w w w . e l s e v i e r . c o m / l o c a t e / p a i n

PAIN�

154 (2013) 1966–1972

A prospective twin cohort study of disability pensions dueto musculoskeletal diagnoses in relation to stability and change in pain

0304-3959/$36.00 � 2013 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.pain.2013.05.029

⇑ Corresponding author at: Finnish Institute of Occupational Health, Topeliuk-senkatu 41 a A, 00250 Helsinki, Finland. Tel.: +358 43 825 1392; Fax: +358 30 4742008.

E-mail address: annina.ropponen@ttl.fi (A. Ropponen).

Annina Ropponen a,b,⇑, Pia Svedberg c, Eija Kalso d,e, Markku Koskenvuo f, Karri Silventoinen g,Jaakko Kaprio f,h,i

a Finnish Institute of Occupational Health, Helsinki, Finlandb School of Medicine, University of Eastern Finland, Kuopio, Finlandc Division of Insurance Medicine, Department of Clinical Neuroscience, Karolinska Institutet, Stockholm, Swedend Pain Clinic, Department of Anesthesiology and Intensive Care Medicine, Helsinki University Central Hospital, Helsinki, Finlande Institute of Clinical Medicine, University of Helsinki, Helsinki, Finlandf Department of Public Health, University of Helsinki, Helsinki, Finlandg Population Research Unit, Department of Sociology, University of Helsinki, Helsinki, Finlandh Department of Mental Health and Substance Abuse Services, National Institute for Health and Welfare, Helsinki, Finlandi Institute for Molecular Medicine, University of Helsinki, Helsinki, Finland

Sponsorships or competing interests that may be relevant to content are disclosed at the end of this article.

a r t i c l e i n f o

Article history:Received 14 March 2013Received in revised form 13 May 2013Accepted 17 May 2013

Keywords:Disability pensionLow backMusculoskeletal diagnosesPainProspective cohort studySick leave

a b s t r a c t

Pain is known to play an important role in the pathway to becoming work disabled, in particular foraward of disability pensions (DP) due to musculoskeletal diagnoses (MSD). This prospective cohort studyinvestigated MSD-related pain stability and/or changes as predictors for DP during a 23-year follow-up.Additionally confounding factors were examined to elucidate whether familial effects (including geneticsand family background) or socioeconomic status, other pain, or use of medication would affect the asso-ciations between pain and DP. Data were available on 11,224 twins (4399 complete pairs) born before1958 surveyed through questionnaires about background factors and musculoskeletal (low back, neck,and shoulder) pain impairing work ability in 1975 and 1981. The follow-up data were collected from pen-sion registers until 2004. Cox proportional hazards regression models were used. During the 23-year fol-low-up, 508 DPs due to MSD, 166 DPs due to osteoarthritis (OA), and 162 DPs due to low back diagnoses(LBD) were granted. Musculoskeletal pain impairing work ability both measured at 1 time point and 6years apart, and either 1 pain location or multiple locations, predicted increased risk for DP due toMSD, OA, and LBD. The associations were independent of familial confounding factors and of severalinfluential background factors, including headache; migraine; use of analgesics, hypnotics, or tranquilliz-ers; life satisfaction; and education and marital status. This study concluded that musculoskeletal painimpairing work ability is an early and direct predictor for DP due to MSD, OA, and LBD.

� 2013 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.

1. Introduction

Decreased work ability due to poor health is a major publichealth problem [5,11]. Musculoskeletal diagnoses (MSD) accountfor most of the absences from work, although stress-related ill-nesses are growing in importance. Both direct and indirect costsof MSDs to the individuals, employers, and society are high[5,11], and chronic pain plays an important role in the often longpathway from onset of disease to becoming work disabled. Forexample, the ranking of risk factors for disease burden in the Glo-

bal Burden of Disease Study 2010 indicates that occupational lowback pain has a stable high ranking as a risk factor, although it iswithout attributable deaths [28].

It is known that musculoskeletal pain impairing work capacityat 1 time point 30 years earlier was a strong predictor both for dis-ability pension (DP) due to MSD and DP due to low back diagnoses(LBD) in Finnish twin data, with around 24,000 twins followed upfor 30 years [36,39]. Furthermore, accumulation of health prob-lems, ie, comorbidity due to various illnesses and symptoms pre-dicted later DP due to MSD and LBD [36,39]. However, the effectof pain (both the number of musculoskeletal pain locations, andpain located in the neck, shoulder region, or back area) was a sig-nificant predictor even when chronic conditions and other factorssuch as socioeconomic status and body mass index were taken intoaccount. Fibromyalgia-associated symptoms, including neck pain

A. Ropponen et al. / PAIN�

154 (2013) 1966–1972 1967

and stiffness, are believed to be strong predictors for DP due toMSD [31]. Studies of pain measured at one time point have beensupported by results of a Swedish twin cohort showing that painwas a strong predictor for DP due to MSD and DP in general. How-ever, the Swedish data, with almost 17,000 twin individuals, had alimited follow-up time, around 6 years [38]. Thus, in all of thesetwin cohort studies [31,36,38,39], pain showed a direct effect onthe risk for DP, ie, present even in within-pair analyses. Such anal-yses with a co-twin control method adjust for genetics and sharedenvironment given that both pain and DP have a moderate geneticcomponent that is a potential confounder of the association be-tween pain and future DP [4,7–9,34,46]. The co-twin control designis a strong extension for the case-control design as the same-sexedtwins are matched for age and sex by definition. Furthermore, thedata of discordant twins allow for the examination of evidence insupport for or against causal associations in the comorbidity oftraits. In line with twin studies, a recent study of about 6000 Finn-ish municipal workers with 8 years of follow-up revealed thatchronic pain, regardless of whether accompanied by chronic condi-tions (ie, unspecified pain not related to underlying disease), was astrong predictor for DP, even when accounting for psychosocialand physical working conditions and socioeconomic status [41].All of these studies indicate that pain as a predictive factor forDP needs to be studied with a prolonged follow-up (whether beingan early predictor), taking into account variability of pain withtime and controlling for genetics and family environment. If thesubjective experience of pain (ie, presence or absence of workimpairing pain) has causal importance for DP, then preventionand effective treatment of pain should be central to our efforts toprevent early retirement due to work disability [29].

In addition to pain as a characteristic of MSD and potential workdisability, a recent meta-analysis has shown a consistent influenceof work-related psychosocial factors on MSD [25]. Subsequently,psychosocial factors at work and in private life have been shownto influence the risk of DP [10,24,26,44]. In particular, low job con-trol [20,23] and high job strain [1,6,32] are associated with the riskof DP, also due to MSD.

The purpose of this study was to investigate self-reported MSD-related pain stability and/or changes in 2 time points 6 years apartas predictors for DP with a 23-year follow-up in a population-based cohort of Finnish twins who did not have malignancies orrheumatoid arthritis. Another aim was to investigate whether theassociation would be influenced by confounding factors such asfamilial effects or socioeconomic status, the presence of other pain,or use of medication.

2. Methods

2.1. Sample

The survey data were gathered at 2 time points, 1975 and 1981(response rates 89% and 84%, respectively), by mailed question-naires with questions on sociodemographic, health, and pain-re-lated factors. All same-sex twin pairs, that is 24,043 twinindividuals inidentified through the Finnish twin cohort [14], weresent the questionnaire. The selection criteria for this study in-cluded having not retired before the second questionnaire in1981, being a resident in Finland, and being born before 1958. Fur-thermore, twins with incomplete data on pain were excluded. Ascancer and some rheumatic diseases may themselves cause painand also may potentially affect the risk for DP, we excluded thoseregistered with rheumatoid arthritis and other inflammatory dis-eases (International Classification of Diseases [ICD]-10: M05-M14, M30-M35, M45-M46) in the Special Refund Category datain the Drug Imbursement Register (KELA) until the end of 1981.

In addition, those with malignancies (ICD-10: malignant neo-plasms C00-C97) from the Finnish cancer registry until the endof 1981 were excluded, as has been done before [31] (Fig. 1).Hence, the final sample included 11,224 twin individuals; of these4734 were twins without information from the co-twin (or the co-twin did not fulfill all the inclusion criteria), 1159 complete mono-zygotic (MZ) and 2086 dizygotic (DZ) pairs.

2.2. Follow-up

Using the unique personal identification codes of all Finnish cit-izens, the records of DP were linked to the baseline data. Informa-tion on DP was obtained from the official Finnish pension registers[8]. For the diagnoses of DP, the ICD was used. ICD-10 codes for DPdue to MSD were M00-M99, and DP due to LBD was restricted toICD-10 rubrics M45-M54, and DP due to OA to M05-M19. Equiva-lent codes for ICD-8 and ICD-9 were used for earlier time periods.The follow-up time of this study was from the date of the question-naire in 1981 to the date when DP was awarded, or until the indi-vidual began to receive old age pension, or to the date of death/emigration, or to December 31, 2004 (maximum follow-up of 23years), whichever date was earliest. Information on mortality andmigration was obtained from the Population Register Centre of Fin-land to take into account censoring and the end of follow-up.

2.3. Questionnaire data

Pain occurring between 1975 and 1981 was considered as a riskfactor for musculoskeletal pain and was inquired via 3 items on ahistory of pain in the neck, shoulder, or low back. The questionwas, ‘‘During the last years have you had pains in the back, shoul-ders or neck that make it difficult for you to work.’’ Response alter-natives were ‘‘yes’’ or ‘‘no’’ for each pain location. The responses tothese 3 items were used to calculate a summary pain score (0 to 3locations) [39]. Further, the summary pain score was used to eval-uate any change between 1975 and 1981 and categorized into 4classes, either being stable no pain (no pain at either time point),stable any pain (ie, pain remaining stable regardless of the numberof pain locations), decreased (any change from 1975 to fewer loca-tions of pain in 1981, including no pain at all), or increased (anychange from 1975, including no pain, to more pain locations in1981).

Background factors from 1981 included education (evaluatedby the question, ‘‘What kind of an education have you had, andwhat occupational degrees have you taken’’ with response alterna-tives in 9 categories and converted into years of education) [19];marital status (single, widowed, divorced, and cohabitation); andfrequency of use of analgesics, hypnotics, and/or tranquilizers fromthe reported number of days that subjects had used such medica-tion during the previous year. The question was, ‘‘During the lastyear, on how many days altogether have you used the followingtypes of medicines,’’ and the response alternatives were: havenot used, fewer than 10 days, 10 to 59 days, 60 to 180 days, andmore than 180 days. The frequencies of use of analgesics, hypnot-ics, and/or tranquilizers were subdivided into 3 categories: no use,infrequent use (1 to 59 days per year of any of these 3 medica-tions), and frequent use (60 or more days per year of medication)[12,37]. For the analyses, those with missing data on hypnoticsand tranquilizers were coded as a missing category to be includedinto the models. In addition, the possible presence of migraine wasinvestigated with the following question: ‘‘Have you ever been toldby a doctor that you have migraine’’ (yes/no), and the frequency ofheadaches was investigated with the question, ‘‘Do you have head-aches,’’ with the following response alternatives: daily, severaltimes/week, once/week, once/month, several times/year, once/year, not at all. Life satisfaction was used as a proxy for depression

Fig. 1. Flowchart for the sample definition based on The Finnish Twin Cohort.

1968 A. Ropponen et al. / PAIN�

154 (2013) 1966–1972

as it has been shown to predict depression and mortality due tosuicide [18]. Life satisfaction was measured with a 4-item scaleon levels of interest, happiness, easiness, and loneliness of life [2]and analyzed as continuous values in a range of 4 to 20, with alow score designating the most satisfied. Thus each single-unit in-crease of life satisfaction score represented a change from beingmore satisfied to less satisfied. The correlation coefficients be-tween each item and the life satisfaction sum score were 0.63 to0.80 (P < .001), and the Cronbach alpha was 0.74 in an earlier studyof the twin cohort [17].

2.4. Statistical methods

Hazard ratios (HR) with 95% confidence intervals (CI) werecomputed using the Cox proportional hazards model with follow-up time in days, and DP due to MSD, OA, or LBD as the outcomevariables. Stata statistical software, version 12.1, was used to per-form the statistical analyses. All of the analyses were age adjustedand clustered on pair identity to take into account the sampling de-sign, including twin pairs rather than independent individuals, toadjust standard errors. To account for the difference betweenmen and women, all HRs were stratified by sex to allocate menand women their own baseline hazard. The proportional hazardsassumptions were graphically tested by observing that the log-log curves for the categories of risk factors were acceptably paral-lel. Because 3 definitions of DP for outcome were used, we alsotested whether the inclusion of DP due to LBD or OA while study-ing DP due to MSD or vice versa would affect the results. Therewere no major differences in results (data not shown), hence wechose to include all individuals in the follow-up to maximize thestatistical power. Therefore, we report the results for DP due toMSD, including also DP due to LBD and OA. When we analyzedDP due to LBD as the outcome, DP due to MSD (other than LBD,but including OA) is censored (in addition to all other DP diagno-ses). When DP due to OA is the outcome, all other diagnoses,including other MSD and LBD, are censored.

To account for potential pain confounders, ie, factors poten-tially associated both with the outcome and with pain as riskfactors, we adjusted the models for the following variables: fre-quency of headache; a history of physician-diagnosed migraine;use of analgesics, hypnotics, and tranquillizers; life satisfaction;education; and marital status. After adjusting for these factorsindividually, we included them all into one model, ie, a fullmodel. None of the variables stood out as having a major effect

on the outcome, nor did they significantly attenuate the associa-tion between pain and DP, hence we report the full model only. Inaddition, the interaction between the different pain factors andage at the time of DP grant (dichotomized into 2 equal groups,the cut-off point was 55 years) was tested in the models. Thestatistical significance of the interaction term was tested by thelog-likelihood ratio test.

As the sample included twin pairs, the co-twin control settingwas applied through the use of discordant twin pairs. In this set-ting, twin pairs discordant for DP were investigated; ie, 1 twinhad a DP whereas the other twin had not been granted a DP duringthe follow-up period. Conditional Cox proportional hazards regres-sion models were used to analyze the discordant pairs by analyzingthe follow-up time to DP in relation to the follow-up time of theco-twin. In the conditional Cox models, each twin pair was strati-fied to allow them to have their own baseline hazard. This providesthe possibility to investigate whether the associations betweenpain and DP would be attributable to familial factors shared ornot by the twin pair. In the optimal case, analyzing MZ and DZ dis-cordant pairs separately would give the possibility of estimatingthe role of family background and genetic factors in the associa-tions. Because MZ twins are genetically identical at the sequencelevel, one may assume that if genetic factors play a role in the asso-ciation between risk factors and DP, then this association should beseen within DZ but not within MZ twins. However, even in theanalyses including MZ and DZ twins together, as was the case inthis study (so as to maximize the power in the analyses), it wouldbe possible to determine whether the association were to be iden-tified within both MZ and DZ pairs. In these analyses, the role ofenvironmental factors not related to family background may bemore pronounced [21].

3. Results

During the mean follow-up time of 19 years (SD 6.5, median 23years), 508 individuals were granted a DP due to MSD, including166 DPs due to OA and 162 DPs due to LBD. There were almostequal amounts of DPs for men (49%) and women (51%). The meanage at DP grant was 54 years for DP due to MSD and LBD, and 56years for DP due to OA. The frequencies and percentages for painlocations and selected background factors are presented in Table 1.

Musculoskeletal pain impairing work ability, irrespective ofwhether measured at 1 time point or 2 time points over the yearsor whether one considered a single pain location or several

Table 1Frequencies and percentages of risk and background factors for DP for MSD, OA, and LBD among 11,224 twin individuals.

Risk factors DP due to MSD (n = 508) DP due to OA (n = 166) DP due to LBD (n = 162)

Pain locations in 1981Neck 177 35% 52 31% 67 41%Shoulder 182 36% 50 30% 61 38%Back 215 42% 60 36% 73 45%

Number of pain locations in 1981None 250 49% 92 55% 71 44%1 pain location 78 15% 24 14% 29 18%2 pain locations 44 9% 12 7% 14 9%3 pain locations 136 27% 38 23% 48 30%

Change in pain locations between 1975 and 1981None 173 34% 67 40% 49 30%Stable 159 31% 42 25% 53 33%Increased 99 19% 32 19% 38 23%Decreased 77 15% 25 15% 22 14%

Background factorsAge at baseline in years: mean (SD) 42 (9.1) 45 (8.9) 42 (8.6)Education in years: mean (SD) 6.6 (1.9) 6.7 (1.8) 6.7 (2.1)Life satisfaction 8.7 (2.9) 8.8 (3.1) 8.6 (2.9)Women 261 51% 104 63% 65 40%Married 384 75% 122 74% 132 81%

Use of hypnotics and/or tranquilizersNo 400 79% 134 81% 126 78%Infrequent use (1 to 59 days/year) 34 7% 10 6% 10 6%Frequent use (over 60 days/year) 8 2% 1 1% 4 2%Missing 66 13% 21 13% 22 14%

Use of analgesics 196 39% 64 39% 54 33%No 173 34% 63 38% 59 36%Infrequent use (1 to 59 days/year) 255 50% 73 44% 83 51%Frequent use (over 60 days/year) 80 16% 30 18% 20 12%

HeadacheDaily 17 3% 6 4% 6 4%Several times/week 23 5% 8 5% 4 2%Once/week 61 12% 18 11% 22 14%Once/month 116 23% 30 18% 38 24%Several times/year 106 21% 36 22% 31 19%Once/year 70 14% 24 15% 25 16%Not at all 111 22% 43 26% 35 22%

Migraine 1981 (yes) 77 15% 25 16% 26 16%

DP = disability pensions; MSD = musculoskeletal diagnoses; OA = osteoarthritis; LBD = low back diagnoses.

Table 2Cox proportional HRs with 95% CIs of pain for those granted DP due to MSD, OA, and LBD while adjusting for age, headache, use of analgesics, hypnotics or tranquilizers, lifesatisfaction, migraine, education, and marital status and stratified for sex.

Risk factors DP due to MSD* DP due to OA� DP due to LBD�

Pain locations in 1981 HR 95% CI HR 95% CI HR 95% CINeck 1.92 1.57–2.36 1.48 1.02–2.15 2.75 1.93–3.92Shoulder 2.13 1.73–2.61 1.36 0.92–2.01 2.45 1.71–3.51Back 2.27 1.87–2.76 1.69 1.17–2.43 2.54 1.81–3.56

Number of pain locations in 1981None 1.00 Reference 1.00 Reference 1.00 Reference1 pain location 1.57 1.21–2.04 1.36 0.85–2.18 2.04 1.31–3.172 pain locations 2.07 1.48–2.90 1.44 0.76–2.73 2.34 1.27–4.313 pain locations 2.74 2.15–3.49 1.73 1.10–2.72 3.73 2.43–5.71

Change in pain locations 1975 to 1981None 1.00 Reference 1.00 Reference 1.00 ReferenceStable 3.07 2.43–3.89 1.88 1.23–2.88 3.74 2.45–5.69Increased 1.71 1.32–2.22 1.36 0.86–2.15 2.34 1.51–3.64Decreased 1.47 1.12–1.94 1.21 0.75–1.95 1.47 0.89–2.45

HR = hazard ratio; CI = confidence interval; DP = disability pensions; MSD = musculoskeletal diagnoses; OA = osteoarthritis; LBD = low back diagnoses.* DP due to MSD includes DP due to OA and LBD and censors all other DP diagnoses.� The model for DP due to OA censors all other DP diagnoses.� The model for DP due to LBD censors all other DP diagnoses.

A. Ropponen et al. / PAIN�

154 (2013) 1966–1972 1969

locations, predicted an increased risk for DP due to MSD, OA, andLBD in the 23 years of follow-up (Table 2). The association wasindependent from familial confounding (Table 3) because it wasobserved within all twin pairs discordant for DP, ie, the point esti-

mates remained significant and predicted higher risk for DP (HR1.81 to 4.12) as compared with the point estimates of the wholecohort (HR 1.58 to 3.52). In addition, the risk remained high (HR1.47 to 3.07) for DP due to MSD when several influential factors

Table 3Cox proportional HRs with 95% CIs of pain for those granted DP due to MSD during follow-up.

Risk factors Wholecohort*

All discordant twin pairs(n = 381 pairs)�

Monozygotic twin pairs(n = 107 pairs)�

Dizygotic twin pairs(n = 274 pairs)�

Pain locations in 1981 HR 95% CI HR 95% CI HR 95% CINeck 2.53 2.10, 3.04 2.06 1.33, 3.18 1.75 0.71, 4.31 2.15 1.31, 3.53Shoulder 2.72 2.25, 3.28 1.86 1.27, 2.73 1.00 0.50, 2.00 2.35 1.46, 3.80Back 2.76 2.31, 3.30 2.69 1.72, 4.20 1.50 0.71, 3.16 3.40 1.93, 6.00

Number of painlocations in 1981

None 1.00 Reference 1.00 Reference 1.00 Reference 1.00 Reference1 pain location 1.73 1.34–2.23 2.78 1.62–4.78 3.39 0.86–12.72 2.80 1.54–5.072 pain locations 2.33 1.69–3.21 2.02 1.02–4.00 0.50 0.14–1.77 3.76 1.47–9.573 pain locations 3.65 2.94–4.53 2.55 1.50–4.33 1.95 0.69–5.51 3.10 1.60–6.02

Change in painlocations 1975 to1981

None 1.00 Reference 1.00 Reference 1.00 Reference 1.00 ReferenceStable 3.82 3.07–4.77 4.12 2.05–8.29 1.99 0.59–6.65 5.35 2.20–13.02Increased 2.05 1.60–2.62 2.38 1.48–3.83 1.97 0.77–5.05 2.65 1.49–4.69Decreased 1.58 1.21–2.06 1.81 1.00–3.26 3.78 1.11–12.83 1.48 0.71–3.08

HR = hazard ratio; CI = confidence interval; DP = disability pensions; MSD = musculoskeletal diagnoses.* Adjusted for age and stratified for sex.� Conditional Cox proportional hazards regression model stratified for pair identity.

1970 A. Ropponen et al. / PAIN�

154 (2013) 1966–1972

were controlled for, eg, frequency of headache; use of analgesics,hypnotics, or tranquillizers; life satisfaction; migraine; education;and marital status (Table 2). The age at DP due to MSD or LBD dis-played a significant interaction with change in pain locations be-tween 2 time points (P < .001) and age at DP due to MSD withback pain (P = .042). The risk of DP due to MSD and LBD was higheramong those who were age 55 years or younger at the granting ofDP when compared with those over 55 years of age.

4. Discussion

Pain has been shown to be an important predictor for perma-nent work incapacity in earlier studies [36,39,41]. The present pro-spective cohort study was designed to shed further light into theassociation between pain and DP by collecting 2 measures of painseparated by 6 years among over 11,000 twins. Our results indicateclearly that musculoskeletal pain impairs work ability and is astrong predictor for DP due to MSD over a long follow-up period.The association between pain and DP was similar in magnitudeand direction irrespective of whether pain was measured at 1 timepoint or at 2 times separated by an interval of 6 years, pain was re-stricted to 1 location or present in multiple (up to 3) locations, orwhere the pain was located (ie, in neck, shoulders, or back). Hencesuch a simple pain measure as a dichotomous question could beadopted for early identification of those at risk for premature pen-sioning due to illness. These results are in agreement with earlierstudies of pain and DP due to MSD [31,36,38,39,41].

Our study has the unique possibility of controlling for the famil-ial effects through the use of the co-twin control design. This de-sign provided us with a powerful tool to analyze discordant twinpairs, ie, individuals within a twin pair who differed in DP and pain,by the optimal matching of case and control subjectss as they weresame-sexed twin pairs. Our results showed clear independencyfrom familial confounding, not only for pain measured at 1 timepoint, but also for repeated measures 6 years apart. This suggeststhat pain has a direct effect on the risk of DP due to MSD. The asso-ciations observed were also independent from various influentialfactors including headachel use of analgesics, hypnoticsm or tran-quillizersl life satisfactionl migrainel educationl and marital status.These results are in agreement with earlier studies of the Finnishtwin cohort [36,39], but also with a study of Swedish twins [38]and other samples [41]. However, some indications exist that the

stability of pain and a higher number of locations might be associ-ated with a greater risk for DP due to MSD or LBD, but less to DPdue to OA. This might reflect the influence of widespread pain[31,35], which could merit further investigation with a larger sam-ple. In the present study, the smaller number of cases in the anal-yses of DP due to OA or LBD should be noted.

As we had access to pain data from 2 time points, our resultsadd to the available knowledge of musculoskeletal pain, ie, painhas been reported to be relatively stable over the years [13], andthe influence of pain at 1 time point vs 2 time points is similar indirection and magnitude [38]. Furthermore, our results supportthe idea of a dose-response relationship [13] for pain and the riskof DP: the risk increased slightly as the numbers of musculoskele-tal pain locations increased. However, we had information only on3 musculoskeletal pain locations, and that may have limited thedose-response effect, but on the other hand we accounted for theinformation of headache, migraine, cancer, and rheumatoid arthri-tis. We found that a decrease in pain between the 2 time pointspredicted increased risk for DP. It is possible that this finding mightbe related to the fact that we only measured pain in a dichotomouson/off fashion and accounted any change from more to less pain inthis category as a decrease. This means that those with the maxi-mum number of 3 pain locations at baseline could only either re-main stable or decrease between 2 time points. An even largersample size with access to more detailed assessment of pain isneeded to elaborate whether this finding is related to the ceiling ef-fect of having 3 dichotomous categories or to true decrease in pain.To investigate the potential modifying effect of age, we also testedthe interactions between age and pain using a median split of age.The results indicate that the risk of DP due to MSD and LBD is high-er among those who were 55 years old or younger at the grantingof DP when compared to those over 55 years of age. However,these kinds of interaction analyses require a larger sample size todemonstrate effects, so the fact that we saw no such interaction ef-fect on DP due to OA should be interpreted with caution, as shouldthe interaction effect results of DP due to MSD and LBD. Althoughthe effect of age has been taken into account in previous studies[36,39,41], this appears to be one of the first to reveal this kindof interaction between pain and age.

In this study, we excluded subjects with malignancies andsome rheumatic diseases because these conditions were consid-ered themselves to be associated with pain and also potentially

A. Ropponen et al. / PAIN�

154 (2013) 1966–1972 1971

to affect the risk for DP. Hence one may assume that our resultsshould represent musculoskeletal pain that is not attributable toany severe underlying (often chronic) condition. Furthermore, aswe included headache, migraine, and the frequency of use ofanalgesics, hypnotics, and tranquillizers into our models, it is rea-sonable to assume also that other sources of pain and pain-relatedbehavior were fairly well captured in this study. The spread ofpain from one location to being widespread pain throughout thebody has been postulated to be due to a maladaptive/dysfunc-tional stress system [27], but until now there have been veryfew studies investigating DP that have accounted for the effectsof having several pain locations or pain-related symptoms[31,40]. Additionally, both depression and anxiety have beenshown to be associated with musculoskeletal pain [33,43]. Anxi-ety may be a predictor of pain, whereas depression is claimedto be linked with the natural course of pain [30,45] and poten-tially to be a consequence of pain [15,43]. In this study, we usedlife satisfaction as a measure for depression when we adjusted themodels. The life satisfaction scale used here has been shown tohave high specificity (88%), sensitivity (87%), and negative predic-tive value (approximately 100%), but also to carry a strong associ-ation with depressive symptoms [18]. Although a direct measureof depression would have been preferred, this was not availablefor this study.

The strengths of this study include the long prospective follow-up from 1981 to 2004, which made it possible to use detailed sur-vey data from 2 time points separated by 6 years. This design pro-vided the relatively rare possibility to measure stability or changesin musculoskeletal pain locations with respect to impaired work-ing capacity. Another additional strength of this large cohort isits access to high-quality DP registry data, including diagnosesand the date when DP was granted. A limitation of this studymay be related to the measure used to assess pain, which wasrather simple. However, as we had a prospective design with a longfollow-up and the associations between pain and risk of DP werefound to be strong and independent of various confounding factors,we believe that this kind of measure would be useful in surveystargeted for primarily healthy populations, and in primary care.It should provide a simple tool for screening of persons at risk inan occupational health care setting or for investigators interestedin early prevention of DP. Another limitation related to the assess-ment of pain is the lack of detailed information about pain inten-sity. With regard to DP, measurement of the intensity of painshould be elucidated because pain intensity has been shown tohave associations with anxiety [15], depression [22], and poorsleep [3], all factors known to predict a higher risk for DP due toMSD [16,42]. Furthermore, although our data set was relativelylarge, with over 11,000 individuals, we lacked the power to runthe analyses for MZ and DZ twin pairs separately. Furthermore, apotential limitation of this study may be related to the specificwelfare systems and society. Hence, we assume that these resultsmay be applicable to other Nordic countries that have similar wel-fare systems and societies, but maybe less so to other countries.The conclusion of this study is that musculoskeletal pain impairingwork ability is an early and direct predictor for DP due to MSD.

5. Conflict of interest

Any of the authors have no conflicts of interest.

Acknowledgements

This work was supported by the Social Insurance Institution,Finland, for the TwinKela Project, and by the Academy of FinlandCenter of Excellence in Complex Disease Genetics (Grant Nos.213506 and 129680) for the Finnish Twin Cohort studies. A.R. is

supported by an Academy of Finland Researcher Grant (122080);P.S. by Grants from the Karolinska Institutet: Public and Interna-tional Health Research Network, the Strategic Research Programin Epidemiology, and the Centre for Health Care Science; and K.S.by an Academy of Finland Researcher Grant (266592).

References

[1] Ahola K, Gould R, Virtanen M, Honkonen T, Aromaa A, Lönnqvist J.Occupational burnout as a predictor of disability pension: a population-based cohort study. Occup Environ Med 2009;66:284–90.

[2] Allardt E. About dimension of welfare: an explanatory analysis of thecomparative Scandinavian survey: University of Helsinki research group ofcomparative sociology research. Helsinki: University of Helsinki; 1973.

[3] Alsaadi SM, McAuley JH, Hush JM, Maher CG. Prevalence of sleep disturbancein patients with low back pain. Eur Spine J 2011;20:737–43.

[4] Battie MC, Videman T, Levalahti E, Gill K, Kaprio J. Heritability of low back painand the role of disc degeneration. PAIN� 2007;131:272–80.

[5] Bevan S, Quadrello T, McGee R, Mahdon M, Vavrovsky A, Barham L. Fit forwork? Musculoskeletal disorders in the European workforce. London: TheWork Foundation; 2009.

[6] Canivet C, Choi B, Karasek R, Moghaddassi M, Staland-Nyman C, Östergren PO.Can high psychological job demands, low decision latitude, and high job strainpredict disability pensions? A 12-year follow-up of middle-aged Swedishworkers. Int Arch Occup Environ Health 2013;86:307–19.

[7] Fejer R, Hartvigsen J, Kyvik KO. Heritability of neck pain: a population-basedstudy of 33 794 Danish twins. Rheumatology 2006;45:589–94.

[8] Harkonmäki K, Silventoinen K, Levälahti E, Pitkäniemi J, Huunan-Seppälä A,Klaukka T, Koskenvuo M, Kaprio J. The genetic liability to disability retirement:a 30-year follow-up study of 24,000 Finnish twins. PLoS One 2008;3:e3402.

[9] Hartvigsen J, Nielsen J, Kyvik KOHM, Fejer R, Vach W, Iachine I, Leboeuf-Yde C.Heritability of spinal pain and consequences of spinal pain: a comprehensivegenetic epidemiologic analysis using a population-based sample of 15,328twins ages 20–71 years. Arthr Rheum 2009;61:1343–51.

[10] Haukenes I, Mykletun A, Knudsen A, Hansen H-T, Maeland J. Disability pensionby occupational class—the impact of work-related factors: the Hordalandhealth study cohort. BMC Pub Health 2011;11:406.

[11] OECD. High-Level Forum. Sickness, disability and work: keeping on track in theeconomic downturn—background paper. Organisation for economic co-operation and development directorate for employment, labour and socialaffairs, 2009.

[12] Hublin C, Partinen M, Koskenvuo M, Kaprio J. Sleep and mortality: apopulation-based 22-year follow-up study. Sleep 2007;30:1245–53.

[13] Kamaleri Y, Natvig B, Ihlebaek CM, Benth JS, Bruusgaard D. Change in thenumber of musculoskeletal pain sites: a 14-year prospective study. PAIN�

2009;141:25–30.[14] Kaprio J, Koskenvuo M. Genetic and environmental factors in complex

diseases: the older Finnish twin cohort. Twin Res 2002;5:358–65.[15] Knaster P, Estlander AM, Karlsson H, Kaprio J, Kalso E. Temperament traits and

chronic pain: the association of harm avoidance and pain-related anxiety. PLoSOne 2012;7:e45672.

[16] Knudsen AK, Øverland S, Aakvaag HF, Harvey SB, Hotopf M, Mykletun A.Common mental disorders and disability pension award: seven year follow-upof the HUSK study. J Psychosom Res 2010;69:59–67.

[17] Koivumaa-Honkanen H, Honkanen R, Viinamäki H, Heikkilä K, Kaprio J,Koskenvuo M. Self-reported life satisfaction and 20-year mortality in healthyFinnish adults. Am J Epidemiol 2000;152:983–91.

[18] Koivumaa-Honkanen H, Kaprio J, Honkanen R, Viinamaki H, Koskenvuo M. Lifesatisfaction and depression in a 15-year follow-up of healthy adults. SocPsychiatry Psychiatr Epidemiol 2004;39:994–9.

[19] Koskenvuo M, Kaprio J, Langinvainio H, Romo M, Sarna S. Psychosocial andenvironmental correlates of coronary-prone behavior in Finland. J Chron Dis1981;34:331–40.

[20] Krokstad S, Westin S. Disability in society—medical and non-medicaldeterminants for disability pension in a Norwegian total county populationstudy. Soc Sci Med 2004;58:1837–48.

[21] Kujala UM, Kaprio J, Koskenvuo M. Modifiable risk factors as predictors of all-cause mortality: the roles of genetics and childhood environment. Am JEpidemiol 2002;156:985–93.

[22] Kwak J, Kim HK, Kim T, Jang SH, Lee KH, Kim MJ, Park SB, Han SH. Theprevalence and characteristics of depression in work-related musculoskeletaldisease. Ann Rehabil Med 2012;36:836–40.

[23] Lahelma E, Laaksonen M, Lallukka T, Martikainen P, Pietilainen O,Saastamoinen P, Gould R, Rahkonen O. Working conditions as risk factors fordisability retirement: a longitudinal register linkage study. BMC Pub Health2012;12:309.

[24] Laine SGD, Virtanen M, Oksanen T, Vahtera J, Elovainio M, Koskinen A, Pentti J,Kivimäki M. Job strain as a predictor of disability pension: the Finnish publicsector study. J Epidemiol Commun Health 2009;63:24–30.

[25] Lang J, Ochsmann E, Kraus T, Lang JW. Psychosocial work stressors asantecedents of musculoskeletal problems: a systematic review and meta-analysis of stability-adjusted longitudinal studies. Soc Sci Med2012;75:1163–74.

1972 A. Ropponen et al. / PAIN�

154 (2013) 1966–1972

[26] Leinonen T, Pietilainen O, Laaksonen M, Rahkonen O, Lahelma E, MartikainenP. Occupational social class and disability retirement among municipalemployees—the contribution of health behaviors and working conditions.Scand J Work Environ Health 2011;37:464–72.

[27] Light KC, Vierck CJ. HPA axis and sympathetic influences on pain and fatigue infibromyalgia, chronic fatigue, and overlapping functional pain syndromes. In:Mayer EA, Bushnell MC, editors. Functional pain syndromes: presentation andpathophysiology. Seattle: IASP; 2009.

[28] Lim SS, Vos T, Flaxman AD, Danaei G, Shibuya K, Adair-Rohani H, Amann M,Anderson HR, Andrews KG, Aryee M, Atkinson C, Bacchus LJ, Bahalim AN,Balakrishnan K, Balmes J, Barker-Collo S, Baxter A, Bell ML, Blore JD, Blyth F,Bonner C, Borges G, Bourne R, Boussinesq M, Brauer M, Brooks P, Bruce NG,Brunekreef B, Bryan-Hancock C, Bucello C, Buchbinder R, Bull F, Burnett RT,Byers TE, Calabria B, Carapetis J, Carnahan E, Chafe Z, Charlson F, Chen H, ChenJS, Cheng AT, Child JC, Cohen A, Colson KE, Cowie BC, Darby S, Darling S, DavisA, Degenhardt L, Dentener F, Des Jarlais DC, Devries K, Dherani M, Ding EL,Dorsey ER, Driscoll T, Edmond K, Ali SE, Engell RE, Erwin PJ, Fahimi S, Falder G,Farzadfar F, Ferrari A, Finucane MM, Flaxman S, Fowkes FG, Freedman G,Freeman MK, Gakidou E, Ghosh S, Giovannucci E, Gmel G, Graham K, GraingerR, Grant B, Gunnell D, Gutierrez HR, Hall W, Hoek HW, Hogan A, Hosgood HD3rd, Hoy D, Hu H, Hubbell BJ, Hutchings SJ, Ibeanusi SE, Jacklyn GL, Jasrasaria R,Jonas JB, Kan H, Kanis JA, Kassebaum N, Kawakami N, Khang YH, KhatibzadehS, Khoo JP, Kok C, Laden F, Lalloo R, Lan Q, Lathlean T, Leasher JL, Leigh J, Li Y,Lin JK, Lipshultz SE, London S, Lozano R, Lu Y, Mak J, Malekzadeh R, Mallinger L,Marcenes W, March L, Marks R, Martin R, McGale P, McGrath J, Mehta S,Mensah GA, Merriman TR, Micha R, Michaud C, Mishra V, Hanafiah KM,Mokdad AA, Morawska L, Mozaffarian D, Murphy T, Naghavi M, Neal B, NelsonPK, Nolla JM, Norman R, Olives C, Omer SB, Orchard J, Osborne R, Ostro B, PageA, Pandey KD, Parry CD, Passmore E, Patra J, Pearce N, Pelizzari PM, Petzold M,Phillips MR, Pope D, Pope CA 3rd, Powles J, Rao M, Razavi H, Rehfuess EA, RehmJT, Ritz B, Rivara FP, Roberts T, Robinson C, Rodriguez-Portales JA, Romieu I,Room R, Rosenfeld LC, Roy A, Rushton L, Salomon JA, Sampson U, Sanchez-Riera L, Sanman E, Sapkota A, Seedat S, Shi P, Shield K, Shivakoti R, Singh GM,Sleet DA, Smith E, Smith KR, Stapelberg NJ, Steenland K, Stockl H, Stovner LJ,Straif K, Straney L, Thurston GD, Tran JH, Van Dingenen R, van Donkelaar A,Veerman JL, Vijayakumar L, Weintraub R, Weissman MM, White RA, WhitefordH, Wiersma ST, Wilkinson JD, Williams HC, Williams W, Wilson N, Woolf AD,Yip P, Zielinski JM, Lopez AD, Murray CJ, Ezzati M, AlMazroa MA, Memish ZA. Acomparative risk assessment of burden of disease and injury attributable to 67risk factors and risk factor clusters in 21 regions, 1990–2010: a systematicanalysis for the global burden of disease study 2010. Lancet 2012;380:2224–60.

[29] Linton SJ, Shaw WS. Impact of psychological factors in the experience of pain.Phys Ther 2011;91:700–11.

[30] Mallen CD, Peat G, Thomas E, Dunn KM, Croft PR. Prognostic factors formusculoskeletal pain in primary care: a systematic review. Br J Gen Pract2007;57:655–61.

[31] Markkula R, Kalso E, Huunan-Seppala A, Koskenvuo M, Koskenvuo K, Leino-Arjas P, Kaprio J. The burden of symptoms predicts early retirement: a twin

cohort study on fibromyalgia-associated symptoms. Eur J Pain2011;15:741–7.

[32] Mäntyniemi A, Oksanen T, Salo P, Virtanen M, Sjösten N, Pentti J, Kivimäki M,Vahtera J. Job strain and the risk of disability pension due to musculoskeletaldisorders, depression or coronary heart disease: a prospective cohort study of69 842 employees. Occup Environ Med 2012;69:574–81.

[33] Naliboff BD, Rhudy JL. Anxiety in functional pain disorders. In: Mayer EA,Bushnell MC, editors. Functional pain syndromes: presentation andpathophysiology. Seattle: IASP Press; 2009. p. 185–213.

[34] Narusyte J, Ropponen A, Silventoinen K, Alexanderson K, Kaprio J, SamuelssonÃ, Svedberg P. Genetic liability to disability pension in women and men: aprospective population-based twin study. PLoS One 2011;6:e23143.

[35] Overland S, Harvey SB, Knudsen AK, Mykletun A, Hotopf M. Widespread painand medically certified disability pension in the Hordaland Health Study. Eur JPain 2012;16:611–20.

[36] Pietikäinen S, Silventoinen K, Svedberg P, Alexanderson K, Huunan-Seppälä A,Koskenvuo K, Koskenvuo M, Kaprio J, Ropponen A. Health-related andsociodemographic risk factors for disability pension due to low backdisorders: a 30-year prospective Finnish twin cohort study. J Occup EnvironMed 2011;53:488–96.

[37] Raiha I, Kemppainen H, Kaprio J, Koskenvuo M, Sourander L. Lifestyle, stress,and genes in peptic ulcer disease: a nationwide twin cohort study. Arch InternMed 1998;158:698–704.

[38] Ropponen A, Narusyte J, Alexanderson K, Svedberg P. Stability and change inhealth behaviours as predictors for disability pension: a prospective cohortstudy of Swedish twins. BMC Pub Health 2011;11:68.

[39] Ropponen A, Silventoinen K, Svedberg P, Alexanderson K, Koskenvuo K,Huunan-Seppälä A, Koskenvuo M, Kaprio J. Health-related risk factors fordisability pensions due to musculoskeletal diagnoses: a 30-year Finnish twincohort study. Scand J Pub Health 2011;39:839–48.

[40] Russell IJ. Future perspectives in generalised musculoskeletal pain syndromes.Best Pract Res Clin Rheumatol 2011;25:321–31.

[41] Saastamoinen P, Laaksonen M, Kaaria SM, Lallukka T, Leino-Arjas P, RahkonenO, Lahelma E. Pain and disability retirement: a prospective cohort study. PAIN�

2012;153:526–31.[42] Salo P, Oksanen T, Sivertsen B, Hall M, Pentti J, Virtanen M, Vahtera J, Kivimäki

M. Sleep disturbances as a predictor of cause-specific work disability anddelayed return to work. Sleep 2010;33:1323–31.

[43] Schatzberg AF. Chronic pain and depression. In: Mayer EA, Bushnell MC,editors. Functional pain syndromes: presentation andpathophysiology. Seattle: IASP; 2009. p. 215–26.

[44] Stattin M, Jarvholm B. Occupation, work environment, and disability pension:a prospective study of construction workers. Scand J Pub Health2005;33:84–90.

[45] Vranceanu A-M, Barsky A, Ring D. Psychosocial aspects of disablingmusculoskeletal pain. J Bone Joint Surg 2009;91:2014–8.

[46] Young EE, Lariviere WR, Belfer I. Genetic basis of pain variability: recentadvances. J Med Genet 2012;49:1–9.