Study protocol for Review C – Version 1.8 26th July 2013
C1
Review of scientific published literature on infant feeding and 1
development of atopic and autoimmune disease: 2
Review C – Maternal and infant diet 3
4
Robert Boyle, Clinical Senior Lecturer, Section of Paediatrics, Imperial College London 5
Vanessa Garcia-Larsen, Post-Doctoral Research Associate, Respiratory Epidemiology and 6
Public Health, National Heart and Lung Institute, Imperial College London 7
Jo Leonardi-Bee, Associate Professor of Community Health Sciences, University of 8
Nottingham 9
Tim Reeves, Research Support Librarian, Faculty of Medicine, Imperial College London 10
11
12
Imperial Consultants, 13
58 Princes Gate, 14
Exhibition Road, 15
London 16
SW7 2PG 17
18
Study protocol for Review C – Version 1.8 26th July 2013
C2
19
Introduction ............................................................................................................................ 4 20
Key words .............................................................................................................................. 5 21
Review question(s) ................................................................................................................. 6 22
Inclusion criteria:.................................................................................................................... 6 23
Types of studies to be included .......................................................................................... 6 24
Participants/population ....................................................................................................... 6 25
Context................................................................................................................................ 7 26
Interventions/exposures ...................................................................................................... 7 27
Comparator(s)/control ...................................................................................................... 10 28
Search strategy ..................................................................................................................... 11 29
Study Outcomes [identical to Review A] ............................................................................. 12 30
Atopic outcomes: .............................................................................................................. 12 31
Autoimmune outcomes: .................................................................................................... 14 32
Study selection and Data Extraction .................................................................................... 15 33
Study selection .................................................................................................................. 15 34
Data extraction .................................................................................................................. 16 35
Risk of bias (quality) assessment ......................................................................................... 16 36
Review level bias .............................................................................................................. 16 37
Study level bias ................................................................................................................. 17 38
Strategy for data synthesis.................................................................................................... 17 39
Data extraction .................................................................................................................. 18 40
Planned subgroup analyses ............................................................................................... 19 41
Graphical exploration of heterogeneity ............................................................................ 19 42
Review registration .............................................................................................................. 20 43
Study protocol for Review C – Version 1.8 26th July 2013
C3
Dissemination of findings .................................................................................................... 20 44
45
46
Study protocol for Review C – Version 1.8 26th July 2013
C4
Introduction 47
This is one of 3 systematic reviews being undertaken as part of a review of the scientific 48
literature on infant feeding and development of atopic and autoimmune diseases 49
commissioned by the UK Food Standards Agency. Atopic conditions such as asthma, 50
eczema, rhinoconjunctivitis and food allergy appear to have increased in prevalence in recent 51
decades in many countries, and are now the leading causes of chronic illness during 52
childhood in the UK 1, 2
. The apparently rapid changes in disease prevalence, combined with 53
data from migration studies, suggest that early life environmental factors may be important 54
modulators of atopic disease risk. Similar findings apply to the autoimmune diseases type I 55
diabetes mellitus and Crohn’s disease, which also appear to have increased in prevalence in 56
some countries 3. Significant attention has focussed on dietary exposures in relation to these 57
immune-mediated atopic and autoimmune diseases for 2 reasons – first the temporal 58
association between rises in atopic/autoimmune conditions and changing dietary exposures in 59
the relevant populations; second the gut associated lymphoid tissue is our largest collection of 60
immune tissue, and our most mature immune organ at the time of birth 4. Hence early enteral 61
exposures are likely to be especially potent modulators of immune development and risk of 62
immune-mediated disease. Although there are a large number of observational studies, some 63
intervention trials and several systematic reviews in this area, they tend to focus on one 64
specific area of diet and a limited number of immune outcomes. The purpose of these 3 65
systematic reviews is to comprehensively assess the existing literature regarding the 66
relationship between maternal and infant dietary exposures and a child’s risk of any of the 67
common atopic and autoimmune disease, in order to inform UK Department of Health 68
feeding guidance for mothers and their infants. These protocols have been developed by the 69
authors, but have been modified after independent expert peer review and reviews and 70
Study protocol for Review C – Version 1.8 26th July 2013
C5
meetings with members of the UK Food Standards Agency and the UK Scientific Advisory 71
Committee on Nutrition. The specific outcomes of interest for these reviews, chosen due to 72
their high prevalence in the UK population, and described in more detail below, are: Food 73
allergy, Eczema, Asthma, Allergic rhinitis, Allergic conjunctivitis, Allergic sensitisation, 74
Type 1 diabetes mellitus, Coeliac disease, Inflammatory bowel disease, Autoimmune thyroid 75
disease, Juvenile rheumatoid arthritis, Vitiligo, Psoriasis. 76
Key words 77
Infant; diet; maternal; pregnancy; lactation; supplement; nutrient; allergy; atopy; asthma; 78
eczema; food allergy; sensitisation; rhinitis; conjunctivitis; autoimmune; diabetes; crohn; 79
inflammatory bowel disease; coeliac; thyroiditis; juvenile arthritis; vitiligo; psoriasis; 80
systematic review 81
82
Study protocol for Review C – Version 1.8 26th July 2013
C6
Review question(s) 83
C1. Does exposure to specific dietary patterns, food groups or nutrients during the first year 84
of life, influence children’s future risk of atopic disease, allergic sensitisation or autoimmune 85
disease. 86
C2. Does maternal exposure to specific dietary patterns, food groups or nutrients during 87
pregnancy or lactation, influence children’s future risk of atopic disease, allergic sensitisation 88
or autoimmune disease. 89
90
Inclusion criteria: 91
Types of studies to be included 92
We will include randomised controlled trials (RCT), quasi RCT, and where necessary 93
prospective cohort or longitudinal studies, retrospective cohort studies, nested case-control 94
studies or other case control studies. We will take a hierarchical approach to study design, 95
such that where data are absent or limited from certain types of studies, we will include lower 96
level study designs. So where high quality intervention studies are lacking, we will include 97
prospective cohort studies; where high quality prospective cohort studies are lacking we will 98
include retrospective cohort studies; where high quality retrospective cohort studies are 99
lacking we will include data from nested case-control studies, and where these are lacking we 100
will use other types of case control studies. We will not include non-comparative studies, or 101
non-human studies. 102
Participants/population 103
Infants between the age of 0 and the end of the 12th
postpartum month (question C1) and their 104
mothers during pregnancy and lactation up to the end of the 24th
month postnatally (question 105
Study protocol for Review C – Version 1.8 26th July 2013
C7
C2). If infants are characterised as high or standard risk for allergic disease based on family 106
history, this information will be recorded so that subgroup analysis can be made. 107
Context 108
Our primary analyses will exclude studies in which participants were defined by a disease 109
state - eg pregnant women with specific nutritional deficiencies, infants born prematurely 110
(<31 weeks gestation) or other groups clearly representing <5% of the UK population, since 111
the results of this review should apply to the general UK population. We will include studies 112
of specific ethnic groups. Studies where subject eligibility is defined on the basis of a family 113
history of allergy will be included, since this applies to a majority of infants in the UK. 114
Studies restricted to populations at specific genetic risk of autoimmune disease (eg by HLA 115
type) will also be included, since it is difficult to undertake studies of autoimmune disease 116
prevention in the general population due to low prevalence. 117
Interventions/exposures 118
Review question #C1: Exposure of infants in the first 12 months, to specific dietary patterns, 119
dietary components (as groups eg vegetables, fruits, nuts; and individually) or nutrients/food 120
supplements (eg. antioxidants, vitamin D, omega-3 and omega-6 fatty acids, vitamin B3, 121
probiotics). For probiotics, we will include studies where they are given as supplements, or in 122
infant formula. Where possible, the exposures will be measured as a continuous variable 123
rather than categorically. Because of the variety in diet, we will in principle, attempt to 124
classify foods according to their nutritional properties and similarities. We will classify the 125
foods following the European Food Consumption Survey Method, recently used by VGL in a 126
large epidemiological survey of dietary risk factors for allergic diseases across Europe 6. We 127
will also consider the UK Department of Health proposed classification for fruits and 128
vegetables (www.nhs.uk/5aday) in our interpretation of the data, and the UK Reference 129
Study protocol for Review C – Version 1.8 26th July 2013
C8
Nutrient Intake. Whenever possible, foods will be classified as indicated in the list below. 130
This method allows for more ambiguous classifications to be included. 131
Hard fruits- apple, pear, peach 132
Oily fruits- olive, avocado 133
Berries (any) 134
Nectarines- nectarine, apricot 135
Citrus- orange, kiwi, lemon, mandarin, grapefruit 136
Dried fruits- raisin, prune, dates 137
Fresh fruit juice 138
Tropical fruits- mango, pine-apple, banana 139
Tinned fruits 140
Melon/watermelon 141
Other fruits- plum, cherries, grapes, fig, rhubarb 142
‘Any fruits’ (to include studies which only included a general question on fruit intake) 143
Other classifications of fruits: ‘flavanoid-rich fruits’ – these will be included as a 144
separate group if the publication has grouped fruits as such without separating 145
individual intake (e.g. apples, grapes, berries) 146
Leafy vegetables- lettuce, spinach, chard, fenugreek, herb, wild greens 147
Fruity vegetables- artichokes, tomato, cucumber, okra, aubergine, capers 148
Root vegetables- carrot, parsnips, turnip, ginger, radish, taro, beetroot 149
Cabbage- cauliflower, coleslaw, brussel sprouts, broccoli, cabbage 150
Stalk- celery, asparagus 151
Allium vegetables- leek, onion, garlic, shallots, 152
Pickled vegetable 153
Study protocol for Review C – Version 1.8 26th July 2013
C9
‘any vegetables (as in ‘any fruits’) 154
Specific classifications of vegetables e.g. ‘red coloured vegetables’ 155
Sweet potatoes, parsnips, swedes and turnips 156
Pulses and beans (baked beans, haricot beans, kidney beans, cannellini beans, butter 157
beans or chickpeas) 158
Nuts 159
Red meat 160
Fish and seafood 161
Dairy products and derivatives 162
Cereals 163
For studies of overall dietary pattern, such as Mediterranean or organic diet, we will look in 164
detail at the definition and only consider meta-analysis where at least 50% of the definition is 165
shared across studies. 166
- inclusion criteria - studies where infant dietary exposure is characterised during the first 12 167
months of life, in such a way that it can be classified for determining the exposure(s) of 168
interest as above; or an intervention study. We will classify the exposures according to how 169
they were measured, i.e. what type of dietary questionnaire or assessment. In general, we will 170
not include studies of blood/plasma/serum/urine levels of circulating nutrients without an 171
assessment of dietary intake, since the purpose of this review is to inform dietary advice for 172
pregnant and lactating mothers and their infants. The single exception is vitamin D, since the 173
primary source of vitamin D is sunlight exposure, so that estimation of vitamin D status based 174
on dietary assessment alone is unreliable. We will include studies of vitamin D status in 175
blood, measured as 25-hydroxy vitamin D in nmol/L or ng/ml, where 2.5nmol/L = 1 ng/ml. 176
Study protocol for Review C – Version 1.8 26th July 2013
C10
Review question #C2: Exposure of mothers during pregnancy and/or lactation, to specific 177
dietary patterns, dietary components (as groups eg vegetables, fruits, nuts; and individually) 178
or nutrients/food supplements (eg. antioxidants, vitamin D, omega-3 and omega-6 fatty acids, 179
vitamin B3, probiotics). Where possible, the exposures will be measured as a continuous 180
variable rather than categorically. We will follow the same classification proposed above to 181
classify the dietary exposure during pregnancy and lactation up to 24 months. 182
- inclusion criteria - studies where maternal dietary exposure is characterised during 183
pregnancy and/or lactation, in such a way that it can be classified for determining the 184
exposure(s) of interest as above; or an intervention study. We will classify the exposures 185
according to how they were measured, i.e. what type of dietary questionnaire or assessment. 186
In general, we will not include studies of blood/plasma/serum/urine levels of circulating 187
nutrients without an assessment of dietary intake, since the purpose of this review is to inform 188
dietary advice for pregnant and lactating mothers and their infants. The single exception is 189
vitamin D, since the primary source of vitamin D is sunlight exposure, so that estimation of 190
vitamin D status based on dietary assessment can be unreliable. We will include studies of 191
vitamin D status in blood, measured as 25-hydroxy vitamin D in nmol/L or ng/ml, where 192
2.5nmol/L = 1 ng/ml. 193
194
Comparator(s)/control 195
All comparators will be included, including studies which compare different doses or forms 196
of an exposure eg different doses/levels of vitamin D, or different probiotic interventions. In 197
the case of studies that only document food intake as reported frequency, we will aim to 198
categorise groups for binary comparisons e.g. ‘at least weekly intake’ vs. ‘never’; ‘at least 199
daily intake’ vs. ‘weekly or less frequently’. 200
Study protocol for Review C – Version 1.8 26th July 2013
C11
Search strategy 201
We will search for eligible studies in MEDLINE, EMBASE, Cochrane, Web of Science and 202
LILACS with no specified start date. We will include peer reviewed publications, and also 203
include proceedings and abstracts presented in scientific conferences in the last 3 years, if 204
they have not subsequently been published as a peer reviewed publication. We will search for 205
studies in progress, or completed but unpublished studies using 206
http://apps.who.int/trialsearch/, and will contact international experts in the field of 207
nutritional exposures in relation to allergy and autoimmune disorders, including where 208
appropriate pharmaceutical or food industry representatives, to identify important 209
unpublished work. We will review the bibliography of eligible studies for possible additional 210
publications, and will include all eligible publications, regardless of the language. Where 211
necessary, and where feasible within the limited timescale of this project, the authors of 212
eligible or potentially eligible studies will be contacted by the research team to obtain any 213
data that might not be available in the abstract/publication. Potentially eligible studies are 214
studies which have clearly recorded both an exposure and an outcome of interest, but have 215
not reported an analysis of the relationship between these. 216
The MEDLINE search strategy is at the end of this document, as an Appendix. 217
We will separately search for existing systematic reviews which cover any of the same 218
exposure(s)/outcome(s) as these, and were published since 1st January 2011. We will quality 219
assess such existing systematic reviews using the revised AMSTAR criteria 5. We will not 220
duplicate any existing systematic reviews with revised AMSTAR score ≥32, but will instead 221
summarise the findings of such reviews and include the summary in our final report. As far as 222
possible, these data will be presented using Cochrane Summary of Findings tables, generated 223
using GradePro. For these pre-existing, high quality systematic reviews, we will also 224
Study protocol for Review C – Version 1.8 26th July 2013
C12
summarise any eligible publications identified in the searches, which were published 225
subsequent to the relevant systematic review, in our final report. 226
Study Outcomes [identical to Review A] 227
We have selected atopic and autoimmune outcomes on the basis of their population 228
prevalence in children and young adults in the UK or other affluent nations. We have 229
included diseases with a prevalence of at least 1 in 1000, in children/adolescents or young 230
adults (aged <40 years), but have not included rarer diseases 6. We have not included 231
pernicious anaemia or adult-onset rheumatoid arthritis despite a high prevalence in middle 232
aged or elderly people, because their prevalence in young people is lower than 1 in 1000, and 233
prospective studies of infant feeding in relation to diseases of older adults are unlikely to 234
have been undertaken. For all outcome measures, age at assessment will be grouped as 1-4 235
years, 5-14 years, 15-24 years, 25-44 years, 45-64 years and ≥65 years. Where studies report 236
the same outcome at different timepoints within one of these frames, we will use the 237
timepoint which has the most complete dataset ie lowest percentage of missing data as the 238
primary assessment point. For each outcome measure in this review, there is more than one 239
possible method of assessment. We have therefore included our preferred method of 240
assessment for each outcome, which is the a priori ‘primary outcome measure’, assessed at 241
the optimal age as defined above. We will however document all relevant outcomes measured 242
using different assessment tools, in each included study. This will allow for meta-analysis of 243
different studies where they have used similar outcome measures. 244
Atopic outcomes: 245
1. Asthma - defined as either ‘asthma’, ‘infantile wheeze’ or similar, using parent/self report, 246
doctor diagnosis, a validated questionnaire, scoring system or objective measure such as 247
Study protocol for Review C – Version 1.8 26th July 2013
C13
bronchial hyper-reactivity, forced vital capacity, peak expiratory flow rate or reversible 248
airways obstruction using forced expiratory volume in 1 second. 249
Primary assessment: parent or self-report using a validated questionnaire such as the 250
International Study of Asthma and Allergies in Childhood questionnaire 7, at 5-14 years. 251
Where multiple measures are used, cumulative incidence of wheezing will be used 252
preferentially. 253
2. Eczema – defined using parent/self report, doctor diagnosis, a validated questionnaire, 254
scoring system or objective measure. 255
Primary assessment: parent or self-report using a validated questionnaire such as the UK 256
adaptation of Hanifin and Rajka criteria 8 at 1-4 years. Where multiple measures are used, 257
cumulative incidence of eczema will be analysed preferentially, but point prevalence will also 258
be reported. 259
3.Allergic Rhinitis – defined using parent/self report, doctor diagnosis, a validated 260
questionnaire, scoring system or objective measure. 261
Primary assessment: parent or self-report using a validated questionnaire such as the 262
International Study of Asthma and Allergies in Childhood questionnaire 7, at 5-14 years. 263
Where multiple measures are used, cumulative incidence will be analysed preferentially. 264
4. Allergic Conjunctivitis - defined using parent/self report, doctor diagnosis, a validated 265
questionnaire, scoring system or objective measure. 266
Primary assessment: parent or self-report using a validated questionnaire, at 5-14 years. 267
Where multiple measures are used, cumulative incidence will be analysed preferentially. 268
5. Food allergy - defined by double blind placebo controlled food challenge, by open food 269
challenge, by medical diagnosis or by self/parent report. 270
Study protocol for Review C – Version 1.8 26th July 2013
C14
Primary assessment: challenge-proven food allergy, assessed at 1-4 years. Where multiple 271
measures are used, cumulative incidence will be analysed preferentially. 272
6. Allergic sensitisation – to an inhalant, an ingestant, or both – defined as positive skin prick 273
test and/or specific IgE test to the relevant allergen using recognised methodologies and 274
scoring criteria 9. 275
Primary assessment: sensitisation to at least one inhalant or ingestant, assessed at 5-14 years 276
or older. Where multiple measures are used, point prevalence will be analysed preferentially. 277
7. Total IgE – measured using a recognised technology such as ImmunoCAP (ThermoFisher, 278
Massachusets). 279
Autoimmune outcomes: 280
1. Type I diabetes mellitus – defined as a medical diagnosis, or a surrogate marker such as 281
autoantibodies against insulin, GAD65, IA-2 or the ZnT8 transporter in the first 3 years of 282
life. 283
Primary assessment: medical diagnosis of type 1 diabetes mellitus using the 1999 WHO 284
recommendations for diagnosis and classification of diabetes mellitus 10
or similar. Where 285
multiple measures are used, cumulative incidence will be analysed preferentially. 286
2. Coeliac disease – defined by characteristic histological features (intraepithelial 287
lymphocytes, crypt hyperplasia and villous atrophy) with improvement in symptoms and 288
histology after institution of a gluten free diet, a medical diagnosis, or a surrogate marker 289
such as IgA tissue transglutaminase or IgA endomysial antibodies. 290
Primary assessment: medical diagnosis of coeliac disease using a histological diagnosis. 291
Where multiple measures are used, cumulative incidence will be analysed preferentially. 292
3. Inflammatory bowel disease (Crohn's disease or Ulcerative colitis) – defined as a medical 293
diagnosis. 294
Study protocol for Review C – Version 1.8 26th July 2013
C15
Primary assessment: medical diagnosis using a histological diagnosis. Where multiple 295
measures are used, cumulative incidence will be analysed preferentially. 296
4. Autoimmune thyroid disease (Graves' disease or Hashimoto's thyroiditis) - defined as a 297
medical diagnosis. 298
Primary assessment: medical diagnosis using serology and thyroid function testing. Where 299
multiple measures are used, cumulative incidence will be analysed preferentially. 300
5. Juvenile rheumatoid arthritis – defined as a medical diagnosis. 301
Primary assessment: medical diagnosis using the 2001 revised International League of 302
Associations for Rheumatology (ILAR) classification criteria 11
. Where different time-points 303
are reported, then the cumulative incidence to the latest reported time-point will be used 304
preferentially. Where multiple measures are used, cumulative incidence will be analysed 305
preferentially. 306
6. Vitiligo - defined as a medical diagnosis. 307
Primary assessment: medical diagnosis using the Vitiligo European Task Force 2007 criteria 308
or similar 12
. Where multiple measures are used, cumulative incidence will be analysed 309
preferentially. 310
7. Psoriasis - defined as a medical diagnosis. 311
Primary assessment: medical diagnosis. Where multiple measures are used, cumulative 312
incidence will be analysed preferentially. 313
314
Study selection and Data Extraction 315
Study selection 316
Two members of the research team (RB and VGL) will independently review titles and 317
abstracts of identified studies. The full text of the paper will also be independently assessed 318
Study protocol for Review C – Version 1.8 26th July 2013
C16
by RB and VGL, and will be assessed for eligibility against the inclusion criteria. Any 319
discrepancies will be resolved through discussions with the research team and, as appropriate, 320
the study sponsor (UK Food Standards Agency, FSA). Electronic records will be kept 321
regarding included and excluded studies for audit purposes, specifying reasons for any 322
exclusion, and these details will be included in the final report. Full text articles will be 323
reviewed in duplicate (by two research team members - RB and VGL), and studies for 324
inclusion will be selected – any discrepancies will be resolved through discussions with the 325
research team and the FSA, as appropriate. The reasons for the exclusion of any relevant 326
studies will be recorded, however ineligible studies will not be analysed further. 327
Data extraction 328
A pilot of the data extraction form will be undertaken using a minimum of 5 papers, after 329
which the extraction form will be amended/updated as necessary. The data extraction form 330
will be used to extract the relevant data fields from each included study independently (by 331
two research team members - RB and VGL), and where appropriate data will be entered into 332
Stata IC 12 statistical software for meta-analysis. 333
334
Risk of bias (quality) assessment 335
Review level bias 336
Publication bias will be assessed using funnel plots and Egger's test. Where asymmetry is 337
evident on the funnel plot, a trim and fill analysis will be used. Possible causes for asymmetry 338
other than publication bias (eg between study heterogeneity) will also be considered. Where 339
significant population based cohorts or randomised controlled trials have assessed mode of 340
infant feeding but not reported relevant atopic or autoimmune outcomes, we will consider 341
Study protocol for Review C – Version 1.8 26th July 2013
C17
contacting authors for original datasets if atopic or autoimmune outcome assessments appear 342
to have been made. 343
Study level bias 344
The risk of bias in included RCTs will be assessed using the Cochrane Collaboration Risk of 345
bias tool, which includes sequence generation, allocation concealment, blinding, incomplete 346
outcome data, and selective outcome reporting, and other bias 13
. RCTs will be considered at 347
low risk of bias where the risk of bias is judged to be low for all key domains of the Cochrane 348
Risk of bias tool. The risk of bias in included cohort and case control studies will be assessed 349
using the National Institute for Clinical Excellence methodological checklist for cohort and 350
case-control studies respectively, which includes considerations of subject selection, 351
assessment of exposure and outcome, and measures to assess confounding 14. Studies will be 352
considered at low risk of bias where most of the criteria in the checklist are addressed, and 353
those that are not addressed or not reported are judged unlikely to change the study findings. 354
For both RCTs and cohort studies, a level of <20% loss to follow up for atopic/autoimmune 355
outcomes will generally be accepted as representing low risk of bias from incomplete 356
outcome data, if there are no other features to suggest increased risk of bias. For all studies, a 357
summary Table of Study Characteristics will be presented for each relevant exposure and 358
outcome, which will include a summary of each study's risk of bias, in addition to the 359
population characteristics, methods used for assessing exposure and for outcome assessment. 360
Strategy for data synthesis 361
Where appropriate, meta-analysis will be undertaken. If meta-analysis is deemed 362
inappropriate, individual study results will be summarised and a balanced conclusion made. 363
Separate analyses will be undertaken for each disease outcome, for each group of similar 364
outcome assessment methods for any given disease, and for each intervention/exposure. 365
Study protocol for Review C – Version 1.8 26th July 2013
C18
Results for randomised or quasi-randomised controlled trials, prospective cohort or 366
longitudinal studies, or where appropriate retrospective cohort studies, nested case-control 367
studies or other case-control studies will be reported separately for each comparison. 368
Data extraction 369
Data will be extracted either using raw values, crude estimates of effect (including odds 370
ratios, risk ratios, incidence rate ratios, hazard ratios, mean differences) or as adjusted 371
estimates of effect. Adjusted estimates of effect will be used in preference, where available. 372
Random effect meta-analyses will be performed to allow for the anticipated heterogeneity 373
between the studies. 374
Heterogeneity 375
Heterogeneity will be quantified using I2. We will explore reasons for heterogeneity using 376
subgroup analyses based on study level factors. Where extreme levels of heterogeneity are 377
detected (I2>75%), we will perform sensitivity analyses to assess the effect of excluding 378
outliers, and re-consider whether quantitative data synthesis is appropriate. Where possible, 379
meta-regression and subgroup analysis will be used to explore sources of heterogeneity 380
arising from study characteristics - ordered forest plots and graphical methods will be used to 381
further investigate potential effects of continuous confounders on study effects. Individual 382
patient data analysis will not be undertaken. 383
Data analysis 384
Data from individual studies will be pooled using the generic inverse variance method. 385
Pooled results for binary outcomes will be presented as relative risks s with 95% confidence 386
intervals and 2-sided p values, and also expressed as risk differences where possible. Pooled 387
results for continuous outcomes measured using similar scales will be presented as mean 388
differences with 95% confidence intervals and 2-sided p values. However, where different 389
Study protocol for Review C – Version 1.8 26th July 2013
C19
scales are pooled across studies, we will report results using standardised mean differences. 390
P<0.05 will be considered statistically significant. Relevant results will be presented in 391
Summary of Findings tables similar to those used by the Cochrane Collaboration 15
. 392
All analyses will be performed using STATA IC 12. 393
Planned subgroup analyses 394
1. High study quality - high quality RCT or cohort studies as defined above will be separately 395
analysed. 396
2. Increased disease risk - studies of populations at increased risk for atopic or autoimmune 397
disease will be separately analysed - for example infants with a family history of atopic or 398
autoimmune disease. 399
3. Type of data - unadjusted versus adjusted data. Factors that we expect to be adjusted for 400
within studies: siblings (parity or birth order or family size); gender; age at outcome 401
assessment; disease risk based on family history; maternal or household smoking (asthma 402
outcomes); maternal age; maternal education or socioeconomic status; duration of 403
breastfeeding or exclusive breastfeeding. 404
4. Quality of dietary assessment - use of a validated food frequency/nutritional exposure 405
assessment tool, or in the case of intervention studies, documented compliance with the study 406
intervention for over 80% of intended doses. Studies which meet these criteria will be 407
analysed separately. 408
5. Quality of dietary exposure - studies of nutrient intake/supplementation in the natural form 409
of the nutrient, as opposed to specific supplementation, will be analysed separately. Some 410
studies have shown stronger effects for foods in their natural form than when given as 411
specific nutritional supplements 17
. 412
Graphical exploration of heterogeneity 413
Study protocol for Review C – Version 1.8 26th July 2013
C20
1. Study year (average year of assessment/birth for study population) 414
2. Average age of study population at examination/assessment 415
416
Review registration 417
This systematic review will be registered with the International Prospective Register of 418
Systematic Reviews (www.crd.york.ac.uk/Prospero), prior to selecting any studies from the 419
search results. This review protocol has been revised following peer review by the UK Food 420
Standards Agency, the UK Scientific Advisory Committee on Nutrition, independent experts 421
Professor Graham Devereux and Dr Carina Venter, and the Lancet. 422
423
Dissemination of findings 424
The findings of this review will inform the Food Standards Agency review of infant feeding 425
which will in turn inform the revised Department of Health guidance on infant feeding in the 426
UK. The reviews will be submitted for publication as peer reviewed manuscripts in academic 427
journals, and presented at national and international conferences. A summary of the findings 428
will be sent to relevant stakeholders such as charities, health and educational institutions 429
involved in advising on or supporting infant feeding in the UK. 430
431
Study protocol for Review C – Version 1.8 26th July 2013
C21
References 432
1. Gupta R, Sheikh A, Strachan DP, Anderson HR. Time trends in allergic disorders in 433
the UK. Thorax. 2007; 62(1): 91-6. 434
2. Gupta R, Sheikh A, Strachan DP, Anderson HR. Burden of allergic disease in the UK: 435
secondary analyses of national databases. Clinical & Experimental Allergy. 2004; 34(4): 520-436
6. 437
3. Bach JF. The effect of infections on susceptibility to autoimmune and allergic 438
diseases. The New England journal of medicine. 2002; 347(12): 911-20. 439
4. Jones CA, Vance GH, Power LL, Pender SL, Macdonald TT, Warner JO. 440
Costimulatory molecules in the developing human gastrointestinal tract: a pathway for fetal 441
allergen priming. The Journal of allergy and clinical immunology. 2001; 108(2): 235-41. 442
5. Kung J, Chiappelli F, Cajulis OO, Avezova R, Kossan G, Chew L, et al. From 443
Systematic Reviews to Clinical Recommendations for Evidence-Based Health Care: 444
Validation of Revised Assessment of Multiple Systematic Reviews (R-AMSTAR) for 445
Grading of Clinical Relevance. The open dentistry journal. 2010; 4: 84-91. 446
6. Garcia-Larsen V, Luczynska M, Kowalski ML, Voutilainen H, Ahlstrom M, Haahtela 447
T, et al. Use of a common food frequency questionnaire (FFQ) to assess dietary patterns and 448
their relation to allergy and asthma in Europe: pilot study of the GA2LEN FFQ. European 449
journal of clinical nutrition. 2011; 65(6): 750-6. 450
7. Cooper GS, Stroehla BC. The epidemiology of autoimmune diseases. Autoimmunity 451
reviews. 2003; 2(3): 119-25. 452
8. Asher MI, Keil U, Anderson HR, Beasley R, Crane J, Martinez F, et al. International 453
Study of Asthma and Allergies in Childhood (ISAAC): rationale and methods. The European 454
Study protocol for Review C – Version 1.8 26th July 2013
C22
respiratory journal : official journal of the European Society for Clinical Respiratory 455
Physiology. 1995; 8(3): 483-91. 456
9. Williams HC, Burney PG, Pembroke AC, Hay RJ. Validation of the U.K. diagnostic 457
criteria for atopic dermatitis in a population setting. U.K. Diagnostic Criteria for Atopic 458
Dermatitis Working Party. The British journal of dermatology. 1996; 135(1): 12-7. 459
10. Host A, Andrae S, Charkin S, Diaz-Vazquez C, Dreborg S, Eigenmann PA, et al. 460
Allergy testing in children: why, who, when and how? Allergy. 2003; 58(7): 559-69. 461
11. Definition and diagnosis of diabetes mellitus and intermediate hyperglycemia : 462
report of a WHO/IDF consultation. Geneva, Switzerland: World Health Organisation/ 463
International Diabetes Federation; 2006. 464
12. Petty RE, Southwood TR, Manners P, Baum J, Glass DN, Goldenberg J, et al. 465
International League of Associations for Rheumatology classification of juvenile idiopathic 466
arthritis: second revision, Edmonton, 2001. The Journal of rheumatology. 2004; 31(2): 390-2. 467
13. Taieb A, Picardo M. The definition and assessment of vitiligo: a consensus report of 468
the Vitiligo European Task Force. Pigment cell research / sponsored by the European Society 469
for Pigment Cell Research and the International Pigment Cell Society. 2007; 20(1): 27-35. 470
14. Higgins JP, Altman DG, Gotzsche PC, Juni P, Moher D, Oxman AD, et al. The 471
Cochrane Collaboration's tool for assessing risk of bias in randomised trials. BMJ. 2011; 343: 472
d5928. 473
15. Methods for development of NICE public health guidance. London: National Institute 474
for Health and Clinical Excellence; 2006. 475
16. Higgins JP, Green S. Cochrane Handbook for Systematic Reviews of Interventions 476
Version 5.1.0.: The Cochrane Collaboration; 2011. 477
Study protocol for Review C – Version 1.8 26th July 2013
C23
17. Wood LG, Garg ML, Smart JM, Scott HA, Barker D, Gibson PG. Manipulating 478
antioxidant intake in asthma: a randomized controlled trial. The American journal of clinical 479
nutrition. 2012; 96(3): 534-43. 480