases andcontrols for at least a year since 1950, and partly on information about thedwellings and their locations. A clear exposure-response relationship for lungcancer risk was not established however; relative risks showed a decline in thehighest exposure category, after an initial increase.

Because of the lack of direct information on the relationship between lung cancerrisk and indoor exposure to radon daughters, estimates of risk have largely beenbased on studies of miners, exposed on average to higher levels of radon daughtersthan those typically found in homes. The difficulties with this approach lie bothin the uncertainty of extrapolation, and the many differences between the miningand domestic situations - for example, aerosol characteristics and breathing patternsof exposed persons. Two examples of risk assessment are the 'Time SinceExposure1 statistical model, developed by the Committee on the Biological Effectsof Ionising Radiations (NRC, 1988), which was based on the mortality experienceof three groups of uranium miners in the USA and Canada, and a group ofSwedish iron ore miners; and the assessment of lifetime lung cancer risk publishedby the International-Commission'on Radiological Protection (ICRP, 1987), whichwas derived in part from results of epidemiological studies of uranium miners inColorado, Ontario and Czechoslovakia. (The NRC report gives a detailed reviewof studies of miners' mortality from lung cancer in relation to radon daughterexposure, published at that time. Another review, of comparable depth, is givenin Volume 43 of the series of monographs on the evaluation of carcinogenic risksto humans published by the International Agency for Research on Cancer (IARC,1988).)

The problem of estimating individual exposures due to indoor radon is avoided inso-called correlation studies, where regional or national death rates are correlatedwith background radiation levels. Edling et al (1982) reported significant positivecorrelations between age-standardized lung cancer rates in males and females andestimated background radiation (with cosmic radiation excluded) in the 24 countiesof Sweden. As the authors pointed out, the correlation may have arisen becauseurban environments have greater air pollution, as well as higher levels ofbackground radiation, than rural environments. A further difficulty ininterpretation would be caused by possible differences in the smoking habits of thecounty populations.

Several of the studies of miners mentioned have shown increased mortality fromcancers at sites other than the lung. In particular, excess stomach cancer, and aslight excess of lymphomas, were reported by Radford and St Clair Renard (1984)in Swedish iron-ore miners, and Morrison et al (1985) reported excess mortalitydue to oral cavity cancer, and salivary gland cancer, among Canadian fluorsparminers. Exposure-response relationships were not examined, and the causalinvolvement of radon daughter exposure remained unproven. Recent correlationstudies (Henshaw et al, 1990; Eatough and Henshaw, 1990), in which nationalincidence rates of disease were examined in relation to mean indoor radonconcentrations obtained in national surveys, demonstrated significant correlations formyeloid leukaemia, kidney cancer, melanoma and prostate cancer. However, itwas subsequently pointed out (Butland et al, 1990) that national records of cancerincidence and of radon levels may have doubtful accuracy in some cases.

An occupational group, whose mortality had not (by the mid 1980s) been studiedin relation to radon daughters, were coalminers. Measurements of radon gas in12 British coalmines (Duggan et al, 1968) had given an estimated median value of20 mWL. (A description of radioactive quantities and units of measurement isgiven in Appendix 1.) At this level of exposure, a man working underground forone year would acquire a cumulative exposure of 0.2 WLM - far less thanexposures in uranium and iron-ore mines. For example, Radford and St ClairRenard (1984) gave a figure of 4.8 WLM as the average annual exposure forSwedish iron-ore miners. Since results of the NRPB survey of British homes hadshown average indoor levels of radon close to those in British coalmines, it

therefore appeared that coalminers were an appropriate group on which to studythe health effects of low-level radon daughter exposure.

Such a study, using the accumulated resources of the British National Coal Board'sPneumoconiosis Field Research (PFR) was proposed to NRPB, in 1987. As partof PFR, the Coal Board, through the Institute of Occupational Medicine (IOM),had maintained a long-term study of the mortality of approximately 30000 miners,employed at 24 collieries during the 1950s (Miller and Jacobsen, 1985). Theproposed study was to be based on a subgroup of these men, augmented byanother group of miners, all of whom had worked at 10 collieries (of the 24),where levels of radon daughters (and also of thoron daughters) had been measuredin the 1970s. (Much less is currently known r about the health effects of thorondaughters; the associated risk of lung cancer is reckoned to be about one-fifththat of radon daughters (James, 1988). In a study of Norwegian niobium minersexposed to radon and thoron daughters, Solli et al (1985) reported excess lungcancer incidence compared to the Norwegian male population.)

Exploratory analysis (by the IOM) of lung cancer mortality at the 10 collierieswhere radioactivity had been measured, showed that age-adjusted colliery-specificdeath rates had a statistically significant Spearman rank correlation coefficient withaverage radon daughter levels. To allow further exploration of this preliminaryresult, NRPB agreed to support the proposed study, which began in November1988, and whose findings are now presented. The objective was to studyrelationships between exposure to low levels of radon and thoron daughters at 10British collieries and subsequent mortality, in particular from lung cancer.

5

2. MEN AND CHARACTERISTICS STUDIED

2.1 The Pneumoconiosis Field Research

The present research project sprang from the PFR. For a full account, seeJacobsen (1981); what follows is a brief summary, with particular reference toaspects relevant to the present study.

In 1952, the National Joint Pneumoconiosis Committee invited the National CoalBoard to undertake a programme of research into the causes of pneumoconiosis.Twenty-five collieries were chosen to participate from all parts of the Britishcoalfield - five from Scotland, six from South Wales, and thirteen from theregions of England, from Kent to Northumberland. The selection was not arandom one; instead an effort was made to cover a range of dust conditionsthought to be relevant to pneumoconiosis.

The two main 'arms' to the research were the medical surveys, and theenvironmental monitoring programme. Between 1953 and 1958, each of the 25collieries was visited by one of two mobile X-ray units based at Cardiff andEdinburgh. This constituted the first round of medical surveys. Second andthird rounds (1958-63, 1963-68) were carried out at 24 of the original 25collieries, fourth and fifth rounds (1968-73, 1973-78) at 10 of these 24, and asixth round (1978) at two of these 10. From the second survey onwards, aquestionnaire on respiratory symptoms and smoking habits was administered. Theaim throughout this series of surveys was to examine all coalminers currentlyemployed at time of survey.

During the 1970s, a special study of miners and ex-miners (the so-calledFollow-up Study) was undertaken at the 24 collieries surveyed in the second round.Medical examinations were conducted, which coincided with the fifth routine PFRsurvey at seven of the 24 collieries, and with the sixth at one other. At theremaining 16 collieries, separate surveys were organized. Participants in theFollow-up Study (i.e. men invited to attend Follow-up Survey) were a sample ofmen seen at first survey some 20 years earlier.

As the medical surveys were proceeding, a detailed programme of dust samplingwas being carried out at the same collieries. Colliery populations were stratifiedinto occupational groups on the basis of exposure to coalmine dust. An'Attendance Records System' was devised to keep a record of times worked byindividual men within occupational groups. During the 1960s, the AttendanceRecords System was computerized, and records of times worked were transferred tomagnetic tape. Some summarization of data was carried out during this transfer:times worked within occupational groups were totalled within 'Inter Survey Periods'(ISPs), which were the approximately five-year periods between successive PFRsurveys. Following computerization, records of times worked within occupationalgroups were kept on a quarterly basis.

(Some useful terminology may be introduced at this point. PFR surveys arereferred to throughout the report by the letters PFR followed by the round numberof the survey. PFR1, PFR2 and PFR3 together constitute Phase One of theField Research; PFR4, PFR5 and PFR6, Phase Two. The 10 collieries wherePhase Two surveys were conducted are collectively known as Phase Two collieries.The shorthand term 'ISP' has already been introduced; the period between PFR1

and PFR2 is called ISP1, between PFR2 and PFR3, ISP2, and so on. ISPOrefers to a man's coalmining experience prior to PFR1.)

Following PFR3, the Attendance Records System was maintained at the 10 PhaseTwo collieries only; and after PFR5, at nine of these 10, one colliery havingclosed in June 1973, shortly after PFR5. Although the final rounds of medicalsurveys were either the fifth (at eight of the 10 Phase Two collieries) or the sixth(at the other two collieries), the Attendance Records System continued for sometime after PFR5 or 6. Thus, 'ISPS' Attendance Records times (spanning between11 and 16 quarters) are available (for men in the present study) at nine collieries,'ISP6* times (two to five quarters) at seven, and 'ISP7' times (one quarter) at one.With the exception-of ISPS at the two collieries where PFR6 was held, these 'ISPs'do not refer to inter-survey .periods, and are regarded here merely as labels,indicating an approximate-period during which time was worked. The latest yearfor which Attendance Records information is available from PFR computer files is1980.

The purpose of the Attendance Records System was to record times worked inoccupational groups at the PFR collieries; times worked at PFR collieries duringISPO, or at non-PFR collieries during any ISP, could not be recorded by thisSystem. An attempt was therefore made to estimate these times by administeringoccupational history questionnaires at each round of routine PFR surveys to menfalling into one of two categories:

i) those attending a PFR survey for the first time;

ii) those attending their second or later survey, but who had worked outside thePFR colliery since they had last been seen.

Information recorded on occupational history forms was summarised, andsubsequently computerised, in the form of time worked, within ISP, in six broadcategories of coalmining activity. Time outside mining was also included, andcategorised as 'noxious' or 'non-noxious1, in respect of likely exposure to airbornesubstances which could conceivably be associated with risk of lung disease.

Occupational history questionnaires were in fact administered to a third group ofmen, but only during Phase Two of the PFR. These were men whose timeinformation (both Attendance Records and occupational history) gathered duringPhase One was not transferred to computer in the 1960s, because of the thenunmanageable size of the data processing task. At that time, the decision wastaken not to process time-worked data for men who had attended only PFR1 or 2(but not both) in Phase One, the so-called 'Singletons'. These men, it was felt,could not contribute to a study of radiological change over the first three surveys,and their time records were dispensed with. Any Singletons attending PFR4, 5 or6, or Follow-up Survey, were therefore asked to complete an occupational historyquestionnaire. The resulting data were used to fill gaps in their time records forISPs 0, 1 and 2.

2.2 Men participating in the Present Study

The present study evolved from earlier PFR-based research into coalminers"mortality - see, for example, Miller and Jacobsen (1985). These authors studied31647 miners who attended PFR1 at 24 of the 25 participating collieries wheresecond and third survey rounds were later conducted. (PFR1 attenders comprisedabout 90% of the coalminers employed at the 24 collieries at time of survey.)

By 1970, identifying information for the 31647 men had been sent to the Office ofPopulation Censuses and Surveys (OPCS) in England and Wales, and the RegistrarGeneral's office in Scotland, in order that searches for their records could be madein the National Health Service Central Register (NHSCR). By methods which aredescribed in more detail below, the IOM was notified of deaths which hadoccurred in this group of men between PFR1 and the date of search. Details ofsubsequent deaths were supplied to the IOM as they occurred, and the notificationprocedure is still active at the present time (1991). Of the 31647 men, 15188who attended PFR1 at one of the 10 Phase 2 colllieries were included asparticipants in the present study. The restriction was necessary, becauseradioactivity data were available only for Phase 2 collieries. A further 4230 menwho attended PFR2, .but not. PFR1, at. the same 10 collieries were also included;their identifying information was sent .to OPCS and the Registrar General's officein 1988. Inclusion of this second group (giving a total cohort of 19418 men)boosted the power of the study, and also increased the proportion of men forwhom smoking information was available, since the PFR respiratory symptomsquestionnaire was not introduced until PFR2.

2.3 Mortality Data

Notifications to the IOM of deaths in the study population were accompanied bycopy death certificates, with the underlying cause of death coded according to theInternational Classification of Diseases (ICD) (World Health Organization [WHO]1957; WHO, 1967; WHO, 1977). Three revisions of the ICD came into forceduring the study period, the Seventh, Eighth and Ninth in 1958, 1968 and 1979respectively. However, the earliest deaths in the subset of the study groupconsidered for the present report - some exclusions were made because ofunreliable or missing data - occurred in 1958. It was only necessary, therefore,to express the causes of death of interest in terms of the Seventh, Eighth andNinth Revisions. These causes, with ICD codings, are given in Table 2.1.

With two exceptions (oesophageal and laryngeal cancer), excess mortality from eachof the causes in the Table has been found in previous studies of radon daughterexposure. The two extra causes were included because it seemed plausible thatthe anatomical sites concerned would be liable to some degree of exposure tocoalmine dust, and therefore, potentially, to radon and thoron daughter atomsattached to dust particles. It should be noted that, of all the causes listed, onlylung cancer has been demonstrably linked to radon daughter exposure. Evidenceof association derived from correlation studies in particular must be regarded withspecial caution.

2.4 Radon and Thoron Daughter Exposures

Men's exposures to radon and thoron daughters were estimated using data from twosources. First, the Attendance Records System and Occupational HistoryQuestionnaires of the PFR gave estimates of times worked within occupationalgroups and six broad classes of coalmining activity, respectively. These werecombined with data from the second source, a survey of radiation levels carried outat the 10 Phase Two collieries (and one other non-PFR colliery, whose data werenot used) during the 1970s and 1980.

During the early 1970s, a draft 'Euratom' directive on ionising radiation was inpreparation, and it was the need to assess the implications of this directive forBritish coalmines which prompted the radiation survey. One hundred and

8

sixty-one measurements of radon and thoron daughter levels were made at 11collieries between April 1972 and June 1980, of which 146 were at PFR collieries.A method developed by Ogden (1974) was used to obtain the measurements. Ithad the advantage of using the Mining Research Establishment's type 113Arespirable dust sampler, which was a routine PFR device. Furthermore, from asingle dust sample, estimates of both radon and thoron daughter levels could beobtained, these levels referring to shift averages, and not to specific points in timeduring shifts. (However, the averages were heavily weighted towards theshift-end. According to Ogden, the first five hours of a seven-hour shift had acumulative weight of only one quarter.)

The results of the'^survey-; were reported win detail by Crawford and Edlin (1982),and part of their data is ;reproduced in-this, report in Appendix 2. As well asboth radon and thoron - daughter levels, these authors gave the location undergroundwhere samples were obtained, the barometric pressure, weight of dust in thesample, length of shift, and the distance underground travelled by the ventilatingairstream from the foot of the colliery shaft to the sampling site (known as theventilation distance). Of these, only radioactivity levels, ventilation distance, andthe coal seam in which samples were obtained, are given in Appendix 1. Oneother variable is listed - the ventilation quantity (m3 sec"1) at the sampling site.This parameter was not measured by Crawford and Edlin; the quoted values arecoalface annual averages for the years in question. These were obtained fromannual PFR reports of environmental conditions at research collieries, and were notavailable for some of the sites where radioactivity samples were taken, and, ofcourse, for the non-PFR colliery (colliery H in Appendix 2).

2.5 Smoking Habit

A questionnaire on respiratory symptoms was clerk-administered at every round ofPFR surveys, except PFR1. A copy is given in Appendix 3. Five of the sixitems on smoking referred to current habit; the sixth was asked only of currentnon-smokers, to establish if they had smoked at any time prior to survey. Thequestionnaire therefore allowed the identification of three groups at each PFRsurvey: non-smokers (these men had also never smoked prior to survey),ex-smokers (current non-smokers, who had, however, smoked at some time priorto survey) and smokers (a group which included both cigarette and pipe smokers).In the latter group, further subdivision by amount smoked was carried out; detailsare given below.

Table 2.1 Causes of death examined in relation to radon andthoron daughter exposure.

Cause of death7th Rev.

ICD Coding

8th Rev. 9th Rev.

Malignant neoplasm of:

oral cavitysalivary glandoesophagusstomachlarynxtrachea, bronchus and

lungbone

Ma1i gnant me 1anoma

Malignant neoplasm of:

prostatekidney

Leukaemia (excludingchronic lymphoid)

140-148142150151161

162196

140-149142150151161

162170

140-149142150151161

162170

190

177180

172

185189

172

185189

204 204-207 204-208(excl. 204.0) (excl. 204.1) (excl. 204.1)

10

11

3. METHODS

3.1 Establishing Vital Status

Following dispatch to OPCS of identifying information on the members of the studygroup, searches for their NHS records in the Central Register were carried out byOPCS staff.

Copies of death certificates for men who had already died were dispatched to theIOM. Records of men who were alive were 'flagged1 with a symbol identifyingthe present mortality study. As these men died during the course of the study,the presence of flagged NHSCR records ensured that copy death certificates weresent to the IOM. The underlying cause of death appearing on certificates wascoded according to the ICD. Seventh or Eighth Revision codings were used fordecedents who had attended PFR1; otherwise, the Ninth Revision was used.

By August 1990, the vital status of approximately 4% of the study population wasstill unknown, or uncertain. An attempt was made to reduce this percentageusing the letter forwarding service of the Department of Social Security (DSS).Letters to selected study group members of unknown vital status were prepared,requesting that they notify the IOM of their current address. These were sent tothe DSS, who forwarded them, whenever possible, to addresses currently held atDSS. (The IOM were not given the addresses to which forwarded letters weresent.) In addition, dates of death were supplied to the IOM for men who haddied according to DSS records. As a result of this exercise, approximately 100 ofthese men were confirmed alive. Updated identifying information (ie. new orcorrected addresses) was sent to OPCS, so that a fresh search could be made fortheir records in the NHSCR. In statistical analysis, these men were regarded asalive up to 31/12/1989. Updated information on vital status (ie. dates of death)was also prepared, for men confirmed dead by DSS, and dispatched to OPCS.However, computer files of mortality information had been 'frozen' for purposes ofstatistical analysis before copy death certificates, with ICD codings, had begun toreach the IOM, and hence, in analysis, they were regarded as 'untraced', andexcluded.

The DSS were unable to forward letters to approximately 300 men, either becausethey could not identify them, or because they did not hold an address. TheNational Concessionary Fuel office of the British Coal Corporation were asked tosearch their own records, and also Pension Fund records, for these men, andwhere possible, to forward letters on the lOM's behalf. For cost reasons, thesearch was limited to a random sample of 50 men, in the first instance. Thiswork is currently in progress. It has not been possible to include in the presentanalysis any of the results of these searches.

According to OPCS, about one-fifth of the 4% of the study group whose vitalstatus was unknown at August 1990, had left the UK. Dates of embarkationwere available, but it has not been possible to process these for the presentanalysis. It would have been preferable to include the men, and regard the dateof embarkation as a date of loss to follow-up. They have, however, beenexcluded from statistical analysis.

12

3.2 The Calculation of Radon and Thoron Daughter Exposures

3.2.1 Outline of the method of calculation

The total time worked in coalmining by each man in the study group withtime-worked information, was partitioned into several 'risk categories'. Averageradon and thoron daughter levels were assigned to each category, and cumulativeexposures calculated by summing the products: average level by time worked, overall categories. This calculation was carried out by ISP, so that men's cumulativeexposures were finally available in roughly five-year "packages', corresponding toISPs. (However, ISPO exposures were of variable calendar-time duration,depending on the date prior to PFR1 when men began work in coalmining.)

3.2.2 Definition of risk categories

Attendance Records time at each colliery (i.e. time worked within PFROccupational Groups, and recorded in the Attendance Records System) waspartitioned first into surface and underground time. The latter was furthersubdivided by coal seam. Two other underground categories introduced were:'all seams', which included time spent in mobile jobs, not specific to single seams;and "pit bottom', which covered time spent in jobs at or near the foot of themain shaft. The complete list of these categories is given in Table 3.1. (Thedecision to calculate exposures from working times within coal seams is discussed inAppendix 4, together with some related issues.)

Occupational History time (i.e. men's own estimates of time worked given atinterview during routine PFR surveys, or at Follow-up Survey) was divided intofour categories: time worked underground at the research colliery (the collierywhere the survey was being conducted); time worked on the surface at theresearch colliery; time worked underground at non-PFR collieries; and timeworked on the surface at non-PFR collieries.

3.2.3 Assignment of radioactivity levels to risk categories

Attendance Records time allocated to a seam where radioactivity measurements hadbeen made was assigned the mean radon and thoron daughter level for the seam.Time in seams with no measurements was assigned the mean level for the researchcolliery. The two remaining categories of underground Attendance Records time(i.e. 'all seams' and 'pit bottom') were also assigned the colliery mean.Attendance Records time spent on the surface was assigned estimated outdoorconcentrations of radon and thoron daughters, the same value of each being usedat all 10 collieries. Estimates were provided by the National RadiologicalProtection Board (NRPB). It was considered unnecessary to distinguish indoorfrom outdoor surface work, since most 'indoor' industrial jobs would have beencarried out in well-ventilated sheds. Table 3.1 also gives the radon and thorondaughter levels specific to Attendance Records time categories.

Time worked underground and on the surface at the research colliery which wasrecorded on Occupational History questionnaires, was assigned the colliery meanlevel and the estimated outdoor concentration, respectively. Time worked atnon-PFR collieries was treated in the same way, the rationSle being that many ofthe men "migrating in' to PFR collieries would have come from neighbouringcollieries working the same coal seams.

13

3.2.4 Partitioning of Attendance Records time

Times worked by individual men within Occupational Groups were available oncomputer file, by ISP. For ISPs 1 and 2, the data were held as total timeswithin ISP; times worked during ISPs 4 to 7 were available quarterly. ISP3times were available quarterly for some Occupational Groups, and as ISP3 totalsfor the remainder. Before partitioning times into risk categories, quarterlyinformation was totalled within ISPs 3 to 7 to give a uniform structure to the data.

Extensive PFR documentation (the so-called "History of Faces') gave thecorrespondence between Occupational Groups (between 342 and 914 per colliery)and seams worked, during PFR. These data were computerized, and programswritten to accumulate time worked in seams.

Some Occupational Group codes appearing in the computer files of AttendanceRecords time were not listed in the History of Faces. Further investigation ofPFR documentation, and discussions with former PFR Investigators (staff who hadbeen stationed at research collieries to oversee dust sampling) established that someof these unlisted codes had been devised locally to cover such contingencies as'away on course at Technical College' or 'temporarily attached to another works'.However, there remained a few codes which could not be accounted for, and timeworked in these Occupational Groups was allocated to an 'unknown' category.

Units of measurement on Attendance Records files were 'numbers of normal shifts';overtime was recorded separately, as hours worked. Conversion factors from shiftsto hours were 7.5 for ISPs 1 and 2, and 7.25 for ISPs 5, 6 and 7, at allcollieries. There was no single date on which the shift length was reduced at allcollieries; 1st January 1973 has been taken as representative. The conversionfactor for shifts worked in ISPs encompassing this date (ISP3 or ISP4 at allcollieries) was taken as the mean of 7.5 and 7.25, weighted according to theproportions of the ISP falling before and after 1/1/73.

3.2.5 Partitioning of estimated working times derivedfrom Occupational Histories

These data were held on computer file as times worked within six categories ofcoalmining activity, by ISP. For the present study's purpose, the categories werecollapsed into two: time underground, and time on surface. Also, time workedat PFR collieries was distinguished from time at non-PFR collieries, thus allowingworking time to be partitioned into the four risk categories already mentioned.

All ISPO times, and times in later ISPs which had been obtained at Phase 2surveys or at Follow-up Survey in order to fill gaps caused by the disposal ofSingletons' data, were held on computer file in units of years worked. A factorof 1740 working hours per year was used to convert to hours (48 weeks of 7£hour shifts, with eight public holidays). Times in ISP1 and ISP2 which wereobtained at Phase 1 surveys (which referred exclusively to non-PFR collieries) wereheld on computer file in units of hours worked. No conversion was required.

3.2.6 Calculation of exposures - an example

Details of the calculations' described above are presented here for a randomlyselected Study Group member at Colliery F.

14

Table 3.2 shows times worked by this man within Occupational Groups, by ISP,essentially as held on computer file. (Most men in the study did not have sucha long employment record; the example was chosen with this feature in mind.)Total times worked (column 5) were calculated by applying a conversion factor toNormal Shifts (column 3) and adding overtime, if any. ISP3 at this collieryended in April 1972; shift lengths of 7.5 and 7.25 hours were assumed for ISP3and 4 respectively. The right hand column shows the risk categories into whichtime worked was allocated. Table 3.1 gives a key to the codes.

For this man, Occupational History time was recorded only for ISPO. Table 3.3shows that he accumulated time in two of the four available risk categories duringthis period. The results -of accumulating Attendance Records time within riskcategories, and combining with. Occupationalr History time, are shown in Table 3.4.The final step in exposure calculation was to multiply times within risk categoriesby the assigned radon and thoron daughter levels, shown in Table 3.1. Resultsare given in Table 3.5, both by ISP, and cumulatively.

3.3 Smoking Habit

For the present statistical analysis, men's smoking habits at each PFR survey atwhich they provided valid data were coded according to the following scheme:

Smoking habit Code

Non-smoker 1Ex-smoker 2Pipe smoker 3

Cigarette, or cigarette and pipe smoker:

Equivalent of 1 to 5 cigarettes per day 4" 6 to 10 " " " 511 11 to 20 " " " 611 21 to 30 " " " 711 31 to 40 " " " 8" over 41 " " " 9

Note that, since a code was calculated for each survey at which a man providedinformation, his code could change throughout the follow-up period. Details ofthe implementation of the coding scheme are given in Appendix 5.

3.4 Definition of Risk Period

Only men with adequate radon and thoron daughter exposure data, and data onsmoking habit, were included in analyses of mortality; and this restriction hadimplications for setting the starting date of the risk period for individual men. Ingeneral, a man entered the period of risk at the earliest PFR survey by which allthe non-mortality information essential for his inclusion in the analysis had beengathered. The implications are considered first for time-worked information, andthen for smoking.

The fact that Singletons' time-worked information gathered during Phase One wasdiscarded prior to computerization, has already been mentioned. However, it was

15

possible to include some Singletons in analyses. At three of the 10 researchcollieries (P, Q and V), computerization of time-worked information gathered inPhase One was in fact completed before the decision to discard data was made.Singletons' time information at these collieries was therefore available, and theywere, in principle, considered to be at risk of dying following their firstattendance. Singletons included in analyses from the other seven collieries werethose who subsequently attended a Phase Two survey. At this second attendance,an Occupational History was taken, upon which estimates of time worked duringISPO, 1 and 2 were based. These men were therefore considered at risk of dyingin the period after their earliest Phase Two attendance. Non-Singletons atcollieries other than P, Q and V were considered at risk following their secondattendance in Phase One, . since it was the fact of their second attendance whichguaranteed the retention of-their, time-worked, data, and hence their inclusion.

The requirement that smoking information be available for men included inmortality analyses implied that they were not considered at risk of dying until theirattendance at a PFR survey at which a respiratory symptoms questionnaire wasadministered. In general therefore, the start of follow-up was no earlier thanPFR2. At colliery P, where an administrative error resulted in no valid smokingdata being obtained at PFR2, the start was no earlier than PFR3. For somemen - for example, Singletons who attended PFR1, or any Singletons at colliery P- the start of follow-up was the earliest attendance in Phase Two.

These considerations are summarized in Table 3.6, which shows dates of entry forvarious subgroups of the study population.

Dates of exit were either dates of death, or, for survivors, the end of the studyperiod, 31/12/89.

3.5 Statistical Methods

3.5.1 Person-years at risk

Associations between radon and • thoron daughter exposure and mortality from thetwo commonest cancers found in the study population, lung cancer and stomachcancer, were investigated first by the person-years-at-risk method (Breslow andDay, 1987). A description of the implementation of the method in the presentstudy follows.

Each calendar year in which each member of the study group was at risk of deathwas added to the cells of two tables of person-years, one for radon daughterexposure, the other for thoron. Person-years tables were five-dimensional, andwere indexed by calendar time, age, smoking habit, colliery of employment andexposure (radon or thoron). Only one aspect of radon and thoron daughterexposure history was used in analyses, namely cumulative exposure, considered as atime-dependent variable, and lagged by 10 years. No attempt was made toestimate a cumulative exposure to tobacco smoke, since respiratory symptomsquestionnaire items referred to current smoking habit only, apart from a singlequestion to identify ex-smokers. Current smoking habit was therefore used inanalyses as a time-dependent variable lagged by five years.

Tabulation for a given man was carried out by determining values of the fiveclassifying variables pertaining at the start of each calendar year during which hewas at risk. For example, tabulation of year 1970-71 required determining his

16

age on 1/1/70, his cumulative exposure to 1/1/60, and his smoking habit on 1/1/65.Colliery of employment was not time varying, and calendar time required nocalculation. Categorized versions of these variables were then derived, and theperson-year added to the appropriate cell of each of the two tables. Table cellswere defined thus:

Calendar time: 1954-59, 1960-69, 1970-79, 1980-89;

Age: -34, 35-44, 45-54, 55-64, 65-74, 75-;

Lagged smoking habit: non-smoker, ex -smoker, pipe smoker, cigarette smoker1-10 per day, 11-20 per day, more than 20 per day;

Colliery: C, F, K, P, Q, T, V, W, X, Y;

Lagged cumulative exposure toradon or thoron daughters (WL hour): 0-, 25-, 50-, 100-, 200-, 400-,

800-.

Tabulation of part -years was carried out only at the beginning and end of theperiod. Values of classifying variables were those pertaining at the date of entry,and 1st January of the calendar year of exit, respectively.

Given the structure of the exposure data (Table 3.5 shows a typical record ofexposure), it was necessary to interpolate to obtain estimates of cumulativeexposure to arbitrary time points between dates of entry and exit. Linearinterpolation gave estimated cumulative exposure to the start of any calendar yearwithin the record of exposure. Cumulative exposure to dates prior to the earliestdate in the record of exposure was not estimated; thus, person-years were notallocated during the first 10 years of any man's record of exposure. Cumulativeexposure to dates later than the end of the record of exposure was assumed equalto the final, highest, value in the record of exposure.

Similar procedures were used to estimate men's smoking habit, as one of ninecategory-codes, at arbitrary time points. For calendar years between PFR surveysat which valid questionnaire responses were obtained, data from the nearest surveyprovided estimates. Smoking categories for years before the earliest, or after thelatest attendance at a PFR survey, were assumed equal to those pertaining at theearliest or latest attendance respectively.

Numbers of deaths from the two causes of interest were also tabulated, by thesame variables classifying person-years tables. Lagged cumulative exposures andsmoking categories were estimated at the start of the calendar year in which deathoccurred. Division of tables of numbers of observed deaths by tables ofperson-years, gave tables of cause-specific death rates.

The joint influence of exposure (radon daughter or thoron daughter) and otherfactors, such as age and smoking, upon death rates, was examined by fittingstatistical models to tables of observed deaths and person -years. Numbers ofdeaths in table cells were assumed to follow the Poisson distribution with meansequal to the product of the underlying death rate for the cell (per 100000person-years) and the person-years-at-risk. Underlying death rates were assumedto be linear functions of explanatory variables, on the natural log scale. The

17

discrepancy between within-cell death rates predicted by statistical models, andthose observed, was measured by the so-called deviance statistic, given by theexpression,

Deviance = 2 £ [y log ^ - (y - /*) j

where y denotes numbers of deaths within table cells, and fi the number of deathspredicted by the regression model. Summation is over the cells of theperson-years table. Explanatory variables were almost always treated ascategorical; for example radon daughter exposure was represented by six binaryvariables indexing categories ^25, ^50, ^100, ^200, ^400, ^800 WL hour.Interactions between, explanatory variables were ^represented by products of binaryindicators. The statistical significance of the effect of a variable, or aninteraction, on death rates was assessed by computing the reduction in the devianceobtained by adding the variable to an existing regression model. Under thehypothesis of no effect, this reduction was assumed to have a Chi-squaredistribution. Degrees of freedom were given by the number of binary indicatorsrepresenting the variable or interaction.

The model-fitting strategy adopted was first to fit an additive model (ie. maineffects only) comprising the stratifying variables age, calendar time, lagged smokinghabit and colliery. All six two-factor interactions were then added, singly, andany found to be significant at the 5% level was included. Then, main effects forlagged radon or thoron daughter exposure, considered as categorical variables, wereadded; where predicted death rates, on inspection, showed trends with exposure,these were tested by fitting one-degree-of-freedom terms. For this purpose,values were assigned to each exposure category, equal to the mean exposure overall person-years (or part-years) allocated to table cells in the category. Effectsof two-factor interactions between exposure and stratifying variables, taken one at atime, were also investigated. In these tests, exposure was considered both as acategorical variable, and as a continuous variable.

Two tables are presented for each cause of death considered in relation to eachexposure: an analysis of deviance table, and a table of predicted death rates.The latter shows death rates estimated by statistical models which contain stratifyingvariables, and significant interactions between them, together with exposurevariables. To display the effect of a variable or interaction, other variables inthe model were given arbitrary values. These values were:

Calendar time: 1970-79

Age: 55-64

Lagged smoking habit: 1-10 cigarettes per day

Colliery: C

Lagged cumulative exposure toradon or thoron daughters (WL hour): 0-25.

For example, in models with no interaction between stratifying variables andexposure, the effect of exposure was summarized by a one-way table of predictedrates, other variables being given the arbitrary values listed above.

18

The person-years method was also used to compare death rates in the study group,from all causes and from lung cancer, with those in the general male populationsof the geographical regions in which the research collieries were located.Regional population figures, and numbers of deaths, both by 10-year age group,were obtained for each year of the risk period from OPCS publications (1974onwards), Annual Reports of the Registrar General for England and Wales(pre-1974), and Annual Reports of the Registrar General for Scotland. Fromthese data, general population death rates specific to age, calendar year, andregion, were calculated. The total person-years lived by the study population wastabulated, exactly as described above, but by age, calendar year, and researchcolliery only (smoking and radon or thoron daughter exposure were not used inthis classification). . Numbers of. deaths expected in the study population ifregional rates had applied,; were calculated: by multiplying person-years at riskwithin table cells, by the death rate appropriate to the age, calendar year, andcolliery classifying the cells, and summing over all cells. Division of the observednumber of deaths by the total expected yielded standardized mortality ratios(SMRs). Age-specific SMRs were also calculated; for this purpose, numbers ofobserved deaths were tabulated by age at death. Statistical significance of SMRswas assessed by noting whether 95% confidence intervals enclosed unity. Gardnerand Altman (1989) discuss the construction of confidence intervals for SMRs.

3.5.2 Case-referent studies

Associations between radon and thoron daughter exposure and mortality from all 11causes of death listed in Table 2.1 were investigated in a series of case-referentstudies. The idea of selecting cases and referents from a larger cohort was firstsuggested by Thomas, in an addendum to a paper by Liddell et al (1977); a fullaccount is given by Breslow and Day (1987).

Men who died (cases) from one of the causes of death being investigated weregrouped into strata according to values of four variables: age at death (recordedas an integer); quinquennium of calendar time when death occurred; smokinghabit (coded as one of nine categories) pertaining at the birthday before death,lagged by five years; and colliery. Each stratum defined a set of potentialreferents, who were men. at risk at the age of stratum cases, and matched to themin respect of colliery, of employment, lagged smoking category (in nine categories)and calendar time (in quinquennia). For example, a stratum containing casesfrom colliery C who died aged 55 during 1960-64 and who smoked 6-10 cigarettesper day on their 50th birthday, defined as potential referents, all men from collieryC at risk aged 55 during 1960-64, who smoked 6-10 cigarettes on their 50thbirthday. Continuing the example, men who died aged 55 from a cause otherthan that of stratum cases were not excluded as potential referents. Neither weremen who died after age 55 from the cause being investigated. Similar rules tothose of the person-years tabulation were followed with regard to extrapolation andinterpolation of cumulative exposures and smoking habit. In particular, cases andpotential referents whose exposures could not be estimated were excluded fromanalysis.

Not all potential referents were included in case-referent comparisons. Referentsfor lung cancer and stomach cancer were randomly selected within strata,case-referent ratios being, respectively, 1 to 4, and 1 to 10. For other causes ofdeath, all potential referents were chosen.

Comparison between cases and referents was, with one exception, of cumulativeexposure to radon and thoron daughters, lagged by 10 years, to the last birthday

19

before the cases' age. The single exception was the comparison of leukaemiacases and referents, where a lag of only two years was used. Lung cancer casesand referents were also compared in respect of a combined measure of dose,derived from the formula (O'Riordan, NRPB, Personal Communication, 1989):

dose in milli sieverts (mSv) = 10 x lagged (10 years) radondaughter exposure (WLM)+ 3Va x lagged (10 years) thorondaughter exposure (WLM).

Associations between case-referent status and exposure (or dose) were investigatedfirst by descriptive methods.. Within each stratum, the difference between casemeans and referent means. was computed. 'The unweighted average of thesedifferences over all- strata was then calculated, with an estimate of standard error,as follows:

d =

' • • • < » - ' [ ££^ ]where d denotes within-stratum difference, n denotes the number of strata, andsummation is over all strata. Tests of significance were carried out by referringd/s.e.(d) to the "t1 distribution on n-1 degrees of freedom. These statistics werealso computed for groups of strata defined by age, quinquennium, colliery ofemployment and lagged smoking category.

Analyses were also carried out using multiple conditional logistic regression analysis(Breslow and Day, 1987). Within strata, the logarithm of the odds of being acase was assumed to be a linear function of lagged cumulative exposure to radonor thoron daughters. Regression coefficients were estimated by the method ofmaximum likelihood; they measured the change in odds on the log scale expectedfrom a unit increase in exposure. (When death rates are comparatively low, asin the present study, ratios of odds are good approximations to relative risks.)Likelihood ratio tests were used to assess statistical significance.

Since cases and referents were matched, the effect of exposure was automaticallyadjusted for the effects, if any, of age, calendar time, colliery and smoking habit.Indeed, the effects of these variables on mortality could not be estimated in thisanalysis. However, it was possible to investigate interaction between exposure andstratifying variables. For example, eight binary variables were defined, indexingthe smoking categories, ex-smoker, pipe smoker, cigarette smoker 1-5 per day,6-10, 11-20, 21-30, 31-40, and over 40. Products of these binary variables withexposure were computed, and a regression model fitted, which included an exposureterm, and all eight product terms. A likelihood ratio test of the joint significanceof the coefficients of the product terms, allowing for an overall effect of exposure,yielded a test of significance of the interaction of smoking with exposure. Thecoefficient of the exposure term yielded the relative risk per unit of exposure, fornon-smokers. Addition of this coefficient to that of any product, yielded relativerisks within other smoking categories. Interaction between exposure and collieryof employment was investigated similarly.

To examine possible variations in relative risk with age and calendar time, acategorized version of age in five-year bands, and categorical quinquennia, were

20

used instead of the one-degree-of-freedom terms, integral age and mid-year ofquinquennia. This allowed an exploratory analysis of interaction, and avoided theassumption that log odds ratios per unit increase in exposure would vary linearlywith age or calendar time. If fitted odds ratios within categories showed evidenceof trends, their significance was tested using one-degree-of-freedom tests.

Interactions were investigated for four causes of death only: lung, stomach,oesophageal and prostate cancer. Conditional logistic analyses were carried outusing the PECAN program (Storer, 1984).

21

Table 3.1 Risk categories, by colliery, with alpha-numeric codes, numbers ofOccupational Groups (OGs) per category, and assigned mean levelsof radon and thoron daughters (mWL) and numbers ofmeasurements on which the means are based.

Colliery Risk category

C Unknown (U)Surface (SU)

*Warwickshire Thick Coal (W)

F Unknown (U)Surface (SU)Al 1 seams (A)Six feet (6)Nine feet (9)*Seven feet (7)Four feet (4)

*Nine feet lower (9L)Pit bottom (PB)

K Unknown (U)Surface (SU)Al 1 seams (A)*Flockton (F)*Beeston (B)Pit bottom (PB)

P Unknown (U)Surface (SU)All seams (A)*Parrot (P)Splint (S)

*Kailblades (K)South coal (SO)Coronation (C)Little coal (L)Smithy (SM)Pit bottom (PB)

Q Unknown (U)Surface (SU)Al 1 seams (A)*High main (H)*Waterloo (W)Second Waterloo (W2)High Hazel (Ha)Main Bright (M)Low Bright (L)Pit bottom (PB)

OGs

1516349

2419112772595658554

2813461213833

34204619516716217781192

222632473131375963

radon

0.544.00

_

0.540.980.980.981.070.980.800.98_

0.548.517.609.428.51

_

0.543.263.073.263.633.263.263.263.263.26_

0.5419.0520.306.9319.0519.0519.0519.0519.05

Number ofthoron measure-

ments

0.364.34 10

_

0.360.600.600.600.53 60.600.73 30.60

_

0.364.395.15 43.62 44.39_

0.363.713.42 63.714.30 33.713.713.713.713.71

_

0.366.666.87 504.33 36.666.666.666.666.66

22

Table 3.1 Continued

Colliery Risk category

T Unknown (U)Surface (SU)Al 1 seams (A)*Harvey (Towneley) (H)*Busty (BU)Brockwell (BR)

V Unknown (U)Surface (SU)All seams (A)*Meadow vein (M)Yard (Y)*01d coal (0)*Big vein (B)Pit Bottom (PB)

W Unknown (U)Surface (SU)All seams (A)Pumpquart (P)

*Big vein (B)Gras (G)

X Unknown (U)Surface (SU)All seams (A)*Dunsil (D)*Beamshaw (B)Barnsley (BA)

*Meltonfield (M)Fenton (F)Newhill (N)Barnsley and Dunsil (BD)Pit bottom (PB)

Y Unknown (U)Surface (SU)All seams (A)High main (H)Hut ton (HU)*Low main (L)C seam (C)*Yard (Y)Maudlin (M)Low main and Maudlin (LM)

OGs

1326397611276

202244361271371443

1516444425413

31203724227191127117482

141851168288110

131355

radon

0.546.465.148.676.46

_

0.542.000.802.002.053.002.00

_

0.541.451.451.451.45

_

0.541.070.622.501.070.701.071.071.071.07

_

0.545.315.315.315.905.315.225.315.31

Number ofthoron measure-

ments

0.363.503.42 53.63 33.50_

0.361.340.97 31.341.14 112.43 31.34

_

0.361.551.551.55 61.55_

0.363.072.98 44.05 23.072.53 33.073.073.073.07_

0.363.883.883.882.30 23.884.12 133.883.88

* Risk categories with measurements

23

Table 3.2 Times worked within Occupational Groups for a randomly selectedmember of the study group at colliery F.

I SP OCSN*

1 0004103510320102

2 10341035103202110173201910481045

3 103210351041105620201056105705601056

4 1056105610561056105610561056105610561056

5 10561056105610564444888810561056105610561056

Normalshifts

506.0230.09.0

467.7

128.0425.2358.04.09.024.027.0100.0

4.0273.010.047.12.0

610.410.05.0

74.5

298.470.053.061.357.147.632.043.053.881.8

15.030.237.949.05.02.057.045.027.052.057.0

Overt ime(hours)

035030

011100001

Total timetworked(hours)

3795176068

3538

960320026863068

' 180202751

3020487535315

45787538559

2163508384444414345232312390593

1092192753553614

413326196377413

Risk£Category

6666

PB6666U46

669PBSUPBPB6PB

PBPBPBPBPBPBPBPBPBPB

PBPBPBPBSUUPBPBPBPBPB

Table 3.2 Continued

24

ISP

6

*t

OGSN*

105610561056

105610561056

Occupational

Normal Overtimesh i f t s (hours)

47.010.010.0

49.752.160.1

Group Serial Number.Conversion factors from shifts to hours for

Total t i m e tworked(hours)

3417272

360378436

ISPs 3 and 4 were

Risk*Category

PBPBPB

PBPBPB

7.5 and 7.25respectively.See Table 3.1 for full names of risk categories.

25

Table 3.3 Times worked in ISPO within categories of coalmining activity, andobtained by Occupational History questionnaire, for a randomlyselected member of the study group at colliery F.

Category of Timecoalmining Years workedactivity worked in hours

Colliery F:

coalface, coal-getting 2.5 4350

coalface, non-coal-getting 0.3 522

Other collieries:

surface 0.4 696

RiskCategory

Under-ground(researchcol 1 iery)

Under-ground(researchcol 1 iery)

Surface(non-PFRcol 1 iery)

Table 3.4 Total times worked (1000s of hours) within risk categories, by ISP, for arandomly selected member of the study group at colliery F.

ISP

0

1

2

3

4

5

6

Occupational History Time: Attendance Records time*

Colliery F non-PFR collieries

Surface Under- Surface Under- 'Unknown1 Surface All '6' '9' '7' '4' '9L1 'PB1

ground ground seams

4.87 0.70

9.16

0.18 6.73 0.20 0.96

0.01 2.12 0.07 5.56

5.79

0.01 0.04 3.17

1.17

ro

* Full names of risk categories are given, in Table 3.1.

27

Table 3.5 Cumulative exposures to radon and thoron daughters (WL hours),by ISP, for a randomly selected member of the study group atcolliery F.

ISPExposures

Radon

0

1

2

3

4

5

6

5

8

7

7

5

3

1

.15

.98

.73

.60

.67

.13

.15

within ISP

Thoron

3

5

4

4

3

1

0

.17

.50

.73

.65

.47

.92

.70

Cumulative exposures

Radon

5

14

21

29

35

38

39

.15

.13

.86

.46

.13

.26

.41

Thoron

3

8

13

18

21

23

24

.17

.67

.40

.05

.52

.44

.14

Table 3.6 Start dates of follow-up, by subgroup of the study population.

Earliest survey attended

Col 1 iery

P

Q,v

Others

PFR1

Singleton

*Earliest Phase Twosurvey attended

tEarliest Phase Twosurvey attended

^Earliest Phase Twosurvey attended

Non-Singleton

First attendanceafter PFR2

Second attendancein Phase One

Second attendancein Phase One

PFR2

Singleton

Earliest Phase Twosurvey attended

PFR2

Earliest Phase Twosurvey attended

Non-Singleton

PFR3

PFR2

PFR3

* Three of these men, exceptionally, had their time worked in Phase One estimated froma Phase Two Occupational History.

f One man had his time worked in Phase One estimated from a Phase Two Occupational History.

f Attendance records in Phase One were computerized for one man.

29

4. RESULTS

4.1 Vital Status

The first searches in NHSCR for attenders at PFR1 were carried out in 1970, andfor attenders at PFR2, in 1988. By August 1990 there remained a group of 774men (4.0% of the study population) whose vital status was unknown, or uncertain.No trace in NHSCR had been found of 541 men; 138 men had emigrated; for72 men, a date of death obtained from colliery sources had not been confirmed byOPCS; and 23 men were the subject of ongoing inquiries.

Of the 613 men (541 + 72) either untraced or with an unconfirmed date of death,letters for forwarding were sent to the DSS for 612. (One man was omitted inerror.) Because of insufficient, or absent, identifying information at DSS, 297letters could not be forwarded; to the 172 which were, 118 replies were received.Of these, 113 were from the men who status was being sought. DSS alsoconfirmed that 143 of the 612 men had died, and supplied dates of death for allbut 17 of them.

As a result of this exercise, updated identifying information for the 113 men foundto be alive was sent to OPCS in April 1991; dates of death for 126 decedents(143-17) were sent in June 1991.

As discussed above in Section 3.1, attempts are currently being made to reducefurther the number of men of unknown vital analysis, using the records of theBritish Coal Corporation Pension Scheme. No results are yet available.

By the end of April 1991, when computer files of mortality information were'frozen' for statistical analysis, 762 men were of unknown or uncertain vital status,according to OPCS. (Routine inquiries, and searches for individual men, hadreduced the number from 774 during the interim). One hundred and thirteen ofthese men were then formally assumed to be alive as a result of DSS tracing,leaving a group of 649, made up as follows: 139 embarkees, 430 untraced inNHSCR, 56 with an unconfirmed date of death, and 24 the subject of continuinginquiry. These men made up 3.3% of the study population, and they wereexcluded from statistical analysis.

4.2 Exclusions from Statistical Analysis on grounds of Missingor Unreliable Data

As discussed in the preceding section, 649 of the 19418 men for whom searcheswere made in NHSCR were excluded from statistical analysis, on grounds ofunknown vital status. Of the remainder, five men whose reported dates of deathapparently preceded their final PFR attendance were excluded. Another man wasexcluded, whose information on time worked during Phase One was obtained byquestionnaire at a Phase Two survey, but who had no record of attendance duringPhase Two. Mortality data suitable for analysis were therefore available for 18763men.

Of the 19418 men who attended PFR1 and 2, smoking data were available for14417. Any men who reported .being a non-smoker at a PFR survey, having

30

previously reported being an ex- or current smoker, were excluded. Twohundred and seventy-two such exclusions left a group of 14145 men with validdata.

Fifteen thousand, three hundred and twelve attenders at PFR1 and 2 hadinformation on time worked. Exclusions from this group were as follows: 222men who had worked at least 10000 hours in risk categories with no assignedexposure level; 127 men with no record of time worked during any of ISPs 0, 1or 2; seven men whose recorded time worked during any ISP exceeded one halfof the duration of the ISP (i.e. they had apparently been working, on average, 12hours per day, every day, during the ISP). A group of 14956 men with validdata on time worked .remained after these exclusions.

The set of men with valid data on vital status, and smoking, and time worked inrisk categories with assigned exposure levels, numbered 12398. Furthercross-checks, between types of data, excluded 37 more men. Fourteen hadimpossibly large times worked in ISPO - their estimated ages at start of miningwere apparently under 7 years; nineteen men had attended a PFR survey, but hadno record of time worked in either the ISP preceding or following; and finally,the reported date of death for four men preceded a PFR survey at which they hadapparently provided information on smoking habit. Following this final step,12361 men had data suitable for analysis.

4.3 Summary Statistics for the Study Group

The median year of entry to the study for the 12361 men with validated data was1960. The frequency distribution is given in Figure 4.1; the minimum year,lower quartile, upper quartile, and maximum were 1958, 1959, 1962 and 1978respectively. The earliest median entry-dates were at collieries Q (1959) and V(1958) where Singletons' data were processed (Table 4.1). Singletons' data werealso processed at colliery P, but the absence of valid smoking data from PFR2 atthat colliery implied a later median date of entry (1964). Approximately 400men entered in 1970 or later; information on time worked in Phase One forthese men was estimated from an Occupational History obtained at a Phase Twosurvey.

The median age at date of entry was 44, varying from 38 at colliery Q, to 50, atcolliery C. The age distribution (minimum age, lower quartile, upper quartile,maximum = 15, 34, 53 and 87 respectively, see Figure 4.2) shows approximately100 men entering the study after their 65th birthday. For these men, the riskperiod began at Follow-up Survey, which they attended after leaving the coalminingindustry.

The most populous colliery was Y, with 1447 men participating; the leastpopulous, W, with 654 men. The total number of men at risk during the studyperiod increased from approximately 1700 in 1959 to 11000 in 1968, and haddeclined to 6500 by the start of 1989 (Figure 4.3).

The proportion of non-smokers (i.e. men who did not smoke, and had neversmoked) observed at each of PFR2 to 6 was roughly constant, at between 12% and14% (Table 4.2). The proportion of ex-smokers observed increased from 6% atPFR2 to 22% at PFR6, and it appears that the new ex-smokers were recruitedmainly from men smoking 6-10 cigarettes per day, and, to a lesser extent, 11-20per day. However, a comparison of the numbers of men in Table 4.2 withnumbers of men at risk at the same dates (Figure 4.3) shows that those giving

31

information on smoking habits formed only a part of the alive and at-risk cohort,at any stage of the study period. At PFR4, 5 and 6, they were the minority.Thus, the results of Table 4.2 do not provide a complete description of theevolution of the cohort's smoking habit with time. A complete description can beobtained for a sub-cohort of 'long-term survivors' in the industry. Table 4.3shows equivalent data for 2942 men who attended PFR2, 3, 4 and 5. The samepatterns are evident, with the trends being slightly less marked.

Mean estimated cumulative exposures to radon and thoron daughters, to end datesof ISPs (or dates of death), are shown in Table 4.4, by ISP (0 to 7). Theseaverage values are very low. For example, 3832 men who were still working inthe research collieries~ during. ISP5, which was the latest ISP for which theAttendance Records System was maintained^ at all 10 collieries, had an estimatedmean cumulative exposure to radon daughters, at that time (1978-80), of 268 WLhour (1.58 WLM). Their average thoron daughter exposure was 167 WL hour(0.98 WLM). Although average exposures were low, there was considerablevariation in the individual values. Boxplots, summarising the distributions ofcumulative exposures to end-dates of ISPs (Figures 4.4 and 4.5) showed markedlyskewed distributions for all ISPs, except 7. For ISPs 1 to 6 the upper quarter ofthe distributions of radon daughter exposure ranged from 200-300 WL hour to1500-1800 WL hour. For thoron daughters, the range was 150-250 WL hour to500-650 WL hour. (The 'box' in these plots indicates quartiles and median;lines above and below the box extend to the upper and lower 5% points.Individual data points in the tails of the distribution are shown by dots; in Figures4.4, 4.5, 4.8 and 4.9 these have merged to give continuous heavy lines.)

The distributions of total cumulative exposure to radon and thoron daughters for all12361 members of the study group (i.e. cumulative exposure to the end-date ofthe latest ISP in which time was worked, or to date of death) showed, asexpected, the long tails already seen in the boxplots of ISP-specific exposures(Figures 4.6 and 4.7). However, when the same exposures were viewed bycolliery (Figures 4.8 and 4.9) the skewness disappeared. These boxplots alsoillustrate the large differences between collieries in total cumulative exposure.

Results relevant to considerations of exposure reliability are given in Appendix 6.

4.4 External Comparisons of Mortality - All Causes of Death,and Lung Cancer

There were 5852 deaths in the study group of 12361 men with valid data, betweendates of entry and the end of the study (31/12/1989). Five hundred andtwenty-one of these were from lung cancer. Comparison with regional deathrates gave an all-causes SMR of 96, and a lung cancer SMR of 87 (Table 4.5).The latter was statistically significantly lower than 100; the 95% confidenceinterval was 80-95. All causes SMRs within ten-year age categories variedbetween 91 and 114, lung cancer SMRs between 80 and 111.

4.5 The Relationship between Radon and Thoron DaughterExposure and Cause-Specific Death Rates

4.5.1 Lung cancer

Lagged radon daughter exposures could not be estimated for 15 of the 521 lung

32

cancer deaths, which occurred within 10 years of the start of exposure records.Analyses were therefore based on 506 deaths. Figure 4.10 shows lung cancerdeath rates per 100000 person-years by radon daughter exposure category and ageat death. (Five men under 45 died of lung cancer; rates for this age grouphave not been plotted. Tables of person-years and numbers of deaths from lungcancer and stomach cancer are given in Appendix 7.) There is an indication ofan increasing trend in men aged over 75, although error bars are wide, butotherwise there is no evidence of an exposure response relationship. Regressionanalysis of death rates upon age, smoking habit, calendar time period and collieryshowed that the effects of these factors were statistically significant. Of all sixtwo-factor interactions, only that between age and calendar time was statisticallysignificant. After allowing. ,for these veffects, -differences in death rate betweenradon daughter exposure categories were not statistically significant (Chi square (X2)on 6 df = 1.83, see Table 4.6). Furthermore, none of the interactions betweenexposure and the four stratifying variables was statistically significant. Death ratesestimated by the statistical model summarized in Table 4.6 show the effects of ageand smoking (see Table 4.7). The lack of any trend with exposure is clear.

The results of a comparison of men who died from lung cancer (cases) to menstill alive at the same age (referents) are shown in Table 4.8. Thirteen casescould not be matched, leaving 493 cases for analysis. The average case-referentdifference was negligible (-1 WL hour), and differences within age and calendartime groups were small compared to standard errors, with no evidence of anytrends. Differences within collieries were also unremarkable, the only statisticallysignificant result occurring at colliery Y, where cases had 35 WL hour moreexposure than referents ('t' on 68 df = 2.24). Non-smoking cases and cases whosmoked fewer than 10 cigarettes per day had higher exposures than referents: thedifference for smokers of 6-10 cigarettes daily (54 WL hour) was 2.9 times itsstandard error. In the higher smoking categories, and in pipe and ex-smokers,the sign of the difference was reversed. Table 4.9 shows results of conditionallogistic regression analysis. The effect of radon daughter exposure overall was notstatistically significant. There was also no indication of a varying effect by age,calendar time, or colliery. However there was evidence of a differential effect ofexposure by smoking category (P < 0.05). Fitted relative risk parameters (Table4.10) showed a slight tendency to increase with age, but a test for trend was notstatistically significant. The apparent decreasing trend with calendar time was alsonon-significant. Relative risks were raised for non- and light smokers (i.e. fewerthan 10 per day), but only significantly so for smokers of 6-10 per day. Inhigher smoking categories, increased radon daughter exposure was apparentlyassociated with reduced risk of lung cancer mortality.

Lung cancer death rates per 100000 person-years by thoron daughter exposure(lagged by 10 years) and age at death are shown in Figure 4.11. (There wasonly a single person-year of observation in the highest exposure category at age45-54, and the corresponding point has not been shown in the Figure.) In theoldest and youngest age groups, rates increase with increasing exposure, but thereis no indication of any trends in other groups. Regression analysis (Table 4.11)showed that differences between exposure categories were not statistically significant(X2 on 5 df = 0.50), after allowing for the effects of age, smoking, calendar timeand colliery. First order interactions between exposure and stratifying variableswere also not significant. The death rates predicted by the model of Table 4.11are shown in Table 4.12. Rates vary only slightly with thoron daughter exposurecategory, and there is no sign of any trend.

Comparisons of thoron daughter exposure between men who died of lung cancerand their matched referents are given in Table 4.13. The overall difference was

33

small (2 WL hour), and there was little variation by age or calendar time, apartfrom a relatively large difference (37 WL hour, 't' on 30 df = 2.54) for men aged80 years and over. Within collieries, differences were well below two standarderrors, except at colliery Y, where 26 WL hour more exposure was recorded forcases ('t1 on 61 df = 2.36). The variation across smoking categories showed apattern previously observed for radon daughter exposure: positive differences fornon- and 'light1 smokers becoming negative in the higher smoking categories.Conditional logistic regression analysis (Table 4.14) showed that, overall, thorondaughter exposure did not have a statistically significant effect upon lung cancerdeath rates. There was also no evidence that relative risks varied between agegroups, calendar time periods, collieries or smoking categories. Fitted relative riskparameters (Table 4.15) .tended to increase with age, and decrease with calendartime, but neither of these trends was statistically significant.

Cases and referents were also compared in respect of a combined dose measure ofradon and thoron daughters. The mean dose for both cases and referents was19.7 mSv (Table 4.16), which would represent approximately 15 years absorption ofthe estimated mean effective dose equivalent per year received by the UKpopulation from indoor and outdoor sources, according to a recent report (Wrixonet al, 1988). Case-referent differences within age groups, calendar time groups,and collieries were not statistically significant, except at colliery Y, where cases hadabsorbed 2.6 mSv more than referents ('t' on 68 df = 2.27). The pattern ofdifferences by smoking category was, not surprisingly, similar to that observed forboth exposure indices - positive differences among non- and light smokers,negative in heavier smokers. The largest difference observed was for non-smokers(7.4 mSv); but only the difference for smokers of 6-10 cigarettes per day (3.6mSv) was statistically significant ('t1 on 79 df = 2.85). Conditional logisticregression analysis (Table 4.17) showed that, overall, dose did not have astatistically significant effect upon lung cancer risk. There was also no evidenceof differential effects of dose with age, calendar time period, or colliery; but theinteraction with smoking habit was statistically significant (Chi square on 8 df =17.47, P < 0.05). Relative risks exceeding unity were estimated fornon-smokers, and smokers of 1-5 and 6-10 cigarettes per day (Table 4.18), butonly in the latter category was the risk significantly raised (1.36, 95% confidenceinterval 1.04-1.77).

4.5.2 Stomach cancer

There were 219 deaths from stomach cancer; exposures could not be estimated for18. Death rates per 100000 person-years are shown in Figure 4.12, by age andradon daughter exposure category. (One man under 45 died from stomachcancer; rates for this age group have not been plotted.) There are noindications of increasing trends in death rate with increasing exposure. As in thelung cancer analysis, the statistical significance of differences between exposurecategories was assessed after allowing for possible effects of age, smoking habit,calendar time and colliery. The results are summarized in Table 4.19. Ageand calendar time both had statistically significant effects, and there was evidenceof real differences between smoking categories, which varied with calendar time (X2

on 10 df = 19.02, P < 0.05). After allowing for these effects, differencesbetween radon daughter exposure categories could have arisen by chance (X2 on 6df = 9.72). Also, none of the interactions between exposure and the fourstratifying variables was statistically significant. Estimated death rates (Table 4.20)showed an increasing trend (with fluctuations) with increasing exposure, but a testfor trend was not statistically significant (X2 on 1 df = 0.04).

34

Comparisons in lagged radon daughter exposure between men who died of stomachcancer (cases) and their matched referents are given in Table 4.21. Of 201 caseswith estimated exposure, four could not be matched. The average differencebetween all cases and referents was negligible (-4 WL hour). Differences withinage and calendar time strata did not show any trends, and with one exception,were not statistically significant (3 cases who died in 1960-64 had on average 46WL hour more exposure than 23 matched referents; 't* on 2 df = 5.61, P <0.05). Differences within collieries were likewise unremarkable. However, therewas evidence of variation with smoking habit: cases who smoked 11-20 cigarettesper day had 37 WL hour more exposure than referents ('t1 on 61 df = 2.10), andex-smoking cases 42 WL hour more ('t' on 27 df = 2.42). Other differenceswere not significant, and, with one exception, negative. Logistic regressionanalysis (Table 4.22) showed that overall, there was no statistically significant effectof radon daughter exposure, and -no evidence of interaction with age, calendartime, colliery or smoking habit. Regression coefficients are given in Table 4.23.

Figure 4.13 shows stomach cancer death rates by lagged thoron daughter exposureand age at death. Increasing trends with exposure are discernible, to someextent, in the two oldest age groups, but regression analysis showed that differencesbetween exposure categories could have arisen by chance (X2 on 5 df = 4.33, seeTable 4.24). Also, none of the interactions between exposure and stratifyingfactors was statistically significant. Fitted death rates within exposure categories(Table 4.25) showed a slight tendency to increase with increasing exposure, but atest for trend was not statistically significant (X2 on 1 df = 1.72).

Case-referent differences were examined by age, calendar time period, colliery andsmoking habit (Table 4.26). The difference over all matched sets was small (2WL hour), and no age group, calendar time period or colliery showed statisticallysignificant differences. There was also no evidence of trends with time-relatedfactors. Cases who smoked 11-20 cigarettes per day had 17 WL hour moreexposure than referents ('t1 on 61 df = 2.02), but differences in other smokingcategories were not statistically significant. Conditional logistic regression analysis(Table 4.27) showed no statistically significant effects. Fitted parameters are givenin Table 4.28.

4.5.3 Other causes of death

Case-referent differences in radon and thoron daughter exposure for eight othercauses of death are given in Table 4.29. Exposure could not be estimated forone of the 17 cases of buccal cavity cancer, three of the 42 oesophageal cancercases, two of the 79 prostate cancer cases, and the single case of salivary glandcancer. In addition, one of the two cases of laryngeal cancer could not bematched, and seven of the 77 cases of prostate cancer with estimated exposures.Approximate 't1 statistics are large for bone cancer, but they are not statisticallysignificant (there were only two deaths from this cause). Otherwise, 't1 statisticsare all below critical values, the largest ones (negative) occurring for oesophagealcancer.

Conditional logistic regression analysis (Table 4.30) gave estimated relative risks ofdeath from oesophageal cancer of 0.66 (95% confidence interval 0.47-0.91) perWLM radon daughter exposure, and 0.47 (0.26-0.85) per WLM thoron daughterexposure. For other causes of death, there was no evidence of any associationbetween death rates and either exposure measure.

35

Variations in relative risk per WLM with age, calendar time and smoking categorywere examined for oesophageal cancer and prostate cancer. Differences inrelative risk of oesophageal cancer mortality per WLM between four 10-year agecategories were not statistically significant for both radon and thoron daughterexposure. However, fitted relative risk parameters showed a decreasing trend,which was significant for both exposures (P < 0.05). For radon daughterexposure, the estimated relative risk per WLM at age 55 was 1.40; risks declinedby an estimated factor of 0.58 per 10 years of age. (One man died from thiscause at age 41; the next deaths occurred at age 54.) The estimated relativerisk per WLM thoron daughter exposure at age 55 was 3.49; risks declined by anestimated factor of 0.28 per 10 years of age. Of the three factors considered,only age was shown -to influence-1 the relative risk of oesophageal cancer mortality.Differences between five calendar time periods and four smoking categories werenot statistically significant. Finally, - there was no evidence that the relative risk ofprostate cancer mortality varied with any of the three factors.

§

D

5000

4000

orcff.op

Men3000

i-f 2000U)

or*

Q.

1000

0 / /1960 1964 1968 1972 1976 1980

Year of entry

i

O55'

o

o

2000

1500 -

Vr-f

a.

Men1000

o>-h

(T>

o«-*

500

0

20 30 40 50 60 70 80

Age at date of entry

12000

10000

c3

1

8000

Men atRisk

6000

S.•<S

4000

2000

01960 1965 1970 975 1980 1985 990 1995

Year

c

g- DO

= 8(§.•0sr oCo r*

crr-f

5'v>

n

Ic5T

X•oo

D.O

Cumulativeexposure toradondaughters(WL hour)

2000

1800 -

1600 -

1400

1200

1000

800

600

400

200

00 3 4

ISP

7

"Tfl

<J5'

I 700

O. CO

«-ien -

a a

cr«-f

o'C/>

n

ie

nx•o

Cumulativeexposure tothorondaughters(WL hour)

600

500

400

300

200

100

o33 0 I I

0 7

ISP

§3000

2500 -

2000 -

o

Ic

«§fDX•oos

I

OQ

53

Men 1500 -

1000 -

500 -

00 500 1000 1500 2000

Cumulative exposure to radon daughters (WL hour)

§

3000

2500 -

<T

O3

2000 -

os.oc3c_5T

Men1500 -

S)

nx"8 1000 -

o33 500 -

OP

00 100 200 300 400 500 600 700

Cumulative exposure to thoron daughters (WL hour)

3

o

2 ao. 9Sio

c

n o3 3

g. »5' of

Cumulativeexposure toradondaughters(WL hour)

oc

X•ao

2000

1800

1600

1400

1200

1000

800

600

400

200

0

1i

K Q T

Colliery

V W X Y

/ \J\J

3c3 600VO

£» 5000 g3 -o. Cumulatives oD.F exposure to(S. «. thoron ^QQq | daughters•^ (WL hour)^ 0.

2. 1 300f l

8. 2005

1 100cwr^

1 n

i i i

•

I__

1

•

^f^

•I

1

•

1Lrp

i

t

-

-

•

-

Oin

IF K P Q T V W X Y

Colliery

<g. 1800

J-A

o

1500o o* f^O &3 C3 c 3IS OQ (to0. P4

(& fl>£3 i-t OO fi38 « g 12°o— X O>§'"0 "*1' 1 a Lung Cancer death" 3 8 rate (per 100000% o ^ person years) QQQ

8| 1

§" w "°

9. f" I 60°

3 ' oa °US

5: tn"? 300

I I I? *<o* n>

&3 &3

CO CA

-

Radon daughter exposurecategory (WL hour)1 0-2 25-3 50-4 100-5 200-6 400-7 800-

-

_

1 C

J^Wrk0 L^ —

(

C)

°TUi) I O C

cpc)

)(

)

o v? Radon daughter 1 2 3 4 5 6 7 1 2 3 4 5 6 7 1i exposure category

S g-^3 Age 45-54 55-64

)C

2

) c

3

(

•\)

4

(

5

(

)

)(

s

X

(

)

(,

()

•N,

)

(

)

-

-

-

)

-

_

6 J

6 7 1 2 3 4 5 6 7

65-74 75+

era

n

o aO W3 C

a »•

8 *" osr x

3*3

I*

o.rt

& Lung Cancer death3 rate (per 100000g person years)

s- ^ "o/D P n3 -i. . po >->K' 6) OO rt f^3 • O

O

l?i5'^3 *<

CT M6) «

|-

vO O$v s Thoron daughterexposure category.

Age

1800

1600

1400

1200

800

600

400

200

0

Thoron daughter exposurecategory (WL hour)123456

-

-

_

-

JL 1 mmCD (p H'H'

0-25-50-100-200-400-

C

)

) ()( 5 1 1(pY

f

C)

(

X

)(

(

)

)

C

— _ ,_,y . 1 2 3 4 5 1

45-54

C

)

)( )(

(

)

(

)

s,

2 3 4 5 6 1 2 3 4 5 6 1 2 3 4

55-64 65-74 75 +

)c

-

-

~

)

-

-

_

-

5 6

31<TQ°

§

not-hQ.'ftOO> s

BIMzo3

ften

S-oo. CB1-1

3

aooo

3. m^

O

en

I*

4vS> O§>• 3

1600

1400

1200

1000

Stomach Cancer deathrate (per 100000person years) 800

600

400

200

0

Radon daughter exposurecategory (WL hour)1 0-2 25-3 50-4 100-5 200-6 400-7 800-

-0 -e

C) C)

o

o

C)o

o

Radon daughterexposure category

Age

1 2 3 4 5 6 7

45-54

1 2 3 4 5 6 7

55-64

1 2 3 4 5 6 7

65-74

1 2 3 4 5 6 7

75 +

3

r>O3 C

P. C/5

~ O

CL 3" &j« f? 2.8 ":r. *3*8IOn

a.n>

ws•o

• { ?

8?§& s(/I

5 w-o5- rn os. oo i3

!!3-0-o •<

3"KO Ot^> 2

1600

1400

1200

1000

Stomach Cancer deathrate (per 100000person years) 800

600

400

200

Thoron daughter exposurecategory (WL hour)1 0-2 25-3 50-4 100-5 200-6 400-

Thoron daughter ^exposure category 1 2 3 4 5

°

00

O

Age 45-54

1 2 3 4 5 6

55-64

1 2 3 4 5 6

65-74

1 2 3 4 5 6

75 +

49

Table 4.1 Numbers of men, median years of entry, median ages atentry, by colliery.

Col 1 iery

C

F

K

P

Q

T

V

W

X

Y

Number ofmen

1394

884

1157

1155

1228

1350

1762

654

1330

1447

Year ofentry

1960

1962

1962

1964

1959

1960

1958

1961

1961

1959

Age at dateof entry

50

46

46

44

38

45

39

43

43

45

Al1 men 12361 1960 44

Table 4.2 Summary statistics on smoking habit, by PFR survey, for five overlappingsub-cohorts of the study group

Survey

*

t

2

3

4

5

6t

Date

1958-62

1964-68

1970-73

1973-77

1978

Includes

Col 1 iery

No . o f me ngivinginformation non-on smoking

10706

8629

4883

4755*

518

ex-miners

P, V only

14

12

13

13

13

seen at Fol

ex-

6

11

15

21

22

low-up

Smoking group (%)

pipe cigarettes per day

1-5 6-10 11-20 21-30 31-40

6 5 21 36 11 1

6 3 17 38 11 1

7 3 11 36 12 2

7 4 10 30 12 3

5 4 6 28 16 5

Survey

Meancigarettesper day

41-

0 10.8

0 10.8

1 10.7

1 9.9

1 11.2

I/Io

Table 4.3 Summary statistics on smoking habit, by PFR survey, for 2942 menproviding smoking information at PFR2, 3, 4, 5

Survey Date

2 1958-62

3 1964-68

4 1970-73

5 1973-77

non-

16

14

14

13

ex-

5

11

15

20

pipe

5

5

6

7

Smoking group (%)

cigarettes per day

1-5

4

4

3

3

6-10

20

16

11

9

11-20

36

38

36

31

21-30

12

10

12

12

31-40

2

1

2

3

41-

0

0

1

1

Meancigarettesper day

11.2

10.6

10.6

10.2

Table 4.4 Mean cumulative exposures to radon and thoron daughters to end-dates of the listed ISPs,for eight over lapping sub-cohorts of the study group (standard deviations in brackets).

ISP End-date of ISP No. of men withtime worked in

0 1954-56

1 1958-62

2 1964-68

3 1970-73

4 1973-77

5 1978-80

6* 1979-80

7t 1980

the ISP

11196

12339

9225

9105

5605

3832

1762

221

Mean cumulative exposure to end-date of ISP (WL hour)

Radon daughters

152

178

209

238

253

268

258

140

(205)

(230)

(248)

(274)

(282)

(303)

(311)

( 56)

Thoron daughters

100

116

136

156

161

167

152

156

( 94)

(104)

(106)

(114)

(111)

(112)

(110)

( 68)

* Collieries C, F, P, Q, V, W, Y only

t Col 1iery P only

Table 4.5 Numbers of observed and expected deaths in the study group from all causes and lung cancer,with SMRs and 95% confidence intervals in brackets. (Expected numbers are based on regionaldeath rates).

Age

15-24

25-34

35-44

45-54

55-64

65-74

75-

Al 1 ages

obs

5

28

120

488

1504

2342

1365

5852

All causes

exp

5.1

24.5

121.6

534.9

1595.5

2298.9

1511.7

6092.2

Cause of death

SMR

98

114

99

91

94

102

90

96

(95%

(32,

(76,

(82,

(83,

(90,

(98,

(86,

(94,

C.I.)

229)

165)

118)

100)

99)

106)

95)

98)

obs

0

0

6

47

162

208

98

521

Lung cancer

exp

0.0

0.4

6.7

55.4

203.7

245.5

88.1

599.7

SMR

-

0

90

85

80

85

111

87

(95%

( o,

(33,

(62,

(68,

(74,

(90,

(80,

C.I.)

922)

195)

113)

93)

97)

136)

95)

Person-yearsat risk

4826

22547

48961

71784

74562

44022

11928

278629

54

Table 4.6 Analysis of deviance of lung cancer death rate, in relation toradon daughter exposure.

Factor

Age (A)

+ Smoking habit

+ Calendar time (T)

+ Col 1 iery

+ A.T

+ Radon

Residual

Total

df

5

5

2

9

10

6

3797

3834

Deviance

488

208

6

17

22

1

1123

1869

55***

72***

.28*

.58*

.90*

.83

.79

.65

* P < 0.05; *** P < 0.001

55

Table 4.7 Estimated lung cancer death rates* per 100,000 person-years, by:-

(a) Age and calendar time period.(b) Smoking habit (lagged by five years).(c) Colliery.(d) Cumulative exposure to radon daughters, lagged by 10 years

(WL hour).

(a) Calendar time x Age

-3435-4445-5455-6465-7475-

1960-691

15902101468

1970-791

1676236600822

1980-89141

44252593

1104

(b) Smoking habit

non-ex-pipecigarettes: 1-10

11-2021-

(c) Colliery

(d) Radon daughterexposure (WL hour)

CFKPQTVWXY

0-25-50-

100-200-400-800-

1187137236303438

236150215164282286173134228268

236275276269238230204

In calculating these rates, a 'baseline1 of men at colliery C aged 55-64 during1970-79, smoking 1-10 cigarettes per day, with a lagged cumulative exposureof 0-25 WL hour radon daughters was used. Their estimated death rate was236 per 100,000 person-years.

56

Table 4.8 Mean radon daughter exposure (WL hour) of men who died oflung cancer (cases) and matched survivors (referents), withcase-referent differences and, in brackets, differences dividedby estimated standard errors; by age, calendar time, colliery,and smoking category.

FrequenciesSub-group

Al 1 men

Age39-4445-4950-5455-5960-6465-6970-7475-7980-86

Calendar time1960-641965-691970-741975-791980-841985-89

Col 1 ieryCFKPQTVWXY

SmokingNon-Ex-PipeCigarettes: 1-5

6-1011-2021-3031-4041-

MeansDifference

Strata

467

511196479119835631

45286112102111

69263638466254155269

3313026801988784

Cases

493

511206684127876132

45488119111117

77263740496756155274

331312681

2159484

Referents

1897

204078262330500334226107

16207343463422446

29810314015819026620356199284

1012211989

317857360194

Case

275

95166240169201312301370382

180204262282318274

216534131468893241006764301

337326278258401249210217153

Referent

275

91188249173200322290390335

175187249286334276

22255418150895356986561267

230380281207348265221257210

<'

-1

4-23-9-42

-1011-2047

51712-5-16-2

-6-2-5-4-6-3221335

107-54-35154-16-11-40-57

t')

(-0.10)

( 0.20)(-0.31)(-0.20)(-0.24)( 0.13)(-0.47)( 0.62)(-0.47)( 1.50)

( 0.10)( 1-02)( 0.64)(-0.31)(-0.55)(-0.15)

(-0.65)(-0.68)(-0.13)(-0.35)(-0.08)(-1.51)( 0.25)( 0.24)( 0.98)( 2.24)

( 1-28)(-0.83)(-0.09)( 1.53)( 2.88)(-1.14)(-0.73)(-0.34)(-1.76)

57

Table 4.9 Results of conditional logistic regression analyses of lung cancerdeath rates upon radon daughter exposure.

Likelihood ratioModel

Radon

RADON ,

RADON ,

RADON ,

RADON ,

daughter

within

within

within

wi thin

exposure (RADON)

age groups

calendar time periods

col 1 ieries

smoking categories

stat ist

0

3

1

8

17

ic* (X2)

.02

.39

.54

.62

.81*

df

1

8

5

9

8

For model 'RADON', the statistic tests the overall effect of radon daughterexposure. For the other models, the statistic tests differences in effect ofexposure between categories.

P < 0.05

58

Table 4.10 Relative risks of lung cancer mortality per increase of 1 WLMradon daughter exposure; by age, calendar time, colliery andsmoking category.

Sub-group Relative risk (95% confidence interval)

All men

Age39-4445-4950-5455-5960-6465-6970-7475-7980-86

Calendar time1960-641965-691970-741975-791980-841985-89

Col 1 ieryCFKPQTVWXY

SmokingNon-Ex -PipeCigarettes: 1-5

6-1011-2021-3031-4041-

0.99

1.580.810.940.930.990.931.100.981.35

1.201.201.090.940.970.97

0.830.191.010.861.000.760.831.563.841.44

3.400.871.061.811.430.910.860.830.01

(0.90,

(0.02,(0.38,(0.58,(0.62,(0.74,(0.80,(0.85,(0.81,(0.80,

(0.08,(0.76,(0.84,(0.76,(0.83,(0.80,

(0.49,(0.00,(0.72,(0.37,(0.89,(0.56,(0.23,(0.03,(0.28,(0.97,

(0.12,(0.69,(0.69,(0.80,(1.06,(0.78,(0.65,(0.45,(0.00,

1.10)

127.70)1.73)1.53)1.40)1.34)1.08)1.42)1.17)2.29)

17.47)1.90)1.43)1.18)1.14)1.17)

1.41)48.51)1.43)2.01)1.13)1.03)2.98)95.26)53.61)2.12)

96.75)1.11)1.64)4.09)1.93)1.05)1.14)1.52)

35.11)

59

Table 4.11 Analysis of deviance of lung cancer death rate, in relation tothoron daughter exposure.

Factor

Age (A)

+ Smoking habit

+ Calendar time (T)

+ Col 1 iery

+ A.T

+ Thoron

Residual

Total

df

5

5

2

9

10

5

3668

3704

Deviance

488.

208.

6.

17.

23.

0.

1119.

1863.

12***

74***

74*

12*

36*

50

33

91

* P < 0.05; *** P < 0.001.

60

Table 4.12 Estimated lung cancer death rates* per 100,000 person-years, by:-

(a) Age and calendar time period.(b) Smoking habit (lagged by five years).(c) Colliery.(d) Cumulative exposure to thoron daughters, lagged by 10 years

(WL hour).

(a) Calendar time x Age

-3435-4445-5455-6465-7475-

1960-691

181042341598

1970-791

1889265655893

1980-89131

502886561201

(b) Smoking habit

non-ex-pipecigarettes: 1-10

11-2021-

(c) Colliery

(d) Thoron daughterexposure (WL hour)

CFKPQTVWXY

0-25-50-100-200-400-

1299153265341496

265188230190268312213167282299

265244247240255277

In calculating these rates, a 'baseline' of men at colliery C aged 55-64 during1970-79, smoking 1-10 cigarettes per day, with a lagged cumulative exposureof 0-25 WL hour thoron daughters was used. Their estimated death ratewas 265 per 100,000 person-years.

61

Table 4.13 Mean thoron daughter exposure (WL hour) of men who died oflung cancer (cases) and matched survivors (referents), withcase-referent differences, and, in brackets, differences divided byestimated standard errors; by age, calendar time, colliery, andsmoking category.

MeansSub-group

All men

Age39-4445-4950-5455-5960-6465-6970-7475-7980-86

Calendar t i m e1960-641965-691970-741975-791980-841985-89

Col 1 ieryCFKPQTVWXY

SmokingNon-Ex-PipeCigaret tes : 1-5

6-1011-2021-3031-4041-

DifferenceCase

180

5999

122118145194202235272

148154179188194175

23233

2121643111726968

167218

216182196177220173159151107

Referent

179

62100123125144195201234236

144141169191197175

23934

2141693131896867

151192

160202197149197179159154175

Ct'!

2

-3-1-1-6

1-1

10

37

31310-3-4

0

-7-1-2-5-2

-1711

1626

57-20-12922-50

-2-68

)

( 0.35)

(-0.23)(-0.02)(-0.06)(-0.72)( 0.14)(-0.14)( 0.12)( 0.01)( 2 .54)

( 0.08)( L42)( 0 .94)(-0.36)(-0.32)(-0.06)

(-0.63)(-0.71)(-0.12)(-0.36)(-0.08)(-1.58)( 0.15)( 0.09)( L46)( 2.36)

( 1.35)( -0 .77)(-0.04)( 1.72)( 2 .42)(-0.78)( 0.02)(-0.05)(-1.93)

62

Table 4.14 Results of conditional logistic regression analyses of lung cancerdeath rates upon thoron daughter exposure.

Likelihood ratioModel

Thoron

THORON ,

THORON ,

THORON ,

THORON ,

daughter

within

within

within

within

exposure (THORON)

age groups

calendar time periods

col 1 ieries

smoking categories

stat ist

0

4

2

9

13

ic+ (X2)

.06

.39

.75

.90

.48

df

1

8

5

9

8

For model 'THORON', the statistic tests the overall effect of thoron daughterexposure. For the other models, the statistic tests differences in effect ofexposure between categories.

63

Table 4.15 Relative risks of lung cancer mortality per increase of 1 WLMthoron daughter exposure; by age, calendar time, colliery andsmoking category.

Sub-group Relative risk (95% confidence interval)

Al 1 men

Age39-4445-4950-5455-5960-6465-6970-7475-7980-86

Calendar time1960-641965-691970-741975-791980-841985-89

Col 1 ieryCFKPQTVWXY

SmokingNon-Ex -PipeCigarettes: 1-5

6-1011-2021-3031-4041-

1.02

1.540.900.910.701.010.911.071.072.10

1.221.601.280.880.990.94

0.850.051.030.881.000.590.701.161.771.68

3.260.751.102.811.780.870.920.730.01

(0.84,

(0.00,(0.14,(0.27,(0.32,(0.57,(0.64,(0.68,(0.68,(0.84,

(0.03,(0.70,(0.78,(0.59,(0.68,(0.63,

(0.53,(0.00,(0.52,(0.43,(0.71,(0.33,(0.11,(0.04,(0.80,(0.98,

(0.12,(0.41,(0.50,(0.75,(1.03,(0.65,(0.55,(0.16,(0.00,

1.25)

280.92)5.80)3.07)1.53)1.79)1.32)1.67)1.67)5.27)

46.48)3.67)2.10)1.31)1.45)1.42)

1.36)625.08)2.04)1.81)1.42)1.07)4.37)36.35)3.89)2.87)

88.15)1.37)2.43)10.48)3.05)1.17)1.55)3.26)90.55)

64

Table 4.16 Mean dose (mSv) received by men who died of lung cancer(cases) and matched survivors (referents), with case-referentdifferences, and, in brackets, differences divided by estimatedstandard errors; by age, calendar time, colliery, and smokingcategory.

Sub-groupMeans

Case

Al 1 men

Age39-4445-4950-5455-5960-6465-6970-7475-7980-86

Calendar time1960-641965-691970-741975-791980-841985-89

Col 1 ieryCFKP

QTVwXY

SmokingNon-Ex-PipeCigarettes: 1-5

6-1011-2021-3031-4041-

19

61116121422212627

131518202219

1732811582275722

242220182718151511

.7

.8

.7

.5

.2

.7

.1

.7

.4

.8

.5

.0

.9

.2

.5

.5

.2

.8

.5

.8

.4

.4

.2

.3

.1

.0

.1

.8

.2

.6

.9

.0

.5

.7

.1

DifferenceReferent ( '

19

61317121422212724

131318202319

1732812582475619

162620152419161815

.7

.6

.0

.0

.6

.6

.7

.0

.5

.3

.1

.8'

.0

.6

.5

.7

.8

.9

.8

.2

.8

.6

.1

.2

.6

.5

.7

.3

.4

.1

.3

.1

.1

.1

.8

-0.

0.-1.-0.-0.0.-0.0.

-1.3.

0.1.0.-0.-1.-0.

-0.-0.-0.-0.-0.-2.0.0.0.2.

7.-3.-0.3.3.

-1.-0.-2.-4.

0

245416725

429302

5134421156

452660647

t 1 )

(-0

( o(-0(-0(-0( o(-0( o(-0( 1

( o( 1( o(-0(-0(-0

(-0(-0(-0(-0(-0(-1( o( o( 1( 2

( 1(-0(-0( 1( 2(-1(-0(-0(-1

.04)

.13)

.28)

.18)

.31)

.14)

.43)

.55)

.42)

.66)

.10)

.10)

.69)

.32)

.52)

.14)

.64)

.69)

.13)

.35)

.08)

.52)

.23)

.20)

.26)

.27)

.30)

.83)

.09)

.57)

.85)

.10)

.61)

.31)

.87)

65

Table 4.17 Results of conditional logistic regression analyses of lung cancerdeath rates upon a combined dose measure of radon and thorondaughters.

Model

Dose

Dose, within age groups

Dose, within calendar time periods

Dose, within collieries

Dose, within smoking categories

Likelihood ratiostatistic! (X2)

0.01

3.49

1.77

9.26

17.47*

df

1

8

5

9

8

t For model 'dose1, the statistic tests the overall effect of dose. For the othermodels, the statistic tests differences in the effect of dose between categories.

* P < 0.05

66

Table 4.18 Relative risks of lung cancer mortality per increase of 10 mSvcombined radon and thoron daughter; by age, calendar time,colliery and smoking category.

Sub-group Relative risk (95% confidence interval)

All men

Age39-4445-4950-5455-5960-6465-6970-7475-7980-86

Calendar time1960-641965-691970-741975-791980-841985-89

Col 1 ieryCFKPQTVWXY

SmokingNon-Ex-PipeCigarettes: 1-5

6-1011-2021-3031-4041-

1.00

1.270.840.950.921.000.941.080.991.32

1.151.191.090.950.980.97

0.880.241.010.901.000.790.851.312.241.34

2.480.891.051.661.360.920.890.850.02

(0.91,

(0.04,(0.43,(0.62,(0.65,(0.77,(0.81,(0.86,(0.83,(0.84,

(0.13,(0.80,(0.86,(0.78,(0.84,(0.82,

(0.60,(0.00,(0.75,(0.49,(0.90,(0.61,(0.30,(0.07,(0.63,(0.98,

(0.21,(0.72,(0.72,(0.84,(1.04,(0.81,(0.70,(0.50,(0.00,

1.08)

43.73)1.65)1.45)1.32)1.29)1.10)1.34)1.18)2.08)

9.89)1.77)1.37)1.16)1.14)1.16)

1.29)25.36)1.36)1.65)1.11)1.03)2.40)24.79)7.99)1.84)

28.64)1.10)1.53)3.29)1.77)1.04)1.14)1.47)13.65)

67

Table 4.19 Analysis of deviance of stomach cancer death rates, in relation toradon daughter exposure.

Factor

Age (A)

+ Smoking habit (S)

+ Calendar time (T)

+ Colliery

+ S.T.

+ Radon

Residual

Total

df

5

5

2

9

10

6

3797

3834

Deviance

205.44***

1.51

7.64*

4.60

19.02*

9.72

747.39

995.32

P < 0.05; *** P < 0.001

68

Table 4.20 Estimated stomach cancer death rates* by 100,000 person-years,by:-

(a) Age.(b) Smoking habit (lagged by five years) and calendar time

period.(c) Colliery.(d) Cumulative exposure to radon daughters, lagged by 10 years

(WL hour).

(a) Age -3435-4445-5455-6465-7475-

121857134219

(b) Calendar time period x smoking habit

non-ex-pipecigarettes: 1-10

11-2021-

(c) Colliery

(d) Radon daughterexposure (WL

1960-69112102

1748133

CFKP

QTVWXY

0-hour) 25-

50-100-200-400-800-

1970-79484359577972

571054943684978529644

57735088

11311252

1980-89574350502855

In calculating these rates, a 'baseline' of men at colliery C aged 55-64 during1970-79, smoking 1-10 cigarettes per day, with a lagged cumulative exposureof 0-25 WL hour radon daughters was used. Their estimated death rate was57 per 100,000 person-years.

69

Table 4.21 Mean radon daughter exposure (WL hour) of men who died ofstomach cancer (cases) and matched survivors (referents), withcase-referent differences and, in brackets, differences dividedby estimated standard errors; by age, calendar time, colliery,and smoking category.

Sub-groupFrequencies Means

Strata Cases Referents Case ReferentDifference

('t')

Al1 men 195 197 1735 265 269 -4 (-0.27)

Age44-4950-5455-5960-6465-6970-7475-7980-83

Calendar time1960-641965-691970-741975-791980-841985-89

CollieryCFKPQTVWXY

SmokingNon-Ex-PipeCigarettes: 1-5

6-1011-2021-3031-

69213350392413

33543463830

331717141720296

2121

25281683362203

69213350392413

69213351402413

598820030247133422457

148152241262241265418252

135147224233243331389266

1361630-2-6729-14

( 0.73)( 0.20)( 0.32)( 0.78)(-0.10)(-1.65)( L25)(-0.24)

33543463830

33544473830

23327389416326254

163205242315273291

118232247292287298

46-27-523-14-7

( 5.61)(-0.72)(-0.16)( 0.83)(-0.53)(-0.20)

331717141720296

2121

331718141721296

2121

30913515313712418724746204193

248484661957743941087761291

24255448169904403997362262

7-61926

-130-994-229

( 0.45)(-1.93)( 0.39)( 1.93)(-0.88)(-0.34)( L04)( 0.31)(-0.36)( 1-45)

1-56-1011-2021-3031-

25281683362203

25281783363203

235251131532846231544

348294225170212269281156

366252261288206232365417

-1842-35

-118637-84

-261

(-0.41)( 2.42)(-0.70)(-0.93)( 0.53)( 2.10)(-1.09)(-0.88)

70

Table 4.22 Results of conditional logistic regression analysis of stomach cancerdeath rates upon radon daughter exposure.

Likelihood ratioModel

Radon

RADON ,

RADON,

RADON,

RADON,

daughter

within

within

within

within

exposure (RADON)

age groups

calendar time periods

col 1 ier ies

smoking categories

stat ist

0

6

3

8

12

ic* (X2)

.02

.75

.62

.88

.34

df

1

7

5

9

7

For model 'RADON', the statistic tests the overall effect of radon daughterexposure. For the other models, the statistic tests differences in effect ofexposure between categories.

71

Table 4.23 Relative risks of stomach cancer mortality per increase of 1 WLMradon daughter exposure; by age, calendar time, colliery andsmoking category.

Sub-group

All men

Age44-4950-5455-5960-6465-6970-7475-7980-83

Calendar time1960-641965-691970-741975-791980-841985-89

CollieryCFKPQTVWXY

SmokingNon-Ex-PipeCigarettes: 1-5

6-1011-2021-3031-

Relative risk

1.01

1.151.291.111.390.990.741.330.89

7.300.890.991.280.900.96

1.200.001.212.770.930.871.943.130.481.55

0.911.480.840.671.161.680.780.57

(95% confidence interval)

(0.86,

(0.37,(0.20,(0.67,(0.83,(0.71,(0.52,(0.71,(0.41,

(0.20,(0.61,(0.72,(0.85,(0.60,(0.65,

(0.54,(0.00,(0.72,(0.54,(0.77,(0.47,(0.36,(0.01,(0.01,(0.72,

(0.62,(0.84,(0.47,(0.33,(0.54,(0.97,(0.53,(0.11,

1.19)

3.58)8.36)1.84)2.35)1.36)1.03)2.49)1.94)

266.05)1.31)1.36)1.93)1.35)1.42)

2.70)1.33)2.04)14.27)1.13)1.60)10.53)

1171.86)22.19)3.32)

1.33)2.61)1.49)1.36)2.48)2.92)1.16)2.97)

72

Table 4.24 Analysis of deviance of stomach cancer death rates, in relation tothoron daughter exposure.

Factor

Age

+ Smoking habit (S)

+ Calendar t ime (T)

+ Col 1 iery

+ S.T

+ Thoron

Residual

df

5

5

2

9

10

5

3668

Deviance

205.48***

1.51

7.64*

4.59

19.03*

4.33

773.60

Total 3704 1016.18

* P < 0.05; *** P < 0.001

73

Table 4.25 Estimated stomach cancer death rates* by 100,000 person-years,by:-

(a) Age.(b) Smoking habit (lagged by five years) and calendar time

period.(c) Colliery.(d) Cumulative exposure to thoron daughters, lagged by 10 years

(WL hour).

(a) Age -3435-4445-5455-6465-7475-

132783191309

(b) Calendar time period x smoking habit

1960-69non-ex-pipecigarettes: 1-10

11-2021-

(c) Colliery

(d) Thoron daughterexposure (WL hour)

162149

110611750

CFKPQTVWXY

0-25-50-100-200-400-

1970-7971638583

115106

8312476597778114638768

83696986

11177

1980-89846573724183

In calculating these rates, a 'baseline' of men at colliery C aged 55-64 during1970-79, smoking 1-10 cigarettes per day, with a lagged cumulative exposureof 0-25 WL hour thoron daughters was used. Their estimated death ratewas 83 per 100,000 person-years.

74

Table 4.26 Mean thoron daughter exposure (WL hour) of men who died ofstomach cancer (cases) and matched survivors (referents), withcase-referent differences, and, in brackets, differences divided byestimated standard errors; by age, calendar time, colliery, andsmoking category.

MeansSub-group

All men

Age44-4950-5455-5960-6465-6970-7475-7980-83

Calendar time1960-641965-691970-741975-791980-841985-89

CollieryCFKPQTVWXY

SmokingNon-Ex -PipeCigarettes: 1-5

6-1011-2021-3031-

DifferenceCase

182

8083152162181206256188

140154181197189187

268302382212732097479149210

22319418113916218517071

Referent

180

78101138148174223249200

98152173191201188

260342291903162156875158190

227168211172160168193154

(V)

2

2-1715147

-177

-12

42286

-13-1

7-49

31-44-664-920

-426-30-33217-23-83

( 0.26)

( 0.12)(-0.73)( 0.77)( 0.96)( 0.67)(-0.97)( 0.42)(-0.33)

( 2.74)( 0.15)( 0.52)( 0.54)(-0.81)(-0.06)

( 0.46)(-2.00)( 0.37)( 2.00)(-0.86)(-0.41)( 0.95)( 0.27)(-0.48)( 1-37)

(-0.20)( L90)(-0.96)(-0.69)( 0.21)( 2.02)(-0.75)(-0.78)

75

Table 4.27 Results of conditional logistic regression analysis of stomach cancerdeath rates upon thoron daughter exposure.

Likelihood ratioModel

Thoron

THORON

THORON

THORON

THORON

daughter

, within

, within

, within

, within

exposure (THORON)

age groups

calendar time periods

col 1 ieries

smoking categories

stat ist

0

5

3

8

8

ic* (X2)

.34

.08

.10

.68

.23

df

1

7

5

9

7

For model 'THORON1, the statistic tests the overall effect of thoron daughterexposure. For the other models, the statistic tests differences in effect ofexposure between categories.

76

Table 4.28 Relative risks of stomach cancer mortality per increase of 1 WLMthoron daughter exposure; by age, calendar time, colliery andsmoking category.

Sub-group

All men

Age44-4950-5455-5960-6465-6970-7475-7980-83

Calendar time1960-641965-691970-741975-791980-841985-89

Col 1 ieryCFKP

QTVWXY

SmokingNon-Ex -PipeCigarettes: 1-5

6-1011-2021-3031-

Relative risk

1.10

1.130.201.731.941.220.801.170.74

15.411.251.241.290.770.98

1.180.001.442.430.820.742.312.370.761.77

0.921.930.660.621.081.940.730.31

(95% confidence interval)

(0.80,

(0.07,(0.01,(0.51,(0.70,(0.64,(0.44,(0.51,(0.22,

(0.14,(0.51,(0.65,(0.64,(0.40,(0.43,

(0.58,(0.00,(0.51,(0.60,(0.47,(0.23,(0.20,(0.02,(0.28,(0.63,

(0.40,(0.77,(0.28,(0.20,(0.41,(0.92,(0.31,(0.01,

1.52)

19.75)3.10)5.92)5.35)2.36)1.44)2.69)2.51)

1667.53)3.05)2.37)2.62)1.47)2.22)

2.43)1.27)4.05)9.92)1.44)2.37)26.50)335.96)2.05)4.98)

2.11)4.83)1.57)1.93)2.85)4.10)1.73)9.81)

Table 4.29 Mean radon (Rn) and thoron (Tn) daughter exposure (WL hour) of men who died from thelisted causes (cases) and matched survivors (referents) with case-referent differencesand, in brackets, differences divided by estimated standard errors.

Cause of death

Malignant neoplasmof:

oral cavity

oesophagus

larynx

bone

Malignant melanoma

Malignant neoplasmof:

prostate

kidney

Leukaemia (excl.chronic lymphoid)

FrequenciesI CD code(9th rev.) Strata Cases Referents

140-149 16 16 424

150 39 39 739

161 1 1 38

170 2 2 40

172 4 4 56

185 70 70 1243

189 14 14 295

204-208 21 21 334(excl. 204.1)

MeansDifference

RnTnRnTnRnTnRnTn

RnTn

RnTnRnTn

RnTn

Case

27619126416323582137125

139141

268191261178

272185

Referent

259170348198884311124110

133134

257185219159

289178

1721-84-35-649-229

1315

77

1164219

-187

('t')

( 0.31)( 0.95)(-1.72)(-1.75)( - )( - )( 7.25)( 3.93)

( 0.36)( 0.35)

( 0.51)( 0.49)( 1.50)( 1.12)

(-0.27)( 0.24)

78

Table 4.30 Results of conditional logistic regression analyses of cause-specificdeath rates upon radon and thoron daughter exposure.

(a) Likelihood ratio statistics (X2 on 1 df).(b) Relative risks per WLM, with 95% confidence intervals in

brackets.

Cause ofdeath

Malignant neoplasmof:

oral cavity

oesophagus

larynx

bone

Malignant melanoma

Malignant neoplasmof:

prostate

kidney

Leukaemia (excl.chronic lymphoid)

ICD code(9th rev.)

140-149

150

161

170

172

185

189

204-208(excl. 204.1)

(a)X2

RnTn

RnTn

RnTn

RnTn

RnTn

RnTn

RnTn

RnTn

00

75

11

00

00

00

21

00

.08

.78

.51**

.89*

.61

.73

.19

.22

.58

.53

.12

.16

.25

.05

.18

.12

11

00

00

33

1611

11

22

01

Relat

.07

.62

.66

.47

.62

.23

.92

.97

.22

.08

.05

.10

.34

.36

.92

.16

(0(0

(0(0

(0(0

(0(0

(0(0

(0(0

(0(0

(0(0

(b)ive risk

.65,

.53,

.47,

.26,

.25,

.01,

.01,

.01,

.01,

.01,

.80,

.68,

.65,

.42,

.65,

.49,

14

00

14

23901917

3178011415

11

813

12

.76)

.99)

.91)

.85)

.57)

.08)

.36)

.70)

.24)

.46)

.38)

.78)

.47)

.46)

.32)

.74)

* P < 0.05;. P < 0.01

79

5. DISCUSSION

5.1 Lung Cancer

Comparison with regional male population death rates showed lower than expectednumbers of deaths from lung cancer, and also from all causes of death, a resultwhich may in part reflect the so-called healthy worker selection effect (Fox andCollier, 1976). It cannot be concluded from this result that the low levels ofradioactivity in these mines presented no hazard to health. What can be safelystated is that, evenwif.vunderground'vradonn .and;;.thoron daughters were responsible forsome lung cancer., deaths, these .were .not ^sufficiently numerous to have increasedthe death rate beyond that of-the; general population. This does not preclude thepossibility that radioactivity has contributed to lung cancer mortality.

This question was addressed by two types of mortality analysis, both carried outentirely within the study group. In the event, neither of these - theperson-years-at-risk analysis or the case-referent studies - demonstrated a positiveassociation between radioactivity exposure and lung cancer mortality, which helduniformly throughout the study group. (In case-referent studies, an interactionbetween exposure and smoking was demonstrated, which is discussed in more detailbelow.)

Although no overall associations with exposure were demonstrated, it is of interestto examine the range of relative risks which were consistent with the data; thefollowing Table shows estimated relative risks, with 95% confidence limits inbrackets. (Estimates were obtained from person-years analysis by attaching arepresentative value to each exposure category, as explained under methods, andfitting a single degree of freedom for trend. Results from the two types ofanalysis agree reasonably well.)

Person-years Case-referentanalysis studies

Radon daughters (per WLM) 0.96 (0.88, 1.04) 0.99 (0.90, 1.10)

Thoron daughters per WLM) 1.02 (0.86, 1.22) 1.02 (0.84, 1.25)

Dose (per 10 mSv) - 1.00 (0.91, 1.08)

The upper limits to confidence intervals may be taken as estimates of themaximum relative risk consistent with the data. For example, the risk of lungcancer mortality for a miner with a lagged cumulative exposure to radon daughtersof 10 WLM, relative to a miner with no exposure, would not exceed 1.48(1.0410). The corresponding maximum risk associated with 3 WLM exposure tothoron daughters would be 1.82 (1.223).

Relative risks predicted by the BEIR IV Committee's Time Since Exposure (TSE)model (Committee on the Biological Effects of Ionizing Radiations, 1988) areconsistent with estimates obtained in the present study. Thus, the TSE model

80

predicts the risk of lung cancer mortality at a given age in a radon daughterexposed group relative to a non-exposed group, by:

1 + 0.03 (w, + | w2) if age <55;

1 + 0.025 (w, + £ w2) if 55 < age < 64;

1 + 0.01 (w, + £ w2) if age > 65,

where w, is the cumulative exposure acquired between 5 and 15 years before thegiven age, and w2 is the exposure acquired more than 15 years before. This isa more complex^ modeluthan;,,any, usedi. int,the.,«present study, and some simplifyingassumptions were. made. ;to allow, comparison:.....-, An age of 60 was chosen forpurposes of illustration,-! and ^.cumulative-exposure was assumed to have beenacquired at a constant rate from birth. Therefore, if the total cumulativeexposure to age 60 is w (in units of WLM), then

w, = Ve w,

w2 = I w,

and the TSE predicts a relative risk of

1 + 0.025 (Vs w + i x ? w),

which equals

1 + 0.035 w.

Both the person-years and case-referent methods, as implemented in the presentstudy, predict the relative risk at age 60 in the form

exp ( 5 /e w/3)

where (3 denotes the fitted regression coefficient, and the factor 5/e accounts forthe 10-year lag.

The comparison of results (present study versus TSE) is given in Figures 5.1 and5.2, which show risks estimated in the person-years analysis and case-referent studyrespectively. The TSE estimate falls within the 95% limits shown in bothFigures. Figure 5.1 shows that, on the basis of this study, the risk of lungcancer mortality for a man aged 60, with an accumulated total exposure of 10WLM relative to a man of the same age, with no exposure, would not exceed1.40. The TSE estimate of relative risk is 1.14. The upper limit in Figure 5.2is rather greater (2.14), due to the use of only four referents per case, instead ofall available referents.

The case-referent studies of radon daughter exposure and of the combined dosemeasure each showed associations between lung cancer mortality which varied withsmoking category. The interaction between thoron daughter exposure and smokingwas not statistically significant, but estimated relative risks showed a pattern ofvariation similar to that shown for the other two variables. (In passing, it isworth remarking on the strength of the association between smoking habit and lungcancer death rates shown in person-years analysis (Table 4.7). Doll and Peto(1978) in their study of British doctors, have shown that annual lung cancerincidence in smokers who began smoking between the ages of 16 and 25 is

81

described by a function of years elapsed since 22 and daily cigarette consumption.This suggests that consideration of both duration of smoking and daily consumptionis appropriate in the assessment of exposure-response. In the present study, sinceage at starting smoking was not recorded, daily consumption only was used toquantify exposure. Nevertheless, a clear exposure-response was demonstrated,which attests to the quality of smoking information in PFR data. A final pointof interest is the elevated rate found among pipe smokers, and also, to a lesserextent, among ex-smokers.) The pattern of interaction found in case-referentstudies was of increased relative risks in non- and light smokers, which declinedwith increasing cigarette consumption, becoming less than unity for heavy smokers.Risks for pipe smokers were only slightly elevated, and for ex-smokers, werereduced below unity,; close <,to those.= for..heavyt, smokers. Tests for trend, in non-and cigarette smokers i. were .statistically ..significant for all three measures ofradioactivity.

The negative association in heavy smokers presents difficulties of interpretation,biologically, and contrasts markedly with most published work. For example,Whittemore and Macmillan (1983) reported a multiplicative interaction betweensmoking and radon daughter exposure in Colorado Plateau uranium miners; thatis, increased relative risks due to each factor, when multiplied together, gaverelative risks for both exposures combined. A slightly less than multiplicativeinteraction was reported by Hornung and Meinhardt (1987), for the same cohort.Cumulative exposures in these uranium miners were of course much higher thanthose of the present study group, and it is possible that patterns of interaction atvery much lower exposure levels could be different. Studies of Swedish metalminers, whose exposures were lower than those of uranium miners, but still higherthan those of the coalminers studied here, have given mixed results. Damber andLarsson (1985), in a case-referent study of lung cancer in iron mining areas ofNorthern Sweden, reported a multiplicative interaction between cumulative cigaretteconsumption and years underground; Radford and St Clair Renard (1984)concluded that the risks of both exposures combined additively, from their cohortstudy of iron miners, and Axelson and Sundall (1978) reported an apparentlyprotective effect of smoking among miners in a case-referent study of zinc andlead miners, a finding which was consistent with an additive or sub-additiveinteraction. Different forms of interaction are also suggested by results ofcase-referent studies-of indoor exposure. Edling et al (1984, 1986), in a study oflung cancer in a rural area of Sweden, reported results consistent with amultiplicative interaction between smoking and exposure. (Indoor levels of radondaughters summarized in Edling et al (1984) were close to underground levelsmeasured in PFR collieries.) However, a pattern of interaction resembling that ofthe present study was reported by Schoenberg et al (1990) in a case control studyof lung cancer and indoor radon exposure, in New Jersey women. In smokers ofless than 25 cigarettes per day, lung cancer risks increased with radon exposure,while in the heaviest smoking category (^ 25 per day) risks apparently declinedwith increasing exposure. These authors speculated that a thickened mucus layerin heavy smokers might protect the bronchial epithelium from alpha radiation, andthereby weaken the effect of exposure. It seems unlikely that this mechanismcould account for the pattern observed in the present study, namely, the tendencyof the elevated risks in non- and lighter smokers to become decidedly less thanunity in higher smoking categories. The strength of the decreasing trend suggeststhat some bias may be operating, perhaps in combination with a real effect, but ithas not been possible, within the limits of the present study, to explore the findingany further.

82

Relative risks reported in Chapter 4 were all 'colliery-adjusted1 - that is, inperson-years analyses, indicator terms for colliery were included in regressionmodels; and in case-referent studies, cases and referents were matched oncolliery. This option was preferred to the alternative of not adjusting forcolliery, on the grounds that unadjusted associations might have been due, in partor entirely, to a relationship between lung cancer mortality and some unknownfactor, whose colliery mean levels were positively associated with colliery meanradioactivity levels. An example of such an unknown confounding factor wouldbe diesel fume. Diesel transport was used at collieries K, Q, T, W, X and Y,where the average radon daughter level was 6.98 mWL, compared to 2.56 mWL atnon-diesel collieries C, F, P and V. Another is quartz, although according to arecent review, ^.evidence*..for.a the •.< carcinogenicity of quartz in man is limited(McDonald, 1989).i . (The 'possibility,. iof such confounding was the main weaknessof the preliminary correlation•.* analysis <which .-provided the impetus for the presentstudy. Incidentally, a repeat of the preliminary analysis, using colliery-specificrelative risks, adjusted for age, calendar time and smoking, showed non-significantSpearman rank correlations with colliery mean levels of radon and thorondaughters, after incorporation of the mortality data from PFR2 attenders.) Adisadvantage of colliery adjustment is that it will have resulted in some loss ofstatistical power, since variation in exposure within collieries was, in general, lessthan that between (Figures 4.8 and 4.9). However, results of limitedperson-years analyses, with adjustment only for age, calendar time and smokinghabit, suggested that the loss will not have been great. Thus, for radondaughters, the estimated relative risk per WLM was 1.02 (95% confidence limits0.97, 1.08); and for thoron daughters, 1.14 (95% confidence limits 0.99, 1.30).Note that these confidence intervals are not substantially narrower than those shownin the Table of Section 5.1. Note also that this analysis suggests a possibleassociation between thoron daughter exposure and lung cancer; but since theapparent relationship disappears after adjustment for colliery, it should not beregarded as supplanting the results of Chapter 4, which are summarized in Section5.1.

5.2 Other Causes of Death

Of the nine other .causes ...of death, analysed -.in this study, only oesophageal cancerwas found to be associated - with exposure .to . radon and thoron daughters. Forboth exposures, the relative risk declined with age, and, averaged over all ages,was less than unity. Health-related job changes may account for this result.Alcohol consumption is known to be associated with oesophageal cancer (IARC,1988); it may be that men with heavy consumption were transferred to surfacework, thereby lowering their radon daughter exposure.

Risks for leukaemia, which has been highlighted in a recent correlation study(Henshaw el al, 1990), did not approach statistical significance.

5.3 Quality of Exposure Data

The rationale underlying the method of exposure estimation is described in detail inAppendix 4. A central assumption is that average radioactivity levels within coalseams were approximately constant during the period when study group memberswere working miners. This is a large assumption to make, since the period inquestion is a long one. Men attending PFR1 in their fifties could have begunmining as early as 1920; men starting their mining work at the time of PFR2could still be working at the end of the study period (31/12/1989). At the start

83

of work on the present project, there was no way of checking the validity of theassumption for all 10 research collieries, since the available radioactivity data atnine of the 10 spanned only short periods of time. However, results fromcolliery Q had been obtained over an interval of eight years, and did support theassumption of constant seam means (see Appendix 6, Tables A6.1, A6.2). Toprovide a further check, 50 measurements of radon and thoron daughters, using atrack etch technique developed at NRPB (Miles, NRPB, Personal communication,1992), were made at two of the research collieries (other than colliery Q) in 1990.The results, which are summarized in Appendix 6, suggest that radioactivity levelsin 1990 were higher than levels in the mid to late 1970s at the two collieriesconcerned. At colliery C, average radon daughter levels were higher by a factorof approximately 4,* and ;thoron-i daughters-_-by-^a factor of 2; at colliery Y, radondaughters were only.-,slightly-i-higher?. but 4horon; daughters were greater by a factorof approximately 3. However, measurements 'from colliery Y were made in adifferent seam to those of the 1970s, which confounds the comparison of timeperiods.

The new data which became available only after statistical analysis of mortality wasessentially complete) raised a question mark over the validity of the basicassumption underlying exposure calculations. The substantial increase in radondaughters at colliery C was of particular concern, since the new measurements weremade in the same seam as previous measurements. Reasons for the increase weresought, and among several possibilities were two changes in mining method whichhad taken place at colliery C since the 1970s (Smith, British Coal ScientificServices, personal communication, 1991). First, the use of a greater extractionthickness in recent years led to a greater proportion of stone and shale, comparedto coal, accumulating in waste areas, and this may have increased radon emanationrates, since the rock had a greater concentration of uranium than the coal.Second, at the time of the 1990 survey, access roadways to two production faceswere in use, which had been driven through waste material created by theadvancing face (so-called 'gob scours'). The walls of the roadways thereforepresented a larger surface area to the coalmine atmosphere than would have beenoffered by roadways driven through rock. Gob scours were not used at colliery Cduring the 1960s and '70s.

If the rise in radioactivity level-at colliery C-can be attributed to changes inmining method which have occurred during the 1980s (and there is no hardevidence that it can) then the assumption of constant levels during the 1970s andbefore is not explicitly contradicted by the new data. Furthermore, since all butone of the causes of death considered in this study were investigated in relation toexposure lagged by 10 years, and since most of the men studied left the industrybefore 1980, radioactivity levels during the 1980s were not a major concern.Nevertheless, the main impact of the 1990 surveys upon the present study has beento cast doubt upon the validity of the assumption of long-term constancy ofradioactivity levels within seams, and hence, upon the reliability of estimatedexposures.

Assuming however, for the moment, that the hypothesis of constant radon andthoron daughter levels within coal seams during the six decades prior to 1980 istrue, estimates of exposure must still be viewed with caution, for the following tworeasons. First, during the course of exposure calculations, it was necessary

84

arbitrarily to assign colliery mean .radioactivity levels to four categories of timeworked underground:

jobs in PFR collieries, not specific to individual seams;

jobs in PFR collieries, specific to seams where radon and thoron daughtershad not been measured;

jobs in PFR collieries, recorded in Occupational Histories;

jobs in non-PFR collieries, recorded in Occupational Histories.

The extent to-frwhichr the ' true exposure, .^experienced by men working thesecategories was adequatelyA'Summarizedi>-by-the PFR colliery mean, cannot be known.It is possible to make some general observations, as follows. Jobs in the firstcategory moved between seams, and assignment of the colliery mean was reasonablein this case. However, its assignment to the second category was more dubious.Real seam differences had been demonstrated (see Appendix 4) and results fromcolliery Q showed they could be large. A preferable option would have been toassign mean values from geologically similar seams, but this would have enlargedthe exposure assessment part of the project beyond the resources available.Occupational Histories made no reference to locations underground, but merely totype of mining activity, and use of the PFR colliery mean for time in the thirdand fourth catetories was the only available option. Its assignment tounderground work at non-PFR collieries seems particularly unsatisfactory, but, onaverage, only 16% of men's total time worked was at these collieries (seeAppendix 6), and the errors resulting from its use may not be unacceptably high.

Second, the existence of a constant seam mean still allows scope for systematicdifferences between locations, within seams. A tendency for radon and thorondaughter levels to increase with ventilation distance (see Appendix 4) suggests thatmen working in return roadways may be exposed to higher levels than menworking in intake roadways. It may be that, over the long periods considered inthis study, men who spent a greater proportion of their time in return roadwayswill have acquired a higher exposure. No account was taken of this possibility inexposure estimation.

The above discussion has focussed on possible weaknesses in the exposure estimates.It should be recalled however that the time-worked information on which theseestimates are partly based, has proved its reliability in numerous studies of thehealth effects of exposure to coalmine dust. In particular, full confidence can beplaced in estimates of time worked underground and on surface, obtained fromPFR data. This is an important consideration, in view of the fact that thecontrast between surface and underground radioactivity levels contributes in largemeasure to the variation in exposure estimates at collieries C, K, P, Q, T and Y.

In the present study, no attempt was made to estimate exposures to radioactivityacquired indoors outside working hours. Estimates of radon daughter exposurewould have been based on published county mean values of radon gasconcentrations, and assumed average occupancy values (Wrixon et al, 1988).Estimation of thoron daughter exposure would have been impossible since countymean levels were not available.) It was felt that such estimates would have beentoo unreliable for use in mortality analyses, and instead, an attempt was made tocalculate the approximate size of the indoor contribution to total cumulativeexposure to radon daughters. Details are given in Appendix 8.

85

The results showed that the average proportion of total exposure acquired duringworking time might be expected roughly to vary between 5% at colliery X to 50%at colliery Q. For four collieries (F, V, W and X) the proportion did notexceed 10%. Figure 4.8 shows that cumulative exposures at these collieries werevery low. In general, the fact that working times were substantially less thanestimated times spent indoors, was the main reason for low values of theseproportions. Although working-time exposure represented only a part of totalexposure and at some collieries, clearly a very small part), relative risks willprobably not have been subject to any major bias because of the omission of theindoor component. This is because there is no reason to suspect a strongcorrelation between the two sources of exposure, allowing for colliery differences,and the effects ofoage.*,,: (The. inverse^relationship. between time spent at work andtime at home suggests .ithe•?possibility^of ;>a- negative correlation, but this is unlikelyto be strong in view of" the high-variability of- indoor levels. Colliery mean radonlevels do not show a significant correlation with county mean levels - Spearman'srank correlation coefficient = -0.38, compared with a critical 5% value of 0.65.Thus, if the two components are regarded as separate variables in regressionanalysis either of death rates, or in conditional logistic analysis), what has beenestimated in the present study is the effect on mortality, of exposure gained duringworking time, unadjusted for the effect of indoor exposure. It is unlikely thataddition of the indoor exposure to a regression model would substantially alter thevalue of the risk parameter for working-time exposure, unless the two exposurevariables were themselves associated. It is possible that estimated risks associatedwith the two exposures would differ, but this in itself would be a finding.

5.4 Completeness of Follow-up; Exclusions from Analysison grounds of Missing or Unrealiable Data

Three-point-three percent of the study population were excluded from statisticalanalysis because of unknown vital status. Some of these men had emigrated, andwere for practical purposes 'lost to follow-up'; that is, their dates and causes ofdeath will never be known, unless they return to the UK in the future. Thisgroup made up 0.7% of the study population. The remaining percentage (2.6%)is not so high as to allow the possibility, of a serious bias in the results; effortsto reduce it further are. presently .being made, through the records of the BritishCoal Corporation's pension scheme and concessionary fuel department.

After exclusion of the 3.3%, 18769 men remained. Further exclusions on thegrounds of missing, or inconsistent data, reduced this number by 34%, giving agroup of 12361 men for analysis. The large reduction from the total number ofmen attending PFR1 or 2, to the number included in analyses, was caused largelyby two factors: lack of smoking information at PFR1 (and at PFR2 for onecolliery), and absence of time-worked information for Singletons. Unfortunately,these features of PFR data cannot now be altered - for example, there is no wayof retrieving time records for Singletons. The possibility was considered ofincreasing the study group size in certain analyses, by including men with nosmoking information. Thus, for example, since smoking does not appear to be arisk factor for stomach cancer, it might be argued that no adjustment for its effectwould be required in estimating relative risks due to radon or thoron daughterexposure. However, inhalation being the principal source of entry of radon andthoron daughters to the body, smoking-related changes to lung structure mightaffect dosage, not only to the lung, but to other sites also.

Exclusions on the grounds of missing time-worked information will have beengreater among older men. This is because men close to retirement age at PFR1

86

will not have remained in the industry long enough to attend PFR2, andconsequently will have become Singletons. Although a disproportionately highnumber of deaths will therefore have been excluded, exposure-response relationshipswill not have been biased, since losses will not have been related to exposure.

5.5 Statistical Methods

5.5.1 Treatment of exposure variables, and smoking

In person-yearsi'analyses^ s(internal^andt. external, comparisons) and case-referentstudies, cumulative^exposurew.measures*; were,^continuously updated throughout thestudy period. Linear interpolation iwas-used ,to estimate exposure to dates betweensurveys; since ISPs lasted only five years on average, this procedure will havebeen sufficiently accurate. Cumulative exposure in the period following the endof men's exposure records was set equal to the final exposure. This approachwas appropriate for men whose exposure records ended while the AttendanceRecords System was still operating at their collieries. For most of them, the endof the exposure record will have coincided with their dates of leaving the industry.On the other hand, for men still working at the time when the AttendanceRecords System ended (the late 1970s), the procedure will have causedunderestimation of the latter part of their working-time exposures. This shortfallwill not have biased exposure-response relationships (except possibly in the analysisof leukaemia cases) since a lag of 10 years was applied to cumulative exposure inall analyses. The use of a two-year lag in the analysis of acute leukaemiameant, in effect, that the actual lag varied from two to a maximum of about 12years.

The choice of a 10-year lag for exposure variables was to some extent arbitrary,there being few precedents for a study involving such low exposures. Recentanalyses of US uranium miners' lung cancer mortality have used lags of five and10 years (Whittemore and McMillan, 1983; National Research Council, 1988).For analyses of mortality from leukaemia (excluding chronic lymphoid) a shorter lagwas judged appropriate; Smith and Doll (1982), in a study of 14111 ankylosingspondylitis patients -given a single course of X-ray treatment, reported that theexcess death rate from -leukaemia was greatest three to five years after treatment.

Smoking habit was also treated as a time-dependent variable in both types ofstatistical analysis. A similar approach to that used for exposure variables wasadopted: interpolation gave estimates of smoking habit between surveys, andsmoking category was assumed unchanged following the final PFR attendance.This assumption was of doubtful validity since results from miners who hadattended a sequence of four PFR surveys suggested that tobacco consuption declinedover the period. If this reflected the trend in smoking habit for the completestudy group then, on average, men's smoking habit will have been overestimatedduring the later part of the study period.

A lag of five years was applied to smoking habit, in all analyses. This allowedfor the possibility that men who later became lung cancer cases might havemodified their smoking habit on experiencing the onset of symptoms.

87

5.5.2 Person-years analyses

Tables of death rates by factors of interest were used to analyse associationsbetween radon and thoron daughter exposure and mortality from two causes - lungcancer and stomach cancer. Numbers of deaths from these causes were largeenough to ensure that death rates had reasonable statistical stability. Directinspection of exposure-response relationships, by tabular or graphical means, wascarried out; such descriptive methods helped the interpretation of regressionanalyses.

Calculation of SMRs, using regional population statistics, was carried out only forall-cause mortality,,; and lung: cancer. The purpose of these analyses was toestablish the magnitude-oft the - Jung ,-cancer ..death rate in the study group, and tocompare it to that of the - general population. . However, variations in SMR withexposure category were not examined. Since the external population had anaverage level of exposure comparable to that of the study population, comparisonwith population rates would have offered no information concerning exposure effectsbeyond what would be obtained in analyses carried out entirely within the studygroup.

5.5.3 Case-referent studies

Case-referent studies were carried out of 10 of the 11 causes of death consideredfor this study. Adjustment for the effects on death rates of age, smoking,colliery and calendar time was effected by matching cases to referents; for causeswith small numbers of deaths, this method was preferred to the alternative ofincluding confounding variables in regression models. Referents for lung cancerand stomach cancer cases were randomly sampled from strata defined byconfounding variables. Case-referent ratios were chosen primarily to give datafiles which could be conveniently processed using available software. Thus, lungcancer cases were matched to four referents, and stomach cancer cases to 10referents. Data for 2390 and 1932 subjects respectively were obtained, to whichconditional logistic regression models incorporating second order terms could befitted reasonably quickly. Four referents per case was judged to be the minimumacceptable ratio; the width of confidence limits shown in Figures 5.1 and 5.2suggests that the loss of power in comparison to person-years analysis has not beenvery great. Sampling was not used for other causes of death, since inclusion ofall available referents did not produce unmanageably large datasets for analysis.

5.6 Conclusions and Recommendations for further Analysis

5.6.1 Conclusions

i) The case-referent study of lung cancer showed a statistically significantlyincreased relative risk of mortality per WLM radon daughter exposure (lagged by10 years) in smokers of 6-10 cigarettes per day; and raised relative risks, but notsignificantly so, in non-smokers and smokers of 1-5 cigarettes per day. Relativerisks in heavier smokers were less than unity. Similar patterns of risk wereobserved for lagged thoron daughter exposure, and a combined measure of dose,but differences between smoking categories were significant only for the latter.The result presents difficulties of interpretation, and may be partly due to theoperation of a bias, as yet unidentified. Person-years analyses did not show any

relationship between lung cancer death rates and either of radon or thorondaughter exposure, or a combined measure of dose.

ii) There was an association between each of radon and thoron daughterexposure, and oesophageal cancer; which took the form of an increased risk formen in their fifties, which declined strongly with age. Overall, both associationswere negative. The result may be due to a health-related selection effect.

iii) For none of the other causes of death examined (cancer of the stomach, oralcavity, larynx, bone, prostate, kidney; malignant melanoma, leukaemia [excludingchronic lymphoid]) was there evidence of an association between death rate andexposure to radon .< or,v thorom - daughters:*,;.,..- -

5.6.2 Recommendations for further analysis

To dismiss the finding of a positive association between lung cancer mortality andradon daughter exposure in light smokers, because of the problematical negativeassociation in heavy smokers, would be premature. Further analyses, using moreof the available information on exposure and smoking, might help to clarify thepresent finding. If such analyses were carried out, it would be appropriate toconsider whether the present estimates of exposure could be improved. Results inAppendix 4 suggest that use of ventilation distance might achieve this.Furthermore, if a case-referent design were adopted, so that numbers of subjectswere kept within limits, it would be feasible to consider estimating domesticexposure. Addresses for the study group members are available from PFR files;exposure level could be categorized on the basis of type of dwelling. Even inthe absence of domestic exposure, however, further analyses, perhaps focussed oncollieries C, K, P, Q, T and Y where working-time exposure appears to make agreater contribution to total exposure, would be a worthwhile extension of thepresent work.

o K «O ^5 h_>u ^ Tj

3 - • cs 3n> <->•

H f* a*•^ 3

3 2 r*

PIX•ao

_, T3 93T» fp *3

I | -01 >J 3

S o

o&•

tt 2

r* LA ftj§•*!W O _ro 2, &

3

2

Relat ive risk0- C , .

» of lungcancermortality

0

BEIR IV

Person-years analysis

( 95% limits)

ooVO

1 2 3 4 5 6 7 8 9 1 0

Cumulative exposure to radon daughters to age 60 (WLM)

3Tl

fsto

3 ' o »*5 5> K s.§.-• s 5n. X n 2-o^ O v* eft

OP *"* P *

g ^ l aH n> r+ ^«"* s. -o ar

o | S "c f» {? g(T> ^ S O

=s Si ^ s>.

vO ^

&§|

D3 O

It

Relative riskof lungcancermortality

0u

BEIR IV

Case —referent study

( 95% limits)

7 8 10

Cumulative exposure to radon daughters to age 60 (WLM)

VOo

91

ACKNOWLEDGEMENTS

The author gratefully acknowledges financial support from the NRPB and theCommission of the European Communities; also, helpful advice from Dr RE Eltonand Dr RM Agius of the Department of Community Medicine, EdinburghUniversity; from Dr NP Crawford, Mr JF Hurley and Mr A Bradley of the IOM;from Mr D Page and Dr D Smith of British Coal; and from Mr J Miles, MrMC O'Riordan and Dr B MacGibbon of the NRPB. The continuing enthusiasmof Dr M Jacobsen, formerly of the IOM, is also gratefully acknowledged; also thegraphics work of-Mr B: Turner of the IOM, and the indispensable help of Miss AMcCarron who typed* the-.manuscript .v - Finally, .the author thanks the British CoalCorporation, and the four mining - unions •--National Union of Mine workers, theUnion of Democratic Mineworkers, the National Association of Colliery Overmen,Deputies and Shotfirers, and the British Association of Colliery Management - fortheir interest in the study.

92

93

REFERENCES

Axelson O, Sundell L. (1978). Mining, lung cancer and smoking. ScandinavianJournal of Work, Environment and Health; 4: 46-52.

Axelson O, Edling C, Kling H. (1979). Lung cancer and residency - acase-referent study on the possible impact of exposure to radon and its daughtersin dwellings. Scandinavian Journal of Work, Environment and Health; 5: 10-15.

Boyd JT, Doll .R, Faulds ;jS,.,Leiper..J..,.(1970)... Cancer of the lung in iron ore(haematite) miners: British YjournaL oh Industrials Medicine; 27: 97-105.

Breslow NE, Day NE. (1987). Statistical methods in cancer research. Vol.11 - Thedesign and analysis of cohort studies. Lyon: International Agency for Research onCancer. (IARC Scientific Publications No.82).

Butland BK, Muirhead CR, Draper GJ. (1990). Radon and leukaemia. Letter.Lancet; 335: 1338-1339.

Crawford NP, Edlin DW. (1982). Radon and thoron daughter working levels inBritish coalmines. Edinburgh: Institute of Occupational Medicine. (IOM ReportTM/82/13).

Damber L, Larsson LG. (1985). Underground mining, smoking, and lung cancer: acase-control study in the iron-ore municipalities in Northern Sweden. Journal ofthe National Cancer Institute; 74: 1207-1213.

Doll R, Peto R. (1978). Cigarette smoking and bronchial carcinoma: dose and timerelationships among regular smokers and lifelong non-smokers. Journal ofEpidemiology and Community Health; 32: 303-313.

Duggan MJ, Howell DM, Soilleux PJ. (1968). Concentrations of radon-222 in coalmines in England and Scotland. Letter. Nature; 219: 1149.

Eatough JP, Henshaw DL. (1990). Radon and prostate cancer. Letter. Lancet; 335:1292.

Edling C, Comba P, Axelson O, Flodin U. (1982). Effects of low-dose radiation -a correlation study. Scandinavian Journal of Work, Environment and Health;S(suppl.l): 59-64.

Edling C, Kling H, Axelson O. (1984). Radon in homes - a possible cause of lungcancer. Scandinavian Journal of Work, Environment and Health; 10: 25-34.

Edling C, Wingren G, Axelson O. (1986). Quantification of the lung cancer riskfrom radon daughter exposure in dwellings - an epidemiological approach.Environment International; 12: 55-60.

Fox AJ, Collier PF. (1976). Low mortality rates in industrial cohort studies due toselection for work and survival in the industry. British Journal of Preventive andSocial Medicine; 30: 225-230.

Gardner MJ, Altman DG, eds. (1989). Statistics with confidence - confidenceintervals and statistical guidelines. London: British Medical Journal.

94

Henshaw DL, Eatough JP, Richardson RB. (1990). Radon as a causative factor ininduction of myeloid leukaemia and other cancers. Lancet; 335: 1008-1012.

Hodgson JT, Jones RD. (1990). Mortality of a cohort of tin miners. BritishJournal of Industrial Medicine; 47: 665-676.

Hornung RW, Meinhardt TJ. (1987). Quantitative risk assessment of lung cancer inUS uranium miners. Health Physics; 52: 417-430.

International Agency for Research on Cancer. (1988). Man-made mineral fibres andradon. Lyon: IARC. (IARC Monographs Vol.43).

International Commission 40n,vRadiological^Protection. (1987). Lung cancer risk fromindoor exposures'^to "radon .-daughters:•*••: Oxford: Pergamon. (ICRP Publication 50).

Jacobsen M. (1981). The importance of epidemiology in research onpneumoconiosis. In: Commission of the European Communities. Epidemiology andtechnical and medical prevention of coal miners' pneumoconiosis. Industrial Healthand Safety Seminar, Luxembourg, 4-5 October 1979. Luxembourg: CEC: 5-25.(EUR 6879).

Jacobsen M, Smith TA, Hurley JF, Robertson A, Roscrow R. (1988). Respiratoryinfections in coal miners exposed to nitrogen oxides. Cambridge (MA): HealthEffects Institute. (HEI Research Report No.18).

James AC. (1988). Lung dosimetry. In: Nazaroff WW, Nero AV, eds. Radon andits decay products in indoor air. New York: John Wiley: 259-309.

Kunz E, Sevjc J, Placek V, Horacek J. (1979). Lung cancer in man in relation todifferent time distribution of radiation exposure. Health Physics; 36: 699-706.

Lundin FE, Lloyd JW, Smith EM, Archer VE, Holaday DA. (1969). Mortality ofuranium miners in relation to radiation exposure, hard rock mining and cigarettesmoking - 1950 through September 1967. Health Physics; 16: 571-578.

McDonald JC. (1989): Silica; silicosis, and' lung cancer. British Journal of IndustrialMedicine; 46: 289-291.

Miller BG, Jacobsen M. (1985). Dust exposure, pneumoconiosis and coalminers1

mortality. British Journal of Industrial Medicine; 42: 723-733.

Morrison HI, Semencier RM, Mao Y, Corkhill DA, Dory AB, De Villiers AJ,Stocker H, Wigle DT. (1985). Lung cancer mortality and radiation exposure amongthe Newfoundland fluorspar miners. In: Stocker H, ed. Occupational radiation safetyin mining. Vol.1. Toronto (Ont): Canadian Nuclear Association: 365-368.

Muller J, Wheeler WC, Gentleman JF, Suranyi G, Kusiak P. (1985). Study ofmortality of Ontario miners. In: Stocker H, ed. Occupational radiation safety inmining. Vol.1. Toronto (Ont): Canadian Nuclear Association: 335-343.

National Research Council, Committee on the Biological Effects of IonisingRadiation. (1988). Health risks of radon and other internally depositedalpha-emitters. Washington (DC): National Academy Press. (BEIR IV).

95

Nero AV. (1988). Radon and its decay products in indoor air: an overview. In:Nazaroff WW, Nero AV, eds. Radon and its decay products in indoor air. NewYork: John Wiley: 1-53.

Ogden TL. (1974). A method of measuring the working-level values of mixedradon and thoron daughters in coalmine air. Annals of Occupational Hygiene; 17:23-34.

Qiao YL, Taylor PR, Yao SX, Schatzkin A, Mao BL, Lubin J, Rao JY, McAdamsM, Xuan XZ, Li JY. (1989). Relation of radon exposure and tobacco use to lungcancer among tin miners in Yunnan Province, China. American Journal ofIndustrial Medicine; 16: 511-521.

Radford EP, Renard tKGStCv'(1984).,. Lung cancer in Swedish iron miners exposedto low doses of radon daughters. New England Journal of Medicine; 310:1485-1494.

Ruosteenoja E. (1991). Indoor radon and risk of lung cancer: an epidemiologicalstudy in Finland. Helsinki: Finnish Centre for Radiation and Nuclear Safety.(Report STUK-A99).

Schoenberg JB, Klotz JB, Wilcox HB, Nicholls GP, Gil-del-Real MT, StemhagenA, Mason TJ. (1990). Case-control study of residential radon and lung canceramong New Jersey women. Cancer Research; 50: 6520-6524.

Solli HM, Anderson A, Stranden E, Langard S. (1985). Cancer incidence amongworkers exposed to radon and thoron daughters at a niobium mine. ScandinavianJournal of Work, Environment and Health; 11: 7-13.

Storer BE. (1984). PECAN user's notes. Version 2.3. Wisconsin Clinical CancerCenter, University of Wisconsin.

Strandon E. (1988). Building materials as a source of indoor radon. In: NazaroffWW, Nero AV, eds. Radon and its decay products in indoor air. New York: JohnWiley: 113-130.

Swedjemark GA.. (1979).; Indoor, measurements^, of natural radioactivity in Sweden.In: Commission of the European Communities. Proceedings. Seminar on theradiological burden of man from natural radioactivity in the countries of theEuropean Communities, Le Vesinet, Paris, 4-6 December 1979. Luxembourg: CEC.

Thomas DC. (1977). Addendum to 'Methods of cohort analysis: appraisal byapplication to asbestos mining' by Liddell FDK, McDonald JC, Thomas DC.Journal of the Royal Statistical Society A; 140: 483-485.

Whittemore AS, McMillan A. (1983). Lung cancer mortality among US uraniummines: a reappraisal. Journal of the National Cancer Institute; 71: 489-499.

World Health Organization. (1967). International classification of diseases. Manualof the international statistical classification of diseases, injuries, and causes of death.Based on the recommendations of the eighth Revision Conference, 1965, andadopted by the nineteenth World Health Assembly. Geneva: WHO.

96

World Health Organization. (1977). International classification of diseases. Manualof the international statistical classification of diseases, injuries, and causes of death.Based on the recommendations of the Ninth Revision Conference, 1975, andadopted by the twenty-ninth World Health Assembly. London: HM StationeryOffice.

World Health Organization. (1957). International classification of diseases. Manualof the international statistical classification of diseases, injuries, and causes of death.Based on the Recommendations of the Seventh Revision Conference 1955, andadopted by the Ninth World Health Assembly under the WHO NomenclatureRegulations. Geneva: WHO.

Wrixon AD, Green-BMR.wLomas-PR,,.Miles JCH, Cliff KD, Francis EA, DriscollCMH, James AC, O!Riordan: MC. (1988). Natural radiation exposure in UKdwellings. Didcot (Oxon): National Radiological Protection Board. (NRPB-R190).

97

APPENDIX 1

Quantities of radioactivity and their units of measurement

Measured radon and thoron daughter levels which form the basis of the exposurevariables analysed in this study, were recorded in units of milli-working levels(mWL). The working level (WL)is the traditional unit of energy concentration fordecay products of radon and thoron.

One WL of radon ^daughters-is, present when; the alpha energy per litre of air,released by decay to 210Pb, is 1.3 x 105 MeV. The same definition holds forthoron daughters, except that decay is to 208Pb. The radioactivity concentrationof the daughters is measured by the equilibrium equivalent decay productconcentration (denoted EEDC 2 2 2 for radon daughters, EEDC 2 2 0 for thorondaughters), in units of Becquerels per cubic metre (Bq m~3). The EEDC is aweighted mean of the individual daughter activity concentrations: if each of thedaughters had an activity concentration equal to the EEDC, the total alpha energyreleased would be equal to the energy concentration of the combination actuallypresent. Conversion from energy concentration to EEDC is given by

1 WL = 3740 Bq m~3 (radon daughters);

1 WL = 276 Bq m~3 (thoron daughters),

(Nero, 1988).

The equilibrium factor is the ratio of the EEDC to the activity concentration ofradon (or thoron) gas.

The units of cumulative exposure which were used in statistical analyses wereworking-level hours (WL hour). Thus, 1 hour worked in an energy concentrationof 1 WL gives a cumulative exposure of 1 WL hour. However, relative riskswere presented using the more usual working-level month (1 WLM = 170 WLhour). This was to-allow easier comparisons to published work.

98

Table A2.1 Results of a survey of radon

99

APPENDIX 2

and thoron daughter levels carriedout at 11 British collieries between 1972 and 1980.

Col l ie ry(Sampl ingdates)

C(Dec 1976-Feb 1977)

F(Aug 1976-Sep 1976)

K(Nov 1972-Dec 1972)

Vent i lat iondistance

Seam* (m)

Warwickshire 2803Thick Coal 2796

28921683181719091974294929992938

Seven Feet 167916811683152415231532

Lower Nine 2409Feet 2411

2413

Beeston 548056855575

11300

Flockton 3510360034208500

Vent i lat ionquant i ty

(m3 sec"1)

4.664.664.667.107.107.107.103.683.683.68

18.4018.4018.4016.0016.0016.00

7.307.307.30

11.0011.0011.00

-

8.258.258.25_

Radiat ion

RadonDaughters

5.88.57.11.13.41.12.12.01.17.8

0.72.30.41.10.71.2

1.00.60.8

5.010.26.7

15.8

8.33.86.5

11.8

levels (mWL)

ThoronDaughters

8.58.38.41.53.52.72.42.41.44.3

0.40.60.50.50.60.6

0.70.60.9

3.03.32.65.6

3.73.54 .29.2

100

Table A2.1 Continued

Colliery(Sampl ingdates)

H(Nov 1977-Jan 1978)

P(Apr 1976-May 1976)

Q(Apr 1972-Nov 1972)

Ventilation Ventilationdistance quantity

Seam* (m) (m3 sec"1)

Busty 350035003500280028002800

Til ley 460046004600

(Upcast shaft) 730073007300

(Workshop, Not applicablesurface) Not applicable

Not applicable

Kailblades 323132003186

Parrot 475547414737368536823682

High Main 8900890089009100910091009100910091009100910091009100

-----_

--

---

4.604.604.60

4.704.704.708.708.708.70

4.774.774.774.124.124.124.124.124.124.124.124.124.12

Radiat ion

RadonDaughters

6.56.76.55.66.76.5

6.47.79.4

12.212.512.3

0.30.40.2

3.44.33.2

3.54.31.44.51.23.5

16.913.925.618.623.721.919.421.326.026.116.016.719.6

levels (mWL)

ThoronDaughters

2.83.12.83.43.83.0

4.63.83.9

5.95.85.4

0.20.20.1

4.63.74.6

2.54.03.44.43.13.1

5.75.94.46.26.47.97.66.87.18.06.95.96.5

101

Table A2.1 Continued

Radiation levels (mlVL)Col 1 iery(Sampl ingdates) Seam*

Q High Main(Apr 1972-Nov 1972)

(Jun 1979-Jul 1979)

(Apr 1980-Jun 1980)

(Apr 1972- WaterlooNov 1972)

Vent i lat iondistance(m)

910071007100710071007100710071009500950038003800920075257685767563106517657269886756699467627000676870066774701267837021678910314103147126689771296902

140022003000

Vent i lat ionquant i ty(m3 sec"1)

4.127.807.807.807.807.807.80-

5.095.093.793.795.735.305.305.304.954.954.954.664.664.664.664.664.664.664.664.664.664.664.66--4.664.664.664.66

7.567.567.56

RadonDaughters

25.69.910.519.018.919.016.323.123.122.016.418.013.429.919.432.927.617.425.710.521.413.014.715.014.724.811.111.714.215.816.052.848.619.416.521.219.7

3.38.78.8

ThoronDaughters

5.57.86.88.86.96.47.66.77.28.810.66.25.95.96.87.75.85.94.94.55.85.87.66.29.76.74.84.66.27.16.111.514.75.76.25.17.5

2.84.35.9

102

Table A2.1 Continued

Radiation levels (m\VL)Col 1 iery(Sampl ingdates)

Q(Apr 1972-Nov 1972)

Vent i lat ion

Seam*

(Pit bottom,air fromHigh Main andWaterloo)

distance(m)

5012500

Ventilationquantity Radon(m3 sec"1) Daughters

2.110.2

ThoronDaughters

1.78.1

(Jan 1973-Feb 1973)

Busty

Harvey

329232923292

2560297229722972

2.002.002.00

2.402.402.402.40

10.49.66.0

4.45.24.84.2

3.15.02.8

3.42.82.84.3

(Pit bottom,air fromHarvey)

7.1 3.8

Old Coal(Oct 1973-Mar 1974)

W(Apr 1974)

Big Vein

Meadow Vein

Big Vein

2730279027901440144014601480134015301550700

358035803400

153016201560

259027102620207220722072

8.408.408.405.885.885.8815.8015.8015.8015.80-

11.2011.2011.20

11.7711.7711.77

16.7016.7016.70--_

4.63.03.41.71.60.71.00.53.50.91.7

3.13.72.2

1.70-0

1.62.32.00.80.81.2

0.31.71.71.21.51.90.90.61.00.71.1

3.02.22.1

1.20.21.5

1.71.91.81.31.31.3

103

Table A2.1 Continued

Col 1 iery(Sampl ingdates)

X(Feb 1976)

Y(Apr 1978-Feb 1979)

Vent i lat iondistance

Seam* (m)

Meltonfield 120017001800

Dunsil 1600180018001800

Beamshaw 22002400

Yard 2970296027002460409039203590449544954180301034602870

Low Main 37503230

Vent i lat ionquant i ty(m3 sec"1)

5.946.866.86

8.158.158.158.15

6.996.99

-----

5.405.405.405.405.406.505.45

8.709.20

Radiation 1

RadonDaughters

-0.11.80.3

0.20.40.31.6

3.02.0

6.45.95.01.81.71.36.56.56.47.36.92.89.4

9.02.8

eve Is (mWL)

ThoronDaughters

2.92.72.0

2.62.43.23.7

3.05.1

4.14.65.13.03.83.04.24.74.05.05.22.24.7

2.91.7

* Locations not specific to a seam appear in brackets.

104

105

APPENDIX 3

This Appendix gives the 'PANDA1 questionnaire on respiratory symptoms andsmoking used at routine PFR surveys.

106

S"'" 70° Form B.90I

INSTITUTE OF OCCUPATIONAL MEDICINE

PNEUMOCONIOSIS FIELD RESEARCH

INSTRUCTIONS FOR PANDA 11

A PERSONAL DATA

Date of birth Month January-December - 01-12Enter code for the month with leading zero for January - September

Year Enter the last two digits of the year (i.e. tens and units)

B RESPIRATORY SYMPTOMS QUESTIONNAIRE

Ql, 2. 3 and 4 If YES enter 'Y'. if NO enter 'N'

Qla, 2a. 3a and 4a If answer to previous question is NO then leave blank,otherwise enter 'Y' for YES and 'N' for NO

Q5. 6 and 7 . If YES enter 'Y'. if NO enter 'N1

Q8 If YES enter 'Y' and ask Qs. 8a - 8d« leaving Q8e blankIf NO enter 'N' and ask Q8c. leaving Qs. 8a - 8d blank

Q8a Enter «C' for Cigarettes only•P' for Pipe only•B' for Cigarettes and Pipe

Q8b and 8c Enter number of cigarettes with leading zeroif less than 10 or '00* if none

Q8d Enter 'X* for X ounce only, otherwise enter thenumber of ounces or '0' for none

Q8c Enter «Y' for YES and *N' for NO

Q9 Enter 'Y' for YES and 'H' for NO

Q9a If answer to Q9 is NO leave blank, otherwise enter•A' for Asthma•B' for Bronchitis•C1 for Cold•D1 for Bronchitis and Asthma'F' for InfluenzaS' for some other chest illness•X* for chest injury (not regarded as chest illness).

Do not amend answer to Q9

CIO ANTHROPOHETRIC DATA AND VENTILATORY FUNCTION

Enter leading zeros where appropriate

ADDITIONAL QUESTIONING

If difficulty is experienced in obtaining 'YES* or 'NO* apply the following:' "I know this is difficult but please try to answer 'YES' or 'NO*

I will repeat the question"If the answer is again equivocal, record 'NO'

APU 2009

107

PANDA 11

A PERSONAL DATA

NAME .._ . „„.

COLLIERY LETTER

X.RAY NUMBER t

B RESPIRATORY SYMPTOMS QUESTIONNAIRE °ATE OF BlRTH (MON™ * YEAft) «

1

s

.PREAMBLE: "I am going to ask you some questions about your chest - about cough and spit, forexample. Please try to answer 'Yes* or 'No*. Your answers will be treated confidentially.

COUGH

Q. 1

Q.la

Q. 2

Q.2a

Do you cough like this on most days for as much as 3 months in the year?

Do you cough during the rest of the day? - I don't mean just at the end of your shift. ... ...

Do you cough like this on nost days for as much as 3 months in the year?

PHLEGM

Q. 3

Q.3a

Q. 4

Q.4a

Do you bring up phlegra when you get up or first thing in the morning?...

Do you bring up phlegm 1 iV.e this on most days for as much as 3 months in

Do you bring up phlegm during the rest of the day? - I don't mean just at

Do you bring up phlegc like this on most days for as ouch as 3 months in

BREATKLESSHESSQ. 5

the year?

the end of your shift.

the year? '. .. ...

10

1 1

12

1)

...

...

...

Do you have to walk slower than other people on level ground because of your chest? ...

WHEEZiHGQ. 6 Do you ever have wheezing or whistling in your chest? - I don't mean only

LEATHER

Q. 1

SKOKIKGq. 8Q.8a

Q.8b

Q.8c

Q.saQ.Se

Do you smoke? (If 'Yes'. Q.8a-8d? If •No*. Q.Se) ...

Kow oany cigarettes do you smoke per day on Mondays to Fridays? ...

How many cigarettes do you smoke per day on Saturdays and Sundays? ...

How nany ounces of tobacco do you smoke p«r week? (Record in ounces, x =

Have you ever smoked as nuch as one cigarette .rj>er day for one year? ...

CHEST ILLNESSESQ. 9

Q.fia

when you have a cold.

is

X)

14

IS

IS

17

16

... It

20

21

22

24

2«

IT

tl

In the last 3 years have you had a chest illness that has kept you off work for more than a week?... tf

10

(A^Amtham; B= Bronchitit: C=Cold; D^ Bronchi tin * A*th*>*; F= Influent*; S = Soae other chttt illnettX ** Not • chemt Ulna**)

C ANTHRGPOUETRIC DATAHeight (cos)

Sitting Height (ess)

Weight(kgos) J7

C YEHTILATORY FUHCTIONSecond Blow

Third Blow

Fourth Blow

{F.E.V. 40

F.V.C. 43

{F.E.V.

F.Y.C.

{F.E.V.

F.V.C.

11

34

J7

40

43

4*

40

51

• 9

*

JJ

3«

11

42

45

41

it

54

S7

108

109

APPENDIX 4

The seam-mean method of exposure estimation

One of the most striking features of the data on which calculation of working-timeexposure in this study were based, is the contrast between the superabundance ofinformation on time worked - both durations and locations - and the paucity ofinformation on radon and thoron daughter levels. Data on time worked (ie.Attendance Records and Occupational Histories) comprise some hundreds ofthousands of records; • for coalface workers in particular, the information providedspecifically by Attendance .Records'is extremely detailed. Thus, it is possible toestablish for each. man - employed .at the coalface during the PFR research, thefaces at which he worked, and for how long. Also, such times worked areavailable by inter-survey period in Phase One, and by quarter in Phase Two.

Radioactivity data, on the other hand, are extremely sparse, consisting of 161measurements at 11 collieries, with 15 of these at a non-PFR colliery. Of 896coalfaces listed in the PFR History of Faces, measurements of radon and thorondaughters used in this study were made on only 42; of 42 coal seams listed,measurements were made in only 20. Furthermore, at nine of the 10 researchcollieries, surveys of radioactivity were carried out over short periods of only a fewmonths during the 1970s (see Appendix 1). The one exception was at colliery Q,where a first series of measurements made between April and November 1972, wasfollowed by a second during June and July 1979, and a third between April andJune 1980.

In view of the few radioactivity data available, the strategy adopted for calculationof exposure was to average radon and thoron daughter measurements withinindividual coal seams, and to use the resulting means as estimates of theradioactivity levels applying to any Occupational Group working in the seams.The rationale was that rocks surrounding coal faces in a single seam would oftenhave fairly similar mineralogical compositions. Since radon and thoron gasconcentrations would depend on the ore-types present in the rock (as well as manyother factors - ventilation • regimes, amount of ground water, type of miningmethod etc.), areas located in the same seams might be expected to be closer inradon and thoron daughter level, on average, than areas in different seams. Thismethod was preferred to a simpler scheme, whereby a single mean radioactivitylevel was to be assigned to all underground work at each colliery. Table A4.1shows mean levels of radon and thoron daughters, by seam. Differences betweenseams are small, but in some cases, particularly at colliery Q, sufficiently large tosuggest that seams should be distinguished in exposure calculations. The statisticalsignificance of seam differences (on the log scale) was examined using analysis ofvariance (Tables A4.2 and A4.3); results showed that the differences inradioactivity level were significant at the 0.001 level.

There were difficulties with the method. As mentioned above, only about halfthe seams worked during PFR had radioactivity measurements made in them.Time worked in 'unmeasured seams' was therefore assigned the colliery mean.Also, Attendance Records Time worked in so-called 'Elsewhere underground'Occupational Groups (ie. non-coalface Groups) was often not specific to anyparticular seam, and again this time was assigned the colliery mean. Collierymean values were also used for records of time worked underground which hadbeen obtained from Occupational History questionnaires. This was because the

110

classification of such times referred only to types of coalmining environment, andnot to coal seams.

Exposures calculated using seam means were therefore used in analyses ofmortality. To gauge the effect of using seam means where possible, instead ofcolliery means, exposures were also calculated by the latter method, andcomparisons made. Figures A4.1 and A4.2 show histograms of differencesbetween total cumulative exposures (WL hour) to radon and thoron daughtersrespectively calculated by the two methods. The distributions are closely centredaround zero, with 91% and 99% of differences in radon and thoron daughterexposure respectively, less than 20 WL hour in absolute value. Clearly, themethods give very similar results.

A more sophisticated method of exposure estimation, which was considered althoughnot used, was based on the idea of using regression functions equating meanradioactivity levels to combinations of certain concomitant variables. To beuseful, such concomitants had to show a strong relationship with radon and thorondaughter levels, and had to be available, or estimable, for at least the most heavilyworked coalfaces current during PFR. Regression equations would be calculatedfrom available radioactivity data and values of the chosen concomitants pertainingto the locations where the measurements were made. These equations could thenserve as 'predictors' of radioactivity levels at other locations and other times forwhich no measurements were available, using known values of the concomitantvariables pertaining to these new locations and times. The method was usedsuccessfully by Jacobsen el al (1988) in a study of the incidence of respiratoryinfection in coalminers, in relation to nitrous fume exposure.

Two concomitant variables were considered:

a) ventilation distance - the distance travelled by the ventilating airstream from theintake shaft to the location of interest;

b) ventilation quantity - the volume of air (m3) passing the location in unit time.

Ventilation distance appeared to have a fairly strong relationship with both radonand thoron daughters (Figs A4.3 and A4.4 respectively); regression analysesconfirmed the existence of increasing trends (P < 0.001 for Rn, P < 0.01 forTn), after allowing for differences between seams (Tables A4.4 and A4.5).Associations between ventilation quantity and radioactivity levels appeared to showdecreasing trends which were not particularly convincing (Figs A4.5 and A4.6).The effect on radon daughters was not statistically significant, and on thorondaughters, just reached significance (P < 0.05), after allowing for differencesbetween seams (Tables A4.6 and A4.7).

Despite the strength of its relationship with radioactivity levels, ventilation distancewas judged unsuitable for use as a concomitant variable given the scope of thepresent study. This was because it was not available from routine PFR data.In principle, estimates could have been obtained from historical colliery plans, butthis exercise would have increased the effort required for exposure estimation, atthe expense of resources allocated to mortality analysis. Some consideration wasgiven to using as a surrogate variable for ventilation distance, time taken to travelfrom pit bottom to the coalface, a quantity which was available from routine PFRdocumentation. Two important objections were, first, that travelling time was anunreliable guide to distance travelled, since it contained unknown components ofwalking time and time on locomotives, and secondly that routes taken by miners

Illtravelling to the coalface did not necessarily reflect the distance travelled by theventilating airstream.

Ventilation quantity enjoyed an advantage over ventilation distance, in beingavailable from PFR documentation, as a coalface average figure. However,relationships with radon and thoron daughter levels were weak, and mainly for thisreason, it was decided not to use the variable in a prediction equation.

11£

I 8 ?-3- 3 =r

fan CL

-3 P"i

a" <tCD 3"2 °

re ""ft3 5"V> rr.

|S.3 3

ll3 O

•-h

3 r»ft O

o

siQ. C

^ s3 0» c» n3 fB

Men

9000

8000

7000

6000

5000

4000

3000

2000

1000

-50 -40 -30 -20 -10 0 10 20 30 40 50 60 70 80

Rn daughter exposure (WL hour) - seam mean estimate minus colliery mean estimate

2!era'

§

3 s q2 »

(W O.

V O

s rO

B

1

3 O3 |-h

faD

3

SO rB O

o 2o. £.

Men

9000

8000 -

7000 -

6000

5000

4000

3000

2000

1000

0

~7

-30 -20 -10 0 10 20 30 40

Tn daughter exposure (WL hour) - seam mean estimate minus colliery mean estimate

uu

I?Op"ct-io>

£ 50k>

< g &o

§. I § 40

-• so

ro ~ &e. ^wI 2..3 ' Radon 30£j Hit

3 S. 3 daughters3"! (mWL)

ll" 2°•*"* (B <-Ag _ ^

o ^S 33 O ff

- en 1 r\r» c 100 0 g

TJ S5o o. n>•—• — t 33 2 S>o 3

^ ° 2, 03 ^ "W (U Q_

i i.§n> ^2n & .10

I I I I I I I I I

A

— —A

— —

A

- A -

A

A A&y\

A A A. A Atj A Q

A A A ^

A AA A^ AS A

Ax A A

AA A

^ A A^ A -

JA A A

y^&\ /V^ AA

A^ A ^ A ^A^^^ A A?A\A A^p^A "V^ ^ .— £iii£* A A ^

I I I I I I I I I

0 1500 3000 4500 6000 7500 9000 10500 12000 13500 15000

Ventilation distance (m)

21cl-tn

± 12.5

< g on

§-. | SI» ° 10.03 *"*• p

»»* (BSu >-*Oq

5' 3Si Q 3 rp,S 2.-? Thoron3 S. S daughters y c3 8 |. (mWL)3 5' oo"2. 5. ~

ST " ^§§i 5.03 ^3 Q)

Crt c/j

^f H

° s- n

•8 §O P^ M

5* 55" r*«-*• r*0 3 0 2.5

n """3 „ ttfl> 5? ?•to CD Oen ^ *nc S o

3 D.n & n n

I I I I I I I I I

-

A

A

— —

A

A A

A & A

A A A A

A A A

^e A =A &

A A AAA^ aA AA^A A

A A A A A A AA

A ^ AA A AA

A A A A A %A

A AA A A A

A AAA AAi/>^ AAA A A A AA A

AA A A A

A i_i

1 ^A A

if A AA *V

A1 A

A^H AIA I ! 1 1 1 1 1 1

vǤ- 0 1500 3000 4500 6000 7500 9000 10500 12000 13500 15000

Ventilation distance (m)

uw

cft

% 50in

< 3 00

III 403 <* n

». °w

•g jSf Radon 30

5" 8 3 daughtersg f | (mWL)ft ft (JQ O /~\

(/)

n 5' """2 _ oo3 3-g ft 3

•"* H- ft

"o § 1 105' y (5" ^ 3•a o g

"* ^ ^ft 3 _ ,~8 £° 0§ So- £• r1s— O W• 3 O.

OX5 3C

1. 1 - 1 o

i i i i i i i i i

-

A

- A

A

A AA . An a

~ A ^ A A ~A

A^ |^ A

A A

A

~ A A ^ AA

A A

A A S A A

i ^ AA A

gS A ^ A ^A A $ A A A

2 A ^ ^^S A A A AA A— ^ A 'S A A 23s Q —

1 1 1 1 1 1 1 1 1£1 C*«£ 0 2 4 6 8 10 12 14 16 18 2Ci ftP 3

Ventilation quantity (m 3 sec "')

1 U.U

21(ro"

.*• 195a\ i /_ . ^

5 3 5 ?o ? oX. &> {aC 0. ?j3 O ^^

^n » " 10.0° ~lr(a (jq=T -0 3

•g ^.3 Thoron| 8 3 daughters 7 s

B 1 1 (mWL)f6 (^ 0^^3 C/l

e N«» >— k

5^ j3 jsjr& oo3 3-s"| 5-°"§ ^ s•j C/i Ot~|L " 3

»« ^^ ^S O 3ft) ~4-"^ ^ Crt

S - S o 2 . 50 ff. ->»g 0"3 ;..».<->D- S' S"

N ' W Q

,£> 3

C£3

|| 0.0

i i i i i i i i i

-

A

— -

A

A A

i

A AA A. A A

A A A

A A A S

& AA A A

A x\S> A A AZ i

A

- A AA A A

A .*,

A A ^ A A A AA A

A £&.

A A A ^ A

A A A A AA A A

A ^ A AA A

A AA A S

A A A AA A

A Arr y^\ A

1 1 1 1 1 A| | | |

^ 0 2 4 6 8 10 12 14 16 18 2

Ventilation quantity (m 3 sec "')

118

Table A4.1 Mean radon and thoron daughter levels (standard deviations in brackets)measured at 11 British collieries during the 1970s and 1980, by seam.

Radondaughters

Colliery Seam

C Warwickshire thick coal

F Seven feetNine feet lower

H* BustyTilley

K FlocktonBeeston

P ParrotKailblades

Q High mainWaterloo

T Harvey (Towneley)Busty

V Meadow veinOld coalBig vein

W Big vein

X Duns i 1Beams hawMeltonfield

Y Low mainYard

(mWL)

4

10

67

79

33

206

58

023

1

020

55

.00

.07

.80

.42

.83

.60

.42

.07

.63

.30

.93

.14

.67

.80

.05

.00

.45

.62

.50

.70

.90

.22

(3

(0(0

(0(1

(3(4

(1(0

(8(3

(1(2

(0(1(0

(0

(0(0(0

(4(2

.00)

.67)

.20)

.41)

.50)

.36)

.77)

.43)

.59)

.14)

.15)

.16)

.34)

.85)

.36)

.75)

.63)

.66)

.71)

.96)

.38)

.52)

Thorondaughters No. of(mWL)

4

00

34

53

34

64

33

012

1

242

24

.34

.53

.73

.15

.10

.15

.62

.42

.30

.87

.33

.42

.63

.97

.15

.43

.55

.97

.05

.53

.30

.12

(2

(0(0

(0(0

(2(1

(0(0

(1(1

(0(1

(0(0(0

(0

(0(1(0

(0(0

measurements

.93)

.08)

.15)

.39)

.44)

.72)

.35)

.69)

.52)

.82)

.55)

.65)

.19)

.68)

.51)

.49)

.28)

.59)

.48)

.47)

.85)

.92)

10

63

63

44

63

503

53

3113

6

423

213

* Non-PFR colliery

119

Table A4.2 Analysis of variance of the natural logarithmt of radondaughter levels (mWL).

Source of variation

Col 1 iery

+ Seam (within colliery)

Residual

df

10

11

131

MS

16.20***

1 .08***

0.26

Total t 152 1.36

*** P < 0.001

t Actually, loge (Rn + 0.3 mWL), since a value of -0.2 mWL was recorded asdata.

t Three of the 161 measurements were in a surface workshop. A further fiveunderground measurements were not specific to a seam.

120

Table A4.3 Analysis of variance of the natural logarithmt ofthoron daughter levels (mWL).

Source of variation

Col 1 iery

+ Seam (within colliery)

Residual

Total *

df

9

11

131

152

MS

5

0

0

0

.574***

.289***

.088

.463

*** P < 0.001

t Logg (Tn + 0.3 mWL) was used, in agreement with the transformation usedfor radon daughter levels.

f See footnote to Table A4.2.

121

Table A4.4 Analysis of variance of the natural logarithm t of radondaughter levels (mWL).

Source of variation df MS

Colliery 10 16.19***

+ Seam (within colliery) 11 1.09***

+ Ventilation distance (m) 1 3.18***

Residual 129 0.24

Total * 151 1.37

*** P < 0.001

t See footnote to Table A4.2.

•f- One measurement at colliery T had no ventilation distance recorded.

122

Table A4.5 Analysis of variance of the natural logarithmt ofthoron daughter levels (mWL).

Source of variation

Colliery

+ Seam (within colliery)

+ Ventilation distance (m)

Residual

Total *

df

10

11

1

129

151

MS

5.575***

0 . 290***

0.702**

0.083

0.466

*** P < 0.001 ** P < 0.01

t See footnote to Table A4.3.

t See footnote to Table A4.4.

123

Table A4.6 Analysis of variance of the natural logarithm! of radondaughter levels (mWL).

Source of variation

Col 1 iery

+ Seam (within colliery)

+ Ventilation quantity (m3 sec"1)

Residual

Total £

df

9

10

1

107

127

MS

15

1

0

0

1

.93***

.16***

.79

.24

.43

*** P < 0.001

t See footnote to Table A4.2.

t All nine underground measurements at colliery H could not be assignedventilation quantities. The same applied to 16 non-face measurements at sixPFR collieries.

124

Table A4.7 Analysis of variance of the natural logarithmt ofthoron daughter levels (mWL).

Source of variation

Colliery

+ Seam (within colliery)

+ Ventilation quantity (m3 sec"1)

Residual

Total t

df

9

10

1

107

127

MS

5.534***

0.291***

0.340*

0.084

0.489

*** P < 0.001 * P < 0.05

t See footnote to Table A4.3.

f See footnote to Table A4.6.

125

APPENDIX 5

Smoking Habit - Derivation of Codes

For convenience, items 8 to 8e of the respiratory symptoms questionnaireadministered from PFR2 onwards, and given in full as Appendix 3, are reproducedhere:

Respiratory symptoms'questionnaire "--items on smoking habit

8 Do you smoke? (If Yes, 8a-8d; If No, 8e.)

8a Do you smoke cigarettes, a pipe or both?

8b How many cigarettes do you smoke per day on Mondays to Fridays?

8c How many cigarettes do you smoke per day on Saturdays and Sundays?

8d How many ounces of tobacco do you smoke per week?

8e Have you ever smoked as much as one cigarette per day for one year?

When questionnaire responses obtained at PFR2 and 3 were computerized duringthe 1970s, there was some summarization of data. Items 8 and 8e werecombined to form a new item, '88e', which had three codes - 0 (Non-smoker: 8= No and 8e = No), 1 (Ex-smoker: 8 = No and 8e = Yes), and 3 (Smoker: 8= Yes). Items 8b and 8c on numbers of cigarettes smoked per day were alsocombined to form '8b8c'), by taking a weighted mean of weekday and week-endconsumption. The result was computerized in the following form:

Cigarettes per day Code (item 8b8c)

0 01-5 16-10 2

11-20 321-30 431-40 541-50 651- 7

126

Item 8d, on pipe tobacco consumption was coded thus:

Ounces per week Code (item 8d)

0il2345

>5

01234567

Data obtained during Phase 2 were not summarized before computerization.

As stated in the main report, men's smoking habits at each PFR survey at whichthey provided valid data were coded according to the following scheme:

Smoking habit Code

Non-smoker 1Ex-smoker 2Pipe smoker 3

Cigarette, or cigarette and pipe smoker:

Equivalent of 1 to 5 cigarettes per day 46 to 10 " " " 5

11 11 to 20 " " " 611 21 to 30 " " " 7" 31 to 40 " " " 8

11 " over 41 " " " 9

The key to codes 1, 2 and 3 is given in the following table:

Code Questionnaire responses

123

88e88e88e

PFR 2 or 3

- 0= 1= 3, 8b8c = 0,8d>0

PFR4 onwards

8 =8 =8 -

'No1 ,'No1 ,'Yes1

8e =8e =

, 8a =

'No1

'Yes''Pipe1

Codes 4 to 9 were used for men who smoked cigarettes, or cigarettes and pipes;i.e. men for whom combined items 88e and 8b8c equalled 3, and exceeded zero,respectively (PFR2 and 3); or for whom item 8 equalled 'Yes1 and item 8aequalled "cigarettes' or "both" (PFR4 onwards).

The tobacco consumption of these men was expressed in units of equivalentcigarettes per day, one ounce of pipe tobacco per week being taken as equivalentto 5 cigarettes per day. Numbers of cigarettes smoked, or ounces of pipetobacco, were not available at PFR2 or 3 other than in the coded forms discussedabove; estimated amounts were therefore taken as the mid-points of ranges, with

127

55.5 cigarettes per day, and 6 ounces of pipe tobacco per week being used for thehighest rates of consumption. With these assumptions, tobacco consumption('Equivalent cigarettes') reported at PFR2 or 3 was calculated as:

Equivalent cigarettes = cigarettes per day +5 x ounces of pipe tobacco per week.

Data from later surveys had not been summarized before computerization; tobaccoconsumption was calculated as:

Equivalent;-cigarettes —••>/! x- cigarettes per week-day +2/7 x cigarettes per week-end day +5 x ounces of pipe tobacco per week.

128

129

APPENDIX 6

Results bearing on the reliability of exposures

A6.1 Time Trends in Radioactivity Levels

Estimates of cumulative exposures to radon and thoron daughters for this studywere based entirely on radon and thoron levels measured in the 1970s and reportedby Crawford and Edlin. (.1982).: The method .of: estimation made no allowance forpossible variations • in -level , with calendar time; only at colliery Q were theresufficient data to r allows examination ;of trends1; .over periods greater than a fewmonths. To test the assumption, implicit in the method of estimation, that theradioactivity measurements made in the 1970s could be regarded as representativeof stable long-term seam levels, two surveys of radon and thoron daughters werecarried out jointly by NRPB and British Coal at collieries Y and C during Mayand August 1990 respectively.

Only one other colliery (F) of the 10 Phase 2 collieries was still open at the timethese measurements were made, but it was not included in the 1990 surveys, sinceits closure was imminent. However, at various times during the 1980s,measurements of radon gas activity concentrations were made at several of thePhase 2 collieries. From these data, estimates of radon daughter levels have beenobtained using assumed values of equilibrium factors (Page, British Coal ScientificServices, personal communication, 1990), and these results are also presented inthis Appendix.

Radon daughter levels are given in Table A6.1. The average of 24 measurementsmade in 1990 at colliery Y was 7.4 mWL (s.d. 6.2 mWL), which is close to themean 1978/79 level (over 15 measurements) of 5.3 mWL (s.d. 2.6). Twomeasurements made in 1983 and 1986 were somewhat higher (16.2 and 12.0mWL), but these were made in the upcast airstream (and therefore, presumably, ata long ventilation distance). At colliery C, 26 measurements were made in the1990 survey. The average- radon daughter level was 18.5 mWL (s.d. 7.5), whichis considerably higher than the mean level of 4.0 mWL (s.d. 3.0, 10measurements) measured in 1976/77. The difference is statistically significant, andremains so after adjustment for the effect of ventilation distance, which was twiceas great, on average, in the later survey. Adjustment reduces the ratio ofgeometric means between the two surveys from 5.8 to 4.8, which still represents asubstantial increase. A single measurement (2.7 mWL) was made at this collieryin 1985 in the upcast air, but unlike the corresponding values at colliery Y, it waslower than either of the two survey means. At collieries other than C or Y,levels measured during the 1980s were higher than 1970s levels, except at collieryQ, where the single measurement made in 1985 (16.1 mWL) was close to the 1972mean (17.2 mWL), and rather lower than the 1979/80 mean (21.4 mWL).

Thoron daughter levels measured in 1990 at collieries C and Y (Table A6.2) weregreater on average than those measured in the 1970s by factors of approximately 2and 3£ respectively. The differences were statistically significant. Time trendscould not be examined at other collieries, because thoron daughter levels had notbeen measured.

130

A6.2 Time-worked Information

If the assumption of constant radioactivity levels within coal seams over time is areasonable approximation to reality, then that portion of men's working time knownto have been spent in coal seams where measurements were made may be regardedas the most reliable contribution to the total exposure.

This portion was calculated for each man in the study, and expressed as apercentage of his total-time worked. The frequency distribution of thesepercentages (Figure A6.1) shows that the large majority of men spent only arelatively small proportion of their total working time in such seams; for example,less than 20% of--total-time was worked- in these seams by over two-thirds of themen. The mean percentage was 16%.

Similar calculations were carried out for five other categories of time worked.The full frequency distributions are not given here, but Table A6.3 shows theaverage proportions of total time worked in all six mutually exclusive categories.On average, exactly one half of men's working times was recorded in AttendanceRecords; the average proportion of time worked at non-PFR collieries was 16%.

I

3 008 8 Bv d 21 "S i"2 at £3 X ?•

a T3O ic o•

..O

o(-»•p

M ^ L.

81<! ff.& 3(T> fl>

O "Ow n>o 3

Men

8000

7000 -

6000

.5000

4000 -

3000 -

2000 -

1000 -

00 20 40 60 80 100

Percentage time worked in seam with measurements

Table A6.1 Mean radon daughter levels (mWL) at 10 PFR collieries, derived from three data sources.

Source of data

IOM Report TM/82/13

Colliery

C

F

K

P

Q

T

V

W

X

Y

Dec

Aug

Nov

Apr

AprJun

Jan

Oct

Apr

Feb

Apr

Date

1976-Feb

1976-Sept

1972-Dec

197 6 -May

1972-Nov1979-Jun

1973-Feb

1973-Mar

1974

1976

1978-Feb

1977

1976

1972

1976

19721980

1973

1974

1979

Mean

4.0

1.0

8.5

3.3

17.221.4

6.5

2.0

1.4

1.1

5.3

(s.d.

( 3.0)

( 0.6)

( 3.9)

( 1.2)

( 6.4)(10.8)

( 2.4)

( 1-3)

( 0.6)

( LI)

( 2.6)

) No.

10

9

8

9

3124

8

17

6

9

15

Dec

JanSept

Nov

Jan

Jan

Dec

JulJun

Existing British Coal

Date

1985

19841989-May 1990

-

-

1985

-

1984

1984

1985-May 1986

19831986

Mean

2.7

4.33.2

16.1

9.2

4.7

7.7

16.212.0

(BC)

(s.d.

( - )

(4.9)(2.0)

-

-

( - )

-

(5.6)

(7.1)

(4.3)

( - )( - )

Special BC/NRPBsurveys

) No. Date Mean (s.d.) No.

1 Aug 1990 18.5 (7.5) 26

5 -19

-

-

1

-

5 -

5 -

8 -

1 May 1990 7.4 (6.2) 241

to

133

Table A6.2 Mean thoron daughter levels (mWL) at 10 PFR collieries, derived fromtwo data sources.

Source of data

Colliery

C Dec

F Aug

K Nov

P Apr

Q AprJun

T Jan

V Oct

W Apr

X Feb

Y Apr

IOM

Date

1976-Feb

1976-Sept

1972-Dec

1976-May

1972-Nov1979-Jun

1973-Feb

1973-Mar

1974

1976

1978-Feb

Report TM/82/13

Mean

1977

1976

1972

1976

19721980

1973

1974

1979

4

0

4

3

66

3

1

1

3

3

.3

.6

.4

.7

.6

.8

.5

.3

.6

.1

.9

(2

(0

(2

(0

(1(2

(0

(0

(0

(0

(1

Special British Coal/NRPBsurveys

.(s.d.) No. Date Mean (s.d.)

.9)

-1)

.2)

.8)

.7)

.3)

.8)

.7)

.3)

.9)

.1)

10 Aug 1990 9.1 (8 .6) 25

9 -

8 -

9 -

31 -24

8 -

17 -

6 -

9 -

15 May 1990 14.2 (9 .6) 24

134

Table A6.3 Average proportion (%) of total time work in sixcategories, for 12,361 study group members.

Time-worked category Average proportion

Attendance Records:

Seams with measurements 16

Seams with no measurements,, or 21underground work not specificto seams

Surface 11

Unclassifiable Occupational Groups 2

Occupational Histories:

Research colliery 34

Non-research colliery 16

135

APPENDIX 7

Tables of Person-years and numbers of deaths

Tables A7.1 and A7.2 give data on which Figures 4.10 to 4.13 in the main reportare based. The slight discrepancy between the tables in the person-years total isdue to rounding error.

Table A7.1 Total person-years at risk, lung cancer deaths andstomach cancer deaths, by age at risk and cumulativeexposure to radon daughters, lagged by 10 years.

Age atrisk

< 34

35-44

45-54

55-64

65-74

> 75

All

Cumulat ive exposure to radon daughters(10-year lag) (WL

0-

826300

1120800

751534

295792

109060

22300

31256186

25-

233500

779310

1141782

8126209

4030209

99964

347005524

50-

220300

724920

1367393

147473812

86183615

2480154

4897010034

100-

100100

701621

1260162

128323613

73263218

1862144

426399038

200-

60900

285300

9846143

166463412

114826533

32932711

4472914059

hour)

400-

--

192100

375033

6307106

73863311

22192812

215847432

800-

--

100

127210

329154

2700152

84382

8106298

All

1441100

3804151

600754417

6490615258

4263220788

119199837

231983506201

136

Table A7.2 Total person-years at risk, lung cancer deaths andstomach cancer deaths, by age at risk and cumulativeexposure to thoron daughters, lagged by 10 years.

Age atrisk

< 34

35-44

45-54

55-64

65-74

^ 75

All

Cumulative exposure to thoron(10-year lag)

0-

835800

1105700

1046664

6517188

2203143

42700

390283815

25-

403400

743120

798444

7123185

5569209

1581106

337225424

50-

187400

1258231

14797112

12087319

53711511

125883

479696826

100-

14500

648100

21999174

193914117

80774816

1917155

5800912142

daughters(WL hour)

200-

--

49000

482863

185854417

193899648

59705820

4926120488

400-

--

_--

100

120302

2023141

76673

3993216

All

1441100

3804151

600754417

6490515258

4263220788

119199837

231982506201

137

APPENDIX 8

Indoor exposures for the study group

Since radon gas is present to a varying extent in indoor air (and in outdoor air,but at lower levels), cumulative exposure outside working hours will have beenacquired by the men in the study group. No account was taken of thiscomponent of exposure in analyses of mortality, since no data were available onradon levels in the miners' homes,- or on the proportions of non-working timespent indoors. However, an attempt has been made to calculate a rough estimateof the magnitude of indoor-exposure, using .data from a survey of radon levels in2000 UK homes carried out recently by NRPB (Wrixon et al, 1988).

Mean radon levels reported by Wrixon et al for counties in which the PFRcollieries were situated were converted to radon daughter levels, in mWL, assumingan equilibrium factor of 0.5 (Table A8.1). These were multiplied by appropriatefactors to give an estimate of the mean cumulative exposure to radon daughters,due to indoor radon, acquired by men working at the research collieries.Exposure was calculated up to the average ages, by colliery, for which the latestrecords of exposure were available from PFR data. This was done to allow adirect comparison with total cumulative exposure gained during working time.The factors used in the calculation were estimates of the proportion of time spentindoors by working miners, and also by children (to take account of ages 0 to 15),and were based on occupancy estimates given by Wrixon et al. These authorsreported an average indoor occupancy of 92%, a figure which included both sexes.Since housewives were reported as spending more time indoors (97%), it seemslikely that the male population will spend slightly less time indoors than the overallaverage, say 87%. Miners spending approximately one-third of their workingdays at the colliery could have an occupancy of no more than approximately 67%on working days. The overall occupancy for a working miner was thereforeestimated as

s/7 67% + 2/7 87% - 70%.

The corresponding occupancy for ages 0-15 was taken as 90% (using the overallaverage occupancy).

Cumulative exposures were calculated using the formula

[ (15 x 0.9) + (AGE-15 x 0.7) ] x mWL,

where mWL denotes the county mean level of radon daughters, and AGE, thechosen age to which exposures were calculated for each colliery. The results aregiven in Table A8.2.

Only at colliery Q was the average working time exposure (805 WL hour) higherthan the estimated average indoor exposure (744 WL hour). At collieries F, V,W and X, where measured levels were low (see Table 3.1 of the main report)indoor exposure considerably exceeded working-time exposure, up to a factor of18.5 at colliery X. These large differences are partly due to the fact that mean

138

radon daughter levels at these collieries are lower than estimated indoor levels, andalso that a much greater proportion of men's time is assumed to have been spentindoors, than at the colliery. (A typical year was taken as 1740 working hours inexposure calculations; in the above calculation of indoor exposure, 6132 hours ina working year were assumed to have been spent indoors.)

139

Table A8.1 Mean radon daughter levels (mWL) for the counties inwhich the PFR collieries are situated, with numbers ofmeasurements in brackets.

County mean radonColliery

C

F

K

P

Q

T

V

W

X

Y

County

West Midlands

Mid Glamorgan

West Midlands

Lothian

Not t i nghamsh i re

Durham

Gwent

Dyfed

West Yorkshire

Durham

daughter

2.2

1.4

2.2

2.1

2.2

3.4

2.6

2.5

3.1

3.4

level (mWL)

(94)

(15)

(94)

(31)

(33)

(23)

(20)

(10)

(83)

(23)

140

Table A8.2 Average ages at the ends of PFR records of exposure;estimated average cumulative indoor exposure to radondaughters (WL hour) at these ages; average overallcumulative exposure to radon daughters (WL hour) gainedduring working time.

Colliery

C

F

K

P

Q

T

V

W

X

Y

Age

57

54

54

54

51

52

50

54

53

56

Average cumulativeindoor exposure

(WL hour)

833

503

790

759

744

1159

847

878

1073

1244

Mean working-timecumulative exposure

(WL hour)

216

54

400

147

805

333

88

72

58

272

HEAD OFFICE:

Research Avenue North,Riccarton,Edinburgh, EH14 4AP, United KingdomTelephone: +44 (0)870 850 5131Facsimile: +44 (0)870 850 5132

Email: iom@iom-world.org

Tapton Park Innovation Centre,Brimington Road, Tapton,Chesterfield, Derbyshire, S41 0TZ, United KingdomTelephone: +44 (0)1246 557866Facsimile: +44 (0)1246 551212

Research House Business Centre,Fraser Road, Perivale, Middlesex, UB6 7AQ,United KingdomTelephone: +44 (0)208 537 3491/2Facsimile: +44 (0)208 537 3493

Brookside Business Park, Cold Meece, Stone, Staffs, ST15 0RZ,United KingdomTelephone: +44 (0)1785 764810Facsimile: +44 (0)1785 764811

(A20115) IOM (R) ReportCov art 3/15/06 12:32 PM Page 2