BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

Low alcohol consumption and pregnancy and childhood outcomes: time to change guidelines indicating apparently ‘safe’ levels of alcohol during pregnancy? A systematic review and meta-analyses.

For peer review only Journal: BMJ Open

Manuscript ID bmjopen-2016-015410

Article Type: Research

Date Submitted by the Author: 02-Dec-2016

Complete List of Authors: Mamluk, Loubaba; School of Social and Community Medicine, IEU Edwards, Hannah; University of Bristol, School of Social and Community Medicine Savovic, Jelena; University of Bristol, NIHR CLAHRC West Leach, Verity; University of Bristol Jones, Tim; NIHR CLAHRC West, Moore, Theresa; NIHR CLAHRC West Ijaz , Sharea ; NIHR CLAHRC West Lewis, Sarah; University of Bristol, School of Social and Community Medicine Donovan, Jenny; NIHR CLAHRC West Lawlor, Debbie; Department of Social Medicine, University of Bristol, MRC http://bmjopen.bmj.com/ Integrative Unit Davey Smith, George; University of Bristol, Social Medicine Fraser, Abigail; University of Bristol, MRC Integrative Epidemiology Unit Zuccolo, Luisa; University of Bristol, School of Social and Community Medicine

Primary Subject Epidemiology Heading: on October 2, 2021 by guest. Protected copyright. Secondary Subject Heading: Reproductive medicine

Keywords: PAEDIATRICS, EPIDEMIOLOGY, Maternal medicine < OBSTETRICS

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 1 of 47 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Systematic Review & Metaanalysis 4 5 6 7 Title: Low alcohol consumption and pregnancy and childhood outcomes: time to change 8 9 guidelines indicating apparently ‘safe’ levels of alcohol during pregnancy? A systematic review 10 11 and metaanalyses. 12 13 14 15 Running tittle:For Low alcohol peer consumption reviewand pregnancy and childhood only outcomes. 16 17 18 19 Authors: Loubaba Mamluk PhD1,2,3, Hannah B Edwards MSc2,3, Jelena Savović PhD2,3, Verity 20 21 Leach PhD 2,3, Timothy Jones PhD 2,3, Theresa HM Moore MSc 2,3, Sharea Ijaz PhD 2,3, Sarah 22 23 J. Lewis PhD 2, Jenny L. Donovan PhD 2,3, Deborah A. Lawlor PhD 1,2,3, 24 25 PhD 1,2, Abigail Fraser PhD*1,2, Luisa Zuccolo PhD*1,2 26 27 * Joint Last Authors 28 29 30 31 32 Author addresses: Hannah B Edwards, Jelena Savovic, Verity Leach, Timothy Jones, 33

34 Theresa HM Moore, Sharea Ijaz, and Jenny L. Donovan: NIHR CLAHRC West, University http://bmjopen.bmj.com/ 35 36 Hospitals Bristol, NHS Foundation Trust, 9th Floor, Whitefriars, Lewins Mead, Bristol, BS1 37 38 2NT, United Kingdom. Sarah J. Lewis, Deborah A. Lawlor, George Davey Smith, Abigail 39 40 Fraser, and Luisa Zuccolo: MRC Integrative Epidemiology Unit, School of Social and 41

42 Community Medicine, University of Bristol, Oakfield House, Oakfield Grove, Bristol BS8 2BN, on October 2, 2021 by guest. Protected copyright. 43 44 United Kingdom. 45 46 47 Corresponding author: Loubaba Mamluk, MRC Integrative Epidemiology Unit, School of 48 49 Social and Community Medicine, University of Bristol, Oakfield House, Oakfield Grove, Bristol 50 51 BS8 2BN, United Kingdom. NIHR CLAHRC West, University Hospitals Bristol 52 53 NHS Foundation Trust, 9th Floor, Whitefriars, Lewins Mead Bristol, BS1 2NT. United Kingdom 54 55 T+44 (0)117 3313378, [email protected] 56 57 1 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 2 of 47 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Author affiliations: 4 5 6 1 MRC Integrative Epidemiology Unit at the University of Bristol, Bristol 7 8 2 School of Social and Community Medicine, University of Bristol, Bristol 9 10 3 NIHR CLAHRC West, University Hospitals Bristol NHS Foundation Trust, Bristol 11 12 Word Count: 3257 13 Figures: 3 14 Tables: 3 15 SupplementaryFor Tables: 3peer review only 16 Supplementary Figures 3 17 Number of references: 53 18 19 20 21 What this paper adds 22 23 What is already known on this subject 24 Until recently UK guidelines advised women to avoid drinking alcohol while trying to conceive, and in the 25 first trimester, but at the same time indicated that consumption should be restricted to within “1 to 2 UK 26 27 units, once or twice a week. 28 Despite the guidance on light drinking versus abstinence being the point of confusion for health 29 professionals and pregnant women and contributing to inconsistent guidance and advice, previous 30 31 reviews have reported on a range of “low/moderate” alcohol consumption levels in pregnancy and have 32 not focused specifically on the threshold that current UK guidelines refer to. 33

34 Additionally, they did not specifically seek out study designs that reduce the impact of confounding and http://bmjopen.bmj.com/ 35 other forms of bias on the effect estimates. 36 37 What this study adds 38 We examined the effects of light alcohol consumption during pregnancy using the best quality evidence by 39 only including studies with prospective assessment of exposure and prioritising results adjusted for main 40 confounders. 41 We uniquely sought to include alternative study designs to further improve causal inference alongside

42 on October 2, 2021 by guest. Protected copyright. 43 standard analytical approaches. 44 45 We found very few studies employing either standard or alternative analytical approaches that answered 46 the research question focussed on the specific range of exposure. Pooled estimates from these studies 47 provided some evidence that even light alcohol consumption in pregnancy is associated with risk of 48 49 preterm delivery and small for gestational age. 50 We found very little evidence to support recommendations of a 32g alcohol/week limit with respect to most 51 52 outcomes. 53 54 55 56 57 2 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 3 of 47 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Abstract: 4 5 Objectives: To determine the effects of lowtomoderate levels of maternal alcohol 6 7 consumption in pregnancy on pregnancy and longerterm offspring outcomes. 8 9 Search Strategy: Medline, Embase, Web of Science, and Psychinfo from inception to 1107 10 11 2016. 12 13 Selection Criteria: Prospective observational studies, negative control and quasi 14 15 experimental studiesFor of pregnant peer women estimatingreview effects of lightonly drinking in pregnancy 16 17 (≤32g/week) versus abstaining. Pregnancy outcomes such as birth weight, and features of 18 19 fetal alcohol syndrome were examined. 20 21 Data Collection and Analysis: One reviewer extracted data and another checked extracted 22 23 data. Random effects metaanalyses were performed where applicable, and a narrative 24 25 summary of findings was carried out otherwise. 26 27 Main Results: 24 cohort and two quasiexperimental studies were included. With the 28 29 exception of birth size and gestational age, there was insufficient data to metaanalyse or 30 31 make robust conclusions. Being smallforgestationalage (SGA) and preterm birth odds were 32 33 higher for babies whose mothers consumed up to 32g/week versus none, but estimates for

34 http://bmjopen.bmj.com/ 35 preterm birth included the null value: summary odd ratios (OR) 1—08, 95% confidence intervals 36 37 (CI) (1—02 to 1—14), I2 0%, OR 1—10, 95%CI (0—95 to1—28), I2 60.2%, respectively. 38 39 40 Conclusion: Evidence of the effects of drinking <=32g/w in pregnancy is sparse. As there 41 was some evidence that even light prenatal alcohol consumption is associated with being

42 on October 2, 2021 by guest. Protected copyright. 43 44 SGA and preterm delivery, guidance could advise abstention as a precautionary principle, but 45 46 should explain the paucity of evidence. 47 48 Keywords: Alcohol, pregnancy, systematic review. 49 50 51 52 53 54 55 56 57 3 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 4 of 47 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Strengths and Limitations of this study: 4 5 Strengths 6 7 Completeness of searches with a focused research question aimed at informing alcohol in 8 9 pregnancy guidelines. 10 11 Biases minimised by only including those with prospective assessment of exposure and 12 13 prioritising results adjusted for main confounders. 14 15 For peer review only 16 Unique effort to include alternative study designs to further improve causal inference 17 18 alongside standard analytical approaches. 19 20 Limitations 21 22 Limitation of results on the effects of light drinking in pregnancy from standard analytical 23 24 approaches is bias due to residual confounding. 25 26 The inclusion of only English language studies may have led to missing some studies, 27 28 however there is little evidence that exclusion of nonEnglish language studies leads to 29 30 systematic bias in systematic reviews of conventional medicine. 31 32 We could not pool eligible studies for various reasons (e.g. too few studies, lack of 33

34 standard errors) http://bmjopen.bmj.com/ 35 36 37 38 39 40 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 4 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 5 of 47 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Introduction 4 Alcohol is a known teratogen[1] and the evidence about the risks of heavy alcohol 5 6 consumption during pregnancy on intellectual ability, birth defects, behaviour, fine motor 7 8 skills, and mental health (comprising fetal alcohol spectrum disorder – FASD)[2] is clear and 9 10 compelling.[3] Internationally, clinical guidelines recommend that pregnant women should 11 12 abstain from heavy or “binge” drinking.[4] However, until recently UK guidelines advised 13 14 women to avoid drinking alcohol while trying to conceive, and in the first trimester, but at the 15 For peer review only 16 17 same time indicated that consumption should be restricted to within “1 to 2 UK units, once or 18 19 twice a week”.[5] The UK Chief Medical Officer (CMO) commissioned a review of guidelines 20 21 on alcohol consumption during pregnancy. Based on a review of reviews, the Guidelines 22 23 Development Expert Group has recently proposed a change to guidelines such that women 24 25 should be advised to abstain from alcohol when pregnant and/or trying to conceive,[6] based 26 27 on the precautionary principle (i.e. “better safe than sorry”), in the absence of robust 28 29 evidence. 30 31 32 Our aim was to conduct a comprehensive systematic review and metaanalysis of the 33 literature to determine the effects of lowtomoderate levels of maternal alcohol consumption 34 http://bmjopen.bmj.com/ 35 36 in pregnancy on pregnancy and longer term offspring outcomes. Here we report on alcohol 37 38 consumption of up to two UK units of alcohol up to twice a week (the equivalent of ~ 39 40 32g/week), compared to no alcohol. In the absence of evidence from randomised controlled 41 trials, we examine observational studies of pregnant women from the general population with

42 on October 2, 2021 by guest. Protected copyright. 43 44 prospective assessment of alcohol exposure, to reduce recall bias. In particular, we 45 46 specifically seek out quasiexperimental studies, negative control comparisons, and 47 48 Mendelian randomisation analyses in order to reduce the impact of confounding and 49 50 measurement error on the effect estimates. 51 52 53 54 55 56 57 5 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 6 of 47 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Methods 4 5 Selection strategy and selection criteria 6 A full protocol of this systematic review carried out using PRISMA guidelines[7] is available 7 8 from the PROSPERO systematic review register (registration number CRD4201501594; 9 10 http://www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42015015941). 11 12 In brief, eligible studies were defined as epidemiological studies of pregnant women or 13 14 women trying to conceive with prospective assessment of prenatal alcohol exposure (i.e. 15 For peer review only 16 before birth), sampled from general population. The protocol specifically included studies 17 18 19 using standard analytical approaches (e.g. multivariable regression analysis), as well as 20 21 studies that used innovative analytical methods to improve causal inference, such as (i) 22 23 quasiexperimental studies (for example comparing outcomes before and after 24 25 implementation of new guidelines on alcohol consumption); (ii) negative control studies (e.g. 26 27 comparing the association of offspring outcomes with maternal alcohol consumption to the 28 29 association of the same outcomes with consumption among fathers, under the assumption 30 31 that confounding is likely to be similar but that if there was a direct causal effect of maternal 32 33 consumption on outcomes, maternal associations would be stronger); and Mendelian

34 http://bmjopen.bmj.com/ 35 randomisation studies (using genetic variants associated with alcohol consumption and 36 37 metabolism). We considered these analytical approaches to be the most appropriate in 38 39 terms of their ability to minimise bias from confounding and other sources. Our original 40 41 protocol included studies exploring the effects of prenatal alcohol consumption up to

42 on October 2, 2021 by guest. Protected copyright. 43 83g/week (the commonly used threshold for moderate consumption [810]) versus 44 45 abstinence. Here we have focused specifically on low alcohol consumption, i.e. up to 46 47 32g/week as this was the cutoff specified by the UK guidelines at the time of writing this 48 49 review as being an implicitly “safe” threshold.[5] This specific cut off value has not been 50 51 reviewed and is the main point of discussion as the guideline change from low consumption 52 53 (equating to 1 to 2 UK units, once or twice a week or 32g/week) to abstinence. 54 55 56 57 6 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 7 of 47 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Outcomes included: 1) pregnancy outcomes: still birth; miscarriage; gestational length and 4 5 preterm delivery; hypertensive disorders of pregnancy; gestational diabetes; small for 6 7 gestational age (SGA) and birth size (weight, length, and head circumference); low amniotic 8 9 fluid (oligohydramnios); placenta previa; placental abruption; assisted delivery (including 10 11 vacuum extraction, forceps delivery, Caesarean section); Apgar score at birth; admission to 12 13 neonatal unit; congenital malformations. 2) Features of foetal alcohol spectrum disorder 14 15 (FASD): childhoodFor growth peer restriction; cranium review size and head circumference;only developmental 16 17 delays; behaviour problems; cognitive impairment and intelligent quotient (IQ); facial 18 19 malformations. 20 21 22 Studies were excluded if: there was no quantitative measure of alcohol consumption that 23 24 could be converted to grams of alcohol/week; there was insufficient data to estimate the 25 26 effect size of the association of our predefined low consumption categories versus 27 28 abstinence with any outcome, including studies that analysed alcohol as a continuous 29 30 variable (i.e. assuming the same linear or log linear effect across the entire alcohol 31 32 distribution); the lowest exposure category (compared to nondrinkers) had an upper bound 33

34 exceeding 32g/week, or was unspecified; they were cohort studies of pregnant women with http://bmjopen.bmj.com/ 35 36 alcohol abuse/dependency; they were casecontrol studies or cohort studies with 37 38 retrospective alcohol consumption assessment (e.g. after birth). 39 40 The following databases were searched: MEDLINE, PsycINFO, EMBASE on Ovid; the 41

42 on October 2, 2021 by guest. Protected copyright. Cochrane Library including CENTRAL (the Cochrane Central Database of Controlled Trials) 43 44 on Wiley Interscience; and Science Citation Index, Social Science Citation Index, on Web of 45 46 47 Science from inception to 11 July 2016 (supplementary Table 1). We limited the search to 48 49 English language papers and excluded animal studies, letters, editorials, and conference 50 51 proceedings for which there were no fulltext papers. Searches were tailored to each 52 53 database by investigators. The search focused on published medical literature and did not 54 55 include grey literature. We additionally performed manual searches of the reference lists of: 56 57 7 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 8 of 47 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 (i) all papers included in recent systematic reviews of the effects of prenatal alcohol 4 5 exposure on the outcomes of interest; and (ii) all recent papers citing those reviews. 6 7 Titles and abstracts, and full texts if necessary, were screened independently by two 8 9 reviewers. Discrepancies were discussed and disagreements resolved through consensus. 10 11 We assessed potential for bias in included studies by assessing how well the study adjusted 12 13 for known confounders known to impact on the exposureoutcome associations (namely: 14 15 socioeconomicFor positioning, peer smoking during review pregnancy, maternal only age, and ethnicity. We 16 17 considered the potential for confounding and bias across studies included in the analyses 18 19 and described it narratively alongside summary results. 20 21 22 Data extraction 23 24 Data were extracted into a custom built Microsoft Access database. We extracted the 25 26 following information from each study: title, authors, publication year, country/region, study 27 28 design, population characteristics (sample size, methods of sampling, age distribution, and 29 30 ethnicity), measures of exposure (assessment method; including timing and quantification of 31 32 alcohol consumption, reference group, categories of exposure, and information on unit 33

34 equivalence if stated), outcome assessment methods (including whether this was abstracted http://bmjopen.bmj.com/ 35 36 from medical records, obtained via a research interview and the person reporting the 37 38 outcome e.g. parent, teacher, health professional, researcher or child), model adjustments, 39 40 and study results. If a study reported more than one result for each outcome, we extracted 41

42 all of them (e.g. relative to different timing of exposure, model adjustments, etc.). Information on October 2, 2021 by guest. Protected copyright. 43 44 from each included paper was extracted by one reviewer (LM) and subsequently checked 45 46 for accuracy and completeness by another reviewer (HE).[11] Extraction errors were minimal 47 48 and were resolved through discussion between extractor and checker. 49 50 51 52 53 54 55 56 57 8 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 9 of 47 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Alcohol unit conversion 4 Alcohol consumption in drinks/week was converted into grams/week based on the pure 5 6 ethanol equivalent of one drink, as stated in each individual article, or otherwise inferred 7 8

9 based on the definition of standard drinks in the country where the study took place. 10 11 Data analysis 12 The association of low alcohol use with pregnancy and related outcomes was investigated 13 14 comparing the highest category within the range of 032g/week to abstention (during 15 For peer review only 16 17 pregnancy). In studies providing data across several categories of intake within the 0 18 19 32g/week range, we used the effect estimate for the highest category of intake. If studies 20 21 reported on exposure to alcohol during different trimesters, we included estimates relative to 22 23 the earliest exposure. Similarly, if results were available from both unadjusted and adjusted 24 25 regressions, we prioritised fully adjusted results, as a way of minimising the impact of 26 27 confounding by important factors such as maternal smoking, age, socioeconomic position 28 29 and ethnicity. In case of multiple results from the same cohort (relative to the same 30 31 outcome), we analysed those pertaining to the largest population size (i.e. conducted on the 32 33 least ‘selected’ population as result of exclusions, to minimise selection bias). Results from

34 http://bmjopen.bmj.com/ 35 all studies that fulfilled our inclusion criteria were summarised, together with information 36 37 about the study. Where appropriate, we additionally pooled results for each outcome. 38 39 Authors were not contacted for extra data. 40 41 Individual study estimates were pooled using random effects metaanalysis. When

42 on October 2, 2021 by guest. Protected copyright. 43 continuous outcomes were measured using different scales, we derived and pooled Cohen’s 44 45 d statistics (representing standardised differences in means by level of exposure). Where 46 47 only two studies were available to metaanalyse, results were pooled unless they were very 48 49 different from each other (I2 ≥ 50%).[12] In this case, a narrative summary of findings was 50 51 carried out and results were reported in Table 2. Where a study only reported unadjusted 52 53 results, we kept these separate in the forest plots (subgroup analysis) but then also showed 54 55 overall pooled estimates combining all results. 56 57 9 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 10 of 47 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Planned subgroup analyses by trimester could not be performed due to insufficient number 4 5 of included studies with this information. 6 7 The likelihood of small study bias deriving from publication bias was assessed through visual 8 9 inspection of funnel plots for pooled analyses including ≥4 studies. 10 11 Statistical analyses were carried out using Stata version 13.1 (StataCorp, College Station, 12 13 TX, USA).[13] 14 15 Results For peer review only 16 17 A flowchart of the article review process is shown in Figure 1. A total of 4680 citation records 18 19 were identified from searching the four relevant databases. A manual search of recent 20 21 systematic reviews identified 33 additional articles. After exclusions, 24 prospective studies 22 23 analysed using standard approaches and two quasiexperimental studies were included, 24 25 reporting on 30 outcomes in total. 26 27 Standard analytical approaches 28 29 Pooled estimates for continuous and binary outcomes are presented in Figure 2 and 3 30 31 respectively. 32 33 The outcome reported by the largest number of studies was preterm delivery (9 studies, for a

34 http://bmjopen.bmj.com/ 35 total of 318832 participants, Figure 3A), followed by birthweight and SGA (7 studies each, 36 37 Figures 2 and 3B), and low birthweight (5 studies, Figure 3C). 38 39 The metaanalysis yielded a summary OR of 1—10 (95% CI 0—95; 1—28) for preterm delivery, 40 41 but there was substantial statistical heterogeneity between studies (I2=60—2%), due to a large

42 on October 2, 2021 by guest. Protected copyright. 43 Danish study reporting a protective effect (Figure 3A). There was also modest evidence for 44 45 an increased risk of being SGA (OR 1—08, 95% CI 1—02; 1—14) for a total of 288512 46 47 participants, although this was almost entirely driven by a single US study contributing 95% 48 49 of the participants to this metaanalysis (Figure 3B). The birthweight metaanalysis yielded a 50 51 summary effect of 13—9g (95% CI 13—49g; +3—31g) for offspring of light drinkers versus non 52 53 drinkers (Figure 2). 54 55 56 57 10 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 11 of 47 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Other outcomes were typically reported by a limited number of studies and mostly could not 4 5 be metaanalysed due to clinical heterogeneity in outcome assessment or incompleteness of 6 7 published data (supplementary Figure 1, and Table 2). Based on two studies with data on 8 9 behavioural outcomes, there was little evidence of any effect for internalising symptoms but 10 11 a suggestion that light drinking in pregnancy protected against high externalising behaviour 12 13 scores (OR 0—97, (95%CI 0—93; 1—01, supplementary Figure 1)). However, an additional 14 15 study assessingFor conduct peer problems and hyperactivityreview (in the same only externalising domain) 16 17 reported results in the opposite direction, which could not be metaanalysed due to different 18 19 outcome definitions. 20 21 Table 2 presents results of included studies that did not contribute to the metaanalyses for 22 23 various reasons. There was no strong evidence of association between consuming up to 32 24 25 g/week of alcohol and any of the remaining outcomes excluded from metaanalyses, with 26 27 three exceptions: a very large US study showing increased risk of placental abruption and 28 29 decreased risk of preeclampsia (OR 1—24, 95%CI 1—05; 1—46 and OR 0—82, 95% CI 0—74; 30 31 0—90, respectively)[14], and a single British study reporting better cognitive outcomes in 32 33 children exposed to light maternal drinking in pregnancy.[15] 34 http://bmjopen.bmj.com/ 35 36 For outcomes with a sufficient number of studies (≥4 studies), there was little evidence of 37 38 small study effect based on inspection of funnel plots (supplemental Figures 25), with only 39 40 preterm birth showing some asymmetry due to the three smallest studies showing point 41 estimates in the direction of increased risk (supplementary Figure 2).

42 on October 2, 2021 by guest. Protected copyright. 43 44 Of all included results, only two were unadjusted[16 17], and most of the others were 45 46 adjusted for maternal smoking, age and socioeconomic position (supplementary Table 2). 47 48 Studies that did not adjust for ethnicity were generally conducted in homogenous 49 50 populations. Due to the small number of studies for each outcome, we could not further 51 52 investigate the effect of adjusting for all or some of these confounders. Similarly, there was 53 54 insufficient data to examine the effect of timing of exposure on outcomes. 55 56 57 11 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 12 of 47 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Alternative analytical approaches 4 Two negative control publications[18] based on the same UK cohort met our inclusion 5 6 criteria.[17 19] They investigated the effects of maternal alcohol consumption on childhood 7 8 9 educational achievement[19] and IQ.[17] Offspring exposed to maternal consumption of 10 11 <12g/week of alcohol in the first trimester did not have worse outcomes compared to those 12 13 of mothers who abstaining from alcohol, and a similar pattern was found for paternal alcohol 14 15 consumption.For peer review only 16 17 One further quasiexperimental study, one natural experiment and five Mendelian 18 19 randomisation studies were excluded from the present review because they did not 20 21 specifically test the effect of consuming up to 32g/week in pregnancy versus abstaining. 22 23 These will be included in a forthcoming review focused on estimating the causal effects of 24 25 prenatal alcohol exposure based on alternative study designs and analytical approaches to 26 27 strengthen causal evidence. 28 29 30 31 32 33

34 http://bmjopen.bmj.com/ 35 36 37 38 39 40 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 12 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 13 13

Yes Yes Yes Yes Yes No yes

Yes Yes Yes Yes Yes Adjusted (yes/no) Adjusted

Yes Yes Yes Yes yes yes

11 months 11 - 7.75 years 7.75 Birth Birth Birth Birth Birth Birth 7 years 7 0 Birth Birth Birth Age at outcome outcome at Age assessment assessment (child) Birth Birth Birth Birth 2-14 years years 2-14 inclusive 7.75 years years 7.75 Yes

analysis) - ifficulties Questionnaire (SDQ), (SDQ), Questionnaire ifficulties in the meta the in

included Externalising Externalising

-

- Externalising Externalising -

BMJ Open

em scores, Strengths & Difficulties Questionnaire (SDQ) (SDQ) Questionnaire Difficulties & scores, Strengths em

http://bmjopen.bmj.com/ percentile)

onduct problems onduct th Hyperactivity

centile) centile) percentile) percentile) percentile) percentile)

th th th th eight (g) eight th weight (g) weight th Birth weight (g) weight Birth Low birth weight <2500g weight birth Low birth Preterm weight birth mean Adjusted weight birth Low Preterm birth (<37 weeks) (<37 birth Preterm 10 (lowest SGA w Birth Bir weeks <37 birth Preterm (<2500g) weight Birth 10% with highest probl highest with 10% C report parent weeks) (32-37 birth preterm Moderate weeks) (<32 birth preterm Very weight birth Low delivery Preterm parent report report parent 10% with highest problem scores, Strengths & D & scores, Strengths problem highest with 10% birth Preterm birth Preterm weight birth Low (SDQ) Questionnaire Difficulties & scores, Strengths problem highest with 10% Parent report report Parent behaviour Internalising Boys behaviour Girls Internalising behaviour Externalising Boys behaviour Girls Externalising 10 (< SGA 10 (< SGA scores) deviaon standard −2 < (weight SGA birth Preterm weight Birth Outcomes Outcomes (<2500g) weight Birth 10 (< SGA Customised birth centile centile birth Customised 10 (< SGA Parent report report Parent behaviour Internalising behaviour Externalising

on October 2, 2021 by guest. Protected copyright.

535 535 1,140 1,140 Birthweight 720 720 29,463 2,063 7,052 398 901 11,698 700 5,710 1,493 276,288 4,132 Number in in Number 3,916 analysis analysis prenatal alcohol exposure on pregnancy outcomes (studies (studies outcomes pregnancy on exposure alcohol prenatal

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

For peer review only UK UK UK UK UK UK Denmark USA UK USA UK Denmark UK Denmark Japan USA UK Holland; Country USA Australia Australia 1849 (CBCL) Checklist Behaviour Child ≥60), score (T problems significant Clinically

Prospective studies examining the effects of effects the studies examining Prospective

Albertsen (2004) Albertsen (1997) Lundsberg (1996) Passaro (1995) Shu (1995) Peacock (1991) Olsen Sayal (2007) Sayal Niclasen (2014) Niclasen (2014) Miyake (2011) Salihu (2007) Jaddoe Lundsberg (2015) Lundsberg

31] [27] [28] [29] [30] [16] [ [26] (1989) Brooke [32] SGA :small for gestation age age gestation for :small SGA [22] [23] [14] [25] [21] Nykjaer (2014) (2014) Nykjaer [21] (2010) Robinson [24] (2007) Sayal [26] Table 1. Table Study (year) [20]

Page 13 of 47 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from Page 14 of 47 14 14

Yes Yes Yes Yes Yes Yes Adjusted Adjusted (yes/no) Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

1·39) 1·46) 1·42) 1·42) 3·84) 1·54) 1·54) 1·50)

– – – – – – 1·5)

(95% CI 0·88 CI (95% [1] those unexposed. unexposed. those exposed, compared to 1·85 for those unexposed unexposed those for 1·85 to compared exposed, (p=0·26) unexposed. unexposed. OR 1·05 (95% CI 0·94 – 1·18) 1·18) 0·94– CI 1·05(95% OR Results Results OR 0·90 (95% CI 0·73 – 1·12) 1·12) 0·73– CI 0·90(95% OR OR 1·10 OR OR 1·0 (95% CI 0·7 0·7 CI 1·0 (95% OR OR 1·12 (95% CI 0·86 CI 1·12(95% OR OR 1·08 (95% CI 0·83 CI 1·08(95% OR 0·67 CI 1·61(95% OR OR 1·10 (95% CI 0·78 CI 1·10(95% OR 3225g in those exposed, compared to 3225g in 3225g to compared exposed, those in 3225g unexposed. those Mean gestational age 40·1 (SD 1·9) for those those 1·9) for (SD 40·1 age gestational Mean those for 2·0) (SD 40·1 to compared exposed, unexposed. Mean total anomalies was 2·60 for those those for 2·60 was anomalies total Mean unexposed those for 2·53 to compared exposed, (p=0·28) OR 0·78 (95% CI 0·40 CI 0·78(95% OR 3·7% in those exposed compared to 3·9% in those those in 3·9% to compared exposed in those 3·7% 3414g in those exposed, compared to 3363g in 3363g to compared exposed, those in 3414g Mean craniofacial anomalies was 1·92 for those those 1·92for was anomalies craniofacial Mean OR 1·06 (95% CI 0·74 – 1·50) 1·50) 0·74 – CI 1·06(95% OR

National / Hospital Hospital / National from or Registries, report) maternal Age at outcome outcome at Age (child) assessment (Outcome confirmed via confirmed (Outcome Hospital / National from or Registries, report) maternal Birth (Outcome confirmed via confirmed (Outcome records) medical Birth Birth Birth Birth Birth Birth Birth Birth

Birth Birth Birth Birth analysis) -

in the meta the in outcome

included BMJ Open analysis analysis Estimates are in different different in are Estimates directions directions directions Estimates are in different different in are Estimates Only study with this with study Only Only study with this outcome outcome this with study Only via confirmed (Outcome Only study with this outcome this with study Only Only study with this outcome this with study Only outcome this with study Only Only study with this outcome this with study Only SE/SD not reported not SE/SD SE/SD not reported reported not SE/SD

Only study with this outcome this with study Only SE/SD not reported not SE/SD SE/SD not reported reported not SE/SD Only study with this outcome outcome this with study Only Birth Only study with this outcome this with study Only http://bmjopen.bmj.com/

percentile percentile percentile percentile

th th

Craniofacial anomalies anomalies Craniofacial Oral clefts clefts Oral Mean birthweight (g) in (g) birthweight Mean non-smokers trimester trimester Stillbirth Spontaneous abortion abortion Spontaneous occurring ≤20 weeks weeks ≤20 occurring gestation Admission to NICU to Admission Lowest 10 Lowest 5 minutes at <7 Apgar Lowest 10 Lowest circumference head of Mean birthweight (g) in in (g) birthweight Mean length birth of smokers smokers weeks Gestational age Gestational with infants of cent Per <37 age gestational Total anomalies Total Major malformations Major

on October 2, 2021 by guest. Protected copyright.

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Country (sample size) size) (sample Country details Outcome meta- in included not Reason France, Spain, Denmark, Denmark, Spain, France, (8,453) Portugal Germany, France, Spain, Denmark, Denmark, Spain, France, USA (1,226,685) USA USA (5,324) USA USA (4,496) USA USA (4,496) USA (4,496) USA USA (4,496) USA Netherlands, Scotland, Scotland, Netherlands, France, Spain, Denmark, Denmark, Spain, France, (8,453) Portugal Germany, Germany, Portugal (8,453) Portugal Germany, UK (10,539) UK USA (873) USA USA (4,496) USA For peer review only

Salihu (2011) Salihu Windham (1997) Windham Lundsberg (2015) Lundsberg Lundsberg (2015) Lundsberg (2015) Lundsberg Lundsberg (2015) Lundsberg Ogston (1992) Ogston Passaro (1996) Passaro Ernhart (1989) Ernhart Lundsberg (2015) Lundsberg

[35] Ogston (1992) (1992) Ogston [35] Scotland, Netherlands, [36] Andersen (2012) (2012) Andersen [36] (89,322) Denmark Stillbirth [14] [37] [20] [20] [20] [20] [35] [29] (1992) Ogston [35] Scotland, Netherlands, [34] [20] (2007) Bille [33] (1020) Denmark [36] Andersen (2012) (2012) Andersen [36] (89,322) Denmark in first Miscarriage [34] Ernhart (1989) (1989) Ernhart [34] (873) USA

Prospective studies examining the effects of prenatal alcohol exposure on pregnancy outcomes (results not not (results outcomes pregnancy on exposure alcohol prenatal of effects the examining studies Prospective

Birthweight Birthweight

Stillbirth Stillbirth Admission to to Admission Apgar score Apgar Neonatal Care Intensive (NICU) Unit Height Head circumference Gestational age age Gestational preterm and birth Table 2. Table Malformations Outcome Outcome (year) Study Miscarriage Miscarriage 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 15 15

Yes Yes Yes Yes No No Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes 0·78) on 0·78)

– 0·02 0·02 -

) 99 · 1·32) 0 1·03) 0·90) 1·46)

1·94)

– – – – – –

5 3 0·38 (95% CI CI 0·38(95% · 1·02 (95% CI 0 CI (95%

59 · exposed compared to unexposed. unexposed. to compared exposed compared to 105·7 (SD 16·2) in those unexposed, unexposed, those in 16·2) (SD 105·7 to compared p=0·10. Mean increase of 1·80 points (SE 1·1) for those those for 1·1) (SE 1·80points of increase Mean of increase Mean exposed compared to unexposed. unexposed. to compared exposed to compared exposed those for score KS2 unexposed. OR 1·41 (95% CI CI 1·41(95% OR OR 1·14 (95% CI 0·98 CI 1·14(95% OR OR 0 OR 0·70 CI 0·85(95% OR OR 0·82 (95% CI 0·74 CI 0·82(95% OR OR 1·24 (95% CI 1·05 CI 1·24(95% OR Mean increase of 0·81 points (se 0·8) for those those for 0·8) (se 0·81points of increase Mean OR 0·99 (95% CI 0·86 – 1·16) 1·16) 0·86– CI 0·99(95% OR Mean IQ score 106·4 (SD 16·3) in those exposed, exposed, inthose 16·3) (SD 106·4 IQ score Mean OR 1·11 (95% CI 0·87 – 1·43) 1·43) 0·87 – CI 1·11(95% OR

9 years 9

- 18 months 18 months 7.75- 9 years years 7.75- 9 1·51) 0·96– CI 1·20(95% OR 18 18 months 11 years 7.75 3.9 3.9 years Birth years 7 Birth Birth

with this this with two studies twostudies BMJ Open No SE/SD reported for for reported SE/SD No group reference Only study with this outcome outcome this with study Only years 8 Only study with this outcome this with study Only No SE/SD reported for for reported SE/SD No group reference instead instead Parent report used in analysis analysis in used report Parent Parent report used in analysis analysis in used report Parent instead Later age used in analysis in analysis used Laterage 3.9 years Later age used in analysis in analysis used Laterage Only study with this outcome this with study Only Only two studies with this this with twostudies Only outcome Only outcome Only study with this outcome outcome this with study Only Birth Only study with this outcome this with study Only http://bmjopen.bmj.com/

onduct onduct

evelopment evelopment eclampsia eclampsia - - onduct problems problems onduct IQ, Weschler Weschler IQ, Placenta previa previa Placenta Psychomotor Psychomotor Index, Development Infant of Scales Bayley Development Intelligence Scale for for Scale Intelligence (WISC-III) Children Key Stage 2 scores 2 Stage Key Mental D Mental Index, Bayley Scales of of Scales Bayley Index, Development Infant Hyperactivity teacher report report teacher Questionnaire (SDQ) (SDQ) Questionnaire Strengths & Difficulties &Difficulties Strengths problem scores, scores, problem 10% with highest highest 10% with C teacher report report teacher Questionnaire (SDQ) (SDQ) Questionnaire Strengths & Difficulties &Difficulties Strengths 10% with highest highest 10% with scores, problem Hyperactivity parent report report parent Questionnaire (SDQ) (SDQ) Questionnaire Strengths & Difficulties &Difficulties Strengths problem scores, scores, problem 10% with highest highest 10% with parent report C report parent Questionnaire (SDQ) (SDQ) Questionnaire Strengths & Difficulties &Difficulties Strengths 10% with highest highest 10% with scores, problem Developmental Developmental Pre Disorder Coordination (DCD) Pre Placental abruption Placental problems problems on October 2, 2021 by guest. Protected copyright.

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

w Zealand, Australia, Australia, Zealand, w France, Spain, Denmark, Denmark, Spain, France, (8,453) Portugal Germany, UK (10,558) UK Netherlands, Scotland, Scotland, Netherlands, France, Spain, Denmark, Denmark, Spain, France, (8,453) Portugal Germany, UK (257) UK UK (967) UK Denmark (32,097) Denmark Ne USA (1,226,685) USA UK Ireland, USA (1,226,685) USA For peer review only

(2013)

Sayal (2013) Sayal Parry (1992) Parry Sayal (2007) Sayal Sayal (2007) Sayal Faebo Larsen Larsen Faebo McCarthy Salihu (2011) Salihu Salihu (2011) Salihu

[26] Sayal (2007) (2007) Sayal [26] (525) UK [26] Sayal (2007) (2007) Sayal [26] (1077) UK [15] [40] [26] [26] [39] [38] (2013) [14] [14] (2011) Salihu [14] (1,226,685) USA [40] Parry (1992) (1992) Parry [40] Scotland, Netherlands, (2008) Alati [17] (4,332) UK

related - eclampsia -

[1] Odds Ratio results compare the odds of outcome in those exposed to ≤32g AA per week. SE: Standard error SD: standard deviation SGA :small for gestation age IQ: intelligence quotient. quotient. intelligence IQ: age gestation for :small SGA deviation standard SD: error Standard SE: week. per AA ≤32g to exposed those in outcome of odds the compare results Ratio Odds [1] Cognition

Motor Motor Behaviour /development development Pre Placenta Page 15 of 47 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from Page 16 of 47 16 16 BMJ Open http://bmjopen.bmj.com/ on October 2, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

For peer review only

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 17 17 Different confounding structures structures confounding Different vs. maternal of association the for outcome the with alcohol paternal Limitations Limitations Different confounding structures structures confounding Different vs. maternal of association the for outcome the with alcohol paternal Maternal alcohol consumption consumption alcohol Maternal 16·2); (SD 105·7 mean abstainers: (SD 106·4 mean glass/week: <1 16·3) consumption alcohol Paternal 16·8); (SD 102·2 mean abstainers: (SD 104·0 mean glass/week: <1 16·7) conclusions as presented in the in presented as conclusions paper Maternal alcohol consumption consumption alcohol Maternal 9·1); (SD 102·0 mean abstainers: (SD 102·8 mean glass/week: <1 8·7) consumption alcohol Paternal <1 11); (SD 98·9 mean abstainers: 9·1) (SD 101·1 mean glass/week: Outcomes Outcomes & results of Summary IQ: Wechsler IQ: Wechsler Scale Intelligence Children for (WISC) achievement: Key achievement: scores 2 Stage (standardised)

assessment assessment (child) 8 years years 8

11 years years 11 Academic Intelligence quotient. Intelligence BMJ Open http://bmjopen.bmj.com/

na

na

na na na on October 2, 2021 by guest. Protected copyright. n difference SD: standard deviation IQ: deviation SD:standard difference n paternal paternal - For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Maternal comparison comparison comparison For peer review only

4,332

Country Country size Sample type Study Gene number SNP-rs outcome at Age UK

UK UK 7,062 Maternal-paternal

Quasiexperimental studies examining the effects of prenatal alcohol exposure on pregnancy outcomes. outcomes. on pregnancy exposure alcohol prenatal of the effects examining studies Quasiexperimental Alati (2008) Alati

Study (year) Study(year) [17] Mea MD: intervals CI: Confidence ratio OddsOR: Abbreviations: Table 3. Table (2013) Alati [19] Page 17 of 47 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open Page 18 of 47 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Discussion 4 5 Main findings: We conducted a comprehensive systematic review of the literature aimed at 6 7 examining whether low levels of alcohol drinking in pregnancy have a causal detrimental 8 9 effect on pregnancy and offspring outcomes. Our two main findings are: i) a surprisingly 10 11 limited number of prospective studies specifically addressing the question of whether light 12 13 maternal alcohol consumption (i.e. up to 32g/week (or 4 UK units) has any causal effect 14 15 (adverse or beneficial)For onpeer infant and later review offspring outcomes andonly pregnancy outcomes, and, 16 17 as a result, ii) a paucity of evidence demonstrating a clear detrimental effect, or safe limit, of 18 19 light alcohol consumption on outcomes. The upper limit that we chose to examine here is 20 21 that of the current version of the UK National Institute for Health and Care Excellence (NICE) 22 23 guidelines.[41] The question we have attempted to address is very important given the 24 25 mixed advice that women are given with regards to whether they should abstain completely 26 27 or be allowed light alcohol consumption in pregnancy. The lack of research evidence to 28 29 address this question is notable. 30 31 32 33 Strengths and limitations: Strengths of this review include the completeness of searches with

34 http://bmjopen.bmj.com/ 35 a focused research question aimed at informing alcohol in pregnancy guidelines. In addition 36 37 to observational studies’ biases minimised by only including those with prospective 38 39 assessment of exposure and prioritising results adjusted for main confounders. Another 40 41 strength of this review is the unique effort to include alternative study designs to further

42 on October 2, 2021 by guest. Protected copyright. 43 improve causal inference alongside standard analytical approaches. The main limitation of 44 45 results on the effects of light drinking in pregnancy from standard analytical approaches is 46 47 bias due to residual confounding. Women who drink low amounts of alcohol may be more 48 49 likely to be of higher socioeconomic position, compared to abstainers, at least in developed 50 51 settings in recent years,[42] and both of these characteristics are associated with better 52 53 pregnancy and cognitive outcomes.[43] Maternal smoking and ethnicity are also known 54 55 56 correlates of maternal alcohol use, and risk factors for e.g. low birth weight.[44] Most studies 57 58 included in this review adjusted for at least some of these factors. However, due to the small 59 60 18

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 19 of 47 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 number of studies included for any given outcome, it was impossible to formally investigate 4 5 the effect of incomplete adjustment for some (or all) of these confounders. Additionally, for 6 7 most outcomes, we could not pool eligible studies for various reasons (e.g. too few studies, 8 9 lack of standard errors) and that we limited our review to a number of prespecified 10 11 outcomes including the most common pregnancyrelated outcomes and childhood outcomes 12 13 related to FASD. The inclusion of only English language studies may have led to missing 14 15 some studies,For however therepeer is little evidence review that exclusion ofonly nonEnglish language studies 16 17 leads to systematic bias in systematic reviews of conventional medicine.[4548] 18 19 20 21 Interpretation: This review demonstrates the paucity and poor quality of evidence addressing 22 23 this important question, and the difficulty of designing studies that can 24 25 effectively evaluate the causal impact of low alcohol consumption whilst minimising bias and 26 27 confounding. It also shows the value of reporting measures of effect for meaningful 28 29 categories of the exposure. Whilst many studies reported that associations did not differ from 30 31 linearity prior to providing a single coefficient for the doseresponse association, it is possible 32 33 that statistical power limited the ability to detect nonlinear associations in single studies.

34 http://bmjopen.bmj.com/ 35 Such detail is especially important when there are controversies about the shape of the 36 37 association of interest (linear, U or Jshaped) and/or the existence of safe thresholds. 38 39 Outstanding questions also remain about the effects of maternal alcohol consumption at 40 41 different stages of conception and pregnancy. Alternative analytical approaches such as

42 on October 2, 2021 by guest. Protected copyright. 43 sibling comparisons[49] and the use of instrumental variable approaches[50] as well as 44 45 triangulating the totality of evidence from multiple study types[51] (formally or informally) are 46 47 needed in order to strengthen confidence in the direction and size of any potential causal 48 49 50 relationships. 51 52 53 54 The recently proposed change in the guidelines for alcohol use in pregnancy in the UK to 55 56 complete abstinence, would be an application of the precautionary principle. This review 57 58 confirmed some increased risk of babies being born SGA but little direct evidence of any 59 60 19

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 20 of 47 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 other detrimental effect for maternal drinking up to 32g/week. However, there have been few 4 5 wellconducted studies examining this specific category of exposure. This issue remains of 6 7 great public health importance, with alcohol consumption during pregnancy prevalent in the 8 9 UK, Ireland, New Zealand and Australia and reaching up to 80% of pregnant women.[52] For 10 11 some, the evidence of the potential for harm – mostly coming from animal experiments and 12 13 human studies of effects due to higher levels of exposure will be sufficient to advocate that 14 15 guidelines shouldFor advise peer women to avoid review all alcohol in pregnancy, only while others will wish to 16 17 retain the existing wording of guidelines.[53] 18 19 20 21 22 In conclusion, we found limited evidence for a causal role of light drinking in pregnancy, 23 24 compared to abstaining, on most of the outcomes examined. Despite the distinction between 25 26 light drinking and abstinence being the point of most tension and confusion for health 27 28 professionals and pregnant women and contributing to inconsistent guidance and advice 29 30 now and in the past, our extensive review shows that this specific question is not being 31 32 researched thoroughly enough, if at all. In addition, there has been no evidence regarding 33

34 possible benefits of light alcohol consumption versus absence. Further studies, including http://bmjopen.bmj.com/ 35 36 those using designs that improve causal inference, are required to provide further evidence 37 38 and a better estimation of the likely effects, in particular to inform expectant women who 39 40 choose not to follow the new recommendation for abstinence or are anxious about their 41

42 on October 2, 2021 by guest. Protected copyright. drinking in the very early stages before pregnancy recognition. Formulating guidance on the 43 44 basis of the current evidence is challenging. However, describing the paucity of current 45 46 47 research and explaining that “absence of evidence is not evidence of absence”, appears 48 49 warranted. 50 51 52 53 54 55 56 57 58 59 60 20

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 21 of 47 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Acknowledgments 4 We thank our colleagues from NIHR CLAHRC West who provided support that greatly 5 6 assisted the research. We thank Louise Waters for assistance with article collection and 7 Alison Richards for developing the search strategy for this review. 8 9 10 11 Details of contributors 12 Loubaba Mamluk: Drafting the protocol, preparing the data extraction database, data 13 14 screening and collection (lead reviewer), data analysis, figures, tables, data interpretation, 15 and drafting Forthe manuscript. peer Hannah B Edwards:review screening, data only collection, data quality 16 17 checking, data interpretation, and preparation of final manuscript. Theresa HM Moore: 18 Advising on the protocol, preparing the protocol, screening title and abstract. Timothy Jones: 19 20 Data extraction and full text screening. Sharea Ijaz: Abstract and full text screening. Jelena 21 Savović: protocol development/study design (including the development of searching, article 22 23 screening and data collection strategies), oversaw and managed the review process and the 24 25 review team, provided methodological advice for the review and metaanalysis conduct, and 26 critically revised the manuscript. George Davey Smith: Obtained funding, formulation of 27 28 project plan, review of progress and of the manuscript. Jenny L. Donovan: Obtained funding 29 and contributed to drafting the manuscript, and approved the final version. Verity Leach: 30 31 Data extraction. Deborah A. Lawlor: contributed to the study design, data interpretation and 32 review of earlier manuscript drafts; she approved the final version that has been submitted. 33

34 Sarah J. Lewis: contributed to the study design, data interpretation and review of earlier http://bmjopen.bmj.com/ 35 manuscript drafts; she approved the final version that has been submitted. Abigail Fraser: 36 37 Contributed to the study design, data analysis and interpretation and drafting of manuscript. 38 Luisa Zuccolo: Contributed to the study design, data analysis and interpretation and drafting 39 40 of manuscript. I, Loubaba Mamluk, the corresponding author of this manuscript, certify that I 41 have full access to all the data in the study and had final responsibility for the decision to

42 on October 2, 2021 by guest. Protected copyright. 43 submit for publication. 44

45 46 Transparency declaration 47 48 I, Loubaba Mamluk, the corresponding author of this manuscript, affirm that this manuscript 49 50 is an honest, accurate, and transparent account of the study being reported; that no 51 important aspects of the study have been omitted; and that any discrepancies from the study 52 53 as planned (and, if relevant, registered) have been explained. 54 55 Conflict of interest: 56 57 The authors have nothing to disclose. 58 59 60 21

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 22 of 47 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 This work has not been submitted for publication elsewhere. 4

5 6 Disclosure 7 All authors have completed the ICMJE uniform disclosure form 8 9 at www.icmje.org/coi_disclosure.pdf and declare: no support from any organisation for the 10 11 submitted work; no financial relationships with any organisations that might have an interest 12 in the submitted work in the previous three years; no other relationships or activities that 13 14 could appear to have influenced the submitted work. 15 For peer review only 16 17 Role of funding source: 18 The UK Medical Research Council (MC_UU_12013/1 and MC_UU_12013/5) and the 19 20 University of Bristol provide core funding for the MRC Integrative Epidemiology Unit; LM, 21 DAL, GDS, AF, and LZ are partially or fully funded through this. AF and LZ are UK Medical 22 23 Research Council research fellowships (Grant refs: MR/M009351/1 and G0902144, 24 respectively). LM, JS, HE, VL, TJ, SI,TM and JLD are fully or partially partly funded by 25 26 National Institute for Health Research (NIHR) Collaboration for Leadership in Applied Health 27 Research and Care West (CLAHRC West) at University Hospitals Bristol NHS Foundation 28 29 Trust. DAL (NFSI061110196) and JLD (NFSI051210119) are NIHR Senior 30 Investigators. SL is a senior lecturer funded by HEFCE. The views expressed here are those 31 32 of the author(s) and not necessarily those of the NHS, the MRC, NIHR or the Department of 33 Health. I, Loubaba Mamluk, the corresponding author of this manuscript, certify that I have 34 http://bmjopen.bmj.com/ 35 full access to all the data in the study and had final responsibility for the decision to submit 36 37 for publication. 38 39 40 Transparency declaration 41 The lead author (Loubaba Mamluk) affirms that this manuscript is an honest, accurate, and

42 on October 2, 2021 by guest. Protected copyright. 43 transparent account of the study being reported; that no important aspects of the study have 44 been omitted; and that any discrepancies from the study as planned (and, if relevant, 45 46 registered) have been explained. 47 48 49 50 51 52 53 54 55 56 57 58 59 60 22

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 23 of 47 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 References 4 1. Streissguth AP, Landesman-Dwyer S, Martin JC, Smith DW. Teratogenic effects of alcohol in 5 humans and laboratory animals. Science (New York, N.Y.) 1980;209(4454):353-61 6 7 2. Popova S, Lange S, Shield K, et al. Comorbidity of fetal alcohol spectrum disorder: a systematic 8 review and meta-analysis. Lancet (London, England) 2016;387(10022):978-87 doi: 9 10.1016/s0140-6736(15)01345-8[published Online First: Epub Date]|. 10 3. Chudley AE, Conry J, Cook JL, Loock C, Rosales T, LeBlanc N. Fetal alcohol spectrum disorder: 11 Canadian guidelines for diagnosis. CMAJ : Canadian Medical Association journal = journal de 12 l'Association medicale canadienne 2005;172(5 Suppl):S1-s21 doi: 13 10.1503/cmaj.1040302[published Online First: Epub Date]|. 14 4. International Alliance for Responsible Drinking- Drinking guidelines for pregnancy and breast 15 feeding.For 2016 peer review only 16 5. (NICE) NIFHACE. Antenatal care. CG62. NICE 2008 17 6. Department of Health. Alcohol Guidelines Review – Report from the Guidelines development 18 group to the UK Chief Medical Officers. 2016 19 20 7. Moher D, Liberati A, Tetzlaff J, Altman DG. Preferred reporting items for systematic reviews and 21 meta-analyses: the PRISMA statement. Annals of internal medicine 2009;151(4):264-69 22 8. Hamlyn B OK, Wands S. Infant Feeding 2000. London: The Stationery Office, 2002. 23 9. Henderson J, Gray R, Brocklehurst P. Systematic review of effects of low–moderate prenatal 24 alcohol exposure on pregnancy outcome. Bjog 2007;114(3):243-52 25 10. Patra J, Bakker R, Irving H, Jaddoe VW, Malini S, Rehm J. Dose–response relationship between 26 alcohol consumption before and during pregnancy and the risks of low birthweight, preterm 27 birth and small for gestational age (SGA)—a systematic review and meta-analyses. Bjog 28 2011;118(12):1411-21 29 11. Dissemination CfRa. CRD’s guidance for undertaking reviews in 30 31 health care. York Publishing Services Ltd 2009:32 32 12. Higgins J, Green S. Cochrane handbook for systematic reviews of interventions version 5.1. 0 33 [updated Mar 2011]. The Cochrane collaboration. 2011, 2012. 34 13. StataCorp. Stata Statistical Software: release 12. College Station TSL, 2015. http://bmjopen.bmj.com/ 35 14. Salihu HM, Kornosky JL, Lynch O, Alio AP, August EM, Marty PJ. Impact of prenatal alcohol 36 37 consumption on placenta-associated syndromes. Alcohol 2011;45(1):73-9 doi: 38 http://dx.doi.org/10.1016/j.alcohol.2010.05.010[published Online First: Epub Date]|. 39 15. Sayal K, Draper ES, Fraser R, Barrow M, Davey Smith G, Gray R. Light drinking in pregnancy and 40 mid-childhood mental health and learning outcomes. Arch Dis Child 2013;98(2):107-11 doi: 41 http://dx.doi.org/10.1136/archdischild-2012-302436[published Online First: Epub Date]|.

42 16. Peacock JL, Bland JM, Anderson HR. Preterm delivery: effects of socioeconomic factors, on October 2, 2021 by guest. Protected copyright. 43 psychological stress, smoking, alcohol, and caffeine. Bmj 1995;311(7004):531-5 44 17. Alati R, Macleod J, Hickman M, et al. Intrauterine exposure to alcohol and tobacco use and 45 childhood IQ: findings from a parental-offspring comparison within the Avon Longitudinal 46 Study of Parents and Children. Pediatr Res 2008;64(6):659-66 doi: 47 http://dx.doi.org/10.1203/PDR.0b013e318187cc31[published Online First: Epub Date]|. 48 18. Lipsitch M, Tchetgen ET, Cohen T. Negative controls: a tool for detecting confounding and bias in 49 50 observational studies. Epidemiology (Cambridge, Mass.) 2010;21(3):383 51 19. Alati R, Davey Smith G, Lewis SJ, et al. Effect of prenatal alcohol exposure on childhood academic 52 outcomes: contrasting maternal and paternal associations in the ALSPAC study. PLoS ONE 53 2013;8(10):e74844 doi: http://dx.doi.org/10.1371/journal.pone.0074844[published Online 54 First: Epub Date]|. 55 20. Lundsberg LS, Illuzzi JL, Belanger K, Triche EW, Bracken MB. Low-to-moderate prenatal alcohol 56 consumption and the risk of selected birth outcomes: a prospective cohort study. Ann 57 58 59 60 23

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 24 of 47 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Epidemiol 2015;25(1):46-54.e3 doi: 4 http://dx.doi.org/10.1016/j.annepidem.2014.10.011[published Online First: Epub Date]|. 5 21. Nykjaer C, Alwan NA, Greenwood DC, et al. Maternal alcohol intake prior to and during 6 pregnancy and risk of adverse birth outcomes: evidence from a British cohort. J Epidemiol 7 Community Health 2014;68(6):542-9 doi: http://dx.doi.org/10.1136/jech-2013- 8 202934[published Online First: Epub Date]|. 9 22. Niclasen J, Nybo Andersen AM, Teasdale TW, Strandberg-Larsen K. Prenatal exposure to alcohol, 10 and gender differences on child mental health at age seven years. J Epidemiol Community 11 12 Health 2014;68(3):224-32 doi: http://dx.doi.org/10.1136/jech-2013-202956[published 13 Online First: Epub Date]|. 14 23. Miyake Y, Tanaka K, Okubo H, Sasaki S, Arakawa M. Alcohol consumption during pregnancy and 15 birth Foroutcomes: the peer Kyushu Okinawa review Maternal and Child Health only Study. BMC Pregnancy 16 Childbirth 2014;14:79 doi: http://dx.doi.org/10.1186/1471-2393-14-79[published Online 17 First: Epub Date]|. 18 24. Robinson M, Oddy WH, McLean NJ, et al. Low-moderate prenatal alcohol exposure and risk to 19 child behavioural development: a prospective cohort study. Bjog 2010;117(9):1139-50 doi: 20 http://dx.doi.org/10.1111/j.1471-0528.2010.02596.x[published Online First: Epub Date]|. 21 25. Jaddoe VW, Bakker R, Hofman A, et al. Moderate alcohol consumption during pregnancy and the 22 risk of low birth weight and preterm birth. The generation R study. Ann Epidemiol 23 2007;17(10):834-40 24 25 26. Sayal K, Heron J, Golding J, Emond A. Prenatal alcohol exposure and gender differences in 26 childhood mental health problems: a longitudinal population-based study. Pediatrics 27 2007;119(2):e426-34 28 27. Albertsen K, Andersen AM, Olsen J, Gronbaek M. Alcohol consumption during pregnancy and the 29 risk of preterm delivery. Am J Epidemiol 2004;159(2):155-61 30 28. Lundsberg LS, Bracken MB, Saftlas AF. Low-to-moderate gestational alcohol use and intrauterine 31 growth retardation, low birthweight, and preterm delivery. Ann Epidemiol 1997;7(7):498- 32 508 33 29. Passaro KT, Little RE, Savitz DA, Noss J. The effect of maternal drinking before conception and in 34 early pregnancy on infant birthweight. The ALSPAC Study Team. Avon Longitudinal Study of http://bmjopen.bmj.com/ 35 Pregnancy and Childhood. Epidemiology 1996;7(4):377-83 36 30. Shu XO, Hatch MC, Mills J, Clemens J, Susser M. Maternal smoking, alcohol drinking, caffeine 37 38 consumption, and fetal growth: results from a prospective study. Epidemiology 39 1995;6(2):115-20 40 31. Olsen J, Pereira AdC, Olsen SF. Does maternal tobacco smoking modify the effect of alcohol on 41 fetal growth? American journal of public health 1991;81(1):69-73 32. Brooke OG, Anderson HR, Bland JM, Peacock JL, Stewart CM. Effects on birth weight of smoking,

42 on October 2, 2021 by guest. Protected copyright. 43 alcohol, caffeine, socioeconomic factors, and psychosocial stress. Bmj 1989;298(6676):795- 44 801 45 33. Bille C, Olsen J, Vach W, et al. Oral clefts and life style factors--a case-cohort study based on 46 prospective Danish data. Eur J Epidemiol 2007;22(3):173-81 47 34. Ernhart CB, Sokol RJ, Ager JW, Morrow-Tlucak M, Martier S. Alcohol-related birth defects: 48 assessing the risk. Ann N Y Acad Sci 1989;562:159-72 49 35. Ogston SA, Parry GJ. EUROMAC. A European concerted action: maternal alcohol consumption 50 51 and its relation to the outcome of pregnancy and child development at 18 months. Results-- 52 strategy of analysis and analysis of pregnancy outcome. Int J Epidemiol 1992;21 Suppl 1:S45- 53 71 54 36. Andersen A-MN, Andersen PK, Olsen J, Grønbæk M, Strandberg-Larsen K. Moderate alcohol 55 intake during pregnancy and risk of fetal death. Int J Epidemiol 2012;41(2):405-13 56 37. Windham GC, Von Behren J, Fenster L, Schaefer C, Swan SH. Moderate maternal alcohol 57 consumption and risk of spontaneous abortion. Epidemiology 1997;8(5):509-14 58 59 60 24

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 25 of 47 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 38. McCarthy FP, O'Keeffe LM, Khashan AS, et al. Association between maternal alcohol 4 consumption in early pregnancy and pregnancy outcomes. Obstetrics & Gynecology 5 2013;122(4):830-37 6 39. Faebo Larsen R, Hvas Mortensen L, Martinussen T, Nybo Andersen AM. Determinants of 7 developmental coordination disorder in 7-year-old children: a study of children in the Danish 8 National Birth Cohort. Dev Med Child Neurol 2013;55(11):1016-22 doi: 9 http://dx.doi.org/10.1111/dmcn.12223[published Online First: Epub Date]|. 10 40. Parry GJ, Ogston SA. EUROMAC. A European concerted action: maternal alcohol consumption 11 12 and its relation to the outcome of pregnancy and child development at 18 months. Results-- 13 child development at age 18 months. Int J Epidemiol 1992;21 Suppl 1:S72-8 14 41. NHS Choices.Can I drink alcohol if I’m pregnant? 15 www.nhs.uk/chq/Pages/2270.aspx?CategoryID=54#closeFor peer review. 2014only 16 42. Kelly Y, Sacker A, Gray R, Kelly J, Wolke D, Quigley MA. Light drinking in pregnancy, a risk for 17 behavioural problems and cognitive deficits at 3 years of age? Int J Epidemiol 18 2009;38(1):129-40 doi: http://dx.doi.org/10.1093/ije/dyn230[published Online First: Epub 19 Date]|. 20 43. Bradley RH, Corwyn RF. Socioeconomic status and child development. Annual review of 21 psychology 2002;53(1):371-99 22 44. Phung H, Bauman A, Nguyen T, Young L, Tran M, Hillman K. Risk factors for low birth weight in a 23 socio-economically disadvantaged population: parity, marital status, ethnicity and cigarette 24 25 smoking. Eur J Epidemiol 2003;18(3):235-43 26 45. Morrison A, Polisena J, Husereau D, et al. The effect of English-language restriction on systematic 27 review-based meta-analyses: a systematic review of empirical studies. International journal 28 of technology assessment in health care 2012;28(02):138-44 29 46. Moher D, Pham B, Lawson M, Klassen T. The inclusion of reports of randomised trials published 30 in languages other than English in systematic reviews. Health Technol Assess 2003;7(41):1- 31 90 32 47. Klassen TP, Lawson ML, Moher D. Language of publication restrictions in systematic reviews gave 33 different results depending on whether the intervention was conventional or 34 complementary. Journal of clinical epidemiology 2005;58(8):769-76. e2 http://bmjopen.bmj.com/ 35 48. Jüni P, Holenstein F, Sterne J, Bartlett C, Egger M. Direction and impact of language bias in meta- 36 analyses of controlled trials: empirical study. Int J Epidemiol 2002;31(1):115-23 37 38 49. Richmond RC, Al-Amin A, Smith GD, Relton CL. Approaches for drawing causal inferences from 39 epidemiological birth cohorts: a review. Early human development 2014;90(11):769-80 doi: 40 10.1016/j.earlhumdev.2014.08.023[published Online First: Epub Date]|. 41 50. Burgess S, Davies NM, Thompson SG. Instrumental variable analysis with a nonlinear exposure- outcome relationship. Epidemiology 2014;25(6):877-85 doi:

42 on October 2, 2021 by guest. Protected copyright. 43 10.1097/ede.0000000000000161[published Online First: Epub Date]|. 44 51. Ades A, Sutton A. Multiparameter evidence synthesis in epidemiology and medical decision- 45 making: current approaches. Journal of the Royal Statistical Society: Series A (Statistics in 46 Society) 2006;169(1):5-35 47 52. O'Keeffe LM, Kearney PM, McCarthy FP, et al. Prevalence and predictors of alcohol use during 48 pregnancy: findings from international multicentre cohort studies. BMJ open 49 2015;5(7):e006323 50 51 53. Mather M, Wiles K, O’Brien P. Should women abstain from alcohol throughout pregnancy? 2015 52

53 54 55 56 57 58 59 60 25

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from Page 26 of 47 26 26

BMJ Open http://bmjopen.bmj.com/ on October 2, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml For peer review only

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

27 27 Adjusted’ refers to adjustedrefersAdjusted’ both for ‘ BMJ Open http://bmjopen.bmj.com/

on October 2, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml For peer review only Pooled mean alcohol consumption consumption withalcohol difference low comparing tofor 32g/week) no (up studies).(7 birthweight Pooled smoking and a intervals.smoking measuresocio-economic status.CI: Confidence of Figure 2.

Page 27 of 47 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from Page 28 of 47 28 28

b) b) BMJ Open http://bmjopen.bmj.com/ on October 2, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml For peer review only

a) a) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

29 29 OR: Odds ratio, CI: Confidence intervals. Pooled OR includes both both OR includes Pooled intervals. CI: Confidence ratio, Odds OR:

Odd Ratios for small for gestational age comparing low alcohol consumption (up to 32g/week) to 32g/week) (up consumption alcohol low comparing age gestational for small for Ratios Odd 3b) 3b) studies); studies);

BMJ Open http://bmjopen.bmj.com/ ) with no alcohol consumption (8 consumption alcohol no with ) on October 2, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml For peer review only Adjusted’ refers to adjusted for both smoking and a measure of socio-economic status.a measure socio-economic and of smoking forto adjusted both refersAdjusted’ ‘ Odd Ratios for low birthweight comparing low alcohol consumption (up to 32g/week) with no alcohol consumption (6 studies). (6 consumption alcohol no 32g/week)with to (up consumption alcohol low comparing birthweight low for Ratios Odd 3c) 3c)

Odd Ratios for preterm birth comparing low alcohol consumption (up to 32g/week to (up consumption alcohol low comparing birth preterm for Ratios Odd c) c) Figure 3a) Figure with no alcohol consumption (7 studies) studies) (7 consumption alcohol no with studies, from estimates unadjusted and adjusted Page 29 of 47 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from Page 30 of 47 30 30 Odds ratios for internalising behaviour behaviour ratios internalising Odds for 1b) b)

BMJ Open http://bmjopen.bmj.com/ on October 2, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

For peer review only Odds ratios for externalising behaviour comparing low alcohol consumption (up to 32g/week) with no alcohol consumption (3 studies);alcohol (3 with consumption no (up to 32g/week) consumption alcohol comparing low behaviour ratios externalising Odds for

material Supplementary Supplementary figure 1a) 1a) Supplementary figure comparing low alcohol consumption (up to 32g/week) with no alcohol consumption (2 studies); OR: Odds ratio, CI: Confidence intervals. intervals. ratio, studies); CI: (2 OR: Odds alcohol Confidence consumption towith (up no 32g/week) consumption low alcohol comparing a) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 31 31

analysis for birthweight in grams. in analysisbirthweight for - Funnel plot of studies included in the meta the in studiesincluded of plot Funnel Funnel plot of studies included in the meta-analysis for birthweight <2500g. birthweight meta-analysis for the in studiesincluded of plot Funnel b) d) Supplementary figure 2d. figure Supplementary Supplementary figure 2b. figure Supplementary

BMJ Open

http://bmjopen.bmj.com/ on October 2, 2021 by guest. Protected copyright. alysis for small for gestational age. gestational for small alysis for an - For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml MEDLINE, PsycINFO, EMBASE on Ovid; the Cochrane Library including CENTRAL (the Cochrane Central Database of Controlled Controlled of Central Database Cochrane (the CENTRAL Library including the Cochrane Ovid; PsycINFO,on EMBASE MEDLINE, to 11·07·2016. 1950 For peer review only Search Search Strategy Funnel plot of studies included in the meta-analysis for preterm birth. birth. preterm for meta-analysis the in studiesincluded of plot Funnel Funnel plot of studies included in the meta the in studiesincluded of plot Funnel Supplementary Table 1. 1. Supplementary Table

Supplementary figure 2c. 2c. figure Supplementary a) a) 2a. figure Supplementary c) Page 31 of 47 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from Page 32 of 47 32 32

BMJ Open http://bmjopen.bmj.com/ on October 2, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml For peer review only 23 or/10-19,22 (213050) (213050) or/10-19,22 23 22 20 and 21 (3108) (3108) 21 20 and 22 addict$ or dependence or response$ or susceptibility)).ti,ab. (110843) (110843) or susceptibility)).ti,ab. or response$ or dependence addict$ 21 *Alcohols/ or exp *Alcohol-Related Disorders/ or (alcohol adj3 (misuse or "use" or abuse or or adj3 (misuse or abuse or "use" *Alcohol-Related Disorders/ (alcohol or exp or *Alcohols/ 21 (instrument$ or variant$ or variable$ or variability or variabilities or variance$))).ti,ab. (494904) (494904) oror variability variance$))).ti,ab. or variabilities or variable$ (instrument$ or variant$ polymorphism, genetic/ or (polymorphism$ or ((gene or genes or genetic or genotyp$) adj3 or genotyp$) adj3 or genetic or genes ((gene (polymorphism$ or or polymorphism, genetic/ 20 Genome-Wide Association Study/ or Linkage Disequilibrium/ or genotype/ or phenotype/ or or phenotype/ Disequilibrium/ or genotype/ or Linkage Study/ Association Genome-Wide 20 19 (ADH1B$ or teetotal$ or temperance or nondrink$ or non-drink$).ti,ab. (3093) or non-drink$).ti,ab. (3093) ortemperance nondrink$ teetotal$ (ADH1B$ or 19 or (10558) 18 ((low or light or moderate or abstin$ or abstain$ or pattern$ or behavio?r$) adj3 drink$).ti,ab. drink$).ti,ab. adj3 or behavio?r$) or pattern$ or abstain$ moderateabstin$ ((loworor or light 18 or low or light or moderat$ or abstin$ or abstain$)).ti,ab. (49470) (49470) moderat$ or abstain$)).ti,ab. or or or light abstin$ or low 17 ((alcohol or alcoholic) adj3 (drink$ or exposure or consumption or consume$ or consuming or orconsuming consume$ (drink$ adj3 oror or alcoholic) exposure consumption ((alcohol 17 16 (alcohol dehydrogenase or acetaldehyde dehydogenase).ti,ab. (8557) dehydogenase).ti,ab. or acetaldehyde dehydrogenase (alcohol 16 15 alcohol$.ti. (106954) alcohol$.ti. 15 14 Temperance/ (2430) (2430) Temperance/ 14 13 exp Drinking Behavior/ (57821) Behavior/ (57821) exp Drinking 13 12 Aldehyde Oxidoreductases/ (3672) Aldehyde 12 11 Alcohol dehydrogenase/ (5809) dehydrogenase/ Alcohol 11 10 Ethanol/ (73449) Ethanol/ (73449) 10 9 or/1-8 (1079283) 9(1079283) or/1-8 8 Maternal exposure/ (5534) (5534) exposure/ 8 Maternal or in utero).ti,ab. (588557) (588557) utero).ti,ab. or in f?etal or f?etus f?etal or f?etus 7 (prepregnan$ or conception or preconception or pregnan$ or prenatal$ or pre-natal$ or or or prenatal$ or or pregnan$ pre-natal$ or preconception or conception 7 (prepregnan$ 6 exp Pregnancy Complications/ (343999) Complications/ Pregnancy 6 exp 5 Fetus/ (68424) 5 (68424) Fetus/ 4 exp "embryonic and fetal development"/ (212127) (212127) fetal development"/ "embryonic and 4 exp 3 preconception care/ or prenatal care/ (21644) care/ (21644) or prenatal care/ 3 preconception 2 Pregnant Women/ (5199) (5199) 2 Women/ Pregnant 1 exp pregnancy/ (719661) pregnancy/ 1 exp Trials) on Wiley Interscience; and Science Citation Index, Social Science Citation Index, WebScience. ofon Index, Citation Science Social Science Citation Index, Wiley and Interscience; Trials) on

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 33 33

BMJ Open http://bmjopen.bmj.com/ adj2 (review$ or overview$)).ti,ab. (review$ or overview$)).ti,ab. adj2 on October 2, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml For peer review only 51 (search$ adj4 literature).ab. (30686) literature).ab. (search$51 adj4 (28877) 50 (search strategy or search criteria or systematic search or study selection or data extraction).ab. or data extraction).ab. selection study or search systematic or criteria or search (search strategy 50 (27094) 49 (reference list$ or bibliograph$ or hand search$ or manual search$ or relevant journals).ab. relevantjournals).ab. search$ or search$ manual or hand or list$ (reference 49 or bibliograph$ 48 "review of reviews".ti,ab. (211) reviews".ti,ab. (211) "review of 48 47 ((systematic$ or evidence$ or realist or narrative or literature) or literature) or realist or narrative ((systematic$ or47 evidence$ (171877) 46 (meta analy$ or metaanaly$ or metanaly$ or meta regression).ti,ab. (72390) (72390) regression).ti,ab. meta metaanaly$ or metanaly$ or or (meta46 analy$ 45 meta-analysis as topic/ (13933) (13933) topic/ as meta-analysis 45 44 meta-analysis/ (52850) (52850) meta-analysis/ 44 43 24 not 42 (6498) (6498) 24 not 43 42 42 or/35-41 (7849766) or/35-41 (7849766) 42 41 (rat or rats or mouse or mice).ti. (1123907) (1123907) or mice).ti. (ratmouse 41 rats or or 40 exp rodentia/ (2688128) exp rodentia/ 40 39 exp Models, Animal/ not humans/ (310115) (310115) not humans/ Animal/ exp Models, 39 38 exp Animal Experimentation/ (6477) exp Animal38 Experimentation/ 37 exp Animals, Laboratory/ (730783) (730783) exp Animals, Laboratory/ 37 36 animals/ not humans/ (3889478) animals/ (3889478) 36 not humans/ 35 33 not 34 (3386180) (3386180) 33 not 35 34 34 randomized controlled trial/ or Randomized Controlled Trials as Topic/ or random$.ti,ab. (888907) or Trialsas Topic/ Controlled trial/ or Randomized randomized controlled 34 33 or/25-32 (3417468) or/25-32 (3417468) 33 32 (letter or comment$).ti. (100442) (100442) (letter comment$).ti. 32 or 31 case report/ (1708277) report/ (1708277) case31 30 comment/ (610211) (610211) comment/30 29 Anecdotes as topic/ (4586) (4586) topic/ Anecdotes as29 28 exp historical article/ (325965) (325965) exp historical28 article/ 27 news/ (166329) news/ (166329) 27 26 editorial/ (368458) (368458) editorial/ 26 25 letter/ (861500) letter/ (861500) 25 24 9 and 23 (11403) 23 (11403) 9 and 24

Page 33 of 47 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from Page 34 of 47 34 34 BMJ Open http://bmjopen.bmj.com/ on October 2, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml For peer review only (489662) or amphetamine$ or cocaine$ or heroin or cannabis or marijuana) not (alcohol or drink$)).ti. or alcoholic not (alcohol or marijuana) or cannabis or heroin or cocaine$ or amphetamine$ 78 ((smok$ or drug$ or tobacco or nicotine or cigarette$ or substance$ or methamphetamine$ methamphetamine$ or substance$ or nicotine or cigarette$ or or tobacco ((smok$78 or drug$ 77 limit 76 to english language (3096) language limit 7677 to english 76 43 and 75 (3349) (3349) 75 43 and 76 75 55 or 70 or 74 (3362878) 55or 74 or75 70 (3362878) 74 or/71-73 (291621) or/71-73 (291621) 74 73 guideline$.mp. (291621) (291621) guideline$.mp. 73 72 guidelines as topic/ or practice guidelines as topic/ (114690) guidelines as (114690) topic/ or practice topic/ as guidelines 72 71 exp guideline/ (25834) exp guideline/ 71 70 or/57-69 (2937160) or/57-69 (2937160) 70 69 ((family or sibling or population) adj3 (study or studies)).ti,ab. (115245) (115245) adj3 or studies)).ti,ab. (study sibling or population) or((family 69 68 (mendel$ or natural experiment$).ti,ab. (11082) (11082) or natural (mendel$ 68 experiment$).ti,ab. 67 cross sectional.ti,ab. (183479) crosssectional.ti,ab. 67 analys$ or cohort$)).ti,ab. (870699) (870699) or cohort$)).ti,ab. analys$ 66 ((longitudinal$ or retrospectiv$ or prospectiv$) and (study or studies or studied or review$ or or review$ or studied studies or (study and or prospectiv$) or retrospectiv$ ((longitudinal$ 66 65 ((follow up or observational) adj (study or studies)).ti,ab. (87771) or (87771) studies)).ti,ab. oradj up (study ((follow observational) 65 64 (cohort adj (study or studies or analys$)).ti,ab. (98537) or analys$)).ti,ab. (98537) or studies (cohort (study 64 adj 63 (case control or negative control).ti,ab. (93097) (93097) control).ti,ab. (case or negative 63 control 62 Mendelian Randomization Analysis/ (229) Analysis/ Randomization (229) Mendelian 62 61 cross-sectional studies/ (185407) studies/ (185407) cross-sectional 61 60 exp cohort studies/ (1394149) (1394149) exp cohort 60 studies/ 59 exp case control studies/ (691899) (691899) studies/ exp case59 control 58 ep.fs. (1223653) (1223653) ep.fs. 58 57 epidemiologic studies/ (6077) epidemiologic 57 56 43 and 55 (201) 55 43 and 56 55 or/44-54 (293491) or/44-54 (293491) 55 54 ((multiple treatment$ or indirect or mixed) adj2 comparison).ti,ab. (991) (991) comparison).ti,ab. ormixed) adj2 ortreatment$ indirect ((multiple 54 53 cochrane.jw. (11053) cochrane.jw. (11053) 53 Citation index or bids or cancerlit).ab. (95294) (95294) or cancerlit).ab. or index bids Citation 52 (medline or pubmed or cochrane or embase or psychlit or psyclit or psychinfo or cinahl or science or orscience cinahl or psychinfo or embase or orpsyclit cochrane or psychlit or pubmed (medline 52

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 35 35 BMJ Open http://bmjopen.bmj.com/ on October 2, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml For peer review only

79 77 not 78 (2822) (2822) 77 not 79 78

Page 35 of 47 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from Page 36 of 47 36 36

loyment; body loyment; body pregnancy. Adjusted pregnancy. ion and baby’s sex and ion se during se during

ing (ever) in first trimester, amount of (ever) first in of trimester, amount ing

Confounders Confounders infant sex, parity, bleeding during pregnancy, high blood pressure, high and pregnancy, during sex, bleeding infant parity,

BMJ Open http://bmjopen.bmj.com/

on October 2, 2021 by guest. Protected copyright. up squared), and maternal cigarette smoking. maternal squared),up and cigarette - tified according to period. The model is stratified according to maternal age and parity. and age to isaccording maternal model stratified The to according period. tified For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml pregnancy weight, height, age, parity, ethnicity, salivary cotinine levels, caffeine intake, education, energy intake, energy gestat education, intake, levels, caffeine salivary parity, cotinine age, height, ethnicity, weight, pregnancy ght, preterm birth: maternal age; region of residence; number of children; family structure; maternal education; maternal emp maternal education; structure; family children; residence; number of of region age; maternal birth: ght, preterm - and social class and For peer review only parity, and total alcohol consumption during during pregnancy. consumption total and parity, alcohol well-being in pregnancy. in pregnancy. well-being marijuana use, cocaine use, study cohort, preterm labour, respiratory problem, infant gender, bleeding, nausea/vomiting, hypertension, incompetent cervix, nausea/vomiting, hypertension, bleeding, gender, infant respiratory problem, labour, cohort,study preterm use, use, cocaine marijuana diabetes. gestational asthma, maternal disease,transmitted induction/augmentation, problems, sexually placental maternal age group; high EPDS score; child ethnicity; gestational age group; and birth weight and birth group; weight age ethnicity; gestational child EPDSscore; high group; age maternal models also controlled for gestational age. for gestational controlled models also homeownership, whether currently married, high scores (>12) on the Edinburgh Postnatal Depression Scale, and child gestational age, birth weight and and age, birth child weight gestational Scale, and Postnatal Depression thescores Edinburgh on high (>12) married, currently whether homeownership, gender. Gestational age, sex, maternal height, and parity sex,maternal and age, height, Gestational Unadjusted asvariables”. smoking “dummy entered the model and alcohol parity, and education Age, school Small for gestational age: smoking in month 7, ethnicity, weight, height, month ethnicity, smoking in 7, weight, age: height, forSmall gestational trimester, 7,third in month exercise use in coffee sex, parity, weight,infant 7, ethnicity, in month height, smoking Low birthweight: preeclampsia/eclampsia. in smoking delivery: Preterm placentaland problems. pressure, anomalies, blood high pre-eclampsia/eclampsia, pregnancy, during bleeding employment, Parental age Parental age child’s age at age follow child’s smoking wasstrasmoking Maternal age, smoking, years of schooling, ethnicity, body mass index, infant sex, maternal status, family income, and drug u drug income, and sex,maternal family status, infant ethnicity, mass index, schooling, body years smoking, of Maternal age, at for age delivery. gestational adjusted Birthweight for clustering. Sex, gestational age, intrauterine growth restriction, maternal age, mother’s occupational status, maternal smok maternal status, mother’s occupational maternal age, restriction, growth age, intrauterine Sex, gestational firsttrimester. alcohol consumption in and smoking maternal index; maternal smoking during pregnancy; gestational age; and baby’s gender. baby’sand age; gender. pregnancy; gestational during maternal smoking index; Small for gestational age: maternal age; region of residence; number of children; family structure; maternal education; maternal employment; body mass maternal body employment; maternal education; structure; family of children; residence; number region of maternal age; age: forSmall gestational Low birthLow wei and gender. baby’s pregnancy; during maternal smoking mass index; Maternal pre month 7, height, parity, age, caffeine use in month 7, exercise first 16 weeks, bleeding during pregnancy, high blood pressure, pre-eclampsia/eclampsia, pressure,blood pregnancy, pre-eclampsia/eclampsia, high during bleeding 7, weeks, first month use exercise 16 in caffeine monthage, parity, height, 7, problems. placental and anomalies, Gestational age, parity, smoking income. parity,and age, smoking Gestational Lists of confounders adjusted for by each study each study for by confounders adjusted Lists of

Brooke (1989) Brooke Peacock (1995) Peacock (1991) Olsen Lundsberg (1997) Lundsberg Bille (2007) Bille McCarthy (2013) McCarthy Faebo Larsen (2013)Larsen Faebo Miyake (2014) Miyake Nykjaer (2014) Nykjaer

[32] [40] Parry (1992) [40] Parry (1992) smoking. parity and age, at age sex,mother’s birth, Gestational [16] [31] (1992) [35] Ogston smoking. parity and age, at age sex,mother’s birth, Gestational [30] Shu (1995) (1995) [30] Shu [29] Passaro (1996) (1996) [29] Passaro mass maternal body index. and smoking, infant sex, parity, maternal age, Gestational [37] Windham (1997) [37] Windham interview. consumptionweek before in caffeine cigarette at and and interview, age gestational spontaneous abortion, prior Maternal age, [28] [27] Albertsen (2004) (2004) [27] Albertsen status the household, in occupational pregnancy, during consumption coffee pregnancy, during smoking preterm previous delivery, age, 1 diabetes, Type [33] (2007) [26] Sayal marital status; ownership; parity; home education; of maternal illicit the firstuse in of trimester; highest level drugs and use cannabis smoking, Gender, [25] Jaddoe (2007) [25] Jaddoe low birth weight and infant birth gender; weight and age level,and ethnicity, height, educational parity smoking, mass maternal for body index, Controlled [24] Robinson (2010) (2010) [24] Robinson (and up at child’s age follow pregnancy, in stress father income, the family home, in the biological family presence of maternal education, Maternal age, [14] Salihu (2011) (2011) [14] Salihu ofbirth the and yearinfant, of gender height, prenatalmaternal care, of adequacy education, marital status, smoking, race,parity, Maternal age, [36] Andersen (2012) (2012) [36] Andersen and smoking.consumption Effect of and coffee pregnancy prior consumption to alcohol since in consumption, changes coffee abortions, Number previous of [38] [15] Sayal (2013) (2013) [15] Sayal the first trimester, drugs cannabis otherduring and/or illicit use of smoking, frequency of daily maternal education, of level parity, Maternal age, highest [39] [23] [21] (2014) [22] Niclasen maternal psychological and parental psychiatric pre-pregnancy diagnoses, education, parental confounders: smoking, the following parental Adjusted for [20] Lundsberg (2015) [20] Lundsberg smoke exposure, multivitamin and use, exercise, work,passive smoking, prenatal caffeine, marital BMI,status, ethnicity, education, maternal age, Parity, Study (year) (year) Study Supplementary Table 2. Supplementary Table

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 37 37 BMJ Open http://bmjopen.bmj.com/ on October 2, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

For peer review only

[34] Ernhart (1989) (1989) [34] Ernhart study. yearof race and smoking, Parity,

Page 37 of 47 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 7 7 3339 3339 7 7 6 &7 6 6 6 5 5 5 5 4 4 1 1 # # on page on Reported Reported Page 38 of 47 38 38 up) and reportand up) characteristics (e.g., years considered, - BMJ Open http://bmjopen.bmj.com/ analysis). analysis). - on October 2, 2021 by guest. Protected copyright. language, publicationlanguage, usedstatus) as for criteria eligibility, giving rationale. State the process thefor State selecting studies (i.e., screening, eligibility, includedinreview, systematic and, if applicable, the included inmeta Present electronicfull Present search strategy for least one at database, limits including any used, such thatitcould be Describe all informationDescribe (e.g.,databasessources with dates ofcoverage, contact withstudy authors to identify Specify characteristicsstudySpecify (e.g.,PICOS, length of follow Indicate reviewa Indicate if protocol if whereandexists, it can accessedWebbe (e.g., and, available,address), if provide Provide anexplicit Provide statementof questions being addressed withreference to participants, interventions, comparisons, studyand outcomes, (PICOS).design Describe therationale theDescribe review for in the context of what known.already is Provide a a Provide structured including,summary applicable:as background; data objectives; sources; eligibilitystudy criteria, interventions;andparticipants, study appraisaland methods;synthesis results; limitations; andconclusions Identify the reportasIdentify systematic a review,metaanalysis, or both. Checklist Checklist item repeated. repeated. additional additional in studies) the search date and last searched. registration informationregistration including registration number. implications findings; of implications key systematicreview number.registration For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

1 # 8 2 4 9 7 3 For peer review only6

RISMA 2009 Checklist Checklist 2009 RISMA P Study Study selection Search Search Information Information sources Eligibility Eligibility criteria Protocol andProtocol registration 5 METHODS METHODS Objectives Rationale Rationale INTRODUCTION INTRODUCTION Structured summary ABSTRACT ABSTRACT Title TITLE TITLE Section/topic Section/topic

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

12 12 29 & 30 29 32 32 29 & 30 29 29 & 30 29 11 11 2127 2127 28 28 9 9 9 9 8 &9 8 9 9 6 &7 6 7 &8 7 8 8 39 39 BMJ Open http://bmjopen.bmj.com/ analysis. - specified. specified. - for each metafor on October 2, 2021 by guest. Protected copyright. ) 2 Summarize Summarize the findingsmain theincluding strength of evidence mainfor outcome;each theirconsider relevance to Give results ofadditionalGive analyses, if done (e.g., sensitivity or subgroup analyses, metaregression Item [see 16]). Present results ofPresent any risk assessment biasof acrossof Itemstudies (see 15). Present results ofPresent each metaanalysis done, including confidence intervalsmeasuresand of consistency. For all outcomes For all considered(benefits harms), or for present, each (a)study: simple summary each data for group(b) intervention andeffect estimates confidence intervals,with ideally foresta plot. Present data Present on risk eachbiasand, studyof of if available,outcome any level assessment (see 12).item For each study,For each present characteristics for which data extractedwere (e.g., study followupsize, PICOS, period) and citations.the provide Give numbers ofstudiesGive screened, assessedfor eligibility, and the included inreview, for with reasons exclusions at stage,with floweach aideally diagram. Describe methods Describe of additional (e.g.,analyses oranalyses,sensitivity subgroup metaregression), if done, indicating were pre which Specify assessment any Specify of thatbias risk the may of affect cumulative (e.g.,publicationevidence bias,selective within reporting studies). Describe themethods Describe of handling data and combining results if of studies,done, measuresincluding of consistency I (e.g., State the principal theState measuressummary (e.g., risk difference ratio, in means). Describe methods Describe used for assessing risk of bias individual of (includingstudies specificationwhether of this was at the done study outcomeor level), thishowand information to be is in usedany synthesis. data List and define variablesList all datawere for which sought (e.g., PICOS, funding sources) any and assumptions and made.simplifications Describe methodof Describe data extraction reports (e.g.,from forms,piloted independently,in duplicate) and any processes obtaining for and datafromconfirming investigators. key groups (e.g., key healthcareproviders, users, policyand makers). For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

11 12 21 14 17 13 16 For peer review only23

Data Data collection process 10 Summary of of Summary evidence 24 DISCUSSION DISCUSSION Risk of bias Risk acrossof studies 22 Synthesis resultsSynthesis of Results ofindividualResults studies 20 Risk of bias Risk withinof studies 19 Study Study characteristics 18 Study Study selection RESULTS RESULTS Risk of bias Risk acrossof studies 15 Synthesis resultsSynthesis of Summary measuresSummary Risk of bias Risk inof individual studies Data items Data items

Additional Additional analysis Additional Additional analyses Page 39 of 47 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

16 16 14 14 12 & 13 12 Page 40 of 47 40 40

BMJ Open http://bmjopen.bmj.com/ on October 2, 2021 by guest. Protected copyright. Describe sourcesDescribe of funding for the systematic review (e.g., other supportand supply of data); ofrole fundersforthe Provide a a Provide general interpretationof thethe results in of context other evidence,implicationsand forresearch. future Discuss limitationsDiscuss at study outcomeand (e.g.,risklevel and of bias), at reviewlevel (e.g.,incomplete retrieval of research, reportingidentified bias). systematic systematic review. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

25 27 26 For peer review only

Limitations Funding Funding FUNDING FUNDING Conclusions

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 41 41 7 7 7 6 6 1 7 7 6 NA NA NA 6&7 table 1table table 1table table 1table Figure Figure 1 Page Page No Reported on Reported supplementary supplementary supplementary

BMJ Open http://bmjopen.bmj.com/ Recommendation on October 2, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml For peer review only 9 9 to include Effort available studies,all including contact with authors 8 8 Search including strategy, time period included in the synthesis keyand words 7 7 Qualifications of searchers librarians and(eg, investigators) 6 6 Study population 5 5 of Type designsstudy used 4 4 of exposure Type or intervention used 3 3 Description of study outcome(s) 2 2 Hypothesis statement 1 1 Problem definition 17 Description of relevance appropriatenessor of studies for assembled assessing the 16 Description of any contact with authors 15 ofhandling Method abstracts and unpublished studies 14 ofaddressing Method articles published in languages than other English 13 ofcitations List located those and excluded, including justification 12 hand Use (eg,of searching reference of obtainedlists articles) 11 Search used, software version, andname including special features (eg, used explosion) 10 Databases registriesand searched Item NoItem Reporting of methods includeshould Reporting of search Reporting of search strategy should include Reporting of background includeshould MOOSE Checklist for Meta-analyses of Observational Studies Observational of forMeta-analyses MOOSEChecklist

Page 41 of 47 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from Page 42 of 47 42 42

8 7 19 NA 9&10 2736 Table 13 Table table 2 82 table & Figures 2&3Figures Figures 2&3Figures Figures 2&3Figures supplementary BMJ Open http://bmjopen.bmj.com/ on October 2, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml For peer review only hypothesis to hypothesistested to be regression on possibleregression predictorson results of study Description of statistical(eg, methodscomplete descriptionof randomfixed or effects justification models, whether of the chosen models account for predictorsof study doseresponseresults, models, cumulativeor metaanalysis) in sufficientto detail be replicated Documentation of how dataand classifiedwere coded raters,multiple (eg, blinding and reliability)interrater Rationale for Rationale the selection ofcoding dataand (eg, sound principlesclinical or convenience) appropriate) Assessment ofAssessmentquality, studyincluding blinding of assessors, quality stratification or Assessment ofAssessment confounding (eg, comparability of and cases controls where studies in 28 of Indication statistical ofuncertainty findings 27 ofsensitivity Results testing (eg, subgroup analysis) 26 giving Table descriptive information for each study included 25 Graphic summarizing individual study estimates andoverall estimate 24 ofappropriate Provision graphicsandtables 23 22 of Assessment heterogeneity 21 20 19 18 Reporting of resultsReporting of should include

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 6 Figure 1 Figure supTable supTable 1 6 5 & 5 6 5 5&6 4 3 1 on page # # on page Reported Reported up) and report up) and (e.g.,characteristics years considered, - BMJ Open http://bmjopen.bmj.com/ analysis). analysis). - on October 2, 2021 by guest. Protected copyright. Describe methodDescribe of data extraction reportsfrom (e.g., forms, piloted independently,induplicate) and any processes for andobtaining confirming data investigators.from State the process theState for selecting studies (i.e., screening, eligibility, review, included in systematic and, ifapplicable, included inmeta the Present full electronic full Present strategy for search leastone at includingdatabase, any limits used,suchthat it could be Describe allinformationDescribe (e.g., databasessources with of dates coverage, contact study authors with to identify Specify studySpecify characteristics(e.g., PICOS,length of follow Indicate reviewa Indicate if protocol if andexists, where accessedWebit can be (e.g., and, available,address), if provide Provide an Provide explicit statement of questions being withaddressed reference to participants, interventions, comparisons, outcomes, anddesign study (PICOS). Describe the rationale theDescribe review for thein the of context known.already what is Provide a Provide structured including,summary as applicable:background; dataobjectives; sources; study eligibilitycriteria, andparticipants, interventions; appraisalstudy and synthesis methods; results;limitations; conclusions and Identify the report Identify a as systematicmeta-analysis, review, both. or Checklist Checklist item language, publicationlanguage, status) used forcriteriaas eligibility, rationale. giving repeated. repeated. additional additional in studies)the search dateand last searched. registration registration information registrationincluding number. implications of of implicationskey findings; systematic review number. registration For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

1 # 8 2 4 9 7 3 For peer review6 only

Data Data collection process 10 Study Study selection Search Search Information Information sources Eligibility Eligibility criteria Protocol andProtocol registration 5 METHODS METHODS Objectives Rationale Rationale INTRODUCTION INTRODUCTION Structured summary ABSTRACT ABSTRACT Title TITLE TITLE Section/topic Section/topic Checklist 2009 PRISMA Page 43 of 47 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

Page 44 of 47 Figure Figure 2&3 13-14 Figure Figure 2&3 supfigure1 Figure Figure 2&3 supfigure1 7 Table 1-3 Table 1-3 Figure 1 Figure 8 7 Figure Figure 2&3 supfigure1 Figure Figure 2&3 supfigure1 6 & 6 7 7 & 7 8 BMJ Open http://bmjopen.bmj.com/ analysis. analysis. - specified. specified. - for metafor each on October 2, 2021 by guest. Protected copyright. ) 2 reporting within reporting studies). Give results results ofadditionalGive analyses,ifdone (e.g., sensitivity or subgroup analyses,meta-regression Item [see 16]). Present results ofPresent any assessment of risk acrossof bias studies (see Item 15). Present results ofPresent meta-analysis each done, including confidence intervalsand measuresof consistency. For For all outcomes considered (benefits harms), orfor present, each (a) study: simple summary data for each Present data Present onrisk eachbias studyof of and, if available,any outcome level assessment (see 12). item For each For each presentstudy, characteristicswhich for data were (e.g., studyextracted PICOS, follow-upsize, period) and the provide citations. Give numbersofGive studies assessedscreened, foreligibility, included inand review, the with reasons for exclusions at Describe methodsDescribe of additional analyses (e.g., sensitivity subgroup or analyses, meta-regression), done, if indicating prewerewhich Specify any Specify assessment bias of risk maythat of affect the cumulative (e.g.,evidence publication bias,selective Describe the methods theDescribe of handling data and combining results of studies, if done, including measures of consistency I (e.g., State the principal theState measuressummary (e.g.,risk ratio, in difference means). Describe methodsDescribe used for riskassessing bias individual of of studies (including specification of whetherthis was at the donestudy outcomeor level), howand information this to be is usedany in data synthesis. List and define and List all variables for which data were sought (e.g., PICOS, funding sources)and any assumptions and made.simplifications intervention group (b) intervention group andeffectestimates confidence intervals,with a ideally plot.forest each with each stage, flowa ideally diagram. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

12 11 21 14 17 16 For peer review only 23

Risk of bias Risk acrossof studies 22 Synthesis resultsSynthesis of Results ofindividualResults studies 20 Risk of bias Risk withinof studies 19 Study Study characteristics 18 Study Study selection RESULTS RESULTS Risk of bias Risk acrossof studies 15 Synthesis resultsSynthesis of Summary measuresSummary 13 Risk of bias Risk inof individual studies Data items Data items Additional Additional analysis Additional Additional analyses 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

17 13 13 & 13 14 13&14 13&14 supfigure1 BMJ Open http://bmjopen.bmj.com/ on October 2, 2021 by guest. Protected copyright. Describe Describe offundingsources forthe systematic review (e.g., other supportand supply ofdata); ofrole the funders for Provide a Provide general interpretation of thethe results in context of otherevidence,and forimplications future research. Discuss limitationsDiscuss at and study (e.g.,outcome level andrisk of bias), review-level (e.g.,incompleteat retrieval of research,reportingidentified bias). Summarize Summarize the main findings theincluding strength of evidence for outcome;maineach theirconsider relevance to systematic systematic review. key groups (e.g.,key healthcare providers, users, policyand makers). For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

25 27 For26 peer review only

Funding Funding FUNDING FUNDING Conclusions Limitations Summary of of Summary evidence 24 DISCUSSION DISCUSSION Page 45 of 47 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from Page 46 of 47 6 7 6 6 7 7 6 6 4 1 7 7 13 6&7 table 1table table 1table Figure Figure 1 Page Page No Reported on Reported supplementary supplementary BMJ Open http://bmjopen.bmj.com/ Recommendation on October 2, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml For peer review only Description of relevance appropriatenessor of studies for assembled assessing the hypothesistested to be 9 9 to include Effort available studies,all including contact with authors 8 8 Search including strategy, time period included in the synthesis keyand words 7 7 Qualifications of searchers librarians and(eg, investigators) 6 6 Study population 5 5 of Type designsstudy used 4 4 of exposure Type or intervention used 3 3 Description of study outcome(s) 2 2 Hypothesis statement 1 1 Problem definition 17 16 Description of any contact with authors 15 ofhandling Method abstracts and unpublished studies 14 ofaddressing Method articles published in languages than other English 13 ofcitations List located those and excluded, including justification 12 hand Use (eg,of searching reference of obtainedlists articles) 11 Search used, software version, andname including special features (eg, used explosion) 10 Databases registriesand searched Item NoItem Reporting of methods includeshould Reporting of search Reporting of search strategy should include Reporting of background includeshould MOOSE Checklist for Meta-analyses of Observational Studies Observational of forMeta-analyses MOOSEChecklist 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

8 7 19 7&8 9&10 27-36 Table 1-3 Table table 2 82 table & Figures 2&3Figures Figures 2&3Figures Figures 2&3Figures supplementary BMJ Open http://bmjopen.bmj.com/ on October 2, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml For peer review only ationale forationale the selection ofcoding dataand (eg, sound principles clinical or regression on possibleregression predictorson results of study R Description of statistical(eg, methodscomplete descriptionof randomfixed or effects justification models, whether of the chosen models account for predictorsof study dose-responseresults, models, cumulativeor meta-analysis) in sufficientto detail be replicated Documentation of how dataand classifiedwere coded raters,multiple (eg, blinding and reliability)interrater convenience) appropriate) Assessment ofAssessmentquality, studyincluding blinding of assessors, quality stratification or Assessment ofAssessment confounding (eg, comparability of and cases controls where studies in 18 28 of Indication statistical ofuncertainty findings 27 ofsensitivity Results testing (eg, subgroup analysis) 26 giving Table descriptive information for each study included 25 Graphic summarizing individual study estimates andoverall estimate 24 ofappropriate Provision graphicsandtables 23 22 of Assessment heterogeneity 21 20 19 Reporting of resultsReporting of should include Page 47 of 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

Low alcohol consumption and pregnancy and childhood outcomes: time to change guidelines indicating apparently ‘safe’ levels of alcohol during pregnancy? A systematic review and meta-analyses.

For peer review only Journal: BMJ Open

Manuscript ID bmjopen-2016-015410.R1

Article Type: Research

Date Submitted by the Author: 24-Apr-2017

Complete List of Authors: Mamluk, Loubaba; University of Bristol School of Social and Community Medicine, IEU Edwards, Hannah; University of Bristol, School of Social and Community Medicine Savovic, Jelena; University of Bristol, NIHR CLAHRC West Leach, Verity; University of Bristol Jones, Tim; NIHR CLAHRC West, Moore, Theresa; NIHR CLAHRC West Ijaz , Sharea ; NIHR CLAHRC West Lewis, Sarah; University of Bristol, School of Social and Community Medicine Donovan, Jenny; NIHR CLAHRC West Lawlor, Debbie; Department of Social Medicine, University of Bristol, MRC http://bmjopen.bmj.com/ Integrative Epidemiology Unit Davey Smith, George; University of Bristol, Social Medicine Fraser, Abigail; University of Bristol, MRC Integrative Epidemiology Unit Zuccolo, Luisa; University of Bristol, School of Social and Community Medicine

Primary Subject Epidemiology Heading: on October 2, 2021 by guest. Protected copyright. Secondary Subject Heading: Reproductive medicine

Keywords: PAEDIATRICS, EPIDEMIOLOGY, Maternal medicine < OBSTETRICS

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 1 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Systematic Review & Metaanalysis 4 5 6 7 Title: Low alcohol consumption and pregnancy and childhood outcomes: time to change 8 9 guidelines indicating apparently ‘safe’ levels of alcohol during pregnancy? A systematic review 10 11 and metaanalyses. 12 13 14 15 Running tittle:For Low alcohol peer consumption reviewand pregnancy and childhood only outcomes. 16 17 18 19 Authors: Loubaba Mamluk PhD1,2,3, Hannah B Edwards MSc2,3, Jelena Savović PhD2,3, Verity 20 21 Leach PhD 2,3, Timothy Jones PhD 2,3, Theresa HM Moore MSc 2,3, Sharea Ijaz PhD 2,3, Sarah 22 23 J. Lewis PhD 2, Jenny L. Donovan PhD 2,3, Debbie Lawlor PhD 1,2,3, George Davey Smith PhD 24 25 1,2, Abigail Fraser PhD*1,2, Luisa Zuccolo PhD*1,2 26 27 * Joint Last Authors 28 29 30 31 32 Author addresses: Hannah B Edwards, Jelena Savovic, Verity Leach, Timothy Jones, 33

34 Theresa HM Moore, Sharea Ijaz, and Jenny L. Donovan: NIHR CLAHRC West, University http://bmjopen.bmj.com/ 35 36 Hospitals Bristol, NHS Foundation Trust, 9th Floor, Whitefriars, Lewins Mead, Bristol, BS1 37 38 2NT, United Kingdom. Sarah J. Lewis, Debbie Lawlor, George Davey Smith, Abigail Fraser, 39 40 and Luisa Zuccolo: MRC Integrative Epidemiology Unit, School of Social and Community 41

42 Medicine, University of Bristol, Oakfield House, Oakfield Grove, Bristol BS8 2BN, United on October 2, 2021 by guest. Protected copyright. 43 44 Kingdom. 45 46 47 Corresponding author: Loubaba Mamluk, MRC Integrative Epidemiology Unit, School of 48 49 Social and Community Medicine, University of Bristol, Oakfield House, Oakfield Grove, Bristol 50 51 BS8 2BN, United Kingdom. NIHR CLAHRC West, University Hospitals Bristol 52 53 NHS Foundation Trust, 9th Floor, Whitefriars, Lewins Mead Bristol, BS1 2NT. United Kingdom 54 55 T+44 (0)117 3313378, [email protected] 56 57 1 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 2 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Author affiliations: 4 5 6 1 MRC Integrative Epidemiology Unit at the University of Bristol, Bristol 7 8 2 School of Social and Community Medicine, University of Bristol, Bristol 9 10 3 NIHR CLAHRC West, University Hospitals Bristol NHS Foundation Trust, Bristol 11 12 Word Count: 3802 13 Figures: 3 14 Tables: 3 15 SupplementaryFor Tables: 3peer review only 16 Supplementary Figures:1 17 Supplementary Document:1 18 Number of references: 61 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33

34 http://bmjopen.bmj.com/ 35 36 37 38 39 40 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 2 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 3 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Abstract: 4 5 Objectives: To determine the effects of lowtomoderate levels of maternal alcohol 6 7 consumption in pregnancy on pregnancy and longerterm offspring outcomes. 8 9 Search Strategy: Medline, Embase, Web of Science, and Psychinfo from inception to 1107 10 11 2016. 12 13 Selection Criteria: Prospective observational studies, negative control and quasi 14 15 experimental studiesFor of pregnant peer women estimatingreview effects of lightonly drinking in pregnancy 16 17 (≤32g/week) versus abstaining. Pregnancy outcomes such as birth weight, and features of 18 19 fetal alcohol syndrome were examined. 20 21 Data Collection and Analysis: One reviewer extracted data and another checked extracted 22 23 data. Random effects metaanalyses were performed where applicable, and a narrative 24 25 summary of findings was carried out otherwise. 26 27 Main Results: 24 cohort and two quasiexperimental studies were included. With the 28 29 exception of birth size and gestational age, there was insufficient data to metaanalyse or 30 31 make robust conclusions. Odds of smallforgestationalage (SGA) and preterm birth were 32 33 higher for babies whose mothers consumed up to 32g/week versus none, but estimates for

34 http://bmjopen.bmj.com/ 35 preterm birth included the null value: summary odd ratios (OR) 1—08, 95% confidence intervals 36 37 (CI) (1—02 to 1—14), I2 0%, (7 studies, all estimates were adjusted) OR 1—10, 95%CI (0—95 38 39 2 40 to1—28), I 60%, (9 studies, includes one unadjusted estimates) respectively. The earliest time 41 points of exposure were used in the analysis.

42 on October 2, 2021 by guest. Protected copyright. 43 44 Conclusion: Evidence of the effects of drinking <=32g/w in pregnancy is sparse. As there 45 46 was some evidence that even light prenatal alcohol consumption is associated with being 47 48 SGA and preterm delivery, guidance could advise abstention as a precautionary principle, but 49 50 should explain the paucity of evidence. 51 52 Keywords: Alcohol, pregnancy, systematic review. 53 54 55 56 57 3 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 4 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 4 5 Strengths and Limitations of this study: 6 7 Strengths 8 9 Completeness of searches with a focused research question aimed at informing alcohol in 10 11 pregnancy guidelines. 12 13 Biases minimised by only including those with prospective assessment of exposure and 14 15 For peer review only 16 prioritising results adjusted for main confounders. 17 18 Unique effort to include alternative study designs to further improve causal inference 19 20 alongside standard analytical approaches. 21 22 Limitations 23 24 Limitation of results on the effects of light drinking in pregnancy from standard analytical 25 26 approaches is bias due to residual confounding. 27 28 The inclusion of only English language studies may have led to missing some studies, 29 30 however there is little evidence that exclusion of nonEnglish language studies leads to 31 32 systematic bias in systematic reviews of conventional medicine. 33

34 We could not pool eligible studies for various reasons (e.g. too few studies, lack of http://bmjopen.bmj.com/ 35 36 standard errors) 37 38 39 40 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 4 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 5 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Keywords: Alcohol, pregnancy, systematic review. 4 5 Introduction 6 7 Alcohol is a known teratogen[1] and the evidence about the risks of heavy alcohol 8 9 consumption during pregnancy on intellectual ability, birth defects, behaviour, fine motor 10 11 skills, and mental health (comprising fetal alcohol spectrum disorder – FASD)[2] is clear and 12 13 compelling.[3] Internationally, clinical guidelines recommend that pregnant women should 14 15 abstain fromFor heavy or “binge” peer drinking.[4] review However, until recently only UK guidelines advised 16 17 women to avoid drinking alcohol while trying to conceive, and in the first trimester, but at the 18 19 same time indicated that consumption should be restricted to within “1 to 2 UK units, once or 20 21 twice a week”.[5] The UK Chief Medical Officer (CMO) commissioned a review of guidelines 22 23 on alcohol consumption during pregnancy. Based on a review of reviews, the Guidelines 24 25 Development Expert Group has recently proposed a change to guidelines such that women 26 27 should be advised to abstain from alcohol when pregnant and/or trying to conceive,[6] based 28 29 on the precautionary principle (i.e. “better safe than sorry”), in the absence of robust 30 31 evidence. 32 33

34 Our aim was to conduct a comprehensive systematic review and metaanalysis of the http://bmjopen.bmj.com/ 35 36 literature to determine the effects of lowtomoderate levels of maternal alcohol consumption 37 38 in pregnancy on pregnancy and longer term offspring outcomes. Here we report on alcohol 39 40 consumption of up to two UK units of alcohol up to twice a week (the equivalent of ~ 41

42 32g/week), compared to no alcohol. In the absence of evidence from randomised controlled on October 2, 2021 by guest. Protected copyright. 43 44 trials, we examine observational studies of pregnant women from the general population with 45 46 prospective assessment of alcohol exposure, to reduce recall bias. In particular, we 47 48 specifically seek out quasiexperimental studies, negative control comparisons, and 49 50 Mendelian randomisation analyses in order to reduce the impact of confounding and 51 52 measurement error on the effect estimates. 53 54 55 56 57 5 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 6 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Methods 4 5 Selection strategy and selection criteria 6 A full protocol of this systematic review carried out using PRISMA guidelines[7] is available 7 8 from the PROSPERO systematic review register (registration number CRD4201501594); 9 10 http://www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42015015941). 11 12 In brief, eligible studies were defined as epidemiological studies of pregnant women or 13 14 women trying to conceive with prospective assessment of prenatal alcohol exposure (i.e. 15 For peer review only 16 before birth), sampled from general population. The protocol specifically included studies 17 18 19 using standard analytical approaches (e.g. multivariable regression analysis), as well as 20 21 studies that used innovative analytical methods to improve causal inference, such as (i) 22 23 quasiexperimental studies (for example comparing outcomes before and after 24 25 implementation of new guidelines on alcohol consumption); (ii) negative control studies (e.g. 26 27 comparing the association of offspring outcomes with maternal alcohol consumption to the 28 29 association of the same outcomes with consumption among fathers, under the assumption 30 31 that confounding is likely to be similar but that if there was a direct causal effect of maternal 32 33 consumption on outcomes, maternal associations would be stronger); and Mendelian

34 http://bmjopen.bmj.com/ 35 randomisation studies (using genetic variants associated with alcohol consumption and 36 37 metabolism). We considered these analytical approaches to be the most appropriate in 38 39 terms of their ability to minimise bias from confounding and other sources. Our original 40 41 protocol included studies exploring the effects of prenatal alcohol consumption up to

42 on October 2, 2021 by guest. Protected copyright. 43 83g/week (the commonly used threshold for moderate consumption [810]) versus 44 45 abstinence. Here we have focused specifically on low alcohol consumption, i.e. up to 46 47 32g/week as this was the cutoff specified by the UK guidelines at the time of writing this 48 49 review as being an implicitly “safe” threshold.[5] This specific cut off value has not been 50 51 reviewed and is the main point of discussion as the guideline change from low consumption 52 53 (equating to 1 to 2 UK units, once or twice a week or 32g/week) to abstinence (reference 54 55 group). 56 57 6 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 7 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Outcomes included: 1) pregnancy outcomes: still birth (pregnancy loss after week 24; 4 5 miscarriage; gestational length and preterm delivery (<37 weeks gestation); hypertensive 6 7 disorders of pregnancy; gestational diabetes; small for gestational age (SGA), < 10th 8 9 percentile in weight or <−2 standard deviation scores) and birth size (weight (including low 10 11 birth weight defined as <2500g), length, and head circumference); low amniotic fluid 12 13 (oligohydramnios); placenta previa; placental abruption; assisted delivery (including vacuum 14 15 extraction, forcepsFor delivery, peer Caesarean section);review Apgar score atonly birth; admission to neonatal 16 17 unit; congenital malformations. 2) Features of foetal alcohol spectrum disorder (FASD): 18 19 childhood growth restriction; cranium size and head circumference; developmental delays; 20 21 behaviour problems; cognitive impairment and intelligent quotient (IQ); facial malformations. 22 23 We adopted study specific definitions for all outcomes. 24 25 26 Studies were excluded if: there was no quantitative measure of alcohol consumption that 27 28 could be converted to grams of alcohol/week; there was insufficient data to estimate the 29 30 effect size of the association of our predefined low consumption categories versus 31 32 abstinence with any outcome, including studies that analysed alcohol as a continuous 33

34 variable (i.e. assuming the same linear or log linear effect across the entire alcohol http://bmjopen.bmj.com/ 35 36 distribution); the lowest exposure category (compared to nondrinkers) had an upper bound 37 38 exceeding 32g/week, or was unspecified; they were cohort studies of pregnant women with 39 40 alcohol abuse/dependency; they were casecontrol studies or cohort studies with 41

42 on October 2, 2021 by guest. Protected copyright. retrospective alcohol consumption assessment (e.g. after birth). 43 44 The following databases were searched: MEDLINE, PsycINFO, EMBASE on Ovid; the 45 46 47 Cochrane Library including CENTRAL (the Cochrane Central Database of Controlled Trials) 48 49 on Wiley Interscience; and Science Citation Index, Social Science Citation Index, on Web of 50 51 Science from inception to 11 July 2016 (supplementary Table 1). We limited the search to 52 53 English language papers and excluded animal studies, letters, editorials, and conference 54 55 proceedings for which there were no fulltext papers. Searches were tailored to each 56 57 7 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 8 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 database by investigators. The search focused on published medical literature and did not 4 5 include grey literature. We additionally performed manual searches of the reference lists of: 6 7 (i) all papers included in recent systematic reviews of the effects of prenatal alcohol 8 9 exposure on the outcomes of interest; and (ii) all recent papers citing those reviews. 10 11 Titles and abstracts, and full texts if necessary, were screened independently by two 12 13 reviewers. Discrepancies were discussed and disagreements resolved through consensus. 14 15 We assessedFor potential forpeer bias in included review studies by assessing only how well the study adjusted 16 17 for several main confounders known to impact on the exposureoutcome associations 18 19 (socioeconomic positioning as measured by the individual study, smoking during pregnancy, 20 21 maternal age, and ethnicity). We considered the potential for confounding and bias across 22 23 studies included in the analyses and described it narratively alongside summary results. 24 25 26 Data extraction 27 28 Data were extracted into a custombuilt Microsoft Access database. We extracted the 29 30 following information from each study: title, authors, publication year, country/region, study 31 32 design, population characteristics (sample size, methods of sampling, age distribution, and 33

34 ethnicity), measures of exposure (assessment method; including timing and quantification of http://bmjopen.bmj.com/ 35 36 alcohol consumption, reference group (abstinence), exposure (e.g 12 units or 24 units), 37 38 and information on unit equivalence if stated), outcome assessment methods (including 39 40 whether this was abstracted from medical records, obtained via a research interview and the 41

42 person reporting the outcome e.g. parent, teacher, health professional, researcher or child), on October 2, 2021 by guest. Protected copyright. 43 44 model adjustments, and study results. If a study reported more than one result for each 45 46 outcome, we extracted all of them (e.g. relative to different timing of exposure, model 47 48 adjustments, etc.). Information from each included paper was extracted by one reviewer 49 50 (LM) and subsequently checked for accuracy and completeness by another reviewer 51 52 (HE).[11] Extraction errors were minimal and were resolved through discussion between 53 54 extractor and checker. 55 56 57 8 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 9 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Alcohol unit conversion 4 Alcohol consumption in drinks/week was converted into grams/week based on the pure 5 6 ethanol equivalent of one drink, as stated in each individual article, or otherwise inferred 7 8

9 based on the definition of standard drinks in the country where the study took place. 10 11 Data analysis 12 The association of low alcohol use with pregnancy and related outcomes was investigated 13 14 comparing the highest category within the range of 032g/week to abstention (during 15 For peer review only 16 17 pregnancy). In studies providing data across several categories of intake within the 0 18 19 32g/week range, we used the effect estimate for the highest category of intake. If studies 20 21 reported on exposure to alcohol during different trimesters, we included estimates relative to 22 23 the earliest exposure. This is because for some outcomes, the first trimester tends to be the 24 25 most critical timing/window of exposure [12] [13] and because most studies that only 26 27 reported on one time point reported on exposure in early gestation. Similarly, if results were 28 29 available from both unadjusted and adjusted regressions, we prioritised fully adjusted 30 31 results, as a way of minimising the impact of confounding by important factors such as 32 33 maternal smoking, age, socioeconomic position and ethnicity. In case of multiple results

34 http://bmjopen.bmj.com/ 35 from the same cohort (relative to the same outcome), we analysed those pertaining to the 36 37 largest population size (i.e. conducted on the least ‘selected’ population as result of 38 39 exclusions, to minimise selection bias). Results from all studies that fulfilled our inclusion 40 41 criteria were summarised, together with information about the study. Where appropriate, we

42 on October 2, 2021 by guest. Protected copyright. 43 additionally pooled results for each outcome. Authors were not contacted for extra data. 44 45 Results from Different study designs have been reviewed separately. Individual study 46 47 estimates were pooled using random effects metaanalysis. Where only two studies were 48 49 available to metaanalyse, results were pooled unless they were very different from each 50 51 other (I2 ≥ 50%).[14] In this case, a narrative summary of findings was carried out and 52 53 results were reported in Table 2. Where a study only reported unadjusted results, we kept 54 55 56 57 9 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 10 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 these separate in the forest plots (subgroup analysis) but then also showed overall pooled 4 5 estimates combining all results. 6 7 Planned subgroup analyses by trimester could not be performed due to insufficient number 8 9 of included studies with this information. 10 11 Risk of bias assessment 12 13 The NewcastleOttawa Scale (NOS) [15] was used to assess risk of bias for included 14 15 reports. ThisFor is an eightitem peer questionnaire review assessing the following: only representativeness of 16 17 the exposed cohort; selection of the nonexposed cohort; exposure assessment methods; 18 19 absence of outcome (of interest) at the start of the study; comparability of exposed and non 20 21 exposed groups (with regard to confounding variables); blind assessment of exposure and 22 23 outcome; and length and adequacy of follow up. NOS allocates ‘stars’ for adequate 24 25 methods, but does not specifically advise calculating the sum of allocated stars to give an 26 27 overall score. Scores for quality are not helpful in assessing the effect of risk of bias on a 28 29 metaanalysis so we report each item separately in line with recommended methods [16 17]. 30 31 To be assessed as adequate for comparability of cohorts (risk of confounding) a study had to 32 33 control for the following four prespecified potential confounding factors related to foetal 34 http://bmjopen.bmj.com/ 35 36 development: maternal age, socioeconomic status (SES), ethnicity, and smoking. 37 38 39 40 The likelihood of small study bias, such as publication bias, could not be assessed through 41 visual inspection of funnel plots for pooled analyses as no outcome was assessed by 10+

42 on October 2, 2021 by guest. Protected copyright. 43 44 studies [18]. Statistical analyses were carried out using Stata version 13.1 (StataCorp, 45 46 College Station, TX, USA).[19] 47 48 Results 49 A flowchart of the article review process is shown in Figure 1. A total of 4680 citation records 50 51 were identified from searching the four relevant databases. A manual search of recent 52 53 systematic reviews identified 33 additional articles. After exclusions, 24 prospective studies 54 55 56 57 10 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 11 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 analysed using standard approaches and two quasiexperimental studies were included, 4 5 reporting on 30 outcomes in total. 6 7 Risk of bias 8 9 Six studies (Albertson 2003; Anderson 2012; Bille 2007; Brooke 1989; Lundsberg 2015; 10 11 McCarthy 2013) had low risk of bias for all eight NOS items and were therefore considered 12 13 14 at low risk of bias overall. All studies were judged to be at a low risk of bias for the following 15 For peer review only 16 three NOS items: selection of the nonexposed cohort (always from the same source 17 18 population as the exposed cohort); the absence of outcome at the start of the study; and 19 20 adequate length of follow up for outcome to have occurred. Fourteen studies had adequate 21 22 ascertainment of exposure as these were all based on structured interviews or validated 23 24 records. Objective outcome assessments (assessor unaware of the exposure status) were 25 26 reported in 16 of the studies. For five others either parent selfreport was used (high risk), 27 28 and for the remaining three the method of outcome assessment was not reported (unclear 29 30 risk). Eleven studies did not report enough detail to decide if cohorts were representative of 31 32 the population, therefore only 10 could be judged as low risk. Only four studies did not 33

34 control for the prespecified potential confounding factors, and one did not report enough http://bmjopen.bmj.com/ 35 36 detail to permit judgement. Thus, in the majority of studies (19) the compared groups were 37 38 similar. Nineteen studies had adequate follow up of the cohort (small loss to follow up). Only 39 40 three were judged high risk for this item and two studies presented insufficient information to 41

42 make a clear judgment. Standard analytical approaches on October 2, 2021 by guest. Protected copyright. 43 44 Pooled estimates for continuous and binary outcomes are presented in Figure 2 and 3 45 46

47 respectively. 48 49 Figure 2 presents results for birthweight (7 studies). Figure 3A presents results for preterm 50 51 delivery (9 studies) Figure 3B, presents results for SGA (7 studies), and results for low 52 53 birthweight (6 studies) are given in Figure 3C. 54 55 56 57 11 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 12 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 The metaanalysis yielded a summary OR of 1—10 (95% CI 0—95; 1—28) for preterm delivery, 4 5 but there was substantial statistical heterogeneity between studies (I2=60%), due to a large 6 7 Danish study reporting a protective effect (Figure 3A). Additionally, most studies assessing 8 9 preterm birth had corrected for main confounders know to be associated with preterm birth, 10 11 with the exception of [20] that did not correct for any. There was also modest evidence for 12 13 an increased risk of being SGA (OR 1—08, 95% CI 1—02; 1—14) for a total of 288512 14 15 participants, Foralthough this peer was almost entirely review driven by a single only US study contributing 95% 16 17 of the participants to this metaanalysis (Figure 3B). The birthweight metaanalysis yielded a 18 19 summary effect of 13—49g (95% CI 30—28g; +3—31g) for offspring of light drinkers versus 20 21 nondrinkers (Figure 2). Summary effect for birthweight <2500g was, OR 1—00, 95% CI 0—82; 22 23 1—22 (Figure 3C). 24 25 Other outcomes were typically reported by a limited number of studies and mostly could not 26 27 be metaanalysed due to clinical heterogeneity in outcome assessment or incompleteness of 28 29 published data (supplementary Figure 1, and Table 2). Based on two studies with data on 30 31 behavioural outcomes, there was little evidence of any effect for internalising symptoms but 32 33 a suggestion that light drinking in pregnancy protected against high externalising behaviour 34 http://bmjopen.bmj.com/ 35 36 scores (OR 0—97, (95%CI 0—93; 1—01, supplementary Figure 1)). However, an additional 37 38 study assessing conduct problems and hyperactivity (in the same externalising domain) 39 40 reported results in the opposite direction, which could not be metaanalysed due to different 41 outcome definitions.

42 on October 2, 2021 by guest. Protected copyright. 43 44 Table 2 presents results of included studies that did not contribute to the metaanalyses for 45 46 various reasons. There was no strong evidence of association between consuming up to 32 47 48 g/week of alcohol and any of the remaining outcomes excluded from metaanalyses, with 49 50 three exceptions: a very large US study showing increased risk of placental abruption and 51 52 decreased risk of preeclampsia (OR 1—24, 95%CI 1—05; 1—46 and OR 0—82, 95% CI 0—74; 53 54 0—90, respectively)[21], and a single British study reporting better cognitive outcomes in 55 56 children exposed to light maternal drinking in pregnancy.[22] 57 12 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 13 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 We did not include funnel plots as no outcome was assessed by 10+ studies. 4 5 Of all included results, only two were unadjusted[20 23], and most of the others were 6 7 adjusted for maternal smoking, age and socioeconomic position (supplementary Table 2). 8 9 Studies that did not adjust for ethnicity were generally conducted in homogenous 10 11 populations. Due to the small number of studies for each outcome, we could not further 12 13 investigate the effect of adjusting for all or some of these confounders. Similarly, there was 14 15 insufficient dataFor to examine peer the effect of reviewtiming of exposure on onlyoutcomes. 16 17 18 Alternative analytical approaches 19 Two negative control publications[24] based on the same UK cohort met our inclusion 20 21 criteria.[23 25] They investigated the effects of maternal alcohol consumption on childhood 22 23 educational achievement[25] and IQ.[23] Offspring exposed to maternal consumption of 24 25 <12g/week of alcohol in the first trimester did not have worse outcomes compared to those 26 27 of mothers who abstaining from alcohol, and a similar pattern was found for paternal alcohol 28 29 consumption. 30 31 One further quasiexperimental study, one natural experiment and five Mendelian 32 33 randomisation studies were excluded from the present review because they did not

34 http://bmjopen.bmj.com/ 35 specifically test the effect of consuming up to 32g/week in pregnancy versus abstaining. 36 37 These will be included in a forthcoming review focused on estimating the causal effects of 38 39 prenatal alcohol exposure based on alternative study designs and analytical approaches to 40 41 strengthen causal evidence.

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 13 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

Page 14 of 51 Yes Yes Yes Yes Yes Yes No Yes Yes Yes yes Yes Yes Yes Yes Yes Adjusted Adjusted (yes/no) (yes/no) Yes Yes Yes Yes 14 14

11 months 11 inclusive years 14 - - Birth Birth Birth Birth 7.75 years 7.75 Birth Birth Birth 7.75 years 7.75 0 Birth Birth 2 Birth Age at outcome outcome at Age assessment (child) (child) assessment 7 years years 7 Birth Birth

44 44 - 20 range range

trimester trimester trimester trimester

trimester ithin the gestational age gestational ithinthe etween 5th and 39th weeks of pregnancy of weeks 39th and 5th etween st st st st 1 Late pregnancy pregnancy Late Birth Yes 1st trimester 1st 1 36 weeks & 28 (12.9, pregnancy Throughout 24 (before care antenatal for booking At 1st trimester 1st gestation) gestation) weeks B 1 W trimester 2nd 1 weeks Timing of exposure exposure of Timing At booking for antenatal care (before 24 24 (before care antenatal for booking At gestation) weeks During pregnancy pregnancy During Entire pregnancy pregnancy Entire

analysis) - Behaviour Checklist (CBCL) Checklist Behaviour

in the meta in the

≥60),Child BMJ Open Externalising Externalising

included http://bmjopen.bmj.com/

- Externalising Externalising -

problem scores, Strengths & Difficulties Questionnaire (SDQ) (SDQ) Questionnaire Difficulties & Strengths scores, problem problem scores, Strengths & Difficulties Questionnaire (SDQ) (SDQ) Questionnaire Difficulties & Strengths scores, problem percentile)

- problems onduct th Hyperactivity

centile) centile) percentile) percentile) percentile) percentile) percentile)

th th th th th Birth weight (g) (g) weight Birth Birth weight (g) (g) weight Birth SGA (lowest 10 (lowest SGA (g) weight Birth weeks) (<37 birth Preterm 10% with highest highest 10%with C report parent weight birth Low weeks) (<37 delivery Preterm Preterm birth (<37 weeks) (<37 birth Preterm Birthweight weeks) (<37 birth Preterm score (T problems significant Clinically highest 10%with report parent Birth weight (<2500g) weight Birth 10 (< SGA 10 (< SGA 10 (< SGA report Parent behaviour Internalising behaviour Externalising Outcomes Outcomes 10 (< SGA Low birth weight (<2500g) (<2500g) weight birth Low weeks) (<37 birth Preterm birthweight mean Adjusted (<2500g) weight birth Low weeks) (<37 birth Preterm SGA (weight < −2 standard deviaon scores) scores) deviaon standard −2 (weight < SGA weeks) (<37 birth Preterm weight Birth Preterm birth (<37 weeks) weeks) (<37 birth Preterm Customised birth centile birthcentile Customised 10 (< SGA Moderate preterm birth (32-37 weeks) weeks) (32-37 birth preterm Moderate weeks) (<32 birth preterm Very Parent report report Parent behaviour Internalising Boys behaviour Internalising Girls behaviour Externalising Boys behaviour Externalising Girls on October 2, 2021 by guest. Protected copyright.

number

2

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

10,649 10,649 (SDQ), Questionnaire Difficulties & Strengths scores, problem highest 10%with 2,06 638 901 10,323 1,493 3166 - 1335 10,323 3,907 3,907 Total 4,132 4,132 weight birth Low 535 535 29,463 29,463 weeks) (<37 birth Preterm

For peer review only

- - 64 - 126 - - 202 - 185 185 - 202 325 150 Event 191 315 274 ------23

USA USA UK UK Japan Zealand, New USA Australia UK Australia, UK, Australia, Ireland and Country USA

Prospective studies examining the effects of prenatal alcohol exposure on pregnancy outcomes (studies (studies outcomes pregnancy on exposure alcohol prenatal of effects the examining studies Prospective

Lundsberg (1997) Lundsberg (1995) Shu (1995) Peacock Sayal Sayal (2007) McCarthy (2013) McCarthy (2011) Salihu Sayal (2007) Lundsberg (2015) Lundsberg

Miyake (2014) Miyake

[35] [37] [20] [33] [29] [30] [21] [33] [27] Nykjaer (2014) (2014) Nykjaer [27] UK (2010) Robinson [31] (2007) Jaddoe [32] UK Holland; - Table 1. Table (year) Study [26] [34] Albertsen (2004) (2004) Albertsen [34] Denmark 1488 (1996) Passaro [36] UK 8,886 [28] Niclasen (2014) (2014) Niclasen [28] Denmark - 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

yes yes yes 15 15

Birth Birth Birth

Early pregnancy pregnancy Early During the first 36 weeks gestation weeks 36 the first During BMJ Open http://bmjopen.bmj.com/

Birthweight (g) (g) Birthweight Birth weight (<2500g) weight Birth on October 2, 2021 by guest. Protected copyright.

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 1,140 1,140 8772 For peer review only

Denmark Denmark

Olsen (1991) Olsen

[38] [39] Brooke (1989) (1989) Brooke [39] UK SGA: small for gestation age age gestation for small SGA: Page 15 of 51 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from Page 16 of 51 16 16

(yes/no) (yes/no) Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Adjusted Adjusted Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

99)

· 1·46) 1·50) 3·84) 0 0·90) 1·39) 1·12)

– – – – – – – 1·5)

– 35 35 ·

(95% CI 0·74 CI (95% [1] 59 (95% CI 0 CI 59(95% · those unexposed. unexposed. those OR 1·05 (95% CI 0·94 – 1·18) 1·18) 0·94– CI 1·05(95% OR Results Results OR 1·12 (95% CI 0·86 CI 1·12(95% OR 0·40 CI 0·78(95% OR OR 1·61 (95% CI 0·67 CI 1·61(95% OR OR 0 OR OR 0·82 OR 3225g in those exposed, compared to 3225g 3225g to compared exposed, those in 3225g unexposed. those in 3414g in those exposed, compared to 3363g 3363g to compared exposed, those in 3414g unexposed. those in OR 1·0 (95% CI 0·7 0·7 CI 1·0 (95% OR those 1·9) for (SD 40·1 age gestational Mean those for 2·0) (SD 40·1 to compared exposed, unexposed. OR 1·10 (95% CI 0·88 CI 1·10(95% OR OR 0·90 (95% CI 0·73 CI 0·90(95% OR 3·7% in those exposed compared to 3·9% in 3·9% to compared exposed in those 3·7% OR 1·08 (95% CI 0·83 – 1·42) 1·42) 0·83 – CI 1·08(95% OR OR 1·10 (95% CI 0·78 – 1·54) 1·54) 0·78 – CI 1·10(95% OR OR 1·11 (95% CI 0·87 – 1·43) 1·43) 0·87 – CI 1·11(95% OR OR 1·24 (95% CI 1·05 – 1·46) 1·46) 1·05 – CI 1·24(95% OR

National / Hospital Hospital / National maternal from or Registries, report) Birth Birth Birth Birth Birth Age at outcome assessment assessment outcome at Age (child) Birth Birth Birth (Outcome confirmed via confirmed (Outcome records) medical Birth (Outcome confirmed via confirmed (Outcome National / Hospital Hospital / National maternal from or Registries, report)

analysis) -

outcome

in the meta the in

with this outcome this with

included o studies with this this with studies o BMJ Open Only study with this outcome this with study Only outcome this with study Only Only study with this outcome this with study Only Only study with this outcome outcome this with study Only Birth Only study with this outcome outcome this with study Only Birth Only tw Only outcome Only two studies with this this with twostudies Only outcome Only study with this outcome outcome this with study Only Birth Only study with this outcome outcome this with study Only Birth SE/SD not reported not SE/SD SE/SD not reported not SE/SD

Only study study Only Only study with this with study Only Only study with this outcome outcome this with study Only via confirmed (Outcome Estimates are in different different in are Estimates directions Estimates are in different different in are Estimates directions analysis analysis Pregnancy outcomes Pregnancy http://bmjopen.bmj.com/ oetal alcohol spectrum disorder (FASD) disorder spectrum alcohol oetal

in

percentile percentile percentile percentile Features of f of Features th th

eclampsia eclampsia - - Lowest 10 Lowest Lowest 10 Lowest Placental abruption abruption Placental Placenta previa previa Placenta trimester trimester Per cent of infants with with infants of cent Per <37 age gestational weeks Admission to NICU to Admission malformations Major Apgar <7 at 5 minutes 5 minutes at <7 Apgar Pre Pre of head circumference circumference head of of birth length length birth of Mean birthweight (g) (g) birthweight Mean Mean birthweight (g) in (g) birthweight Mean smokers non-smokers non-smokers Gestational age Gestational Spontaneous abortion abortion Spontaneous occurring ≤20 weeks weeks ≤20 occurring gestation Stillbirth Stillbirth on October 2, 2021 by guest. Protected copyright.

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Country (Total number) number) (Total Country details Outcome meta- in included not Reason France, Spain, Denmark, Denmark, Spain, France, (8,453) Portugal Germany, USA (4,496) USA (4,496) USA USA (4,496) USA New Zealand, Australia, Australia, Zealand, New USA (1,226,685) USA UK Ireland, Netherlands, Scotland, Scotland, Netherlands, Netherlands, Scotland, Scotland, Netherlands, Denmark, Spain, France, (8,453) Portugal Germany, France, Spain, Denmark, Denmark, Spain, France, (8,453) Portugal Germany, UK (10,539) UK USA (5,324) USA USA (1,226,685) USA Denmark (89,322) Denmark For peer review only

Lundsberg (2015) Lundsberg (2015) Lundsberg Lundsberg (2015) Lundsberg McCarthy (2013) McCarthy Salihu (2011) Salihu Ogston (1992) Ogston Ogston (1992) Ogston Passaro (1996) Passaro Windham (1997) Windham Salihu (2011) Salihu Andersen (2012) Andersen

[26] Lundsberg (2015) (2015) Lundsberg [26] (4,496) USA [21] Salihu (2011) (2011) Salihu [21] (1,226,685) USA [42] Ogston (1992) (1992) Ogston [42] Scotland, Netherlands, [40] Andersen (2012) (2012) Andersen [40] (89,322) Denmark in first Miscarriage [26] [26] [26] [30] [21] [42] [42] [36] [41] [21] [40] [26] Lundsberg (2015) (2015) Lundsberg [26] (4,496) USA [21] Salihu (2011) (2011) Salihu [21] (1,226,685) USA Study (year) (year) Study

Prospective studies examining the effects of prenatal alcohol exposure on pregnancy outcomes (results not not (results outcomes pregnancy on exposure alcohol prenatal of effects the examining studies Prospective

eclampsia - Admission to to Admission Malformations Apgar score Apgar Intensive Neonatal (NICU) Unit Care Height Pre Placenta-related Birthweight Gestational age Gestational birth and preterm Miscarriage Table 2. Table Stillbirth Outcome Outcome Head Head circumference

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 17 17

Yes Yes Yes Yes Yes Yes Yes Yes Yes No Yes Yes Yes Yes Yes Yes Yes Yes Yes

1·16) 1·03) 1·50)

1·51)

– – – –

on KS2 score for those exposed compared to to compared exposed those for score KS2 on unexposed. those exposed, compared to 1·85 for those those for 1·85 to compared exposed, those (p=0·26) unexposed Mean increase of 0·81 points (se 0·8) for for 0·8) (se 0·81points of increase Mean inthose 16·3) (SD 106·4 IQ score Mean those exposed compared to unexposed. unexposed. to compared exposed those in 16·2) (SD 105·7 to compared exposed, p=0·10. unexposed, those OR 1·20 (95% CI 0·96 CI 1·20(95% OR those exposed compared to unexposed. unexposed. to compared exposed those OR 0·99 (95% CI 0·86 CI 0·99(95% OR OR 0·85 (95% CI 0·70 CI 0·85(95% OR Mean total anomalies was 2·60 for those those for 2·60 was anomalies total Mean those for 2·53 to compared exposed, (p=0·28) unexposed OR 1·06 (95% CI 0·74 CI 1·06(95% OR Mean increase of 1·80 points (SE 1·1) for for 1·1) (SE 1·80points of increase Mean OR 1·14 (95% CI 0·98 – 1·32) 1·32) 0·98– CI 1·14(95% OR Mean increase of 0·38 (95% CI -0·02 – 0·78) 0·78) – CI -0·02 0·38(95% of increase Mean Mean craniofacial anomalies was 1·92 for 1·92for was anomalies craniofacial Mean

9 years 9

- for gestation age IQ: intelligence quotient. intelligence IQ: age gestation for

18 months 18 months 7.75- 9 years years 7.75- 9 1·94) 1·02– CI 1·41(95% OR 18 18 months years 8 7.75 3.9 3.9 years 7 years 7 Birth Birth Birth Birth

deviation SGA: small SGA: deviation : standard standard : BMJ Open No SE/SD reported for for reported SE/SD No outcome this with study Only reference group group reference group reference No SE/SD reported for for reported SE/SD No Parent report used in analysis analysis in used report Parent instead instead instead Parent report used in analysis analysis in used report Parent Later age used in analysis used Laterage Later age used in analysis in analysis used Laterage 3.9 years Only study with this outcome this with study Only SE/SD not reported not SE/SD SE/SD not reported reported not SE/SD Only study with this outcome this with study Only error SD error http://bmjopen.bmj.com/

onduct onduct

onduct problems problems onduct Key Stage 2 scores scores 2 Stage Key outcome this with study Only 11 years Craniofacial anomalies anomalies Craniofacial Mental Development Development Mental of Scales Bayley Index, Development Infant Psychomotor Psychomotor Weschler IQ, Development Index, Index, Development Infant of Scales Bayley Development for Scale Intelligence (WISC-III) Children Hyperactivity teacher report report teacher Questionnaire (SDQ) (SDQ) Questionnaire Strengths & Difficulties &Difficulties Strengths 10% with highest highest 10% with scores, problem C teacher report report teacher Questionnaire (SDQ) (SDQ) Questionnaire Strengths & Difficulties &Difficulties Strengths problem scores, scores, problem 10% with highest highest 10% with Hyperactivity parent report report parent Questionnaire (SDQ) (SDQ) Questionnaire Strengths & Difficulties &Difficulties Strengths 10% with highest highest 10% with scores, problem parent report C report parent Questionnaire (SDQ) (SDQ) Questionnaire Strengths & Difficulties &Difficulties Strengths problem scores, scores, problem 10% with highest highest 10% with Developmental Developmental Total anomalies Total Disorder Coordination (DCD) Oral clefts Oral problems problems ≤32g AA per week. SE: Standard Standard SE: week. per AA ≤32g on October 2, 2021 by guest. Protected copyright.

)

(1020 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

France, Spain, Denmark, Denmark, Spain, France, (8,453) Portugal Germany, Netherlands, Scotland, Scotland, Netherlands, (4,332) UK France, Spain, Denmark, Denmark, Spain, France, (8,453) Portugal Germany, UK (525) UK UK (1077) UK Denmark (32,097) Denmark USA (873) USA Denmark Denmark For peer review only

Parry (1992) Parry (2008) Alati Sayal (2007) Sayal Sayal (2007) Sayal Faebo Larsen Larsen Faebo Ernhart (1989) Ernhart Bille (2007) Bille

[22] Sayal (2013) (2013) Sayal [22] (10,558) UK [46] Parry (1992) (1992) Parry [46] Scotland, Netherlands, [33] Sayal (2007) (2007) Sayal [33] (257) UK [33] Sayal (2007) (2007) Sayal [33] (967) UK [46] [23] [33] [33] [45] [44] (2013) [43] [44] Ernhart (1989) (1989) Ernhart [44] (873) USA

to exposed those in outcome of odds the compare results Ratio Odds [1] Cognition Cognition

Motor Motor Behaviour /development development

Page 17 of 51 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from Page 18 of 51 18 18 BMJ Open http://bmjopen.bmj.com/ on October 2, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

For peer review only

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 19 19 Different confounding structures structures confounding Different vs. maternal of association the for outcome the with alcohol paternal Different confounding structures structures confounding Different vs. maternal of association the for outcome the with alcohol paternal Limitations Limitations Maternal alcohol consumption consumption alcohol Maternal 16·2); (SD 105·7 mean abstainers: (SD 106·4 mean glass/week: <1 16·3) consumption alcohol Paternal 16·8); (SD 102·2 mean abstainers: (SD 104·0 mean glass/week: <1 16·7) Maternal alcohol consumption consumption alcohol Maternal 9·1); (SD 102·0 mean abstainers: (SD 102·8 mean glass/week: <1 8·7) consumption alcohol Paternal <1 11); (SD 98·9 mean abstainers: 9·1) (SD 101·1 mean glass/week: conclusions as presented in the in presented as conclusions paper IQ: Wechsler IQ: Wechsler Scale Intelligence Children for (WISC) achievement: Key achievement: scores 2 Stage (standardised) Outcomes Outcomes & results of Summary

8 years years 8

assessment assessment (child) 11 years years 11 Academic Intelligence quotient. Intelligence BMJ Open http://bmjopen.bmj.com/

na

na

na na na on October 2, 2021 by guest. Protected copyright. paternal paternal - For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Study type type Study Gene number SNP-rs outcome at Age Maternal comparison comparison comparison For peer review only

Number Number 4,332

Country Country Total UK

UK UK 7,062 Maternal-paternal

Quasiexperimental studies examining the effects of prenatal alcohol exposure on pregnancy outcomes. outcomes. on pregnancy exposure alcohol prenatal of the effects examining studies Quasiexperimental Alati (2008) Alati

Study (year) Study(year) [23] IQ: deviation SD:standard difference Mean MD: intervals CI: Confidence ratio OddsOR: Abbreviations: Table 3. Table (2013) Alati [25] Page 19 of 51 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open Page 20 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Discussion 4 5 Main findings: In this comprehensive systematic review of the literature on the effects of low 6 7 levels of alcohol drinking in pregnancy, the two main findings are: i) a surprisingly limited 8 9 number of prospective studies specifically addressing the question of whether light maternal 10 11 alcohol consumption (i.e. up to 32g/week (or 4 UK units) has any causal effect (adverse or 12 13 beneficial) on infant and later offspring outcomes and pregnancy outcomes, and, as a result, 14 15 ii) a paucity ofFor evidence demonstratingpeer areview clear detrimental effect, only or safe limit, of light alcohol 16 17 consumption on outcomes. The upper limit that we chose to examine here is that of the 18 19 current version of the UK National Institute for Health and Care Excellence (NICE) 20 21 guidelines.[47] The question we have attempted to address is very important given the 22 23 mixed advice that women are given with regards to whether they should abstain completely 24 25 or be allowed light alcohol consumption in pregnancy. The lack of research evidence to 26 27 address this question is notable. 28 29 30 31 Strengths and limitations: Strengths of this review include the completeness of searches with 32 33 a focused research question aimed at informing alcohol in pregnancy guidelines. In addition

34 http://bmjopen.bmj.com/ 35 to observational studies’ biases minimised by only including those with prospective 36 37 assessment of exposure and prioritising results adjusted for main confounders. Another 38 39 strength of this review is the unique effort to include alternative study designs to further 40 41 improve causal inference alongside standard analytical approaches. The main limitation of

42 on October 2, 2021 by guest. Protected copyright. 43 results on the effects of light drinking in pregnancy from standard analytical approaches is 44 45 bias due to residual confounding. SE position is a complex, multifaceted entity. Several 46 47 studies have attempted to adjust for SE position by collecting information on, for example, 48 49 maternal education, familylevel SE position around the time of the pregnancy, home 50 51 addressbased deprivation index etc. Few studies included more than one of these 52 53 measured [22 2931]. Whereas we consider attempting to adjust for at least one of these 54 55 56 characteristics to be a minimum requirement to account for some of the confounding 57 58 introduced by SE position, there remains scope for residual confounding.[48] Given the 59 60 20

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 21 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 strong relationship between SE position and both the exposure (alcohol use in pregnancy) 4 5 and outcomes in this review, any degree of residual confounding is of course an issue when 6 7 interpreting the effect estimates from the observational studies included in this review. 8 9 Women who drink low amounts of alcohol may be more likely to be of higher socio 10 11 economic position, compared to abstainers, at least in developed settings in recent 12 13 years,[49] and both of these characteristics are associated with better pregnancy and 14 15 cognitive outcomes.[50]For Maternalpeer smoking review and ethnicity are also only known correlates of 16 17 maternal alcohol use, and risk factors for e.g. low birth weight.[51] Most studies included in 18 19 this review adjusted for at least some of these factors. However, due to the small number of 20 21 studies included for any given outcome, it was impossible to formally investigate the effect of 22 23 incomplete adjustment for some (or all) of these confounders. Additionally, for most 24 25 26 outcomes, we could not pool eligible studies for various reasons (e.g. too few studies, lack of 27 28 standard errors) and that we limited our review to a number of prespecified outcomes 29 30 including the most common pregnancyrelated outcomes and childhood outcomes related to 31 32 FASD. This also was the case for identifying effects based on time of exposure, which is 33 also a limitation. 34 http://bmjopen.bmj.com/ 35 36 The inclusion of only English language studies may have led to missing some studies, 37 38 however there is little evidence that exclusion of nonEnglish language studies leads to 39 40 systematic bias in systematic reviews of conventional medicine.[5255] 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 Interpretation: This review demonstrates the paucity and poor quality of evidence addressing 45 46 this important public health question, and the difficulty of designing studies that can 47 48 effectively evaluate the causal impact of low alcohol consumption whilst minimising bias and 49 50 confounding. It also shows the value of reporting measures of effect for meaningful 51 52 categories of the exposure. Whilst many studies reported that associations did not differ from 53 54 linearity prior to providing a single coefficient for the doseresponse association, it is possible 55 56 that statistical power limited the ability to detect nonlinear associations in single studies. 57 58 59 60 21

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 22 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Such detail is especially important when there are controversies about the shape of the 4 5 association of interest (linear, U or Jshaped) and/or the existence of safe thresholds. 6 7 Outstanding questions also remain about the effects of maternal alcohol consumption at 8 9 different stages of conception and pregnancy. Alternative analytical approaches such as 10 11 sibling comparisons[56] and the use of instrumental variable approaches[57] as well as 12 13 triangulating the totality of evidence from multiple study types[58] (formally or informally) are 14 15 needed in orderFor to strengthen peer confidence review in the direction and sizeonly of any potential causal 16 17 relationships. 18 19 The recently proposed change in the guidelines for alcohol use in pregnancy in the UK to 20 21 complete abstinence, would be an application of the precautionary principle. This review 22 23 confirmed some increased risk of babies being born SGA but little direct evidence of any 24 25 other detrimental effect for maternal drinking up to 32g/week. However, there have been few 26 27 wellconducted studies examining this specific category of exposure. This issue remains of 28 29 great public health importance, with alcohol consumption during pregnancy prevalent in the 30 31 UK, Ireland, New Zealand and Australia with up to 80% of women consuming some alcohol 32 33 during pregnancy.[59] For some, the evidence of the potential for harm – mostly coming from 34 http://bmjopen.bmj.com/ 35 36 animal experiments and human studies of effects due to higher levels of exposure will be 37 38 sufficient to advocate that guidelines should advise women to avoid all alcohol in pregnancy, 39 40 while others will wish to retain the existing wording of guidelines.[60] Here we found that 41 maternal alcohol consumption of up to 32g/week was associated with an 10% increased risk

42 on October 2, 2021 by guest. Protected copyright. 43 44 of preterm birth (95%CI: 0—95 to1—28). In comparison, light to moderate smoking (<20 45 46 cigarettes per day) is associated with a 22% increased risk of preterm birth (95% CI: 1—13 to 47 48 1—32).[61] 49 50 51 In conclusion, we found limited evidence for a causal role of light drinking in pregnancy, 52 53 compared to abstaining, on most of the outcomes examined. Despite the distinction between 54 55 light drinking and abstinence being the point of most tension and confusion for health 56 57 professionals and pregnant women and contributing to inconsistent guidance and advice 58 59 60 22

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 23 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 now and in the past, our extensive review shows that this specific question is not being 4 5 researched thoroughly enough, if at all. In addition, there has been no evidence regarding 6 7 possible benefits of light alcohol consumption versus absence. Further studies, including 8 9 those using designs that improve causal inference, are required to provide further evidence 10 11 and a better estimation of the likely effects. Formulating guidance on the basis of the current 12 13 evidence is challenging. However, describing the paucity of current research and explaining 14 15 that “absenceFor of evidence peer is not evidence review of absence”, appears only warranted. 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33

34 http://bmjopen.bmj.com/ 35 36 37 38 39 40 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 23

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 24 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Acknowledgments 4 We thank our colleagues from NIHR CLAHRC West who provided support that greatly 5 6 assisted the research. We thank Louise Waters for assistance with article collection and 7 Alison Richards for developing the search strategy for this review. 8 9 10 11 Details of contributors 12 Loubaba Mamluk: Drafting the protocol, preparing the data extraction database, data 13 14 screening and collection (lead reviewer), data analysis, figures, tables, data interpretation, 15 and drafting Forthe manuscript. peer Hannah B Edwards:review screening, data only collection, data quality 16 17 checking, data interpretation, and preparation of final manuscript. Theresa HM Moore: 18 Advising on the protocol, preparing the protocol, screening title and abstract. Timothy Jones: 19 20 Data extraction and full text screening. Sharea Ijaz: Abstract and full text screening. Jelena 21 Savović: protocol development/study design (including the development of searching, article 22 23 screening and data collection strategies), oversaw and managed the review process and the 24 25 review team, provided methodological advice for the review and metaanalysis conduct, and 26 critically revised the manuscript. George Davey Smith: Obtained funding, formulation of 27 28 project plan, review of progress and of the manuscript. Jenny L. Donovan: Obtained funding 29 and contributed to drafting the manuscript, and approved the final version. Verity Leach: 30 31 Data extraction. Deborah Lawlor: contributed to the study design, data interpretation and 32 review of earlier manuscript drafts; she approved the final version that has been submitted. 33

34 Sarah J. Lewis: contributed to the study design, data interpretation and review of earlier http://bmjopen.bmj.com/ 35 manuscript drafts; she approved the final version that has been submitted. Abigail Fraser: 36 37 Contributed to the study design, data analysis and interpretation and drafting of manuscript. 38 Luisa Zuccolo: Contributed to the study design, data analysis and interpretation and drafting 39 40 of manuscript. I, Loubaba Mamluk, the corresponding author of this manuscript, certify that I 41 have full access to all the data in the study and had final responsibility for the decision to

42 on October 2, 2021 by guest. Protected copyright. 43 submit for publication. 44

45 46 Transparency declaration 47 48 I, Loubaba Mamluk, the corresponding author of this manuscript, affirm that this manuscript 49 50 is an honest, accurate, and transparent account of the study being reported; that no 51 important aspects of the study have been omitted; and that any discrepancies from the study 52 53 as planned (and, if relevant, registered) have been explained. 54 55 Conflict of interest: 56 57 The authors have nothing to disclose. 58 59 60 24

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 25 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 This work has not been submitted for publication elsewhere. 4

5 6 Disclosure 7 All authors have completed the ICMJE uniform disclosure form 8 9 at www.icmje.org/coi_disclosure.pdf and declare: no support from any organisation for the 10 11 submitted work; no financial relationships with any organisations that might have an interest 12 in the submitted work in the previous three years; no other relationships or activities that 13 14 could appear to have influenced the submitted work. 15 For peer review only 16 17 Role of funding source 18 The UK Medical Research Council (MC_UU_12013/1 and MC_UU_12013/5) and the 19 20 University of Bristol provide core funding for the MRC Integrative Epidemiology Unit; LM, 21 DAL, GDS, AF, and LZ are partially or fully funded through this. AF and LZ are UK Medical 22 23 Research Council research fellowships (Grant refs: MR/M009351/1 and G0902144, 24 respectively). LM, JS, HE, VL, TJ, SI,TM and JLD are fully or partially partly funded by 25 26 National Institute for Health Research (NIHR) Collaboration for Leadership in Applied Health 27 Research and Care West (CLAHRC West) at University Hospitals Bristol NHS Foundation 28 29 Trust. DAL (NFSI061110196) and JLD (NFSI051210119) are NIHR Senior 30 Investigators. SL is a senior lecturer funded by HEFCE. The views expressed here are those 31 32 of the author(s) and not necessarily those of the NHS, the MRC, NIHR or the Department of 33 Health. I, Loubaba Mamluk, the corresponding author of this manuscript, certify that I have 34 http://bmjopen.bmj.com/ 35 full access to all the data in the study and had final responsibility for the decision to submit 36 37 for publication. 38 39 40 Transparency declaration 41 The lead author (Loubaba Mamluk) affirms that this manuscript is an honest, accurate, and

42 on October 2, 2021 by guest. Protected copyright. 43 transparent account of the study being reported; that no important aspects of the study have 44 been omitted; and that any discrepancies from the study as planned (and, if relevant, 45 46 registered) have been explained. 47 48 Data sharing Statement 49 50 NA 51 52 53 54 55 56 57 58 59 60 25

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 26 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 References 4 1. Streissguth AP, Landesman-Dwyer S, Martin JC, Smith DW. Teratogenic effects of alcohol in 5 humans and laboratory animals. Science (New York, N.Y.) 1980;209(4454):353-61 6 7 2. Popova S, Lange S, Shield K, et al. Comorbidity of fetal alcohol spectrum disorder: a systematic 8 review and meta-analysis. Lancet (London, England) 2016;387(10022):978-87 doi: 9 10.1016/s0140-6736(15)01345-8[published Online First: Epub Date]|. 10 3. Chudley AE, Conry J, Cook JL, Loock C, Rosales T, LeBlanc N. Fetal alcohol spectrum disorder: 11 Canadian guidelines for diagnosis. CMAJ : Canadian Medical Association journal = journal de 12 l'Association medicale canadienne 2005;172(5 Suppl):S1-s21 doi: 13 10.1503/cmaj.1040302[published Online First: Epub Date]|. 14 4. International Alliance for Responsible Drinking- Drinking guidelines for pregnancy and breast 15 feeding.For 2016 peer review only 16 5. (NICE) NIFHACE. Antenatal care. CG62. NICE 2008 17 6. Department of Health. Alcohol Guidelines Review – Report from the Guidelines development 18 group to the UK Chief Medical Officers. 2016 19 20 7. Moher D, Liberati A, Tetzlaff J, Altman DG. Preferred reporting items for systematic reviews and 21 meta-analyses: the PRISMA statement. Annals of internal medicine 2009;151(4):264-69 22 8. Hamlyn B OK, Wands S. Infant Feeding 2000. London: The Stationery Office, 2002. 23 9. Henderson J, Gray R, Brocklehurst P. Systematic review of effects of low–moderate prenatal 24 alcohol exposure on pregnancy outcome. Bjog 2007;114(3):243-52 25 10. Patra J, Bakker R, Irving H, Jaddoe VW, Malini S, Rehm J. Dose–response relationship between 26 alcohol consumption before and during pregnancy and the risks of low birthweight, preterm 27 birth and small for gestational age (SGA)—a systematic review and meta-analyses. Bjog 28 2011;118(12):1411-21 29 11. Dissemination CfRa. CRD’s guidance for undertaking reviews in 30 31 health care. York Publishing Services Ltd 2009:32 32 12. Becker HC, Diaz-Granados JL, Randall CL. Teratogenic actions of ethanol in the mouse: a 33 minireview. Pharmacology Biochemistry and Behavior 1996;55(4):501-13 34 13. Haycock PC. Fetal Alcohol Spectrum Disorders: The Epigenetic Perspective 1. Biology of http://bmjopen.bmj.com/ 35 reproduction 2009;81(4):607-17 36 37 14. Higgins J, Green S. Cochrane handbook for systematic reviews of interventions version 5.1. 0 38 [updated Mar 2011]. The Cochrane collaboration. 2011, 2012. 39 15. Wells G, Shea B, O'Connell D, et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of 40 nonrandomised studies in meta-analyses. 41 http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp. Accessed March 2017

42 16. Moher D, Liberati A, Tetzlaff J, Altman DG. Preferred reporting items for systematic reviews and on October 2, 2021 by guest. Protected copyright. 43 meta-analyses: the PRISMA statement. PLoS medicine 2009;6(7):e1000097 doi: 44 10.1371/journal.pmed.1000097[published Online First: Epub Date]|. 45 17. Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of observational studies in epidemiology: a 46 proposal for reporting. Meta-analysis Of Observational Studies in Epidemiology (MOOSE) 47 group. Jama 2000;283(15):2008-12 48 18. Sterne JA, Sutton AJ, Ioannidis JP, et al. Recommendations for examining and interpreting funnel 49 50 plot asymmetry in meta-analyses of randomised controlled trials. Bmj 2011;343:d4002 51 19. StataCorp. Stata Statistical Software: release 12. College Station TSL, 2015. 52 20. Peacock JL, Bland JM, Anderson HR. Preterm delivery: effects of socioeconomic factors, 53 psychological stress, smoking, alcohol, and caffeine. Bmj 1995;311(7004):531-5 54 21. Salihu HM, Kornosky JL, Lynch O, Alio AP, August EM, Marty PJ. Impact of prenatal alcohol 55 consumption on placenta-associated syndromes. Alcohol 2011;45(1):73-9 doi: 56 http://dx.doi.org/10.1016/j.alcohol.2010.05.010[published Online First: Epub Date]|. 57 58 59 60 26

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 27 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 22. Sayal K, Draper ES, Fraser R, Barrow M, Davey Smith G, Gray R. Light drinking in pregnancy and 4 mid-childhood mental health and learning outcomes. Arch Dis Child 2013;98(2):107-11 doi: 5 http://dx.doi.org/10.1136/archdischild-2012-302436[published Online First: Epub Date]|. 6 23. Alati R, Macleod J, Hickman M, et al. Intrauterine exposure to alcohol and tobacco use and 7 childhood IQ: findings from a parental-offspring comparison within the Avon Longitudinal 8 Study of Parents and Children. Pediatr Res 2008;64(6):659-66 doi: 9 http://dx.doi.org/10.1203/PDR.0b013e318187cc31[published Online First: Epub Date]|. 10 24. Lipsitch M, Tchetgen ET, Cohen T. Negative controls: a tool for detecting confounding and bias in 11 12 observational studies. Epidemiology (Cambridge, Mass.) 2010;21(3):383 13 25. Alati R, Davey Smith G, Lewis SJ, et al. Effect of prenatal alcohol exposure on childhood academic 14 outcomes: contrasting maternal and paternal associations in the ALSPAC study. PLoS ONE 15 2013;For8(10):e74844 peer doi: http://dx.doi.org/10.1371/journal.pone.0074844[published review only Online 16 First: Epub Date]|. 17 26. Lundsberg LS, Illuzzi JL, Belanger K, Triche EW, Bracken MB. Low-to-moderate prenatal alcohol 18 consumption and the risk of selected birth outcomes: a prospective cohort study. Ann 19 Epidemiol 2015;25(1):46-54.e3 doi: 20 http://dx.doi.org/10.1016/j.annepidem.2014.10.011[published Online First: Epub Date]|. 21 27. Nykjaer C, Alwan NA, Greenwood DC, et al. Maternal alcohol intake prior to and during 22 pregnancy and risk of adverse birth outcomes: evidence from a British cohort. J Epidemiol 23 Community Health 2014;68(6):542-9 doi: http://dx.doi.org/10.1136/jech-2013- 24 25 202934[published Online First: Epub Date]|. 26 28. Niclasen J, Nybo Andersen AM, Teasdale TW, Strandberg-Larsen K. Prenatal exposure to alcohol, 27 and gender differences on child mental health at age seven years. J Epidemiol Community 28 Health 2014;68(3):224-32 doi: http://dx.doi.org/10.1136/jech-2013-202956[published 29 Online First: Epub Date]|. 30 29. Miyake Y, Tanaka K, Okubo H, Sasaki S, Arakawa M. Alcohol consumption during pregnancy and 31 birth outcomes: the Kyushu Okinawa Maternal and Child Health Study. BMC Pregnancy 32 Childbirth 2014;14:79 doi: http://dx.doi.org/10.1186/1471-2393-14-79[published Online 33 First: Epub Date]|. 34 30. McCarthy FP, O'Keeffe LM, Khashan AS, et al. Association between maternal alcohol http://bmjopen.bmj.com/ 35 consumption in early pregnancy and pregnancy outcomes. Obstetrics & Gynecology 36 2013;122(4):830-37 37 38 31. Robinson M, Oddy WH, McLean NJ, et al. Low-moderate prenatal alcohol exposure and risk to 39 child behavioural development: a prospective cohort study. Bjog 2010;117(9):1139-50 doi: 40 http://dx.doi.org/10.1111/j.1471-0528.2010.02596.x[published Online First: Epub Date]|. 41 32. Jaddoe VW, Bakker R, Hofman A, et al. Moderate alcohol consumption during pregnancy and the risk of low birth weight and preterm birth. The generation R study. Ann Epidemiol

42 on October 2, 2021 by guest. Protected copyright. 43 2007;17(10):834-40 44 33. Sayal K, Heron J, Golding J, Emond A. Prenatal alcohol exposure and gender differences in 45 childhood mental health problems: a longitudinal population-based study. Pediatrics 46 2007;119(2):e426-34 47 34. Albertsen K, Andersen AM, Olsen J, Gronbaek M. Alcohol consumption during pregnancy and the 48 risk of preterm delivery. Am J Epidemiol 2004;159(2):155-61 49 35. Lundsberg LS, Bracken MB, Saftlas AF. Low-to-moderate gestational alcohol use and intrauterine 50 51 growth retardation, low birthweight, and preterm delivery. Ann Epidemiol 1997;7(7):498- 52 508 53 36. Passaro KT, Little RE, Savitz DA, Noss J. The effect of maternal drinking before conception and in 54 early pregnancy on infant birthweight. The ALSPAC Study Team. Avon Longitudinal Study of 55 Pregnancy and Childhood. Epidemiology 1996;7(4):377-83 56 57 58 59 60 27

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 28 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 37. Shu XO, Hatch MC, Mills J, Clemens J, Susser M. Maternal smoking, alcohol drinking, caffeine 4 consumption, and fetal growth: results from a prospective study. Epidemiology 5 1995;6(2):115-20 6 38. Olsen J, Pereira AdC, Olsen SF. Does maternal tobacco smoking modify the effect of alcohol on 7 fetal growth? American journal of public health 1991;81(1):69-73 8 39. Brooke OG, Anderson HR, Bland JM, Peacock JL, Stewart CM. Effects on birth weight of smoking, 9 alcohol, caffeine, socioeconomic factors, and psychosocial stress. Bmj 1989;298(6676):795- 10 801 11 12 40. Andersen A-MN, Andersen PK, Olsen J, Grønbæk M, Strandberg-Larsen K. Moderate alcohol 13 intake during pregnancy and risk of fetal death. Int J Epidemiol 2012;41(2):405-13 14 41. Windham GC, Von Behren J, Fenster L, Schaefer C, Swan SH. Moderate maternal alcohol 15 consumptionFor and riskpeer of spontaneous review abortion. Epidemiology only 1997;8(5):509-14 16 42. Ogston SA, Parry GJ. EUROMAC. A European concerted action: maternal alcohol consumption 17 and its relation to the outcome of pregnancy and child development at 18 months. Results-- 18 strategy of analysis and analysis of pregnancy outcome. Int J Epidemiol 1992;21 Suppl 1:S45- 19 71 20 43. Bille C, Olsen J, Vach W, et al. Oral clefts and life style factors--a case-cohort study based on 21 prospective Danish data. Eur J Epidemiol 2007;22(3):173-81 22 44. Ernhart CB, Sokol RJ, Ager JW, Morrow-Tlucak M, Martier S. Alcohol-related birth defects: 23 assessing the risk. Ann N Y Acad Sci 1989;562:159-72 24 25 45. Faebo Larsen R, Hvas Mortensen L, Martinussen T, Nybo Andersen AM. Determinants of 26 developmental coordination disorder in 7-year-old children: a study of children in the Danish 27 National Birth Cohort. Dev Med Child Neurol 2013;55(11):1016-22 doi: 28 http://dx.doi.org/10.1111/dmcn.12223[published Online First: Epub Date]|. 29 46. Parry GJ, Ogston SA. EUROMAC. A European concerted action: maternal alcohol consumption 30 and its relation to the outcome of pregnancy and child development at 18 months. Results-- 31 child development at age 18 months. Int J Epidemiol 1992;21 Suppl 1:S72-8 32 47. NHS Choices.Can I drink alcohol if I’m pregnant? 33 www.nhs.uk/chq/Pages/2270.aspx?CategoryID=54#close. 2014 34 48. Galobardes B, Shaw M, Lawlor DA, Lynch JW, Smith GD. Indicators of socioeconomic position http://bmjopen.bmj.com/ 35 (part 1). Journal of epidemiology and community health 2006;60(1):7-12 36 49. Kelly Y, Sacker A, Gray R, Kelly J, Wolke D, Quigley MA. Light drinking in pregnancy, a risk for 37 38 behavioural problems and cognitive deficits at 3 years of age? Int J Epidemiol 39 2009;38(1):129-40 doi: http://dx.doi.org/10.1093/ije/dyn230[published Online First: Epub 40 Date]|. 41 50. Bradley RH, Corwyn RF. Socioeconomic status and child development. Annual review of psychology 2002;53(1):371-99

42 on October 2, 2021 by guest. Protected copyright. 43 51. Phung H, Bauman A, Nguyen T, Young L, Tran M, Hillman K. Risk factors for low birth weight in a 44 socio-economically disadvantaged population: parity, marital status, ethnicity and cigarette 45 smoking. Eur J Epidemiol 2003;18(3):235-43 46 52. Morrison A, Polisena J, Husereau D, et al. The effect of English-language restriction on systematic 47 review-based meta-analyses: a systematic review of empirical studies. International journal 48 of technology assessment in health care 2012;28(02):138-44 49 53. Moher D, Pham B, Lawson M, Klassen T. The inclusion of reports of randomised trials published 50 51 in languages other than English in systematic reviews. Health Technol Assess 2003;7(41):1- 52 90 53 54. Klassen TP, Lawson ML, Moher D. Language of publication restrictions in systematic reviews gave 54 different results depending on whether the intervention was conventional or 55 complementary. Journal of clinical epidemiology 2005;58(8):769-76. e2 56 55. Jüni P, Holenstein F, Sterne J, Bartlett C, Egger M. Direction and impact of language bias in meta- 57 analyses of controlled trials: empirical study. Int J Epidemiol 2002;31(1):115-23 58 59 60 28

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 29 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 56. Richmond RC, Al-Amin A, Smith GD, Relton CL. Approaches for drawing causal inferences from 4 epidemiological birth cohorts: a review. Early human development 2014;90(11):769-80 doi: 5 10.1016/j.earlhumdev.2014.08.023[published Online First: Epub Date]|. 6 57. Burgess S, Davies NM, Thompson SG. Instrumental variable analysis with a nonlinear exposure- 7 outcome relationship. Epidemiology 2014;25(6):877-85 doi: 8 10.1097/ede.0000000000000161[published Online First: Epub Date]|. 9 58. Ades A, Sutton A. Multiparameter evidence synthesis in epidemiology and medical decision- 10 making: current approaches. Journal of the Royal Statistical Society: Series A (Statistics in 11 12 Society) 2006;169(1):5-35 13 59. O'Keeffe LM, Kearney PM, McCarthy FP, et al. Prevalence and predictors of alcohol use during 14 pregnancy: findings from international multicentre cohort studies. BMJ open 15 2015;For5(7):e006323 peer review only 16 60. Mather M, Wiles K, O’Brien P. Should women abstain from alcohol throughout pregnancy? 2015 17 61. Shah NR, Bracken MB. A systematic review and meta-analysis of prospective studies on the 18 association between maternal cigarette smoking and preterm delivery. American journal of 19 obstetrics and gynecology 2000;182(2):465-72 20 21 22 23 24 25 26 27 28 29 30 31 32 33

34 http://bmjopen.bmj.com/ 35 36 37 38 39 40 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 29

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 30 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Figure 1. Flowchart of search strategy including primary reasons for article exclusion. 4 5 Figure 2. Pooled mean difference for birthweight comparing low alcohol consumption (up to 6 32g/week) with no alcohol consumption (7 studies). ‘Adjusted’ refers to adjusted for both smoking 7 and a measure of socio-economic status. CI: Confidence intervals. 8 9 Figure 3a) Odd Ratios for preterm birth comparing low alcohol consumption (up to 32g/week) with 10 11 no alcohol consumption (9 studies); 3b) Odd Ratios for small for gestational age comparing low 12 alcohol consumption (up to 32g/week) with no alcohol consumption (7 studies) 3c) Odd Ratios for 13 low birthweight comparing low alcohol consumption (up to 32g/week) with no alcohol consumption 14 (6 studies). OR: Odds ratio, CI: Confidence intervals. Pooled OR includes both adjusted and 15 For peer review only 16 unadjusted estimates from studies, ‘Adjusted’ refers to adjusted for both smoking and a measure of 17 socio-economic status. 18 19 Supplementary figure 1a) Odds ratios for externalising behaviour comparing low alcohol 20 consumption (up to 32g/week) with no alcohol consumption (3 studies); 1b) Odds ratios for 21 internalising behaviour comparing low alcohol consumption (up to 32g/week) with no alcohol 22 23 consumption (2 studies); OR: Odds ratio, CI: Confidence intervals. 24 25 26 27 28 29 30 31 32 33

34 http://bmjopen.bmj.com/ 35 36 37 38 39 40 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 30 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 31 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 For peer review only 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33

34 http://bmjopen.bmj.com/ 35 36 37 38 39 40 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 Figure 1. Flowchart of search strategy including primary reasons for article exclusion.

48 143x183mm (300 x 300 DPI) 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 32 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 For peer review only 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 Figure 2. Pooled mean difference for birthweight comparing low alcohol consumption (up to 32g/week) with 32 no alcohol consumption (7 studies). ‘Adjusted’ refers to adjusted for both smoking and a measure of socio- 33 economic status. CI: Confidence intervals.

34 297x209mm (300 x 300 DPI) http://bmjopen.bmj.com/ 35 36 37 38 39 40 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 33 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 For peer review only 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 Figure 3a) Odd Ratios for preterm birth comparing low alcohol consumption (up to 32g/week) with no 32 alcohol consumption (9 studies); 3b) Odd Ratios for small for gestational age comparing low alcohol 33 consumption (up to 32g/week) with no alcohol consumption (7 studies) 3c) Odd Ratios for low birthweight comparing low alcohol consumption (up to 32g/week) with no alcohol consumption (6 studies). OR: Odds 34 ratio, CI: Confidence intervals. Pooled OR includes both adjusted and unadjusted estimates from studies, http://bmjopen.bmj.com/ 35 ‘Adjusted’ refers to adjusted for both smoking and a measure of socio-economic status. 36 37 297x209mm (300 x 300 DPI) 38 39 40 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 34 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 For peer review only 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 Supplementary figure 1a) Odds ratios for externalising behaviour comparing low alcohol consumption (up to 32 32g/week) with no alcohol consumption (3 studies); 1b) Odds ratios for internalising behaviour comparing 33 low alcohol consumption (up to 32g/week) with no alcohol consumption (2 studies); OR: Odds ratio, CI: Confidence intervals. 34 http://bmjopen.bmj.com/ 35 297x209mm (300 x 300 DPI) 36 37 38 39 40 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 35 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 PROSPERO International prospective register of systematic reviews 4 5 6 7 Systematic review of the effects of low-moderate prenatal alcohol exposure on 8 pregnancy and childhood outcomes 9 Loubaba Mamluk, Luisa Zuccolo, Theresa Moore, Alison Richards 10 11 12 Citation 13 Loubaba Mamluk, Luisa Zuccolo, Theresa Moore, Alison Richards. Systematic review of the effects of low-moderate 14 prenatal alcohol exposure on pregnancy and childhood outcomes. PROSPERO 2015:CRD42015015941 Available 15 from http://www.crd.york.ac.uk/PROSPERO_REBRANDING/display_record.asp?ID=CRD42015015941For peer review only 16 17 Review question(s) 18 To determine what is known about the effects of prenatal alcohol exposure, corresponding to low-to-moderate levels 19 20 of maternal consumption, on pregnancy outcomes. These include pregnancy complications, delivery outcomes and 21 Fetal Alcohol Syndrome (FAS) features. This will include the assessment of systematic reviews and meta-analyses. A 22 particular focus will be placed on identifying practical and meaningful outcomes of alcohol toxicity during 23 pregnancy. 24 25 Searches 26 Publications will be identified by searching the following major relevant databases: Medline, Embase, web of science 27 and Psychinfo. All databases will be searched from inception. Internet searches will be carried out using Google 28 Scholar. Attempts to identify further studies will be made by examining the reference lists of included studies and of 29 previous reviews. All studies to be restricted to those published in the English language only. 30 31 Types of study to be included 32 1) Prospective studies and systematic reviews/meta-analyses of prospective studies (cohort or case-control studies 33 nested in a cohort). 2) Natural experiments / studies using instrumental variables to improve casual inference, including Mendelian Randomization (MR) studies 3) Sibling comparison studies 4) Parental comparisons 34 http://bmjopen.bmj.com/ 35 36 Condition or domain being studied 37 Pregnancy and delivery outcomes as well as offspring outcomes (from the domains affected by Fetal Alcohol 38 Syndrome (FAS)) 39 40 Participants/ population 41 Pregnant women or women who are trying to become pregnant sampled from the general population. Cohorts of pregnant women with alcohol abuse/dependency will be excluded.

42 on October 2, 2021 by guest. Protected copyright. 43 44 Intervention(s), exposure(s) 45 Inclusion: lowto-moderate levels of prenatal alcohol consumption (up to 10.4 UK units or 83 g/week). 46 47 Exclusion: studies will be excluded if there was no quantitative measure of alcohol consumption that could be 48 converted to UK standard units or grams of alcohol and if there was insufficient data for an (adjusted and/or crude) 49 effect measure of lowmoderate consumption to be extracted. Cohorts of pregnant women with alcohol 50 abuse/dependency will be excluded. 51 52 Comparator(s)/ control 53 women with no or very sporadic alcohol consumption in pregnancy. 54 55 Outcome(s) 56 Primary outcomes 57 Outcomes: (in both children and adults) 58 59 60 Page: 1 / 5 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 36 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 1) Pregnancy complications 4 5 - Intra uterine growth restrictions (IUGR) 6 7 - Miscarriage 8 9 - Premature labour and birth- Gestational age 10 11 - Preeclampsia and gestational hypertension 12 13 - Low amniotic fluid (oligohydramnios) 14 15 - Gestational diabetesFor peer review only 16 17 - Placenta previa 18 19 2) Delivery outcomes 20 21 - Birth weight/ low birth weight/ small for gestational age (SGA) 22 - Still birth 23 24 - Delivery intervention (including vacuum extraction, forceps delivery, 25 26 Caesarean section) 27 28 - Apgar score 29 30 3) FAS features 31 32 - Facial malformation 33

34 - Growth restrictions (height- measurements of growth restriction) http://bmjopen.bmj.com/ 35 36 - Cranium size/ head circumference 37 38 - Developmental delays 39 40 - Behaviour complications 41 - Cognitive impairment / IQ

42 on October 2, 2021 by guest. Protected copyright. 43 44 - Attention scores / Attention deficit and hyperactivity disorder (ADHD) 45 46 Secondary outcomes 47 none 48 49 Data extraction, (selection and coding) 50 Selection of studies: 51 Titles and abstracts will be screened independently by one reviewer (a random selection of 20% will also be screened 52 by a second reviewer independently) with inclusion/exclusion being decided according to prespecified criteria. 53 Discrepancies will be discussed and disagreements resolved through consensus. The full-text of each of the articles 54 identified through screening of titles and abstracts will be obtained in order to determine their inclusion in the review. 55 56 57 Data extraction: 58 59 60 Page: 2 / 5 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 37 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Data extraction will be carried out using Microsoft Access. This will be piloted on a small selection of studies and 4 adjusted as necessary. Relevant data will be documented from each identified study including information on study 5 design and location, population characteristics, exposures studied (including timing of exposure), methods used to 6 ascertain exposures, outcomes studied, method of outcome ascertainment (including person reporting the outcome, 7 whether parent, teacher, health professional, researcher, child&), study results (from both unadjusted and fully 8 adjusted regressions), statistical adjustments etc. Data extraction will be carried out independently by one reviewer 9 and a random selection of 20% will checked by a second reviewer. Discrepancies will be resolved through discussion 10 or referral to a third reviewer. Where necessary, authors will be contacted for additional information. 11 12 Risk of bias (quality) assessment 13 Studies that did not adjust for smoking and maternal education/social class as potential confounders in their final 14 model will be considered to be of low evidence quality. 15 For peer review only 16 Strategy for data synthesis 17 The impact of low-to-moderate alcohol use on pregnancy and related outcomes will be investigated using high-low 18 methods of meta-analysis techniques; however, due to anticipated low number of studies for some outcomes, a formal 19 meta-analysis may not be appropriate for some of the outcomes. Where meta-analysis is not possible, a narrative 20 summary of findings will be carried out. Random-effect meta-analysis will be performed alongside fixed-effect in the 21 presence of high levels of between-studies heterogeneity (measured through I2). Item response theory (IRT) will be 22 used to combining results from different scales if required. The likelihood of small study bias deriving from 23 publication bias will further be assessed through drawing funnel plots. 24 25 Analysis of subgroups or subsets 26 Where the included number of studies in the meta-analysis is large enough, sub-group analyses will be performed for 27 1) studies adjusting for smoking and maternal education/social class; 2) studies reporting separately on the effects of 28 alcohol use in different gestational periods; 3) studies using different exposure/outcome assessments. 29 30 Dissemination plans 31 We anticipate dissemination to regional and national public health directors 32 33 Contact details for further information

34 http://bmjopen.bmj.com/ 35 Dr Mamluk 36 MRC Integrative Epidemiology Unit 37 38 School of Social and Community Medicine 39 40 University of Bristol 41

42 Oakfield House on October 2, 2021 by guest. Protected copyright. 43 44 Oakfield Grove 45 46 Bristol BS8 2BN 47 48 United Kingdom 49 50 [email protected] 51 52 Organisational affiliation of the review 53 none 54 55 Review team 56 Dr Loubaba Mamluk, University of Bristol, UK 57 Dr Luisa Zuccolo, University of Bristol, UK 58 Ms Theresa Moore, University of Bristol, UK 59 60 Page: 3 / 5 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 38 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Ms Alison Richards, University of Bristol, UK 4 5 Collaborators 6 Dr Luisa Zuccolo, University of Bristol, UK 7 Dr Sarah Lewis, University of Bristol, UK 8 Dr Abi Fraser, University of Bristol, UK 9 Professor Jenny Donovan, University of Bristol, UK 10 Professor George Davey Smith, University of Bristol, UK 11 12 Anticipated or actual start date 13 12 January 2015 14 15 Anticipated completionFor date peer review only 16 01 September 2015 17 18 Funding sources/sponsors 19 National Institute of Health Research, Medical Research Council, University of Bristol. UK 20 21 Conflicts of interest 22 None known 23 24 Language 25 English 26 27 Country 28 England 29 30 Subject index terms status 31 32 Subject indexing assigned by CRD 33 Subject index terms 34 http://bmjopen.bmj.com/ 35 Alcohol Drinking; Humans; Pregnancy Outcome; Prenatal Exposure Delayed Effects 36 37 Stage of review 38 Ongoing 39 40 Date of registration in PROSPERO 41 19 January 2015

42 on October 2, 2021 by guest. Protected copyright. 43 Date of publication of this revision 44 25 February 2015 45 46 DOI 47 10.15124/CRD42015015941 48 49 50 Stage of review at time of this submission Started Completed 51 Preliminary searches No Yes 52 Piloting of the study selection process No Yes 53 Formal screening of search results against eligibility criteria No No 54 Data extraction No No 55 Risk of bias (quality) assessment No No 56 Data analysis No No 57 58 59 60 Page: 4 / 5 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 39 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 PROSPERO 4 International prospective register of systematic reviews 5 The information in this record has been provided by the named contact for this review. CRD has accepted this information in good 6 faith and registered the review in PROSPERO. CRD bears no responsibility or liability for the content of this registration record, 7 any associated files or external websites. 8 9 10 11 12 13 14 15 For peer review only 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33

34 http://bmjopen.bmj.com/ 35 36 37 38 39 40 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page: 5 / 5 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from Page 40 of 51 6 7 6 6 7 7 6 6 4 1 7 7 13 6&7 table 1table table 1table Figure Figure 1 Page Page No Reported on Reported supplementary supplementary BMJ Open http://bmjopen.bmj.com/ Recommendation on October 2, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml For peer review only Description of relevance appropriatenessor of studies for assembled assessing the hypothesistested to be 9 9 to include Effort available studies,all including contact with authors 8 8 Search including strategy, time period included in the synthesis keyand words 7 7 Qualifications of searchers librarians and(eg, investigators) 6 6 Study population 5 5 of Type designsstudy used 4 4 of exposure Type or intervention used 3 3 Description of study outcome(s) 2 2 Hypothesis statement 1 1 Problem definition 17 16 Description of any contact with authors 15 ofhandling Method abstracts and unpublished studies 14 ofaddressing Method articles published in languages than other English 13 ofcitations List located those and excluded, including justification 12 hand Use (eg,of searching reference of obtainedlists articles) 11 Search used, software version, andname including special features (eg, used explosion) 10 Databases registriesand searched Item NoItem Reporting of methods includeshould Reporting of search Reporting of search strategy should include Reporting of background includeshould MOOSE Checklist for Meta-analyses of Observational Studies Observational of forMeta-analyses MOOSEChecklist 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

8 7 19 7&8 9&10 27-36 Table 1-3 Table table 2 82 table & Figures 2&3Figures Figures 2&3Figures Figures 2&3Figures supplementary BMJ Open http://bmjopen.bmj.com/ on October 2, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml For peer review only ationale forationale the selection ofcoding dataand (eg, sound principles clinical or regression on possibleregression predictorson results of study R Description of statistical(eg, methodscomplete descriptionof randomfixed or effects justification models, whether of the chosen models account for predictorsof study dose-responseresults, models, cumulativeor meta-analysis) in sufficientto detail be replicated Documentation of how dataand classifiedwere coded raters,multiple (eg, blinding and reliability)interrater convenience) appropriate) Assessment ofAssessmentquality, studyincluding blinding of assessors, quality stratification or Assessment ofAssessment confounding (eg, comparability of and cases controls where studies in 18 28 of Indication statistical ofuncertainty findings 27 ofsensitivity Results testing (eg, subgroup analysis) 26 giving Table descriptive information for each study included 25 Graphic summarizing individual study estimates andoverall estimate 24 ofappropriate Provision graphicsandtables 23 22 of Assessment heterogeneity 21 20 19 Reporting of resultsReporting of should include Page 41 of 51 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

BMJ Open Page 42 of 51

1 2 3 Supplementary Table 1. Search Strategy 1950 to 11·07·2016. MEDLINE, PsycINFO, EMBASE on Ovid; the Cochrane Library including CENTRAL (the Cochrane Central Database of Controlled 4 Trials) on Wiley Interscience; and Science Citation Index, Social Science Citation Index, on Web of Science. 5 1 exp pregnancy/ (719661) 6 2 Pregnant Women/ (5199) 7 8 3 preconception care/ or prenatal care/ (21644)

9 4 exp "embryonic and fetal development"/ (212127) 10 5 Fetus/ (68424) 11 For peer review only 6 exp Pregnancy Complications/ (343999) 12 13 7 (prepregnan$ or conception or preconception or pregnan$ or prenatal$ or pre-natal$ or 14 f?etal or f?etus 15 or in utero).ti,ab. (588557) 16 8 Maternal exposure/ (5534)

17 http://bmjopen.bmj.com/ 9 or/1-8 (1079283) 18 19 10 Ethanol/ (73449)

20 11 Alcohol dehydrogenase/ (5809) 21 12 Aldehyde Oxidoreductases/ (3672) 22 13 exp Drinking Behavior/ (57821) 23 24 14 Temperance/ (2430) 25 15 alcohol$.ti. (106954) on October 2, 2021 by guest. Protected copyright. 26 16 (alcohol dehydrogenase or acetaldehyde dehydogenase).ti,ab. (8557) 27 28 17 ((alcohol or alcoholic) adj3 (drink$ or exposure or consumption or consume$ or consuming 29 or low or light or moderat$ or abstin$ or abstain$)).ti,ab. (49470) 30 18 ((low or light or moderate or abstin$ or abstain$ or pattern$ or behavio?r$) adj3 drink$).ti,ab. 31 (10558) 32 19 (ADH1B$ or teetotal$ or temperance or nondrink$ or non-drink$).ti,ab. (3093) 33 34 20 Genome-Wide Association Study/ or Linkage Disequilibrium/ or genotype/ or phenotype/ or 35 polymorphism, genetic/ or (polymorphism$ or ((gene or genes or genetic or genotyp$) adj3 36 (instrument$ or variant$ or variable$ or variability or variabilities or variance$))).ti,ab. (494904) 37 21 *Alcohols/ or exp *Alcohol-Related Disorders/ or (alcohol adj3 (misuse or "use" or abuse or 38 addict$ or dependence or response$ or susceptibility)).ti,ab. (110843) 39 22 20 and 21 (3108)

40 23 or/10-19,22 (213050) 41 42 43 44 45 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

Page 43 of 51 BMJ Open

1 2 3 24 9 and 23 (11403) 4 25 letter/ (861500) 5 6 26 editorial/ (368458)

7 27 news/ (166329) 8 28 exp historical article/ (325965) 9 29 Anecdotes as topic/ (4586) 10 11 30 comment/ (610211) For peer review only

12 31 case report/ (1708277) 13 32 (letter or comment$).ti. (100442) 14 15 33 or/25-32 (3417468)

16 34 randomized controlled trial/ or Randomized Controlled Trials as Topic/ or random$.ti,ab. (888907)

17 http://bmjopen.bmj.com/ 35 33 not 34 (3386180) 18 36 animals/ not humans/ (3889478) 19 20 37 exp Animals, Laboratory/ (730783)

21 38 exp Animal Experimentation/ (6477) 22 39 exp Models, Animal/ not humans/ (310115) 23 40 exp rodentia/ (2688128) 24 25

41 (rat or rats or mouse or mice).ti. (1123907) on October 2, 2021 by guest. Protected copyright. 26 42 or/35-41 (7849766) 27 43 24 not 42 (6498) 28 29 44 meta-analysis/ (52850)

30 45 meta-analysis as topic/ (13933) 31 46 (meta analy$ or metaanaly$ or metanaly$ or meta regression).ti,ab. (72390) 32 33 47 ((systematic$ or evidence$ or realist or narrative or literature) adj2 (review$ or overview$)).ti,ab. 34 (171877) 35 48 "review of reviews".ti,ab. (211)

36 49 (reference list$ or bibliograph$ or hand search$ or manual search$ or relevant journals).ab. 37 (27094) 38 50 (search strategy or search criteria or systematic search or study selection or data extraction).ab. 39 40 (28877) 51 (search$ adj4 literature).ab. (30686) 41 42 43 44 45 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

BMJ Open Page 44 of 51

1 2 3 52 (medline or pubmed or cochrane or embase or psychlit or psyclit or psychinfo or cinahl or science 4 Citation index or bids or cancerlit).ab. (95294) 5 53 cochrane.jw. (11053) 6 54 ((multiple treatment$ or indirect or mixed) adj2 comparison).ti,ab. (991) 7 8 55 or/44-54 (293491) 9 56 43 and 55 (201) 10 57 epidemiologic studies/ (6077) 11 For peer review only 12 58 ep.fs. (1223653)

13 59 exp case control studies/ (691899) 14 60 exp cohort studies/ (1394149) 15 61 cross-sectional studies/ (185407) 16

17 62 Mendelian Randomization Analysis/ (229) http://bmjopen.bmj.com/ 18 63 (case control or negative control).ti,ab. (93097) 19 64 (cohort adj (study or studies or analys$)).ti,ab. (98537) 20 21 65 ((follow up or observational) adj (study or studies)).ti,ab. (87771)

22 66 ((longitudinal$ or retrospectiv$ or prospectiv$) and (study or studies or studied or review$ or 23 analys$ or cohort$)).ti,ab. (870699) 24 67 cross sectional.ti,ab. (183479) 25 68 (mendel$ or natural experiment$).ti,ab. (11082) on October 2, 2021 by guest. Protected copyright. 26 27 69 ((family or sibling or population) adj3 (study or studies)).ti,ab. (115245)

28 70 or/57-69 (2937160) 29 71 exp guideline/ (25834) 30 72 guidelines as topic/ or practice guidelines as topic/ (114690) 31 32 73 guideline$.mp. (291621) 33 74 or/71-73 (291621) 34 75 55 or 70 or 74 (3362878) 35 36 76 43 and 75 (3349)

37 77 limit 76 to english language (3096) 38 78 ((smok$ or drug$ or tobacco or nicotine or cigarette$ or substance$ or methamphetamine$ 39 40 or amphetamine$ or cocaine$ or heroin or cannabis or marijuana) not (alcohol or alcoholic or drink$)).ti. 41 (489662) 42 43 44 45 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

Page 45 of 51 BMJ Open

1 2 3 79 77 not 78 (2822) 4 5 6 7 8 9 10 11 For peer review only 12 13 14 15 16

17 http://bmjopen.bmj.com/ 18 19 20 21 22 23 24 25 on October 2, 2021 by guest. Protected copyright. 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

BMJ Open Page 46 of 51 Supplementary Table 2. Lists of confounders adjusted for by each study Study (year) Confounders 1 [26] Lundsberg (2015) Parity, maternal age, education, BMI, marital status, ethnicity, caffeine, smoking, exercise, work, prenatal and multivitamin use, passive smoke exposure, marijuana use, 2 cocaine use, study cohort, preterm labour, respiratory problem, infant gender, bleeding, nausea/vomiting, hypertension, incompetent cervix, placental problems, 3 sexually transmitted disease, induction/augmentation, maternal asthma, gestational diabetes. 4 [27] Nykjaer (2014) Maternal pre-pregnancy weight, height, age, parity, ethnicity, salivary cotinine levels, caffeine intake, education, energy intake, gestation and baby’s sex 5 [28] Niclasen (2014) Parental smoking, parental education, parental pre-pregnancy psychiatric diagnoses, and maternal psychological well-being in pregnancy. 6 [29] Miyake (2014) Low birth weight, preterm birth: maternal age, region of residence, number of children, family structure, maternal education, maternal employment, body mass index, 7 maternal smoking during pregnancy, and baby’s gender. 8 Small for gestational age: maternal age, region of residence, number of children, family structure, maternal education, maternal employment, body mass index, maternal 9 smoking during pregnancy, gestational age, and baby’s gender. 10 [45] Faebo Larsen Sex, gestational age, intrauterine growth restriction, maternal age, mother’s occupational status, maternal smoking (ever) in first trimester, amount of maternal smoking 11 (2013) and alcohol consumptionFor in first trimester. peer review only 12 [22] Sayal (2013) Maternal age, parity, highest level of maternal education, daily frequency of smoking, use of cannabis and/or other illicit drugs during the first trimester, 13 homeownership, whether currently married, high scores (>12) on the Edinburgh Postnatal Depression Scale, and child gestational age, birth weight and gender. 14 [30] McCarthy (2013) Maternal age, smoking, years of schooling, ethnicity, body mass index, infant sex, maternal status, family income, and drug use during pregnancy. Adjusted for clustering. 15 Birthweight adjusted for gestational age at delivery. 16 [40] Andersen (2012) Number of previous abortions, coffee consumption, changes in alcohol consumption since prior to pregnancy and smoking. Effect of coffee consumption and smoking was stratified according to period. The model is stratified according to maternal age and parity.

17 http://bmjopen.bmj.com/ 18 [21] Salihu (2011) Maternal age, parity, race, smoking, education, marital status, adequacy of prenatal care, maternal height, gender of the infant, and year of birth 19 [31] Robinson (2010) Maternal age, maternal education, presence of the biological father in the family home, family income, stress in pregnancy, child’s age at follow up (and child’s age at 20 follow-up squared), and maternal cigarette smoking. 21 [32] Jaddoe (2007) Controlled for maternal body mass index, smoking, educational level, height, ethnicity, parity and age and infant gender; birth weight and low birth weight models also 22 controlled for gestational age. [43] Bille (2007) Parental age and social class. 23 [33] Sayal (2007) Gender, smoking, cannabis use and use of illicit drugs in the first trimester, highest level of maternal education, home ownership, marital status, parity, maternal age 24 group, high EPDS score, child ethnicity, gestational age group, and birth weight. 25 [34] Albertsen (2004) Type 1 diabetes, age, previous preterm delivery, smoking during pregnancy, coffee consumption during pregnancy, occupational on October 2, 2021 by guest. Protected copyright. status in the household, parity, and total 26 alcohol consumption during pregnancy. 27 [35] Lundsberg (1997) Small for gestational age: smoking in month 7, ethnicity, weight, height, infant sex, parity, bleeding during pregnancy, high blood pressure, and preeclampsia/eclampsia. 28 Low birthweight: smoking in month 7, height, weight, ethnicity, infant sex, parity, coffee use in month 7, exercise in third trimester, employment, bleeding during 29 pregnancy, high blood pressure, pre-eclampsia/eclampsia, anomalies, and placental problems. Preterm delivery: smoking in month 7, height, parity, age, caffeine use in 30 month 7, exercise first 16 weeks, bleeding during pregnancy, high blood pressure, pre-eclampsia/eclampsia, anomalies, and placental problems. 31 [41] Windham (1997) Maternal age, prior spontaneous abortion, gestational age at interview, and cigarette and caffeine consumption in week before interview. 32 [36] Passaro (1996) Gestational age, infant sex, parity, maternal smoking, and maternal body mass index. 33 [37] Shu (1995) Gestational age, parity, smoking and income. 34 [20] Peacock (1995) Unadjusted 35 [38] Olsen (1991) Age, school education, parity, alcohol and smoking entered the model as “dummy variables”. 36 [42] Ogston (1992) Gestational age at birth, sex, mother’s age, parity and smoking. 37 [46] Parry (1992) Gestational age at birth, sex, mother’s age, parity and smoking. 38 [39] Brooke (1989) Gestational age, sex, maternal height, and parity 39 [44] Ernhart (1989) Parity, smoking, race and year of study. 40 41 42 43 44 45 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

Page 47 of 51 BMJ Open

1 2 3 4 5 Supplementary Table 3. Assessment of studies using the The Newcastle-Ottawa Scale (NOS) 6 7 8 Representative Selection of the Ascertainm Outcome of Comparability Assessment of Follow-up Adequacy Risk of bias 9 ness of the non-exposed ent of interest was not of cohorts on outcome blind/ long enough of follow questions 10 exposed cohort cohort exposure present at start of the basis of the record linkage for outcome to up of 11 For peer reviewstudy design/analysis only occur cohorts 12 Study ID 13 14 ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ 15 Albertson 2003 16

17 ✔ ✔ ✔ ✔ ✔ ✔ http://bmjopen.bmj.com/ ✔ ✔ Anderson 2012 18 19 ✔ ✔ ✔ ✔ ✔ ✔ ✔ ? 20 Bille 2007 21 ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ 22 Brooke 1989 23 24 x ✔ ✔ ✔ ? ✔ ✔ x 25 Erhart on October 2, 2021 by guest. Protected copyright. 26 ✔ ✔ x ✔ ✔ ✔ ✔ ✔ 27 Jaddoe 2007 28 ✔ ✔ ✔ ✔ ✔ x ✔ x 29 Larsen 2013 30 31 Lundsberg (Lisbet) ? ✔ ✔ ✔ ✔ ✔ ✔ ✔ 32 2015 33 Lundsberg (Lisbet) ? ✔ ✔ ✔ ✔ ✔ ✔ ✔ 34 1997 35 36 McCarthy 2013 ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ 37 38 Miyake 2014 ? ✔ x ✔ ✔ ✔ ✔ ✔ 39 40 41 42 43 44 45 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

BMJ Open Page 48 of 51

1 2 3 Niclasen 2013 ? ✔ ? ✔ x x ✔ ? 4 5 6 Nykjaer 2014 x ✔ x ✔ ✔ ✔ ✔ ✔ 7 8 Ogston 1992 ? ✔ ✔ ✔ x ✔ ✔ ✔ 9 10 Olsen 1991 ? ✔ x ✔ ✔ ✔ ✔ ✔ 11 For peer review only 12 13 Parry 1992 ? ✔ ✔ ✔ ✔ x ✔ x 14 15 Passaro 1996 ? ✔ x ✔ ✔ ✔ ✔ ✔ 16

17 http://bmjopen.bmj.com/ Peacock 1995 ? x ? 18 ✔ ✔ ✔ ✔ ✔ 19 20 Robinson 2010 ? ✔ x ✔ ✔ x ✔ ✔ 21 22 Salihu 2011 ? ✔ x ✔ ✔ ? ✔ ✔ 23 24 Sayal 2012 x ? 25 ✔ ✔ ✔ ✔ ✔ ✔ on October 2, 2021 by guest. Protected copyright. 26 27 Sayal 2006 ✔ ✔ x ✔ ✔ x ✔ ✔ 28 29 Shu 1995 ✔ ✔ ✔ ✔ x ✔ ✔ ✔ 30 31 Windham 1997 x 32 ✔ ✔ ✔ ✔ ✔ ✔ ✔ 33 34 35 36 37 38 39 40 41 42 43 44 45 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 6 Figure 1 Figure supTable supTable 1 6 5 & 5 6 5 5&6 4 3 1 on page # # on page Reported Reported up) and report up) and (e.g.,characteristics years considered, - BMJ Open http://bmjopen.bmj.com/ analysis). analysis). - on October 2, 2021 by guest. Protected copyright. Describe methodDescribe of data extraction reportsfrom (e.g., forms, piloted independently,induplicate) and any processes for andobtaining confirming data investigators.from State the process theState for selecting studies (i.e., screening, eligibility, review, included in systematic and, ifapplicable, included inmeta the Present full electronic full Present strategy for search leastone at includingdatabase, any limits used,suchthat it could be Describe allinformationDescribe (e.g., databasessources with of dates coverage, contact study authors with to identify Specify studySpecify characteristics(e.g., PICOS,length of follow Indicate reviewa Indicate if protocol if andexists, where accessedWebit can be (e.g., and, available,address), if provide Provide an Provide explicit statement of questions being withaddressed reference to participants, interventions, comparisons, outcomes, anddesign study (PICOS). Describe the rationale theDescribe review for thein the of context known.already what is Provide a Provide structured including,summary as applicable:background; dataobjectives; sources; study eligibilitycriteria, andparticipants, interventions; appraisalstudy and synthesis methods; results;limitations; conclusions and Identify the report Identify a as systematicmeta-analysis, review, both. or Checklist Checklist item language, publicationlanguage, status) used forcriteriaas eligibility, rationale. giving repeated. repeated. additional additional in studies)the search dateand last searched. registration registration information registrationincluding number. implications of of implicationskey findings; systematic review number. registration For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

1 # 8 2 4 9 7 3 For peer review6 only

Data Data collection process 10 Study Study selection Search Search Information Information sources Eligibility Eligibility criteria Protocol andProtocol registration 5 METHODS METHODS Objectives Rationale Rationale INTRODUCTION INTRODUCTION Structured summary ABSTRACT ABSTRACT Title TITLE TITLE Section/topic Section/topic Checklist 2009 PRISMA Page 49 of 51 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

Page 50 of 51 Figure Figure 2&3 13-14 Figure Figure 2&3 supfigure1 Figure Figure 2&3 supfigure1 7 Table 1-3 Table 1-3 Figure 1 Figure 8 7 Figure Figure 2&3 supfigure1 Figure Figure 2&3 supfigure1 6 & 6 7 7 & 7 8 BMJ Open http://bmjopen.bmj.com/ analysis. analysis. - specified. specified. - for metafor each on October 2, 2021 by guest. Protected copyright. ) 2 reporting within reporting studies). Give results results ofadditionalGive analyses,ifdone (e.g., sensitivity or subgroup analyses,meta-regression Item [see 16]). Present results ofPresent any assessment of risk acrossof bias studies (see Item 15). Present results ofPresent meta-analysis each done, including confidence intervalsand measuresof consistency. For For all outcomes considered (benefits harms), orfor present, each (a) study: simple summary data for each Present data Present onrisk eachbias studyof of and, if available,any outcome level assessment (see 12). item For each For each presentstudy, characteristicswhich for data were (e.g., studyextracted PICOS, follow-upsize, period) and the provide citations. Give numbersofGive studies assessedscreened, foreligibility, included inand review, the with reasons for exclusions at Describe methodsDescribe of additional analyses (e.g., sensitivity subgroup or analyses, meta-regression), done, if indicating prewerewhich Specify any Specify assessment bias of risk maythat of affect the cumulative (e.g.,evidence publication bias,selective Describe the methods theDescribe of handling data and combining results of studies, if done, including measures of consistency I (e.g., State the principal theState measuressummary (e.g.,risk ratio, in difference means). Describe methodsDescribe used for riskassessing bias individual of of studies (including specification of whetherthis was at the donestudy outcomeor level), howand information this to be is usedany in data synthesis. List and define and List all variables for which data were sought (e.g., PICOS, funding sources)and any assumptions and made.simplifications intervention group (b) intervention group andeffectestimates confidence intervals,with a ideally plot.forest each with each stage, flowa ideally diagram. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

12 11 21 14 17 16 For peer review only 23

Risk of bias Risk acrossof studies 22 Synthesis resultsSynthesis of Results ofindividualResults studies 20 Risk of bias Risk withinof studies 19 Study Study characteristics 18 Study Study selection RESULTS RESULTS Risk of bias Risk acrossof studies 15 Synthesis resultsSynthesis of Summary measuresSummary 13 Risk of bias Risk inof individual studies Data items Data items Additional Additional analysis Additional Additional analyses 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

17 13 13 & 13 14 13&14 13&14 supfigure1 BMJ Open http://bmjopen.bmj.com/ on October 2, 2021 by guest. Protected copyright. Describe Describe offundingsources forthe systematic review (e.g., other supportand supply ofdata); ofrole the funders for Provide a Provide general interpretation of thethe results in context of otherevidence,and forimplications future research. Discuss limitationsDiscuss at and study (e.g.,outcome level andrisk of bias), review-level (e.g.,incompleteat retrieval of research,reportingidentified bias). Summarize Summarize the main findings theincluding strength of evidence for outcome;maineach theirconsider relevance to systematic systematic review. key groups (e.g.,key healthcare providers, users, policyand makers). For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

25 27 For26 peer review only

Funding Funding FUNDING FUNDING Conclusions Limitations Summary of of Summary evidence 24 DISCUSSION DISCUSSION Page 51 of 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

Low alcohol consumption and pregnancy and childhood outcomes: time to change guidelines indicating apparently ‘safe’ levels of alcohol during pregnancy? A systematic review and meta-analyses.

For peer review only Journal: BMJ Open

Manuscript ID bmjopen-2016-015410.R2

Article Type: Research

Date Submitted by the Author: 18-May-2017

Complete List of Authors: Mamluk, Loubaba; University of Bristol School of Social and Community Medicine, IEU Edwards, Hannah; University of Bristol, School of Social and Community Medicine Savovic, Jelena; University of Bristol, NIHR CLAHRC West Leach, Verity; University of Bristol Jones, Tim; NIHR CLAHRC West, Moore, Theresa; NIHR CLAHRC West Ijaz , Sharea ; NIHR CLAHRC West Lewis, Sarah; University of Bristol, School of Social and Community Medicine Donovan, Jenny; NIHR CLAHRC West Lawlor, Debbie; Department of Social Medicine, University of Bristol, MRC http://bmjopen.bmj.com/ Integrative Epidemiology Unit Davey Smith, George; University of Bristol, Social Medicine Fraser, Abigail; University of Bristol, MRC Integrative Epidemiology Unit Zuccolo, Luisa; University of Bristol, School of Social and Community Medicine

Primary Subject Epidemiology Heading: on October 2, 2021 by guest. Protected copyright. Secondary Subject Heading: Reproductive medicine

Keywords: PAEDIATRICS, EPIDEMIOLOGY, Maternal medicine < OBSTETRICS

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 1 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Systematic Review & Metaanalysis 4 5 6 7 Title: Low alcohol consumption and pregnancy and childhood outcomes: time to change 8 9 guidelines indicating apparently ‘safe’ levels of alcohol during pregnancy? A systematic review 10 11 and metaanalyses. 12 13 14 15 Running tittle:For Low alcohol peer consumption reviewand pregnancy and childhood only outcomes. 16 17 18 19 Authors: Loubaba Mamluk PhD1,2,3, Hannah B Edwards MSc2,3, Jelena Savović PhD2,3, Verity 20 21 Leach PhD 2,3, Timothy Jones PhD 2,3, Theresa HM Moore MSc 2,3, Sharea Ijaz PhD 2,3, Sarah 22 23 J. Lewis PhD 2, Jenny L. Donovan PhD 2,3, Debbie Lawlor PhD 1,2,3, George Davey Smith PhD 24 25 1,2, Abigail Fraser PhD*1,2, Luisa Zuccolo PhD*1,2 26 27 * Joint Last Authors 28 29 30 31 32 Author addresses: Hannah B Edwards, Jelena Savovic, Verity Leach, Timothy Jones, 33

34 Theresa HM Moore, Sharea Ijaz, and Jenny L. Donovan: NIHR CLAHRC West, University http://bmjopen.bmj.com/ 35 36 Hospitals Bristol, NHS Foundation Trust, 9th Floor, Whitefriars, Lewins Mead, Bristol, BS1 37 38 2NT, United Kingdom. Sarah J. Lewis, Debbie Lawlor, George Davey Smith, Abigail Fraser, 39 40 and Luisa Zuccolo: MRC Integrative Epidemiology Unit, School of Social and Community 41

42 Medicine, University of Bristol, Oakfield House, Oakfield Grove, Bristol BS8 2BN, United on October 2, 2021 by guest. Protected copyright. 43 44 Kingdom. 45 46 47 Corresponding author: Loubaba Mamluk, MRC Integrative Epidemiology Unit, School of 48 49 Social and Community Medicine, University of Bristol, Oakfield House, Oakfield Grove, Bristol 50 51 BS8 2BN, United Kingdom. NIHR CLAHRC West, University Hospitals Bristol 52 53 NHS Foundation Trust, 9th Floor, Whitefriars, Lewins Mead Bristol, BS1 2NT. United Kingdom 54 55 T+44 (0)117 3313378, [email protected] 56 57 1 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 2 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Author affiliations: 4 5 6 1 MRC Integrative Epidemiology Unit at the University of Bristol, Bristol 7 8 2 School of Social and Community Medicine, University of Bristol, Bristol 9 10 3 NIHR CLAHRC West, University Hospitals Bristol NHS Foundation Trust, Bristol 11 12 Word Count: 3802 13 Figures: 3 14 Tables: 3 15 SupplementaryFor Tables: 3peer review only 16 Supplementary Figures:1 17 Supplementary Document:1 18 Number of references: 61 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33

34 http://bmjopen.bmj.com/ 35 36 37 38 39 40 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 2 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 3 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Abstract: 4 5 Objectives: To determine the effects of lowtomoderate levels of maternal alcohol 6 7 consumption in pregnancy on pregnancy and longerterm offspring outcomes. 8 9 Search Strategy: Medline, Embase, Web of Science, and Psychinfo from inception to 1107 10 11 2016. 12 13 Selection Criteria: Prospective observational studies, negative control and quasi 14 15 experimental studiesFor of pregnant peer women estimatingreview effects of lightonly drinking in pregnancy 16 17 (≤32g/week) versus abstaining. Pregnancy outcomes such as birth weight, and features of 18 19 fetal alcohol syndrome were examined. 20 21 Data Collection and Analysis: One reviewer extracted data and another checked extracted 22 23 data. Random effects metaanalyses were performed where applicable, and a narrative 24 25 summary of findings was carried out otherwise. 26 27 Main Results: 24 cohort and two quasiexperimental studies were included. With the 28 29 exception of birth size and gestational age, there was insufficient data to metaanalyse or 30 31 make robust conclusions. Odds of smallforgestationalage (SGA) and preterm birth were 32 33 higher for babies whose mothers consumed up to 32g/week versus none, but estimates for

34 http://bmjopen.bmj.com/ 35 preterm birth were also compatible with no association: summary odd ratios (OR) 1—08, 95% 36 37 confidence intervals (CI) (1—02 to 1—14), I2 0%, (7 studies, all estimates were adjusted) OR 38 39 2 40 1—10, 95%CI (0—95 to1—28), I 60%, (9 studies, includes one unadjusted estimates) 41 respectively. The earliest time points of exposure were used in the analysis.

42 on October 2, 2021 by guest. Protected copyright. 43 44 Conclusion: Evidence of the effects of drinking <=32g/w in pregnancy is sparse. As there 45 46 was some evidence that even light prenatal alcohol consumption is associated with being 47 48 SGA and preterm delivery, guidance could advise abstention as a precautionary principle, but 49 50 should explain the paucity of evidence. 51 52 Keywords: Alcohol, pregnancy, systematic review. 53 54 55 56 57 3 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 4 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 4 5 Strengths and Limitations of this study: 6 7 Strengths 8 9 Completeness of searches with a focused research question aimed at informing alcohol in 10 11 pregnancy guidelines. 12 13 Biases minimised by only including those with prospective assessment of exposure and 14 15 For peer review only 16 prioritising results adjusted for main confounders. 17 18 Unique effort to include alternative study designs to further improve causal inference 19 20 alongside standard analytical approaches. 21 22 Limitations 23 24 Limitation of results on the effects of light drinking in pregnancy from standard analytical 25 26 approaches is bias due to residual confounding. 27 28 The inclusion of only English language studies may have led to missing some studies, 29 30 however there is little evidence that exclusion of nonEnglish language studies leads to 31 32 systematic bias in systematic reviews of conventional medicine. 33

34 We could not pool eligible studies for various reasons (e.g. too few studies, lack of http://bmjopen.bmj.com/ 35 36 standard errors) 37 38 39 40 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 4 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 5 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Keywords: Alcohol, pregnancy, systematic review. 4 5 Introduction 6 7 Alcohol is a known teratogen[1] and the evidence about the risks of heavy alcohol 8 9 consumption during pregnancy on intellectual ability, birth defects, behaviour, fine motor 10 11 skills, and mental health (comprising fetal alcohol spectrum disorder – FASD)[2] is clear and 12 13 compelling.[3] Internationally, clinical guidelines recommend that pregnant women should 14 15 abstain fromFor heavy or “binge” peer drinking.[4] review However, until recently only UK guidelines advised 16 17 women to avoid drinking alcohol while trying to conceive, and in the first trimester, but at the 18 19 same time indicated that consumption should be restricted to within “1 to 2 UK units, once or 20 21 twice a week”.[5] The UK Chief Medical Officer (CMO) commissioned a review of guidelines 22 23 on alcohol consumption during pregnancy. Based on a review of reviews, the Guidelines 24 25 Development Expert Group has recently proposed a change to guidelines such that women 26 27 should be advised to abstain from alcohol when pregnant and/or trying to conceive,[6] based 28 29 on the precautionary principle (i.e. “better safe than sorry”), in the absence of robust 30 31 evidence. 32 33

34 Our aim was to conduct a comprehensive systematic review and metaanalysis of the http://bmjopen.bmj.com/ 35 36 literature to determine the effects of lowtomoderate levels of maternal alcohol consumption 37 38 in pregnancy on pregnancy and longer term offspring outcomes. Here we report on alcohol 39 40 consumption of up to two UK units of alcohol up to twice a week (the equivalent of ~ 41

42 32g/week), compared to no alcohol. In the absence of evidence from randomised controlled on October 2, 2021 by guest. Protected copyright. 43 44 trials, we examine observational studies of pregnant women from the general population with 45 46 prospective assessment of alcohol exposure, to reduce recall bias. In particular, we 47 48 specifically seek out quasiexperimental studies, negative control comparisons, and 49 50 Mendelian randomisation analyses in order to reduce the impact of confounding and 51 52 measurement error on the effect estimates. 53 54 55 56 57 5 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 6 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Methods 4 5 Selection strategy and selection criteria 6 A full protocol of this systematic review carried out using PRISMA guidelines (supplementary 7 8 document)[7] is available from the PROSPERO systematic review register (registration 9 10 number CRD4201501594); 11 12 http://www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42015015941). 13 14 In brief, eligible studies were defined as epidemiological studies of pregnant women or 15 For peer review only 16 women trying to conceive with prospective assessment of prenatal alcohol exposure (i.e. 17 18 19 before birth), sampled from general population. The protocol specifically included studies 20 21 using standard analytical approaches (e.g. multivariable regression analysis), as well as 22 23 studies that used innovative analytical methods to improve causal inference, such as (i) 24 25 quasiexperimental studies (for example comparing outcomes before and after 26 27 implementation of new guidelines on alcohol consumption); (ii) negative control studies (e.g. 28 29 comparing the association of offspring outcomes with maternal alcohol consumption to the 30 31 association of the same outcomes with consumption among fathers, under the assumption 32 33 that confounding is likely to be similar but that if there was a direct causal effect of maternal

34 http://bmjopen.bmj.com/ 35 consumption on outcomes, maternal associations would be stronger); and Mendelian 36 37 randomisation studies (using genetic variants associated with alcohol consumption and 38 39 metabolism). We considered these analytical approaches to be the most appropriate in 40 41 terms of their ability to minimise bias from confounding and other sources. Our original

42 on October 2, 2021 by guest. Protected copyright. 43 protocol included studies exploring the effects of prenatal alcohol consumption up to 44 45 83g/week (the commonly used threshold for moderate consumption [810]) versus 46 47 abstinence. Here we have focused specifically on low alcohol consumption, i.e. up to 48 49 32g/week as this was the cutoff specified by the UK guidelines at the time of writing this 50 51 review as being an implicitly “safe” threshold.[5] This specific cut off value has not been 52 53 reviewed and is the main point of discussion as the guideline change from low consumption 54 55 56 57 6 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 7 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 (equating to 1 to 2 UK units, once or twice a week or 32g/week) to abstinence (reference 4 5 group). 6 7 Outcomes included: 1) pregnancy outcomes: still birth (pregnancy loss after week 24; 8 9 miscarriage; gestational length and preterm delivery (<37 weeks gestation); hypertensive 10 11 disorders of pregnancy; gestational diabetes; small for gestational age (SGA), < 10th 12 13 percentile in weight or <−2 standard deviation scores) and birth size (weight (including low 14 15 birth weight definedFor as <2500g),peer length, reviewand head circumference); only low amniotic fluid 16 17 (oligohydramnios); placenta previa; placental abruption; assisted delivery (including vacuum 18 19 extraction, forceps delivery, Caesarean section); Apgar score at birth; admission to neonatal 20 21 unit; congenital malformations. 2) Features of foetal alcohol spectrum disorder (FASD): 22 23 childhood growth restriction; cranium size and head circumference; developmental delays; 24 25 behaviour problems; cognitive impairment and intelligent quotient (IQ); facial malformations. 26 27 We adopted study specific definitions for all outcomes. 28 29 30 Studies were excluded if: there was no quantitative measure of alcohol consumption that 31 32 could be converted to grams of alcohol/week; there was insufficient data to estimate the 33

34 effect size of the association of our predefined low consumption categories versus http://bmjopen.bmj.com/ 35 36 abstinence with any outcome, including studies that analysed alcohol as a continuous 37 38 variable (i.e. assuming the same linear or log linear effect across the entire alcohol 39 40 distribution); the lowest exposure category (compared to nondrinkers) had an upper bound 41

42 on October 2, 2021 by guest. Protected copyright. exceeding 32g/week, or was unspecified; they were cohort studies of pregnant women with 43 44 alcohol abuse/dependency; they were casecontrol studies or cohort studies with 45 46 47 retrospective alcohol consumption assessment (e.g. after birth). 48 49 The following databases were searched: MEDLINE, PsycINFO, EMBASE on Ovid; the 50 51 Cochrane Library including CENTRAL (the Cochrane Central Database of Controlled Trials) 52 53 on Wiley Interscience; and Science Citation Index, Social Science Citation Index, on Web of 54 55 Science from inception to 11 July 2016 (supplementary Table 1). We limited the search to 56 57 7 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 8 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 English language papers and excluded animal studies, letters, editorials, and conference 4 5 proceedings for which there were no fulltext papers. Searches were tailored to each 6 7 database by investigators. The search focused on published medical literature and did not 8 9 include grey literature. We additionally performed manual searches of the reference lists of: 10 11 (i) all papers included in recent systematic reviews of the effects of prenatal alcohol 12 13 exposure on the outcomes of interest; and (ii) all recent papers citing those reviews. 14 15 Titles and abstracts,For and peer full texts if necessary, review were screened onlyindependently by two 16 17 reviewers. Discrepancies were discussed and disagreements resolved through consensus. 18 19 We assessed potential for bias in included studies by assessing how well the study adjusted 20 21 for several main confounders known to impact on the exposureoutcome associations 22 23 (socioeconomic positioning as measured by the individual study, smoking during pregnancy, 24 25 maternal age, and ethnicity). We considered the potential for confounding and bias across 26 27 studies included in the analyses and described it narratively alongside summary results. 28 29 30 Data extraction 31 32 Data were extracted into a custombuilt Microsoft Access database. We extracted the 33

34 following information from each study: title, authors, publication year, country/region, study http://bmjopen.bmj.com/ 35 36 design, population characteristics (sample size, methods of sampling, age distribution, and 37 38 ethnicity), measures of exposure (assessment method; including timing and quantification of 39 40 alcohol consumption, reference group (abstinence), exposure (e.g 12 units or 24 units), 41

42 and information on unit equivalence if stated), outcome assessment methods (including on October 2, 2021 by guest. Protected copyright. 43 44 whether this was abstracted from medical records, obtained via a research interview and the 45 46 person reporting the outcome e.g. parent, teacher, health professional, researcher or child), 47 48 model adjustments, and study results. If a study reported more than one result for each 49 50 outcome, we extracted all of them (e.g. relative to different timing of exposure, model 51 52 adjustments, etc.). Information from each included paper was extracted by one reviewer 53 54 (LM) and subsequently checked for accuracy and completeness by another reviewer 55 56 57 8 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 9 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 (HE).[11] Extraction errors were minimal and were resolved through discussion between 4 5 extractor and checker. 6 7 8 Alcohol unit conversion 9 10 Alcohol consumption in drinks/week was converted into grams/week based on the pure 11 12 ethanol equivalent of one drink, as stated in each individual article, or otherwise inferred 13 14 based on the definition of standard drinks in the country where the study took place. 15 For peer review only 16 Data analysis 17 18 The association of low alcohol use with pregnancy and related outcomes was investigated 19 20 comparing the highest category within the range of 032g/week to abstention (during 21 22 pregnancy). In studies providing data across several categories of intake within the 0 23 24 32g/week range, we used the effect estimate for the highest category of intake. If studies 25 26 reported on exposure to alcohol during different trimesters, we included estimates relative to 27 28 the earliest exposure. This is because for some outcomes, the first trimester tends to be the 29 30 most critical timing/window of exposure [12] [13] and because most studies that only 31 32 reported on one time point reported on exposure in early gestation. Similarly, if results were 33

34 available from both unadjusted and adjusted regressions, we prioritised fully adjusted http://bmjopen.bmj.com/ 35 36 results, as a way of minimising the impact of confounding by important factors such as 37 38 maternal smoking, age, socioeconomic position and ethnicity. In case of multiple results 39 40 from the same cohort (relative to the same outcome), we analysed those pertaining to the 41

42 largest population size (i.e. conducted on the least ‘selected’ population as result of on October 2, 2021 by guest. Protected copyright. 43 44 exclusions, to minimise selection bias). Results from all studies that fulfilled our inclusion 45 46 criteria were summarised, together with information about the study. Where appropriate, we 47 48 additionally pooled results for each outcome. Authors were not contacted for extra data. 49 50 Results from Different study designs have been reviewed separately. Individual study 51 52 estimates were pooled using random effects metaanalysis. Where only two studies were 53 54 available to metaanalyse, results were pooled unless they were very different from each 55 56 57 9 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 10 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 other (I2 ≥ 50%).[14] In this case, a narrative summary of findings was carried out and 4 5 results were reported in Table 2. Where a study only reported unadjusted results, we kept 6 7 these separate in the forest plots (subgroup analysis) but then also showed overall pooled 8 9 estimates combining all results. 10 11 Planned subgroup analyses by trimester could not be performed due to insufficient number 12 13 of included studies with this information. 14 15 Risk of bias Forassessment peer review only 16 17 The NewcastleOttawa Scale (NOS)[15] was used to assess risk of bias for included reports. 18 19 This is an eightitem questionnaire assessing the following: representativeness of the 20 21 exposed cohort; selection of the nonexposed cohort; exposure assessment methods; 22 23 absence of outcome (of interest) at the start of the study; comparability of exposed and non 24 25 exposed groups (with regard to confounding variables); blind assessment of exposure and 26 27 outcome; and length and adequacy of follow up. NOS allocates ‘stars’ for adequate 28 29 methods, but does not specifically advise calculating the sum of allocated stars to give an 30 31 overall score. Scores for quality are not helpful in assessing the effect of risk of bias on a 32 33 metaanalysis so we report each item separately in line with recommended methods [16 17]. 34 http://bmjopen.bmj.com/ 35 36 To be assessed as adequate for comparability of cohorts (risk of confounding) a study had to 37 38 control for the following four prespecified potential confounding factors related to foetal 39 40 development: maternal age, socioeconomic status (SES), ethnicity, and smoking. 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 The likelihood of small study bias, such as publication bias, could not be assessed through 45 46 visual inspection of funnel plots for pooled analyses as no outcome was assessed by 10+ 47 48 studies [18]. Statistical analyses were carried out using Stata version 13.1 (StataCorp, 49 50 College Station, TX, USA).[19] 51 52 Results 53 A flowchart of the article review process is shown in Figure 1. A total of 4680 citation records 54 55 56 were identified from searching the four relevant databases. A manual search of recent 57 10 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 11 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 systematic reviews identified 33 additional articles. After exclusions, 24 prospective studies 4 5 analysed using standard approaches and two quasiexperimental studies were included, 6 7 reporting on 30 outcomes in total. 8 9 Risk of bias 10 11 Six studies (Albertson 2003; Anderson 2012; Bille 2007; Brooke 1989; Lundsberg 2015; 12 13 McCarthy 2013) had low risk of bias for all eight NOS items and were therefore considered 14 15 at low risk ofFor bias overall. peer All studies were review judged to be at a low only risk of bias for the following 16 17 three NOS items (supplementary Table 3): selection of the nonexposed cohort (always from 18 19 the same source population as the exposed cohort); the absence of outcome at the start of 20 21 the study; and adequate length of follow up for outcome to have occurred. Fourteen studies 22 23 had adequate ascertainment of exposure as these were all based on structured interviews or 24 25 validated records. Objective outcome assessments (assessor unaware of the exposure 26 27 status) were reported in 16 of the studies. For five others either parent selfreport was used 28 29 (high risk), and for the remaining three the method of outcome assessment was not reported 30 31 (unclear risk). Eleven studies did not report enough detail to decide if cohorts were 32 33 representative of the population, therefore only 10 could be judged as low risk. Only four 34 http://bmjopen.bmj.com/ 35 36 studies did not control for the prespecified potential confounding factors, and one did not 37 38 report enough detail to permit judgement. Thus, in the majority of studies (19) the compared 39 40 groups were similar. Nineteen studies had adequate follow up of the cohort (small loss to 41 follow up). Only three were judged high risk for this item and two studies presented

42 on October 2, 2021 by guest. Protected copyright. 43 44 insufficient information to make a clear judgment. 45 46 47 48 49 50 51 52 53 54 55 56 57 11 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 12 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Studies included in the metaanalysis are presented in Table 1. Standard analytical 4 5 approaches pooled estimates for continuous and binary outcomes are presented in Figure 2 6 7 and 3 respectively. Figure 2 presents results for birthweight (7 studies). Figure 3A presents 8 9 results for preterm delivery (9 studies) Figure 3B, presents results for SGA (7 studies), and 10 11 results for low birthweight (6 studies) are given in Figure 3C. 12 13 14 The metaanalysis yielded a summary OR of 1—10 (95% CI 0—95; 1—28) for preterm delivery, 15 For peer review only2 16 but there was substantial statistical heterogeneity between studies (I =60%), due to a large 17 18 Danish study reporting a protective effect (Figure 3A). Additionally, most studies assessing 19 20 preterm birth had corrected for main confounders know to be associated with preterm birth, 21 22 with the exception of [20] that did not correct for any. There was also modest evidence for 23 24 an increased risk of being SGA (OR 1—08, 95% CI 1—02; 1—14) for a total of 288512 25 26 participants, although this was almost entirely driven by a single US study contributing 95% 27 28 of the participants to this metaanalysis (Figure 3B). The birthweight metaanalysis yielded a 29 30 summary effect of 13—49g (95% CI 30—28g; +3—31g) for offspring of light drinkers versus 31 32 nondrinkers (Figure 2). Summary effect for birthweight <2500g was, OR 1—00, 95% CI 0—82; 33

34 1—22 (Figure 3C). http://bmjopen.bmj.com/ 35 36 Other outcomes were typically reported by a limited number of studies and mostly could not 37 38 be metaanalysed due to clinical heterogeneity in outcome assessment or incompleteness of 39 40 published data (supplementary Figure 1, and Table 2). Based on two studies with data on 41

42 behavioural outcomes, there was little evidence of any effect for internalising symptoms but on October 2, 2021 by guest. Protected copyright. 43 44 a suggestion that light drinking in pregnancy protected against high externalising behaviour 45 46 scores (OR 0—97, (95%CI 0—93; 1—01, supplementary Figure 1)). However, an additional 47 48 study assessing conduct problems and hyperactivity (in the same externalising domain) 49 50 reported results in the opposite direction, which could not be metaanalysed due to different 51 52 outcome definitions. 53 54 55 56 57 12 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 13 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Table 2 presents results of included studies that did not contribute to the metaanalyses for 4 5 various reasons. There was no strong evidence of association between consuming up to 32 6 7 g/week of alcohol and any of the remaining outcomes excluded from metaanalyses, with 8 9 three exceptions: a very large US study showing increased risk of placental abruption and 10 11 decreased risk of preeclampsia (OR 1—24, 95%CI 1—05; 1—46 and OR 0—82, 95% CI 0—74; 12 13 0—90, respectively)[21], and a single British study reporting better cognitive outcomes in 14 15 children exposedFor to light peermaternal drinking review in pregnancy.[22] only 16 17 We did not include funnel plots as no outcome was assessed by 10+ studies. 18 19 Of all included results, only two were unadjusted[20 23], and most of the others were 20 21 adjusted for maternal smoking, age and socioeconomic position (supplementary Table 2). 22 23 Studies that did not adjust for ethnicity were generally conducted in homogenous 24 25 populations. Due to the small number of studies for each outcome, we could not further 26 27 investigate the effect of adjusting for all or some of these confounders. Similarly, there was 28 29 insufficient data to examine the effect of timing of exposure on outcomes. 30 31 32 Alternative analytical approaches 33 Two negative control publications[24] based on the same UK cohort met our inclusion

34 http://bmjopen.bmj.com/ 35 criteria (Table 3).[23 25] They investigated the effects of maternal alcohol consumption on 36 37 childhood educational achievement[25] and IQ.[23] Offspring exposed to maternal 38 39 consumption of <12g/week of alcohol in the first trimester did not have worse outcomes 40 41 compared to those of mothers who abstaining from alcohol, and a similar pattern was found

42 on October 2, 2021 by guest. Protected copyright. 43 44 for paternal alcohol consumption. 45 46 One further quasiexperimental study, one natural experiment and five Mendelian 47 48 randomisation studies were excluded from the present review because they did not 49 50 specifically test the effect of consuming up to 32g/week in pregnancy versus abstaining. 51 52 These will be included in a forthcoming review focused on estimating the causal effects of 53 54 prenatal alcohol exposure based on alternative study designs and analytical approaches to 55 56 strengthen causal evidence. 57 13 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

Page 14 of 51 Yes Yes Yes Yes Yes Yes No Yes Yes Yes yes Yes Yes Yes Yes Yes Adjusted Adjusted (yes/no) (yes/no) Yes Yes Yes Yes 14 14

11 months 11 inclusive years 14 - - Birth Birth Birth Birth 7.75 years 7.75 Birth Birth Birth 7.75 years 7.75 0 Birth Birth 2 Birth Age at outcome outcome at Age assessment (child) (child) assessment 7 years years 7 Birth Birth

44 44 - 20 range range

trimester trimester trimester trimester

trimester ithin the gestational age gestational ithinthe etween 5th and 39th weeks of pregnancy of weeks 39th and 5th etween st st st st 1 Late pregnancy pregnancy Late Birth Yes 1st trimester 1st 1 36 weeks & 28 (12.9, pregnancy Throughout 24 (before care antenatal for booking At 1st trimester 1st gestation) gestation) weeks B 1 W trimester 2nd 1 weeks Timing of exposure exposure of Timing At booking for antenatal care (before 24 24 (before care antenatal for booking At gestation) weeks During pregnancy pregnancy During Entire pregnancy pregnancy Entire

analysis) - Behaviour Checklist (CBCL) Checklist Behaviour in the meta in the

≥60),Child ≥60),Child BMJ Open Externalising Externalising

included http://bmjopen.bmj.com/

- Externalising Externalising -

problem scores, Strengths & Difficulties Questionnaire (SDQ) (SDQ) Questionnaire Difficulties & Strengths scores, problem problem scores, Strengths & Difficulties Questionnaire (SDQ) (SDQ) Questionnaire Difficulties & Strengths scores, problem percentile)

- problems onduct th Hyperactivity

centile) centile) percentile) percentile) percentile) percentile) percentile)

th th th th th Birth weight (g) (g) weight Birth Birth weight (g) (g) weight Birth SGA (lowest 10 (lowest SGA (g) weight Birth weeks) (<37 birth Preterm 10% with highest highest 10%with C report parent weight birth Low weeks) (<37 delivery Preterm Preterm birth (<37 weeks) (<37 birth Preterm Birthweight weeks) (<37 birth Preterm score (T problems significant Clinically highest 10%with report parent Birth weight (<2500g) weight Birth 10 (< SGA 10 (< SGA 10 (< SGA report Parent behaviour Internalising behaviour Externalising Outcomes Outcomes 10 (< SGA Low birth weight (<2500g) (<2500g) weight birth Low weeks) (<37 birth Preterm birthweight mean Adjusted (<2500g) weight birth Low weeks) (<37 birth Preterm SGA (weight < −2 standard deviaon scores) scores) deviaon standard −2 (weight < SGA weeks) (<37 birth Preterm weight Birth Preterm birth (<37 weeks) weeks) (<37 birth Preterm Customised birth centile birthcentile Customised 10 (< SGA Moderate preterm birth (32-37 weeks) weeks) (32-37 birth preterm Moderate weeks) (<32 birth preterm Very Parent report report Parent behaviour Internalising Boys behaviour Internalising Girls behaviour Externalising Boys behaviour Externalising Girls on October 2, 2021 by guest. Protected copyright.

number

2

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

10,649 10,649 (SDQ), Questionnaire Difficulties & Strengths scores, problem highest 10%with 2,06 638 901 10,323 1,493 3166 - 1335 10,323 3,907 3,907 Total 4,132 4,132 weight birth Low 535 535 29,463 29,463 weeks) (<37 birth Preterm

For peer review only

- - 64 - 126 - - 202 - 185 185 - 202 325 150 Event 191 315 274 ------23

USA USA UK UK Japan Zealand, New USA Australia UK Australia, UK, Australia, Ireland and Country USA

Prospective studies examining the effects of prenatal alcohol exposure on pregnancy outcomes (studies (studies outcomes pregnancy on exposure alcohol prenatal of effects the examining studies Prospective

Lundsberg (1997) Lundsberg (1995) Shu (1995) Peacock Sayal Sayal (2007) McCarthy (2013) McCarthy (2011) Salihu Sayal (2007) Lundsberg (2015) Lundsberg

Miyake (2014) Miyake

[35] [37] [20] [33] [29] [30] [21] [33] [27] Nykjaer (2014) (2014) Nykjaer [27] UK (2010) Robinson [31] (2007) Jaddoe [32] UK Holland; - Table 1. Table (year) Study [26] [34] Albertsen (2004) (2004) Albertsen [34] Denmark 1488 (1996) Passaro [36] UK 8,886 [28] Niclasen (2014) (2014) Niclasen [28] Denmark - 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

yes yes yes 15 15

Birth Birth Birth

Early pregnancy pregnancy Early During the first 36 weeks gestation weeks 36 the first During BMJ Open http://bmjopen.bmj.com/

Birthweight (g) (g) Birthweight Birth weight (<2500g) weight Birth on October 2, 2021 by guest. Protected copyright.

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 1,140 1,140 8772 For peer review only

Denmark Denmark

Olsen (1991) Olsen

[38] [39] Brooke (1989) (1989) Brooke [39] UK SGA: small for gestation age age gestation for small SGA: Page 15 of 51 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from Page 16 of 51 16 16

(yes/no) (yes/no) Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Adjusted Adjusted Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

99)

· 1·46) 1·50) 3·84) 0 0·90) 1·39) 1·12)

– – – – – – – 1·5)

– 35 35 ·

(95% CI 0·74 CI (95% [1] 59 (95% CI 0 CI 59(95% · those unexposed. unexposed. those OR 1·05 (95% CI 0·94 – 1·18) 1·18) 0·94– CI 1·05(95% OR Results Results OR 1·12 (95% CI 0·86 CI 1·12(95% OR 0·40 CI 0·78(95% OR OR 1·61 (95% CI 0·67 CI 1·61(95% OR OR 0 OR OR 0·82 OR 3225g in those exposed, compared to 3225g 3225g to compared exposed, those in 3225g unexposed. those in 3414g in those exposed, compared to 3363g 3363g to compared exposed, those in 3414g unexposed. those in OR 1·0 (95% CI 0·7 0·7 CI 1·0 (95% OR those 1·9) for (SD 40·1 age gestational Mean those for 2·0) (SD 40·1 to compared exposed, unexposed. OR 1·10 (95% CI 0·88 CI 1·10(95% OR OR 0·90 (95% CI 0·73 CI 0·90(95% OR 3·7% in those exposed compared to 3·9% in 3·9% to compared exposed in those 3·7% OR 1·08 (95% CI 0·83 – 1·42) 1·42) 0·83 – CI 1·08(95% OR OR 1·10 (95% CI 0·78 – 1·54) 1·54) 0·78 – CI 1·10(95% OR OR 1·11 (95% CI 0·87 – 1·43) 1·43) 0·87 – CI 1·11(95% OR OR 1·24 (95% CI 1·05 – 1·46) 1·46) 1·05 – CI 1·24(95% OR

National / Hospital Hospital / National maternal from or Registries, report) Birth Birth Birth Birth Birth Age at outcome assessment assessment outcome at Age (child) Birth Birth Birth (Outcome confirmed via confirmed (Outcome records) medical Birth (Outcome confirmed via confirmed (Outcome National / Hospital Hospital / National maternal from or Registries, report)

analysis) -

outcome

in the meta the in

with this outcome this with

included o studies with this this with studies o BMJ Open Only study with this outcome this with study Only outcome this with study Only Only study with this outcome this with study Only Only study with this outcome outcome this with study Only Birth Only study with this outcome outcome this with study Only Birth Only tw Only outcome Only two studies with this this with twostudies Only outcome Only study with this outcome outcome this with study Only Birth Only study with this outcome outcome this with study Only Birth SE/SD not reported not SE/SD SE/SD not reported not SE/SD

Only study study Only Only study with this with study Only Only study with this outcome outcome this with study Only via confirmed (Outcome Estimates are in different different in are Estimates directions Estimates are in different different in are Estimates directions analysis analysis Pregnancy outcomes Pregnancy http://bmjopen.bmj.com/ oetal alcohol spectrum disorder (FASD) disorder spectrum alcohol oetal

in

percentile percentile percentile percentile Features of f of Features th th

eclampsia eclampsia - - Lowest 10 Lowest Lowest 10 Lowest Placental abruption abruption Placental Placenta previa previa Placenta trimester trimester Per cent of infants with with infants of cent Per <37 age gestational weeks Admission to NICU to Admission malformations Major Apgar <7 at 5 minutes 5 minutes at <7 Apgar Pre Pre of head circumference circumference head of of birth length length birth of Mean birthweight (g) (g) birthweight Mean Mean birthweight (g) in (g) birthweight Mean smokers non-smokers non-smokers Gestational age Gestational Spontaneous abortion abortion Spontaneous occurring ≤20 weeks weeks ≤20 occurring gestation Stillbirth Stillbirth on October 2, 2021 by guest. Protected copyright.

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Country (Total number) number) (Total Country details Outcome meta- in included not Reason France, Spain, Denmark, Denmark, Spain, France, (8,453) Portugal Germany, USA (4,496) USA (4,496) USA USA (4,496) USA New Zealand, Australia, Australia, Zealand, New USA (1,226,685) USA UK Ireland, Netherlands, Scotland, Scotland, Netherlands, Netherlands, Scotland, Scotland, Netherlands, Denmark, Spain, France, (8,453) Portugal Germany, France, Spain, Denmark, Denmark, Spain, France, (8,453) Portugal Germany, UK (10,539) UK USA (5,324) USA USA (1,226,685) USA Denmark (89,322) Denmark For peer review only

Lundsberg (2015) Lundsberg (2015) Lundsberg Lundsberg (2015) Lundsberg McCarthy (2013) McCarthy Salihu (2011) Salihu Ogston (1992) Ogston Ogston (1992) Ogston Passaro (1996) Passaro Windham (1997) Windham Salihu (2011) Salihu Andersen (2012) Andersen

[26] Lundsberg (2015) (2015) Lundsberg [26] (4,496) USA [21] Salihu (2011) (2011) Salihu [21] (1,226,685) USA [42] Ogston (1992) (1992) Ogston [42] Scotland, Netherlands, [40] Andersen (2012) (2012) Andersen [40] (89,322) Denmark in first Miscarriage [26] [26] [26] [30] [21] [42] [42] [36] [41] [21] [40] [26] Lundsberg (2015) (2015) Lundsberg [26] (4,496) USA [21] Salihu (2011) (2011) Salihu [21] (1,226,685) USA Study (year) (year) Study

Prospective studies examining the effects of prenatal alcohol exposure on pregnancy outcomes (results not not (results outcomes pregnancy on exposure alcohol prenatal of effects the examining studies Prospective

eclampsia - Admission to to Admission Malformations Apgar score Apgar Intensive Neonatal (NICU) Unit Care Height Pre Placenta-related Birthweight Gestational age Gestational birth and preterm Miscarriage Table 2. Table Stillbirth Outcome Outcome Head Head circumference

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 17 17

Yes Yes Yes Yes Yes Yes Yes Yes Yes No Yes Yes Yes Yes Yes Yes Yes Yes Yes

1·16) 1·03) 1·50)

1·51)

– – – –

on KS2 score for those exposed compared to to compared exposed those for score KS2 on unexposed. those exposed, compared to 1·85 for those those for 1·85 to compared exposed, those (p=0·26) unexposed Mean increase of 0·81 points (se 0·8) for for 0·8) (se 0·81points of increase Mean inthose 16·3) (SD 106·4 IQ score Mean those exposed compared to unexposed. unexposed. to compared exposed those in 16·2) (SD 105·7 to compared exposed, p=0·10. unexposed, those OR 1·20 (95% CI 0·96 CI 1·20(95% OR those exposed compared to unexposed. unexposed. to compared exposed those OR 0·99 (95% CI 0·86 CI 0·99(95% OR OR 0·85 (95% CI 0·70 CI 0·85(95% OR Mean total anomalies was 2·60 for those those for 2·60 was anomalies total Mean those for 2·53 to compared exposed, (p=0·28) unexposed OR 1·06 (95% CI 0·74 CI 1·06(95% OR Mean increase of 1·80 points (SE 1·1) for for 1·1) (SE 1·80points of increase Mean OR 1·14 (95% CI 0·98 – 1·32) 1·32) 0·98– CI 1·14(95% OR Mean increase of 0·38 (95% CI -0·02 – 0·78) 0·78) – CI -0·02 0·38(95% of increase Mean Mean craniofacial anomalies was 1·92 for 1·92for was anomalies craniofacial Mean

9 years 9

- for gestation age IQ: intelligence quotient. intelligence IQ: age gestation for

18 months 18 months 7.75- 9 years years 7.75- 9 1·94) 1·02– CI 1·41(95% OR 18 18 months years 8 7.75 3.9 3.9 years 7 years 7 Birth Birth Birth Birth

deviation SGA: small SGA: deviation : standard standard : BMJ Open No SE/SD reported for for reported SE/SD No outcome this with study Only reference group group reference group reference No SE/SD reported for for reported SE/SD No Parent report used in analysis analysis in used report Parent instead instead instead Parent report used in analysis analysis in used report Parent Later age used in analysis used Laterage Later age used in analysis in analysis used Laterage 3.9 years Only study with this outcome this with study Only SE/SD not reported not SE/SD SE/SD not reported reported not SE/SD Only study with this outcome this with study Only error SD error http://bmjopen.bmj.com/

onduct onduct

onduct problems problems onduct Key Stage 2 scores scores 2 Stage Key outcome this with study Only 11 years Craniofacial anomalies anomalies Craniofacial Mental Development Development Mental of Scales Bayley Index, Development Infant Psychomotor Psychomotor Weschler IQ, Development Index, Index, Development Infant of Scales Bayley Development for Scale Intelligence (WISC-III) Children Hyperactivity teacher report report teacher Questionnaire (SDQ) (SDQ) Questionnaire Strengths & Difficulties &Difficulties Strengths 10% with highest highest 10% with scores, problem C teacher report report teacher Questionnaire (SDQ) (SDQ) Questionnaire Strengths & Difficulties &Difficulties Strengths problem scores, scores, problem 10% with highest highest 10% with Hyperactivity parent report report parent Questionnaire (SDQ) (SDQ) Questionnaire Strengths & Difficulties &Difficulties Strengths 10% with highest highest 10% with scores, problem parent report C report parent Questionnaire (SDQ) (SDQ) Questionnaire Strengths & Difficulties &Difficulties Strengths problem scores, scores, problem 10% with highest highest 10% with Developmental Developmental Total anomalies Total Disorder Coordination (DCD) Oral clefts Oral problems problems ≤32g AA per week. SE: Standard Standard SE: week. per AA ≤32g on October 2, 2021 by guest. Protected copyright.

)

(1020 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

France, Spain, Denmark, Denmark, Spain, France, (8,453) Portugal Germany, Netherlands, Scotland, Scotland, Netherlands, (4,332) UK France, Spain, Denmark, Denmark, Spain, France, (8,453) Portugal Germany, UK (525) UK UK (1077) UK Denmark (32,097) Denmark USA (873) USA Denmark Denmark For peer review only

Parry (1992) Parry (2008) Alati Sayal (2007) Sayal Sayal (2007) Sayal Faebo Larsen Larsen Faebo Ernhart (1989) Ernhart Bille (2007) Bille

[22] Sayal (2013) (2013) Sayal [22] (10,558) UK [46] Parry (1992) (1992) Parry [46] Scotland, Netherlands, [33] Sayal (2007) (2007) Sayal [33] (257) UK [33] Sayal (2007) (2007) Sayal [33] (967) UK [46] [23] [33] [33] [45] [44] (2013) [43] [44] Ernhart (1989) (1989) Ernhart [44] (873) USA

to exposed those in outcome of odds the compare results Ratio Odds [1] Cognition Cognition

Motor Motor Behaviour /development development

Page 17 of 51 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from Page 18 of 51 18 18 BMJ Open http://bmjopen.bmj.com/ on October 2, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

For peer review only

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 19 19 Different confounding structures structures confounding Different vs. maternal of association the for outcome the with alcohol paternal Different confounding structures structures confounding Different vs. maternal of association the for outcome the with alcohol paternal Limitations Limitations Maternal alcohol consumption consumption alcohol Maternal 16·2); (SD 105·7 mean abstainers: (SD 106·4 mean glass/week: <1 16·3) consumption alcohol Paternal 16·8); (SD 102·2 mean abstainers: (SD 104·0 mean glass/week: <1 16·7) Maternal alcohol consumption consumption alcohol Maternal 9·1); (SD 102·0 mean abstainers: (SD 102·8 mean glass/week: <1 8·7) consumption alcohol Paternal <1 11); (SD 98·9 mean abstainers: 9·1) (SD 101·1 mean glass/week: conclusions as presented in the in presented as conclusions paper IQ: Wechsler IQ: Wechsler Scale Intelligence Children for (WISC) achievement: Key achievement: scores 2 Stage (standardised) Outcomes Outcomes & results of Summary

8 years years 8

assessment assessment (child) 11 years years 11 Academic Intelligence quotient. Intelligence BMJ Open http://bmjopen.bmj.com/

na

na

na na na on October 2, 2021 by guest. Protected copyright. paternal paternal - For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Study type type Study Gene number SNP-rs outcome at Age Maternal comparison comparison comparison For peer review only

Number Number 4,332

Country Country Total UK

UK UK 7,062 Maternal-paternal

Quasiexperimental studies examining the effects of prenatal alcohol exposure on pregnancy outcomes. outcomes. on pregnancy exposure alcohol prenatal of the effects examining studies Quasiexperimental Alati (2008) Alati

Study (year) Study(year) [23] IQ: deviation SD:standard difference Mean MD: intervals CI: Confidence ratio OddsOR: Abbreviations: Table 3. Table (2013) Alati [25] Page 19 of 51 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open Page 20 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Discussion 4 5 Main findings: In this comprehensive systematic review of the literature on the effects of low 6 7 levels of alcohol drinking in pregnancy, the two main findings are: i) a surprisingly limited 8 9 number of prospective studies specifically addressing the question of whether light maternal 10 11 alcohol consumption (i.e. up to 32g/week (or 4 UK units) has any causal effect (adverse or 12 13 beneficial) on infant and later offspring outcomes and pregnancy outcomes, and, as a result, 14 15 ii) a paucity ofFor evidence demonstratingpeer areview clear detrimental effect, only or safe limit, of light alcohol 16 17 consumption on outcomes. The upper limit that we chose to examine here is that of the 18 19 current version of the UK National Institute for Health and Care Excellence (NICE) 20 21 guidelines.[47] The question we have attempted to address is very important given the 22 23 mixed advice that women are given with regards to whether they should abstain completely 24 25 or be allowed light alcohol consumption in pregnancy. The lack of research evidence to 26 27 address this question is notable. 28 29 30 31 Strengths and limitations: Strengths of this review include the completeness of searches with 32 33 a focused research question aimed at informing alcohol in pregnancy guidelines. In addition

34 http://bmjopen.bmj.com/ 35 to observational studies’ biases minimised by only including those with prospective 36 37 assessment of exposure and prioritising results adjusted for main confounders. Another 38 39 strength of this review is the unique effort to include alternative study designs to further 40 41 improve causal inference alongside standard analytical approaches. The main limitation of

42 on October 2, 2021 by guest. Protected copyright. 43 results on the effects of light drinking in pregnancy from standard analytical approaches is 44 45 bias due to residual confounding. SE position is a complex, multifaceted entity. Several 46 47 studies have attempted to adjust for SE position by collecting information on, for example, 48 49 maternal education, familylevel SE position around the time of the pregnancy, home 50 51 addressbased deprivation index etc. Few studies included more than one of these 52 53 measured [22 2931]. Whereas we consider attempting to adjust for at least one of these 54 55 56 characteristics to be a minimum requirement to account for some of the confounding 57 58 introduced by SE position, there remains scope for residual confounding.[48] Given the 59 60 20

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 21 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 strong relationship between SE position and both the exposure (alcohol use in pregnancy) 4 5 and outcomes in this review, any degree of residual confounding is of course an issue when 6 7 interpreting the effect estimates from the observational studies included in this review. 8 9 Women who drink low amounts of alcohol may be more likely to be of higher socio 10 11 economic position, compared to abstainers, at least in developed settings in recent 12 13 years,[49] and both of these characteristics are associated with better pregnancy and 14 15 cognitive outcomes.[50]For Maternalpeer smoking review and ethnicity are also only known correlates of 16 17 maternal alcohol use, and risk factors for e.g. low birth weight.[51] Most studies included in 18 19 this review adjusted for at least some of these factors. However, due to the small number of 20 21 studies included for any given outcome, it was impossible to formally investigate the effect of 22 23 incomplete adjustment for some (or all) of these confounders. Additionally, for most 24 25 26 outcomes, we could not pool eligible studies for various reasons (e.g. too few studies, lack of 27 28 standard errors) and that we limited our review to a number of prespecified outcomes 29 30 including the most common pregnancyrelated outcomes and childhood outcomes related to 31 32 FASD. This also was the case for identifying effects based on time of exposure, which is 33 also a limitation. 34 http://bmjopen.bmj.com/ 35 36 The inclusion of only English language studies may have led to missing some studies, 37 38 however there is little evidence that exclusion of nonEnglish language studies leads to 39 40 systematic bias in systematic reviews of conventional medicine.[5255] 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 Interpretation: This review demonstrates the paucity and poor quality of evidence addressing 45 46 this important public health question, and the difficulty of designing studies that can 47 48 effectively evaluate the causal impact of low alcohol consumption whilst minimising bias and 49 50 confounding. It also shows the value of reporting measures of effect for meaningful 51 52 categories of the exposure. Whilst many studies reported that associations did not differ from 53 54 linearity prior to providing a single coefficient for the doseresponse association, it is possible 55 56 that statistical power limited the ability to detect nonlinear associations in single studies. 57 58 59 60 21

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 22 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Such detail is especially important when there are controversies about the shape of the 4 5 association of interest (linear, U or Jshaped) and/or the existence of safe thresholds. 6 7 Outstanding questions also remain about the effects of maternal alcohol consumption at 8 9 different stages of conception and pregnancy. Alternative analytical approaches such as 10 11 sibling comparisons[56] and the use of instrumental variable approaches[57] as well as 12 13 triangulating the totality of evidence from multiple study types[58] (formally or informally) are 14 15 needed in orderFor to strengthen peer confidence review in the direction and sizeonly of any potential causal 16 17 relationships. 18 19 The recently proposed change in the guidelines for alcohol use in pregnancy in the UK to 20 21 complete abstinence, would be an application of the precautionary principle. This review 22 23 confirmed some increased risk of babies being born SGA but little direct evidence of any 24 25 other detrimental effect for maternal drinking up to 32g/week. However, there have been few 26 27 wellconducted studies examining this specific category of exposure. This issue remains of 28 29 great public health importance, with alcohol consumption during pregnancy prevalent in the 30 31 UK, Ireland, New Zealand and Australia with up to 80% of women consuming some alcohol 32 33 during pregnancy.[59] For some, the evidence of the potential for harm – mostly coming from 34 http://bmjopen.bmj.com/ 35 36 animal experiments and human studies of effects due to higher levels of exposure will be 37 38 sufficient to advocate that guidelines should advise women to avoid all alcohol in pregnancy, 39 40 while others will wish to retain the existing wording of guidelines.[60] Here we found that 41 maternal alcohol consumption of up to 32g/week was associated with an 10% increased risk

42 on October 2, 2021 by guest. Protected copyright. 43 44 of preterm birth (95%CI: 0—95 to1—28). In comparison, light to moderate smoking (<20 45 46 cigarettes per day) is associated with a 22% increased risk of preterm birth (95% CI: 1—13 to 47 48 1—32).[61] 49 50 51 In conclusion, we found limited evidence for a causal role of light drinking in pregnancy, 52 53 compared to abstaining, on most of the outcomes examined. Despite the distinction between 54 55 light drinking and abstinence being the point of most tension and confusion for health 56 57 professionals and pregnant women and contributing to inconsistent guidance and advice 58 59 60 22

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 23 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 now and in the past, our extensive review shows that this specific question is not being 4 5 researched thoroughly enough, if at all. In addition, there has been no evidence regarding 6 7 possible benefits of light alcohol consumption versus absence. Further studies, including 8 9 those using designs that improve causal inference, are required to provide further evidence 10 11 and a better estimation of the likely effects. Formulating guidance on the basis of the current 12 13 evidence is challenging. However, describing the paucity of current research and explaining 14 15 that “absenceFor of evidence peer is not evidence review of absence”, appears only warranted. 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33

34 http://bmjopen.bmj.com/ 35 36 37 38 39 40 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 23

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 24 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Acknowledgments 4 We thank our colleagues from NIHR CLAHRC West who provided support that greatly 5 6 assisted the research. We thank Louise Waters for assistance with article collection and 7 Alison Richards for developing the search strategy for this review. 8 9 10 11 Details of contributors 12 Loubaba Mamluk: Drafting the protocol, preparing the data extraction database, data 13 14 screening and collection (lead reviewer), data analysis, figures, tables, data interpretation, 15 and drafting Forthe manuscript. peer Hannah B Edwards:review screening, data only collection, data quality 16 17 checking, data interpretation, and preparation of final manuscript. Theresa HM Moore: 18 Advising on the protocol, preparing the protocol, screening title and abstract. Timothy Jones: 19 20 Data extraction and full text screening. Sharea Ijaz: Abstract and full text screening. Jelena 21 Savović: protocol development/study design (including the development of searching, article 22 23 screening and data collection strategies), oversaw and managed the review process and the 24 25 review team, provided methodological advice for the review and metaanalysis conduct, and 26 critically revised the manuscript. George Davey Smith: Obtained funding, formulation of 27 28 project plan, review of progress and of the manuscript. Jenny L. Donovan: Obtained funding 29 and contributed to drafting the manuscript, and approved the final version. Verity Leach: 30 31 Data extraction. Deborah Lawlor: contributed to the study design, data interpretation and 32 review of earlier manuscript drafts; she approved the final version that has been submitted. 33

34 Sarah J. Lewis: contributed to the study design, data interpretation and review of earlier http://bmjopen.bmj.com/ 35 manuscript drafts; she approved the final version that has been submitted. Abigail Fraser: 36 37 Contributed to the study design, data analysis and interpretation and drafting of manuscript. 38 Luisa Zuccolo: Contributed to the study design, data analysis and interpretation and drafting 39 40 of manuscript. I, Loubaba Mamluk, the corresponding author of this manuscript, certify that I 41 have full access to all the data in the study and had final responsibility for the decision to

42 on October 2, 2021 by guest. Protected copyright. 43 submit for publication. 44

45 46 Transparency declaration 47 48 I, Loubaba Mamluk, the corresponding author of this manuscript, affirm that this manuscript 49 50 is an honest, accurate, and transparent account of the study being reported; that no 51 important aspects of the study have been omitted; and that any discrepancies from the study 52 53 as planned (and, if relevant, registered) have been explained. 54 55 Conflict of interest: 56 57 The authors have nothing to disclose. 58 59 60 24

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 25 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 This work has not been submitted for publication elsewhere. 4

5 6 Disclosure 7 All authors have completed the ICMJE uniform disclosure form 8 9 at www.icmje.org/coi_disclosure.pdf and declare: no support from any organisation for the 10 11 submitted work; no financial relationships with any organisations that might have an interest 12 in the submitted work in the previous three years; no other relationships or activities that 13 14 could appear to have influenced the submitted work. 15 For peer review only 16 17 Role of funding source 18 The UK Medical Research Council (MC_UU_12013/1 and MC_UU_12013/5) and the 19 20 University of Bristol provide core funding for the MRC Integrative Epidemiology Unit; LM, 21 DAL, GDS, AF, and LZ are partially or fully funded through this. AF and LZ are UK Medical 22 23 Research Council research fellowships (Grant refs: MR/M009351/1 and G0902144, 24 respectively). LM, JS, HE, VL, TJ, SI,TM and JLD are fully or partially partly funded by 25 26 National Institute for Health Research (NIHR) Collaboration for Leadership in Applied Health 27 Research and Care West (CLAHRC West) at University Hospitals Bristol NHS Foundation 28 29 Trust. DAL (NFSI061110196) and JLD (NFSI051210119) are NIHR Senior 30 Investigators. SL is a senior lecturer funded by HEFCE. The views expressed here are those 31 32 of the author(s) and not necessarily those of the NHS, the MRC, NIHR or the Department of 33 Health. I, Loubaba Mamluk, the corresponding author of this manuscript, certify that I have 34 http://bmjopen.bmj.com/ 35 full access to all the data in the study and had final responsibility for the decision to submit 36 37 for publication. 38 39 40 Transparency declaration 41 The lead author (Loubaba Mamluk) affirms that this manuscript is an honest, accurate, and

42 on October 2, 2021 by guest. Protected copyright. 43 transparent account of the study being reported; that no important aspects of the study have 44 been omitted; and that any discrepancies from the study as planned (and, if relevant, 45 46 registered) have been explained. 47 48 Data sharing Statement 49 50 NA 51 52 53 54 55 56 57 58 59 60 25

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 26 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 References 4 1. Streissguth AP, Landesman-Dwyer S, Martin JC, Smith DW. Teratogenic effects of alcohol in 5 humans and laboratory animals. Science (New York, N.Y.) 1980;209(4454):353-61 6 7 2. Popova S, Lange S, Shield K, et al. Comorbidity of fetal alcohol spectrum disorder: a systematic 8 review and meta-analysis. Lancet (London, England) 2016;387(10022):978-87 doi: 9 10.1016/s0140-6736(15)01345-8[published Online First: Epub Date]|. 10 3. Chudley AE, Conry J, Cook JL, Loock C, Rosales T, LeBlanc N. Fetal alcohol spectrum disorder: 11 Canadian guidelines for diagnosis. CMAJ : Canadian Medical Association journal = journal de 12 l'Association medicale canadienne 2005;172(5 Suppl):S1-s21 doi: 13 10.1503/cmaj.1040302[published Online First: Epub Date]|. 14 4. International Alliance for Responsible Drinking- Drinking guidelines for pregnancy and breast 15 feeding.For 2016 peer review only 16 5. (NICE) NIFHACE. Antenatal care. CG62. NICE 2008 17 6. Department of Health. Alcohol Guidelines Review – Report from the Guidelines development 18 group to the UK Chief Medical Officers. 2016 19 20 7. Moher D, Liberati A, Tetzlaff J, Altman DG. Preferred reporting items for systematic reviews and 21 meta-analyses: the PRISMA statement. Annals of internal medicine 2009;151(4):264-69 22 8. Hamlyn B OK, Wands S. Infant Feeding 2000. London: The Stationery Office, 2002. 23 9. Henderson J, Gray R, Brocklehurst P. Systematic review of effects of low–moderate prenatal 24 alcohol exposure on pregnancy outcome. Bjog 2007;114(3):243-52 25 10. Patra J, Bakker R, Irving H, Jaddoe VW, Malini S, Rehm J. Dose–response relationship between 26 alcohol consumption before and during pregnancy and the risks of low birthweight, preterm 27 birth and small for gestational age (SGA)—a systematic review and meta-analyses. Bjog 28 2011;118(12):1411-21 29 11. Dissemination CfRa. CRD’s guidance for undertaking reviews in 30 31 health care. York Publishing Services Ltd 2009:32 32 12. Becker HC, Diaz-Granados JL, Randall CL. Teratogenic actions of ethanol in the mouse: a 33 minireview. Pharmacology Biochemistry and Behavior 1996;55(4):501-13 34 13. Haycock PC. Fetal Alcohol Spectrum Disorders: The Epigenetic Perspective 1. Biology of http://bmjopen.bmj.com/ 35 reproduction 2009;81(4):607-17 36 37 14. Higgins J, Green S. Cochrane handbook for systematic reviews of interventions version 5.1. 0 38 [updated Mar 2011]. The Cochrane collaboration. 2011, 2012. 39 15. Wells G, Shea B, O'Connell D, et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of 40 nonrandomised studies in meta-analyses. 41 http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp. Accessed March 2017

42 16. Moher D, Liberati A, Tetzlaff J, Altman DG. Preferred reporting items for systematic reviews and on October 2, 2021 by guest. Protected copyright. 43 meta-analyses: the PRISMA statement. PLoS medicine 2009;6(7):e1000097 doi: 44 10.1371/journal.pmed.1000097[published Online First: Epub Date]|. 45 17. Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of observational studies in epidemiology: a 46 proposal for reporting. Meta-analysis Of Observational Studies in Epidemiology (MOOSE) 47 group. Jama 2000;283(15):2008-12 48 18. Sterne JA, Sutton AJ, Ioannidis JP, et al. Recommendations for examining and interpreting funnel 49 50 plot asymmetry in meta-analyses of randomised controlled trials. Bmj 2011;343:d4002 51 19. StataCorp. Stata Statistical Software: release 12. College Station TSL, 2015. 52 20. Peacock JL, Bland JM, Anderson HR. Preterm delivery: effects of socioeconomic factors, 53 psychological stress, smoking, alcohol, and caffeine. Bmj 1995;311(7004):531-5 54 21. Salihu HM, Kornosky JL, Lynch O, Alio AP, August EM, Marty PJ. Impact of prenatal alcohol 55 consumption on placenta-associated syndromes. Alcohol 2011;45(1):73-9 doi: 56 http://dx.doi.org/10.1016/j.alcohol.2010.05.010[published Online First: Epub Date]|. 57 58 59 60 26

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 27 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 22. Sayal K, Draper ES, Fraser R, Barrow M, Davey Smith G, Gray R. Light drinking in pregnancy and 4 mid-childhood mental health and learning outcomes. Arch Dis Child 2013;98(2):107-11 doi: 5 http://dx.doi.org/10.1136/archdischild-2012-302436[published Online First: Epub Date]|. 6 23. Alati R, Macleod J, Hickman M, et al. Intrauterine exposure to alcohol and tobacco use and 7 childhood IQ: findings from a parental-offspring comparison within the Avon Longitudinal 8 Study of Parents and Children. Pediatr Res 2008;64(6):659-66 doi: 9 http://dx.doi.org/10.1203/PDR.0b013e318187cc31[published Online First: Epub Date]|. 10 24. Lipsitch M, Tchetgen ET, Cohen T. Negative controls: a tool for detecting confounding and bias in 11 12 observational studies. Epidemiology (Cambridge, Mass.) 2010;21(3):383 13 25. Alati R, Davey Smith G, Lewis SJ, et al. Effect of prenatal alcohol exposure on childhood academic 14 outcomes: contrasting maternal and paternal associations in the ALSPAC study. PLoS ONE 15 2013;For8(10):e74844 peer doi: http://dx.doi.org/10.1371/journal.pone.0074844[published review only Online 16 First: Epub Date]|. 17 26. Lundsberg LS, Illuzzi JL, Belanger K, Triche EW, Bracken MB. Low-to-moderate prenatal alcohol 18 consumption and the risk of selected birth outcomes: a prospective cohort study. Ann 19 Epidemiol 2015;25(1):46-54.e3 doi: 20 http://dx.doi.org/10.1016/j.annepidem.2014.10.011[published Online First: Epub Date]|. 21 27. Nykjaer C, Alwan NA, Greenwood DC, et al. Maternal alcohol intake prior to and during 22 pregnancy and risk of adverse birth outcomes: evidence from a British cohort. J Epidemiol 23 Community Health 2014;68(6):542-9 doi: http://dx.doi.org/10.1136/jech-2013- 24 25 202934[published Online First: Epub Date]|. 26 28. Niclasen J, Nybo Andersen AM, Teasdale TW, Strandberg-Larsen K. Prenatal exposure to alcohol, 27 and gender differences on child mental health at age seven years. J Epidemiol Community 28 Health 2014;68(3):224-32 doi: http://dx.doi.org/10.1136/jech-2013-202956[published 29 Online First: Epub Date]|. 30 29. Miyake Y, Tanaka K, Okubo H, Sasaki S, Arakawa M. Alcohol consumption during pregnancy and 31 birth outcomes: the Kyushu Okinawa Maternal and Child Health Study. BMC Pregnancy 32 Childbirth 2014;14:79 doi: http://dx.doi.org/10.1186/1471-2393-14-79[published Online 33 First: Epub Date]|. 34 30. McCarthy FP, O'Keeffe LM, Khashan AS, et al. Association between maternal alcohol http://bmjopen.bmj.com/ 35 consumption in early pregnancy and pregnancy outcomes. Obstetrics & Gynecology 36 2013;122(4):830-37 37 38 31. Robinson M, Oddy WH, McLean NJ, et al. Low-moderate prenatal alcohol exposure and risk to 39 child behavioural development: a prospective cohort study. Bjog 2010;117(9):1139-50 doi: 40 http://dx.doi.org/10.1111/j.1471-0528.2010.02596.x[published Online First: Epub Date]|. 41 32. Jaddoe VW, Bakker R, Hofman A, et al. Moderate alcohol consumption during pregnancy and the risk of low birth weight and preterm birth. The generation R study. Ann Epidemiol

42 on October 2, 2021 by guest. Protected copyright. 43 2007;17(10):834-40 44 33. Sayal K, Heron J, Golding J, Emond A. Prenatal alcohol exposure and gender differences in 45 childhood mental health problems: a longitudinal population-based study. Pediatrics 46 2007;119(2):e426-34 47 34. Albertsen K, Andersen AM, Olsen J, Gronbaek M. Alcohol consumption during pregnancy and the 48 risk of preterm delivery. Am J Epidemiol 2004;159(2):155-61 49 35. Lundsberg LS, Bracken MB, Saftlas AF. Low-to-moderate gestational alcohol use and intrauterine 50 51 growth retardation, low birthweight, and preterm delivery. Ann Epidemiol 1997;7(7):498- 52 508 53 36. Passaro KT, Little RE, Savitz DA, Noss J. The effect of maternal drinking before conception and in 54 early pregnancy on infant birthweight. The ALSPAC Study Team. Avon Longitudinal Study of 55 Pregnancy and Childhood. Epidemiology 1996;7(4):377-83 56 57 58 59 60 27

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 28 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 37. Shu XO, Hatch MC, Mills J, Clemens J, Susser M. Maternal smoking, alcohol drinking, caffeine 4 consumption, and fetal growth: results from a prospective study. Epidemiology 5 1995;6(2):115-20 6 38. Olsen J, Pereira AdC, Olsen SF. Does maternal tobacco smoking modify the effect of alcohol on 7 fetal growth? American journal of public health 1991;81(1):69-73 8 39. Brooke OG, Anderson HR, Bland JM, Peacock JL, Stewart CM. Effects on birth weight of smoking, 9 alcohol, caffeine, socioeconomic factors, and psychosocial stress. Bmj 1989;298(6676):795- 10 801 11 12 40. Andersen A-MN, Andersen PK, Olsen J, Grønbæk M, Strandberg-Larsen K. Moderate alcohol 13 intake during pregnancy and risk of fetal death. Int J Epidemiol 2012;41(2):405-13 14 41. Windham GC, Von Behren J, Fenster L, Schaefer C, Swan SH. Moderate maternal alcohol 15 consumptionFor and riskpeer of spontaneous review abortion. Epidemiology only 1997;8(5):509-14 16 42. Ogston SA, Parry GJ. EUROMAC. A European concerted action: maternal alcohol consumption 17 and its relation to the outcome of pregnancy and child development at 18 months. Results-- 18 strategy of analysis and analysis of pregnancy outcome. Int J Epidemiol 1992;21 Suppl 1:S45- 19 71 20 43. Bille C, Olsen J, Vach W, et al. Oral clefts and life style factors--a case-cohort study based on 21 prospective Danish data. Eur J Epidemiol 2007;22(3):173-81 22 44. Ernhart CB, Sokol RJ, Ager JW, Morrow-Tlucak M, Martier S. Alcohol-related birth defects: 23 assessing the risk. Ann N Y Acad Sci 1989;562:159-72 24 25 45. Faebo Larsen R, Hvas Mortensen L, Martinussen T, Nybo Andersen AM. Determinants of 26 developmental coordination disorder in 7-year-old children: a study of children in the Danish 27 National Birth Cohort. Dev Med Child Neurol 2013;55(11):1016-22 doi: 28 http://dx.doi.org/10.1111/dmcn.12223[published Online First: Epub Date]|. 29 46. Parry GJ, Ogston SA. EUROMAC. A European concerted action: maternal alcohol consumption 30 and its relation to the outcome of pregnancy and child development at 18 months. Results-- 31 child development at age 18 months. Int J Epidemiol 1992;21 Suppl 1:S72-8 32 47. NHS Choices.Can I drink alcohol if I’m pregnant? 33 www.nhs.uk/chq/Pages/2270.aspx?CategoryID=54#close. 2014 34 48. Galobardes B, Shaw M, Lawlor DA, Lynch JW, Smith GD. Indicators of socioeconomic position http://bmjopen.bmj.com/ 35 (part 1). Journal of epidemiology and community health 2006;60(1):7-12 36 49. Kelly Y, Sacker A, Gray R, Kelly J, Wolke D, Quigley MA. Light drinking in pregnancy, a risk for 37 38 behavioural problems and cognitive deficits at 3 years of age? Int J Epidemiol 39 2009;38(1):129-40 doi: http://dx.doi.org/10.1093/ije/dyn230[published Online First: Epub 40 Date]|. 41 50. Bradley RH, Corwyn RF. Socioeconomic status and child development. Annual review of psychology 2002;53(1):371-99

42 on October 2, 2021 by guest. Protected copyright. 43 51. Phung H, Bauman A, Nguyen T, Young L, Tran M, Hillman K. Risk factors for low birth weight in a 44 socio-economically disadvantaged population: parity, marital status, ethnicity and cigarette 45 smoking. Eur J Epidemiol 2003;18(3):235-43 46 52. Morrison A, Polisena J, Husereau D, et al. The effect of English-language restriction on systematic 47 review-based meta-analyses: a systematic review of empirical studies. International journal 48 of technology assessment in health care 2012;28(02):138-44 49 53. Moher D, Pham B, Lawson M, Klassen T. The inclusion of reports of randomised trials published 50 51 in languages other than English in systematic reviews. Health Technol Assess 2003;7(41):1- 52 90 53 54. Klassen TP, Lawson ML, Moher D. Language of publication restrictions in systematic reviews gave 54 different results depending on whether the intervention was conventional or 55 complementary. Journal of clinical epidemiology 2005;58(8):769-76. e2 56 55. Jüni P, Holenstein F, Sterne J, Bartlett C, Egger M. Direction and impact of language bias in meta- 57 analyses of controlled trials: empirical study. Int J Epidemiol 2002;31(1):115-23 58 59 60 28

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 29 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 56. Richmond RC, Al-Amin A, Smith GD, Relton CL. Approaches for drawing causal inferences from 4 epidemiological birth cohorts: a review. Early human development 2014;90(11):769-80 doi: 5 10.1016/j.earlhumdev.2014.08.023[published Online First: Epub Date]|. 6 57. Burgess S, Davies NM, Thompson SG. Instrumental variable analysis with a nonlinear exposure- 7 outcome relationship. Epidemiology 2014;25(6):877-85 doi: 8 10.1097/ede.0000000000000161[published Online First: Epub Date]|. 9 58. Ades A, Sutton A. Multiparameter evidence synthesis in epidemiology and medical decision- 10 making: current approaches. Journal of the Royal Statistical Society: Series A (Statistics in 11 12 Society) 2006;169(1):5-35 13 59. O'Keeffe LM, Kearney PM, McCarthy FP, et al. Prevalence and predictors of alcohol use during 14 pregnancy: findings from international multicentre cohort studies. BMJ open 15 2015;For5(7):e006323 peer review only 16 60. Mather M, Wiles K, O’Brien P. Should women abstain from alcohol throughout pregnancy? 2015 17 61. Shah NR, Bracken MB. A systematic review and meta-analysis of prospective studies on the 18 association between maternal cigarette smoking and preterm delivery. American journal of 19 obstetrics and gynecology 2000;182(2):465-72 20 21 22 23 24 25 26 27 28 29 30 31 32 33

34 http://bmjopen.bmj.com/ 35 36 37 38 39 40 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 29

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 30 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Figure 1. Flowchart of search strategy including primary reasons for article exclusion. 4 5 Figure 2. Pooled mean difference for birthweight comparing low alcohol consumption (up to 6 32g/week) with no alcohol consumption (7 studies). ‘Adjusted’ refers to adjusted for both smoking 7 and a measure of socio-economic status. CI: Confidence intervals. 8 9 Figure 3a) Odd Ratios for preterm birth comparing low alcohol consumption (up to 32g/week) with 10 11 no alcohol consumption (9 studies); 3b) Odd Ratios for small for gestational age comparing low 12 alcohol consumption (up to 32g/week) with no alcohol consumption (7 studies) 3c) Odd Ratios for 13 low birthweight comparing low alcohol consumption (up to 32g/week) with no alcohol consumption 14 (6 studies). OR: Odds ratio, CI: Confidence intervals. Pooled OR includes both adjusted and 15 For peer review only 16 unadjusted estimates from studies, ‘Adjusted’ refers to adjusted for both smoking and a measure of 17 socio-economic status. 18 19 Supplementary figure 1a) Odds ratios for externalising behaviour comparing low alcohol 20 consumption (up to 32g/week) with no alcohol consumption (3 studies); 1b) Odds ratios for 21 internalising behaviour comparing low alcohol consumption (up to 32g/week) with no alcohol 22 23 consumption (2 studies); OR: Odds ratio, CI: Confidence intervals. 24 25 26 27 28 29 30 31 32 33

34 http://bmjopen.bmj.com/ 35 36 37 38 39 40 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 30 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 31 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 For peer review only 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33

34 http://bmjopen.bmj.com/ 35 36 37 38 39 40 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 Figure 1. Flowchart of search strategy including primary reasons for article exclusion.

48 143x183mm (300 x 300 DPI) 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 32 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 For peer review only 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 Figure 2. Pooled mean difference for birthweight comparing low alcohol consumption (up to 32g/week) with 32 no alcohol consumption (7 studies). ‘Adjusted’ refers to adjusted for both smoking and a measure of socio- 33 economic status. CI: Confidence intervals.

34 297x209mm (300 x 300 DPI) http://bmjopen.bmj.com/ 35 36 37 38 39 40 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 33 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 For peer review only 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 Figure 3a) Odd Ratios for preterm birth comparing low alcohol consumption (up to 32g/week) with no 32 alcohol consumption (9 studies); 3b) Odd Ratios for small for gestational age comparing low alcohol 33 consumption (up to 32g/week) with no alcohol consumption (7 studies) 3c) Odd Ratios for low birthweight comparing low alcohol consumption (up to 32g/week) with no alcohol consumption (6 studies). OR: Odds 34 ratio, CI: Confidence intervals. Pooled OR includes both adjusted and unadjusted estimates from studies, http://bmjopen.bmj.com/ 35 ‘Adjusted’ refers to adjusted for both smoking and a measure of socio-economic status. 36 37 297x209mm (300 x 300 DPI) 38 39 40 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 34 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 PROSPERO International prospective register of systematic reviews 4 5 6 7 Systematic review of the effects of low-moderate prenatal alcohol exposure on 8 pregnancy and childhood outcomes 9 Loubaba Mamluk, Luisa Zuccolo, Theresa Moore, Alison Richards 10 11 12 Citation 13 Loubaba Mamluk, Luisa Zuccolo, Theresa Moore, Alison Richards. Systematic review of the effects of low-moderate 14 prenatal alcohol exposure on pregnancy and childhood outcomes. PROSPERO 2015:CRD42015015941 Available 15 from http://www.crd.york.ac.uk/PROSPERO_REBRANDING/display_record.asp?ID=CRD42015015941For peer review only 16 17 Review question(s) 18 To determine what is known about the effects of prenatal alcohol exposure, corresponding to low-to-moderate levels 19 20 of maternal consumption, on pregnancy outcomes. These include pregnancy complications, delivery outcomes and 21 Fetal Alcohol Syndrome (FAS) features. This will include the assessment of systematic reviews and meta-analyses. A 22 particular focus will be placed on identifying practical and meaningful outcomes of alcohol toxicity during 23 pregnancy. 24 25 Searches 26 Publications will be identified by searching the following major relevant databases: Medline, Embase, web of science 27 and Psychinfo. All databases will be searched from inception. Internet searches will be carried out using Google 28 Scholar. Attempts to identify further studies will be made by examining the reference lists of included studies and of 29 previous reviews. All studies to be restricted to those published in the English language only. 30 31 Types of study to be included 32 1) Prospective studies and systematic reviews/meta-analyses of prospective studies (cohort or case-control studies 33 nested in a cohort). 2) Natural experiments / studies using instrumental variables to improve casual inference, including Mendelian Randomization (MR) studies 3) Sibling comparison studies 4) Parental comparisons 34 http://bmjopen.bmj.com/ 35 36 Condition or domain being studied 37 Pregnancy and delivery outcomes as well as offspring outcomes (from the domains affected by Fetal Alcohol 38 Syndrome (FAS)) 39 40 Participants/ population 41 Pregnant women or women who are trying to become pregnant sampled from the general population. Cohorts of pregnant women with alcohol abuse/dependency will be excluded.

42 on October 2, 2021 by guest. Protected copyright. 43 44 Intervention(s), exposure(s) 45 Inclusion: lowto-moderate levels of prenatal alcohol consumption (up to 10.4 UK units or 83 g/week). 46 47 Exclusion: studies will be excluded if there was no quantitative measure of alcohol consumption that could be 48 converted to UK standard units or grams of alcohol and if there was insufficient data for an (adjusted and/or crude) 49 effect measure of lowmoderate consumption to be extracted. Cohorts of pregnant women with alcohol 50 abuse/dependency will be excluded. 51 52 Comparator(s)/ control 53 women with no or very sporadic alcohol consumption in pregnancy. 54 55 Outcome(s) 56 Primary outcomes 57 Outcomes: (in both children and adults) 58 59 60 Page: 1 / 5 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 35 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 1) Pregnancy complications 4 5 - Intra uterine growth restrictions (IUGR) 6 7 - Miscarriage 8 9 - Premature labour and birth- Gestational age 10 11 - Preeclampsia and gestational hypertension 12 13 - Low amniotic fluid (oligohydramnios) 14 15 - Gestational diabetesFor peer review only 16 17 - Placenta previa 18 19 2) Delivery outcomes 20 21 - Birth weight/ low birth weight/ small for gestational age (SGA) 22 - Still birth 23 24 - Delivery intervention (including vacuum extraction, forceps delivery, 25 26 Caesarean section) 27 28 - Apgar score 29 30 3) FAS features 31 32 - Facial malformation 33

34 - Growth restrictions (height- measurements of growth restriction) http://bmjopen.bmj.com/ 35 36 - Cranium size/ head circumference 37 38 - Developmental delays 39 40 - Behaviour complications 41 - Cognitive impairment / IQ

42 on October 2, 2021 by guest. Protected copyright. 43 44 - Attention scores / Attention deficit and hyperactivity disorder (ADHD) 45 46 Secondary outcomes 47 none 48 49 Data extraction, (selection and coding) 50 Selection of studies: 51 Titles and abstracts will be screened independently by one reviewer (a random selection of 20% will also be screened 52 by a second reviewer independently) with inclusion/exclusion being decided according to prespecified criteria. 53 Discrepancies will be discussed and disagreements resolved through consensus. The full-text of each of the articles 54 identified through screening of titles and abstracts will be obtained in order to determine their inclusion in the review. 55 56 57 Data extraction: 58 59 60 Page: 2 / 5 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 36 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Data extraction will be carried out using Microsoft Access. This will be piloted on a small selection of studies and 4 adjusted as necessary. Relevant data will be documented from each identified study including information on study 5 design and location, population characteristics, exposures studied (including timing of exposure), methods used to 6 ascertain exposures, outcomes studied, method of outcome ascertainment (including person reporting the outcome, 7 whether parent, teacher, health professional, researcher, child&), study results (from both unadjusted and fully 8 adjusted regressions), statistical adjustments etc. Data extraction will be carried out independently by one reviewer 9 and a random selection of 20% will checked by a second reviewer. Discrepancies will be resolved through discussion 10 or referral to a third reviewer. Where necessary, authors will be contacted for additional information. 11 12 Risk of bias (quality) assessment 13 Studies that did not adjust for smoking and maternal education/social class as potential confounders in their final 14 model will be considered to be of low evidence quality. 15 For peer review only 16 Strategy for data synthesis 17 The impact of low-to-moderate alcohol use on pregnancy and related outcomes will be investigated using high-low 18 methods of meta-analysis techniques; however, due to anticipated low number of studies for some outcomes, a formal 19 meta-analysis may not be appropriate for some of the outcomes. Where meta-analysis is not possible, a narrative 20 summary of findings will be carried out. Random-effect meta-analysis will be performed alongside fixed-effect in the 21 presence of high levels of between-studies heterogeneity (measured through I2). Item response theory (IRT) will be 22 used to combining results from different scales if required. The likelihood of small study bias deriving from 23 publication bias will further be assessed through drawing funnel plots. 24 25 Analysis of subgroups or subsets 26 Where the included number of studies in the meta-analysis is large enough, sub-group analyses will be performed for 27 1) studies adjusting for smoking and maternal education/social class; 2) studies reporting separately on the effects of 28 alcohol use in different gestational periods; 3) studies using different exposure/outcome assessments. 29 30 Dissemination plans 31 We anticipate dissemination to regional and national public health directors 32 33 Contact details for further information

34 http://bmjopen.bmj.com/ 35 Dr Mamluk 36 MRC Integrative Epidemiology Unit 37 38 School of Social and Community Medicine 39 40 University of Bristol 41

42 Oakfield House on October 2, 2021 by guest. Protected copyright. 43 44 Oakfield Grove 45 46 Bristol BS8 2BN 47 48 United Kingdom 49 50 [email protected] 51 52 Organisational affiliation of the review 53 none 54 55 Review team 56 Dr Loubaba Mamluk, University of Bristol, UK 57 Dr Luisa Zuccolo, University of Bristol, UK 58 Ms Theresa Moore, University of Bristol, UK 59 60 Page: 3 / 5 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 37 of 51 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 Ms Alison Richards, University of Bristol, UK 4 5 Collaborators 6 Dr Luisa Zuccolo, University of Bristol, UK 7 Dr Sarah Lewis, University of Bristol, UK 8 Dr Abi Fraser, University of Bristol, UK 9 Professor Jenny Donovan, University of Bristol, UK 10 Professor George Davey Smith, University of Bristol, UK 11 12 Anticipated or actual start date 13 12 January 2015 14 15 Anticipated completionFor date peer review only 16 01 September 2015 17 18 Funding sources/sponsors 19 National Institute of Health Research, Medical Research Council, University of Bristol. UK 20 21 Conflicts of interest 22 None known 23 24 Language 25 English 26 27 Country 28 England 29 30 Subject index terms status 31 32 Subject indexing assigned by CRD 33 Subject index terms 34 http://bmjopen.bmj.com/ 35 Alcohol Drinking; Humans; Pregnancy Outcome; Prenatal Exposure Delayed Effects 36 37 Stage of review 38 Ongoing 39 40 Date of registration in PROSPERO 41 19 January 2015

42 on October 2, 2021 by guest. Protected copyright. 43 Date of publication of this revision 44 25 February 2015 45 46 DOI 47 10.15124/CRD42015015941 48 49 50 Stage of review at time of this submission Started Completed 51 Preliminary searches No Yes 52 Piloting of the study selection process No Yes 53 Formal screening of search results against eligibility criteria No No 54 Data extraction No No 55 Risk of bias (quality) assessment No No 56 Data analysis No No 57 58 59 60 Page: 4 / 5 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 38 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 PROSPERO 4 International prospective register of systematic reviews 5 The information in this record has been provided by the named contact for this review. CRD has accepted this information in good 6 faith and registered the review in PROSPERO. CRD bears no responsibility or liability for the content of this registration record, 7 any associated files or external websites. 8 9 10 11 12 13 14 15 For peer review only 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33

34 http://bmjopen.bmj.com/ 35 36 37 38 39 40 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page: 5 / 5 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

Page 39 of 51 BMJ Open

1 2 3 Supplementary Table 1. Search Strategy 1950 to 11·07·2016. MEDLINE, PsycINFO, EMBASE on Ovid; the Cochrane Library including CENTRAL (the Cochrane Central Database of Controlled 4 Trials) on Wiley Interscience; and Science Citation Index, Social Science Citation Index, on Web of Science. 5 1 exp pregnancy/ (719661) 6 2 Pregnant Women/ (5199) 7 8 3 preconception care/ or prenatal care/ (21644)

9 4 exp "embryonic and fetal development"/ (212127) 10 5 Fetus/ (68424) 11 For peer review only 6 exp Pregnancy Complications/ (343999) 12 13 7 (prepregnan$ or conception or preconception or pregnan$ or prenatal$ or pre-natal$ or 14 f?etal or f?etus 15 or in utero).ti,ab. (588557) 16 8 Maternal exposure/ (5534)

17 http://bmjopen.bmj.com/ 9 or/1-8 (1079283) 18 19 10 Ethanol/ (73449)

20 11 Alcohol dehydrogenase/ (5809) 21 12 Aldehyde Oxidoreductases/ (3672) 22 13 exp Drinking Behavior/ (57821) 23 24 14 Temperance/ (2430) 25 15 alcohol$.ti. (106954) on October 2, 2021 by guest. Protected copyright. 26 16 (alcohol dehydrogenase or acetaldehyde dehydogenase).ti,ab. (8557) 27 28 17 ((alcohol or alcoholic) adj3 (drink$ or exposure or consumption or consume$ or consuming 29 or low or light or moderat$ or abstin$ or abstain$)).ti,ab. (49470) 30 18 ((low or light or moderate or abstin$ or abstain$ or pattern$ or behavio?r$) adj3 drink$).ti,ab. 31 (10558) 32 19 (ADH1B$ or teetotal$ or temperance or nondrink$ or non-drink$).ti,ab. (3093) 33 34 20 Genome-Wide Association Study/ or Linkage Disequilibrium/ or genotype/ or phenotype/ or 35 polymorphism, genetic/ or (polymorphism$ or ((gene or genes or genetic or genotyp$) adj3 36 (instrument$ or variant$ or variable$ or variability or variabilities or variance$))).ti,ab. (494904) 37 21 *Alcohols/ or exp *Alcohol-Related Disorders/ or (alcohol adj3 (misuse or "use" or abuse or 38 addict$ or dependence or response$ or susceptibility)).ti,ab. (110843) 39 22 20 and 21 (3108)

40 23 or/10-19,22 (213050) 41 42 43 44 45 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

BMJ Open Page 40 of 51

1 2 3 24 9 and 23 (11403) 4 25 letter/ (861500) 5 6 26 editorial/ (368458)

7 27 news/ (166329) 8 28 exp historical article/ (325965) 9 29 Anecdotes as topic/ (4586) 10 11 30 comment/ (610211) For peer review only

12 31 case report/ (1708277) 13 32 (letter or comment$).ti. (100442) 14 15 33 or/25-32 (3417468)

16 34 randomized controlled trial/ or Randomized Controlled Trials as Topic/ or random$.ti,ab. (888907)

17 http://bmjopen.bmj.com/ 35 33 not 34 (3386180) 18 36 animals/ not humans/ (3889478) 19 20 37 exp Animals, Laboratory/ (730783)

21 38 exp Animal Experimentation/ (6477) 22 39 exp Models, Animal/ not humans/ (310115) 23 40 exp rodentia/ (2688128) 24 25

41 (rat or rats or mouse or mice).ti. (1123907) on October 2, 2021 by guest. Protected copyright. 26 42 or/35-41 (7849766) 27 43 24 not 42 (6498) 28 29 44 meta-analysis/ (52850)

30 45 meta-analysis as topic/ (13933) 31 46 (meta analy$ or metaanaly$ or metanaly$ or meta regression).ti,ab. (72390) 32 33 47 ((systematic$ or evidence$ or realist or narrative or literature) adj2 (review$ or overview$)).ti,ab. 34 (171877) 35 48 "review of reviews".ti,ab. (211)

36 49 (reference list$ or bibliograph$ or hand search$ or manual search$ or relevant journals).ab. 37 (27094) 38 50 (search strategy or search criteria or systematic search or study selection or data extraction).ab. 39 40 (28877) 51 (search$ adj4 literature).ab. (30686) 41 42 43 44 45 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

Page 41 of 51 BMJ Open

1 2 3 52 (medline or pubmed or cochrane or embase or psychlit or psyclit or psychinfo or cinahl or science 4 Citation index or bids or cancerlit).ab. (95294) 5 53 cochrane.jw. (11053) 6 54 ((multiple treatment$ or indirect or mixed) adj2 comparison).ti,ab. (991) 7 8 55 or/44-54 (293491) 9 56 43 and 55 (201) 10 57 epidemiologic studies/ (6077) 11 For peer review only 12 58 ep.fs. (1223653)

13 59 exp case control studies/ (691899) 14 60 exp cohort studies/ (1394149) 15 61 cross-sectional studies/ (185407) 16

17 62 Mendelian Randomization Analysis/ (229) http://bmjopen.bmj.com/ 18 63 (case control or negative control).ti,ab. (93097) 19 64 (cohort adj (study or studies or analys$)).ti,ab. (98537) 20 21 65 ((follow up or observational) adj (study or studies)).ti,ab. (87771)

22 66 ((longitudinal$ or retrospectiv$ or prospectiv$) and (study or studies or studied or review$ or 23 analys$ or cohort$)).ti,ab. (870699) 24 67 cross sectional.ti,ab. (183479) 25 68 (mendel$ or natural experiment$).ti,ab. (11082) on October 2, 2021 by guest. Protected copyright. 26 27 69 ((family or sibling or population) adj3 (study or studies)).ti,ab. (115245)

28 70 or/57-69 (2937160) 29 71 exp guideline/ (25834) 30 72 guidelines as topic/ or practice guidelines as topic/ (114690) 31 32 73 guideline$.mp. (291621) 33 74 or/71-73 (291621) 34 75 55 or 70 or 74 (3362878) 35 36 76 43 and 75 (3349)

37 77 limit 76 to english language (3096) 38 78 ((smok$ or drug$ or tobacco or nicotine or cigarette$ or substance$ or methamphetamine$ 39 40 or amphetamine$ or cocaine$ or heroin or cannabis or marijuana) not (alcohol or alcoholic or drink$)).ti. 41 (489662) 42 43 44 45 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

BMJ Open Page 42 of 51

1 2 3 79 77 not 78 (2822) 4 5 6 7 8 9 10 11 For peer review only 12 13 14 15 16

17 http://bmjopen.bmj.com/ 18 19 20 21 22 23 24 25 on October 2, 2021 by guest. Protected copyright. 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

Page 43 of 51 BMJ Open Supplementary Table 2. Lists of confounders adjusted for by each study Study (year) Confounders 1 [26] Lundsberg (2015) Parity, maternal age, education, BMI, marital status, ethnicity, caffeine, smoking, exercise, work, prenatal and multivitamin use, passive smoke exposure, marijuana use, 2 cocaine use, study cohort, preterm labour, respiratory problem, infant gender, bleeding, nausea/vomiting, hypertension, incompetent cervix, placental problems, 3 sexually transmitted disease, induction/augmentation, maternal asthma, gestational diabetes. 4 [27] Nykjaer (2014) Maternal pre-pregnancy weight, height, age, parity, ethnicity, salivary cotinine levels, caffeine intake, education, energy intake, gestation and baby’s sex 5 [28] Niclasen (2014) Parental smoking, parental education, parental pre-pregnancy psychiatric diagnoses, and maternal psychological well-being in pregnancy. 6 [29] Miyake (2014) Low birth weight, preterm birth: maternal age, region of residence, number of children, family structure, maternal education, maternal employment, body mass index, 7 maternal smoking during pregnancy, and baby’s gender. 8 Small for gestational age: maternal age, region of residence, number of children, family structure, maternal education, maternal employment, body mass index, maternal 9 smoking during pregnancy, gestational age, and baby’s gender. 10 [45] Faebo Larsen Sex, gestational age, intrauterine growth restriction, maternal age, mother’s occupational status, maternal smoking (ever) in first trimester, amount of maternal smoking 11 (2013) and alcohol consumptionFor in first trimester. peer review only 12 [22] Sayal (2013) Maternal age, parity, highest level of maternal education, daily frequency of smoking, use of cannabis and/or other illicit drugs during the first trimester, 13 homeownership, whether currently married, high scores (>12) on the Edinburgh Postnatal Depression Scale, and child gestational age, birth weight and gender. 14 [30] McCarthy (2013) Maternal age, smoking, years of schooling, ethnicity, body mass index, infant sex, maternal status, family income, and drug use during pregnancy. Adjusted for clustering. 15 Birthweight adjusted for gestational age at delivery. 16 [40] Andersen (2012) Number of previous abortions, coffee consumption, changes in alcohol consumption since prior to pregnancy and smoking. Effect of coffee consumption and smoking was stratified according to period. The model is stratified according to maternal age and parity.

17 http://bmjopen.bmj.com/ 18 [21] Salihu (2011) Maternal age, parity, race, smoking, education, marital status, adequacy of prenatal care, maternal height, gender of the infant, and year of birth 19 [31] Robinson (2010) Maternal age, maternal education, presence of the biological father in the family home, family income, stress in pregnancy, child’s age at follow up (and child’s age at 20 follow-up squared), and maternal cigarette smoking. 21 [32] Jaddoe (2007) Controlled for maternal body mass index, smoking, educational level, height, ethnicity, parity and age and infant gender; birth weight and low birth weight models also 22 controlled for gestational age. [43] Bille (2007) Parental age and social class. 23 [33] Sayal (2007) Gender, smoking, cannabis use and use of illicit drugs in the first trimester, highest level of maternal education, home ownership, marital status, parity, maternal age 24 group, high EPDS score, child ethnicity, gestational age group, and birth weight. 25 [34] Albertsen (2004) Type 1 diabetes, age, previous preterm delivery, smoking during pregnancy, coffee consumption during pregnancy, occupational on October 2, 2021 by guest. Protected copyright. status in the household, parity, and total 26 alcohol consumption during pregnancy. 27 [35] Lundsberg (1997) Small for gestational age: smoking in month 7, ethnicity, weight, height, infant sex, parity, bleeding during pregnancy, high blood pressure, and preeclampsia/eclampsia. 28 Low birthweight: smoking in month 7, height, weight, ethnicity, infant sex, parity, coffee use in month 7, exercise in third trimester, employment, bleeding during 29 pregnancy, high blood pressure, pre-eclampsia/eclampsia, anomalies, and placental problems. Preterm delivery: smoking in month 7, height, parity, age, caffeine use in 30 month 7, exercise first 16 weeks, bleeding during pregnancy, high blood pressure, pre-eclampsia/eclampsia, anomalies, and placental problems. 31 [41] Windham (1997) Maternal age, prior spontaneous abortion, gestational age at interview, and cigarette and caffeine consumption in week before interview. 32 [36] Passaro (1996) Gestational age, infant sex, parity, maternal smoking, and maternal body mass index. 33 [37] Shu (1995) Gestational age, parity, smoking and income. 34 [20] Peacock (1995) Unadjusted 35 [38] Olsen (1991) Age, school education, parity, alcohol and smoking entered the model as “dummy variables”. 36 [42] Ogston (1992) Gestational age at birth, sex, mother’s age, parity and smoking. 37 [46] Parry (1992) Gestational age at birth, sex, mother’s age, parity and smoking. 38 [39] Brooke (1989) Gestational age, sex, maternal height, and parity 39 [44] Ernhart (1989) Parity, smoking, race and year of study. 40 41 42 43 44 45 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

BMJ Open Page 44 of 51

1 2 3 4 5 Supplementary Table 3. Assessment of studies using the The Newcastle-Ottawa Scale (NOS) 6 7 8 Representative Selection of the Ascertainm Outcome of Comparability Assessment of Follow-up Adequacy Risk of bias 9 ness of the non-exposed ent of interest was not of cohorts on outcome blind/ long enough of follow questions 10 exposed cohort cohort exposure present at start of the basis of the record linkage for outcome to up of 11 For peer reviewstudy design/analysis only occur cohorts 12 Study ID 13 14 ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ 15 Albertson 2003 16

17 ✔ ✔ ✔ ✔ ✔ ✔ http://bmjopen.bmj.com/ ✔ ✔ Anderson 2012 18 19 ✔ ✔ ✔ ✔ ✔ ✔ ✔ ? 20 Bille 2007 21 ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ 22 Brooke 1989 23 24 x ✔ ✔ ✔ ? ✔ ✔ x 25 Erhart on October 2, 2021 by guest. Protected copyright. 26 ✔ ✔ x ✔ ✔ ✔ ✔ ✔ 27 Jaddoe 2007 28 ✔ ✔ ✔ ✔ ✔ x ✔ x 29 Larsen 2013 30 31 Lundsberg (Lisbet) ? ✔ ✔ ✔ ✔ ✔ ✔ ✔ 32 2015 33 Lundsberg (Lisbet) ? ✔ ✔ ✔ ✔ ✔ ✔ ✔ 34 1997 35 36 McCarthy 2013 ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ 37 38 Miyake 2014 ? ✔ x ✔ ✔ ✔ ✔ ✔ 39 40 41 42 43 44 45 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

Page 45 of 51 BMJ Open

1 2 3 Niclasen 2013 ? ✔ ? ✔ x x ✔ ? 4 5 6 Nykjaer 2014 x ✔ x ✔ ✔ ✔ ✔ ✔ 7 8 Ogston 1992 ? ✔ ✔ ✔ x ✔ ✔ ✔ 9 10 Olsen 1991 ? ✔ x ✔ ✔ ✔ ✔ ✔ 11 For peer review only 12 13 Parry 1992 ? ✔ ✔ ✔ ✔ x ✔ x 14 15 Passaro 1996 ? ✔ x ✔ ✔ ✔ ✔ ✔ 16

17 http://bmjopen.bmj.com/ Peacock 1995 ? x ? 18 ✔ ✔ ✔ ✔ ✔ 19 20 Robinson 2010 ? ✔ x ✔ ✔ x ✔ ✔ 21 22 Salihu 2011 ? ✔ x ✔ ✔ ? ✔ ✔ 23 24 Sayal 2012 x ? 25 ✔ ✔ ✔ ✔ ✔ ✔ on October 2, 2021 by guest. Protected copyright. 26 27 Sayal 2006 ✔ ✔ x ✔ ✔ x ✔ ✔ 28 29 Shu 1995 ✔ ✔ ✔ ✔ x ✔ ✔ ✔ 30 31 Windham 1997 x 32 ✔ ✔ ✔ ✔ ✔ ✔ ✔ 33 34 35 36 37 38 39 40 41 42 43 44 45 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open Page 46 of 51 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 For peer review only 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 Supplementary figure 1a) Odds ratios for externalising behaviour comparing low alcohol consumption (up to 32 32g/week) with no alcohol consumption (3 studies); 1b) Odds ratios for internalising behaviour comparing 33 low alcohol consumption (up to 32g/week) with no alcohol consumption (2 studies); OR: Odds ratio, CI: Confidence intervals. 34 http://bmjopen.bmj.com/ 35 297x209mm (300 x 300 DPI) 36 37 38 39 40 41

42 on October 2, 2021 by guest. Protected copyright. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 6 7 6 6 7 7 6 6 4 1 7 7 13 6&7 table 1table table 1table Figure Figure 1 Page Page No Reported on Reported supplementary supplementary BMJ Open http://bmjopen.bmj.com/ Recommendation on October 2, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml For peer review only Description of relevance appropriatenessor of studies for assembled assessing the hypothesistested to be 9 9 to include Effort available studies,all including contact with authors 8 8 Search including strategy, time period included in the synthesis keyand words 7 7 Qualifications of searchers librarians and(eg, investigators) 6 6 Study population 5 5 of Type designsstudy used 4 4 of exposure Type or intervention used 3 3 Description of study outcome(s) 2 2 Hypothesis statement 1 1 Problem definition 17 16 Description of any contact with authors 15 ofhandling Method abstracts and unpublished studies 14 ofaddressing Method articles published in languages than other English 13 ofcitations List located those and excluded, including justification 12 hand Use (eg,of searching reference of obtainedlists articles) 11 Search used, software version, andname including special features (eg, used explosion) 10 Databases registriesand searched Item NoItem Reporting of methods includeshould Reporting of search Reporting of search strategy should include Reporting of background includeshould MOOSE Checklist for Meta-analyses of Observational Studies Observational of forMeta-analyses MOOSEChecklist Page 47 of 51 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from Page 48 of 51

8 7 19 7&8 9&10 27-36 Table 1-3 Table table 2 82 table & Figures 2&3Figures Figures 2&3Figures Figures 2&3Figures supplementary BMJ Open http://bmjopen.bmj.com/ on October 2, 2021 by guest. Protected copyright. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml For peer review only ationale forationale the selection ofcoding dataand (eg, sound principles clinical or regression on possibleregression predictorson results of study R Description of statistical(eg, methodscomplete descriptionof randomfixed or effects justification models, whether of the chosen models account for predictorsof study dose-responseresults, models, cumulativeor meta-analysis) in sufficientto detail be replicated Documentation of how dataand classifiedwere coded raters,multiple (eg, blinding and reliability)interrater convenience) appropriate) Assessment ofAssessmentquality, studyincluding blinding of assessors, quality stratification or Assessment ofAssessment confounding (eg, comparability of and cases controls where studies in 18 28 of Indication statistical ofuncertainty findings 27 ofsensitivity Results testing (eg, subgroup analysis) 26 giving Table descriptive information for each study included 25 Graphic summarizing individual study estimates andoverall estimate 24 ofappropriate Provision graphicsandtables 23 22 of Assessment heterogeneity 21 20 19 Reporting of resultsReporting of should include 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from 6 Figure 1 Figure supTable supTable 1 6 5 & 5 6 5 5&6 4 3 1 on page # # on page Reported Reported up) and report up) and (e.g.,characteristics years considered, - BMJ Open http://bmjopen.bmj.com/ analysis). analysis). - on October 2, 2021 by guest. Protected copyright. Describe methodDescribe of data extraction reportsfrom (e.g., forms, piloted independently,induplicate) and any processes for andobtaining confirming data investigators.from State the process theState for selecting studies (i.e., screening, eligibility, review, included in systematic and, ifapplicable, included inmeta the Present full electronic full Present strategy for search leastone at includingdatabase, any limits used,suchthat it could be Describe allinformationDescribe (e.g., databasessources with of dates coverage, contact study authors with to identify Specify studySpecify characteristics(e.g., PICOS,length of follow Indicate reviewa Indicate if protocol if andexists, where accessedWebit can be (e.g., and, available,address), if provide Provide an Provide explicit statement of questions being withaddressed reference to participants, interventions, comparisons, outcomes, anddesign study (PICOS). Describe the rationale theDescribe review for thein the of context known.already what is Provide a Provide structured including,summary as applicable:background; dataobjectives; sources; study eligibilitycriteria, andparticipants, interventions; appraisalstudy and synthesis methods; results;limitations; conclusions and Identify the report Identify a as systematicmeta-analysis, review, both. or Checklist Checklist item language, publicationlanguage, status) used forcriteriaas eligibility, rationale. giving repeated. repeated. additional additional in studies)the search dateand last searched. registration registration information registrationincluding number. implications of of implicationskey findings; systematic review number. registration For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

1 # 8 2 4 9 7 3 For peer review6 only

Data Data collection process 10 Study Study selection Search Search Information Information sources Eligibility Eligibility criteria Protocol andProtocol registration 5 METHODS METHODS Objectives Rationale Rationale INTRODUCTION INTRODUCTION Structured summary ABSTRACT ABSTRACT Title TITLE TITLE Section/topic Section/topic Checklist 2009 PRISMA Page 49 of 51 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

Page 50 of 51 Figure Figure 2&3 13-14 Figure Figure 2&3 supfigure1 Figure Figure 2&3 supfigure1 7 Table 1-3 Table 1-3 Figure 1 Figure 8 7 Figure Figure 2&3 supfigure1 Figure Figure 2&3 supfigure1 6 & 6 7 7 & 7 8 BMJ Open http://bmjopen.bmj.com/ analysis. analysis. - specified. specified. - for metafor each on October 2, 2021 by guest. Protected copyright. ) 2 reporting within reporting studies). Give results results ofadditionalGive analyses,ifdone (e.g., sensitivity or subgroup analyses,meta-regression Item [see 16]). Present results ofPresent any assessment of risk acrossof bias studies (see Item 15). Present results ofPresent meta-analysis each done, including confidence intervalsand measuresof consistency. For For all outcomes considered (benefits harms), orfor present, each (a) study: simple summary data for each Present data Present onrisk eachbias studyof of and, if available,any outcome level assessment (see 12). item For each For each presentstudy, characteristicswhich for data were (e.g., studyextracted PICOS, follow-upsize, period) and the provide citations. Give numbersofGive studies assessedscreened, foreligibility, included inand review, the with reasons for exclusions at Describe methodsDescribe of additional analyses (e.g., sensitivity subgroup or analyses, meta-regression), done, if indicating prewerewhich Specify any Specify assessment bias of risk maythat of affect the cumulative (e.g.,evidence publication bias,selective Describe the methods theDescribe of handling data and combining results of studies, if done, including measures of consistency I (e.g., State the principal theState measuressummary (e.g.,risk ratio, in difference means). Describe methodsDescribe used for riskassessing bias individual of of studies (including specification of whetherthis was at the donestudy outcomeor level), howand information this to be is usedany in data synthesis. List and define and List all variables for which data were sought (e.g., PICOS, funding sources)and any assumptions and made.simplifications intervention group (b) intervention group andeffectestimates confidence intervals,with a ideally plot.forest each with each stage, flowa ideally diagram. For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

12 11 21 14 17 16 For peer review only 23

Risk of bias Risk acrossof studies 22 Synthesis resultsSynthesis of Results ofindividualResults studies 20 Risk of bias Risk withinof studies 19 Study Study characteristics 18 Study Study selection RESULTS RESULTS Risk of bias Risk acrossof studies 15 Synthesis resultsSynthesis of Summary measuresSummary 13 Risk of bias Risk inof individual studies Data items Data items Additional Additional analysis Additional Additional analyses 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

BMJ Open: first published as 10.1136/bmjopen-2016-015410 on 3 August 2017. Downloaded from

17 13 13 & 13 14 13&14 13&14 supfigure1 BMJ Open http://bmjopen.bmj.com/ on October 2, 2021 by guest. Protected copyright. Describe Describe offundingsources forthe systematic review (e.g., other supportand supply ofdata); ofrole the funders for Provide a Provide general interpretation of thethe results in context of otherevidence,and forimplications future research. Discuss limitationsDiscuss at and study (e.g.,outcome level andrisk of bias), review-level (e.g.,incompleteat retrieval of research,reportingidentified bias). Summarize Summarize the main findings theincluding strength of evidence for outcome;maineach theirconsider relevance to systematic systematic review. key groups (e.g.,key healthcare providers, users, policyand makers). For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

25 27 For26 peer review only

Funding Funding FUNDING FUNDING Conclusions Limitations Summary of of Summary evidence 24 DISCUSSION DISCUSSION Page 51 of 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60