Preconception exposures and adverse pregnancy, birth and postpartum outcomes: Umbrella review of systematic reviews

Background: Preconception exposures have been associated with adverse pregnancy, birth and postpartum outcomes. However, the reports, statements and guidelines of national and international health organisations vary in what they recommend indi - viduals should monitor, avoid, reduce or practise in the preconception period. Objectives: To synthesise and evaluate the evidence across systematic reviews for associations between exposures before conception and adverse pregnancy, birth and postpartum outcomes. Data sources: MEDLINE, Embase, Epistemonikos (to May 2020) and reference lists of included reviews, without language or date restrictions. Study selection, data extraction and synthesis: Systematic literature reviews of ob - servational and/or interventional studies reporting associations between preconcep - tion exposures in women and/or men of reproductive age and pregnancy, birth or postpartum health outcomes were included. The


| BACKG ROU N D
Adverse pregnancy, birth and postpartum outcomes continue to have a substantial impact on morbidity and mortality.Worldwide, there are around 23 million miscarriages, 1 14.8 million live preterm births, 2 295,000 neonatal deaths due to congenital anomalies 3 and 295,000 maternal deaths due to pregnancy or childbirth complications each year. 46][7] However, a recent review by Schoenaker et al. 8 found that the clinical guidelines, position statements and policy reports of national and international health organisations vary in the preconception exposures they name as being important.Moreover, many of these guidelines, statements and reports fail to define the criteria used to determine exposure eligibility or evaluate the quality and strength of the evidence supporting the associations of these exposures with adverse outcomes.Systematic reviews-considered the gold standard method of retrieving, synthesising and appraising all available, relevant evidence for clinical questions 9 -are commonly used to inform the creation of guidelines and position statements and play an important role in clinical decision-making and evidence-based healthcare. 10While many systematic reviews reporting associations between preconception exposures and adverse pregnancy, birth and postpartum outcomes may exist, these are likely to vary in their scope and quality, and may present conflicting findings. 11For instance, while a systematic review by Rumbold et al. concluded that maternal periconception vitamin supplementation does not affect the risk of stillbirth, 12 a later systematic review reported an association between this exposure and reduced stillbirth risk. 13 offer researchers, clinicians and policymakers an overview of the complete body of systematic review evidence in this area, we synthesised the findings, quality and certainty of the evidence reported in systematic reviews of interventional and/or observational studies investigating the link between preconception exposures and adverse pregnancy, birth and postpartum health outcomes.

| ME THODS
The umbrella review (systematic review of systematic reviews) was prospectively registered with PROSPERO (registration number CRD42020196511) and is reported following PRISMA guidelines. 14view board approval was not required as this is a review of publicly available literature.

| Data sources
We searched MEDLINE, the systematic review repository Epistemonikos and the Cochrane Library databases from their inception to 21 May 2020.We combined relevant medical subject heading (MeSH) terms and word variants for 'preconception,' 'periconception,' 'interpregnancy' and 'contemplating pregnancy' with the Centre for Reviews and Dissemination's Search Strategy (2.1) for retrieving systematic reviews and meta-analyses 15 and a 'sensitivity-and-precision-maximising strategy' for retrieving overviews of systematic reviews. 16This search strategy was developed with a subject librarian for use in MEDLINE (Table S1) and was then interpregnancy weight gain and advanced paternal age are important preconception considerations.High-quality research is required to better understand other exposure-outcome associations.

K E Y W O R D S
GRADE approach, preconception care, pregnancy, pregnancy outcome, systematic review

Study question
What is the certainty of the evidence across systematic reviews for associations between preconception exposures and adverse pregnancy, birth and postpartum outcomes?

What's already known
Preconception exposures have been associated with adverse pregnancy, birth and postpartum outcomes.
However, the clinical guidelines, position statements and policy reports of national and international health organisations vary in what they recommend in the preconception period, creating uncertainty for clinicians, policymakers and patients.

What this study adds
This review has identified high-and moderate-certainty evidence that maternal preconception folate supplementation, body mass index, interpregnancy weight change and physical inactivity as well as advanced paternal age are associated with adverse pregnancy, birth and postpartum outcomes.adapted for use in Epistemonikos and the Cochrane Library.No language, publication date, ethnicity or parity restrictions were applied.Where the full texts of eligible reviews were unavailable, their authors were contacted to request these.The reference lists of eligible reviews were also hand-searched to identify additional relevant articles.

| Study selection and data extraction
The eligibility criteria were informed by the PECOS (population, exposure, comparison, outcomes and study designs) framework presented in Table 1.We included systematic reviews of observational and/or interventional studies.In line with the definition detailed in the Cochrane handbook for systematic reviews, 17 these needed to report: the eligibility criteria used for study inclusion, a pre-defined, specific research question(s), a transparent and systematic methodology (i.e. a reproducible search strategy with specific search terms provided), an analysis (meta-or narrative) including all retrieved, eligible studies and an assessment of the validity (i.e.risk of bias) of all included studies.Reviews were required to include at least two studies, involving adult (≥16 years) women and/or men, that reported an exposure(s) occurring before conception and its association with a child and/or maternal pregnancy, birth or (up to six weeks) postpartum health outcome(s), relative to absence or lower dose(s) of the exposure.
Exposures that continued into the broader periconceptional period-defined as the time before and up to 10 weeks after conception-were eligible if the exposure clearly began before conception and preconception-specific results were not reported separately (e.g.periconceptional folic acid supplementation). 18Maternal weight measured in the first trimester or at the first antenatal appointment was also eligible, as this has been found to be a valid proxy measure of pre-pregnancy weight due to minimal gestational weight gain during this period. 19As specific guidelines exist for exposures and outcomes unique to fertility (e.g.assisted reproduction, sperm count, pregnancy rate), 20 prescription-only medication, 21 and specific patient groups such as women with pregestational diabetes, 22 asthma, depression, epilepsy, human immunodeficiency virus (HIV) and thyroid disorders, 23 these were excluded.
To identify articles of potential relevance, two reviewers independently screened the first 10% of abstracts against the eligibility criteria in a 'pilot phase', 24 to identify any divergences in interpretation of the criteria.Disagreements were discussed and resolved by consensus after consulting the protocol.As the agreement was 92%, the remaining 90% were screened by the lead reviewer.This process was repeated at the full-text review stage, where agreement between both reviewers was 98%.The lead reviewer therefore soloreviewed the remaining texts, coding each as either eligible or ineligible and providing a reason if the latter.The list of exclusion decisions is provided as a supplementary file.A data collection spreadsheet was then piloted by both reviewers using a 10% sample of eligible reviews.No divergences were identified, and the lead reviewer therefore performed the remaining data collection alone, with the completed spreadsheet reviewed by the remaining investigators.
We extracted data on: review publication year, databases searched, eligibility and search restrictions, preconception exposures and eligible outcomes investigated, included primary studies and their design, country (categorised by income according to the World Bank classification 25 ), comparators, covariates and review author-assigned quality rating.We also extracted pooled summary statistics (e.g.estimates of association and uncertainty, corresponding descriptive statistics, the number of participants and events).
Adjusted estimates were prioritised where these were reported without missing information.Where data were missing, review authors were contacted for additional information, which was supplemented as required.

| Overlapping reviews
There were a number of 'overlapping reviews', where multiple systematic reviews addressed identical or near-identical research questions with many of the same primary studies included.To avoid bias through double-counting of outcome data, Cochrane's 'evidencebased decision tool' was used. 11If an eligible Cochrane Review comprehensively covered the available literature for a relevant exposure-outcome association in the same population and setting, only this was included, as Cochrane endeavours not to publish multiple reviews on any given topic. 11If an eligible Cochrane Review either did not exist or was not all-inclusive, all non-overlapping reviews were retained and the overlapping review covering the greatest number of studies for the association in question was used in order to maximise the amount of outcome data included in the review while still avoiding the issue of review overlap. 11

| Quality of included reviews
The methodological quality of included reviews was assessed using the AMSTAR 2 tool, which has been validated for use with systematic reviews of both randomised and non-randomised studies. 26ch review was assigned an overall quality rating of critically low, low, medium or high quality based on its ratings on applicable items of the tool's 16 quality criterion. 26Some of these items, such as whether included studies were described in 'adequate' detail, have been criticised for being susceptible to subjective opinions and biases. 27The methodological quality of the first 10% of included reviews-sorted alphabetically by first author surname-was therefore assessed by two reviewers to identify any divergences in interpretation of the item criteria.Disagreements were discussed and resolved by consensus after consulting the protocol.As inter-rater agreement across the tool's 16 criteria was 82.3% overall, the remaining 90% were reviewed by the lead reviewer alone.

| Quality of evidence in included reviews
We used the GRADE approach 28 -the most widely used systematic framework for assessing evidence quality 29 -to rate each identified exposure-outcome association as being of either high, moderate, low or very low certainty depending on the likelihood that the true effect differs substantially from the reported estimate.As per GRADE guidance, 28 randomised controlled trial evidence was initially assigned a high rating and, due to likely residual confounding, evidence involving observational data was initially rated as low.This rating was then either decreased or increased by a maximum of two levels after accounting for issues relating to: risk of bias, imprecision, inconsistency, indirectness, publication bias, magnitude of effect, dose-response effects and residual confounding. 29Further information on how decisions were made for each of these eight domains can be found in Table S2.The completed GRADE decisions spreadsheet was reviewed by all investigators.For transparency we have presented the decisions made for each review across each GRADE domain in Table S3.

| RE SULTS
The PRISMA flow diagram 30 in Figure 1 shows that we identified 4,004 articles and screened 3,592 abstracts.Of the 661 articles that underwent full-text screening, 592 were ineligible and 22 were removed as they reported only associations covered by more comprehensive or Cochrane Reviews (see Table S4), resulting in 47 reviews being included. 12, Six ligible reviews were identified through citation tracking, [77][78][79][80][81][82] resulting in a final sample of 53 reviews.

| Review characteristics
The characteristics of the 53 reviews, including the number of databases and range of publication years searched, eligibility criteria and preconception exposures reported, are presented in Table S5.

| Methodological quality of included studies and certainty of the evidence
Two included reviews were rated as 'high'-quality, 12,41 two as 'moderate', 38,61 11 as 'low' 36,39,42,50,51,57,60,63,68,73,77 and the remaining 38 as 'critically low' (Table S6).The most common 'critical' methodological concerns were lack of: appropriate methods for combining results (76.9% of relevant reviews-primarily due to a lack of justification for combining data in a meta-analysis and/or combining unadjusted estimates), justification for study exclusions (76.9%), comprehensive literature search (59.6%) and evidence of a priori establishment of review methods (50.0%).Two exposure-outcome associations were rated as 'high'-certainty, 27 as 'moderate', 56 as 'low' and 120 as 'very low' (Table S3).Risk of bias was the most common reason for downgrading evidence certainty.

| Synthesis of results
Figures 2-6 show the exposure-outcome associations, grouped by their exposure domain, along with key information such as whether the exposure continued into the broader periconceptional period and whether the included studies had observational or interventional designs.Associations that were assigned a 'very low' GRADE rating are not included in the figures or in the narrative synthesis below, as very little confidence can be had in these findings. 83A summary of the findings for all 205 associations, regardless of GRADE rating, can be found in Table S7.Of the 205 associations identified, 191 (93.2%) relate to maternal exposures.These include pre-pregnancy body mass index (BMI; 75 associations), interpregnancy or birth intervals (23), interpregnancy weight change (18), age (14), folate supplementation (11), diet and nutrition (10), parity and education (seven each), physical activity (five), abuse or neglect, over-the-counter drugs and environmental exposures (four each), smoking (active or passive exposure), other vitamin supplementation and immigration status (two each), and vaccination, alcohol consumption and ethnicity (one each).The remaining 14 (6.8%) associations are related to the paternal exposures of age (10) and alcohol consumption (four).
For associations relating to BMI, the reference group was 'normal' BMI (20-24.9kg/m 2 ) and underweight, overweight and obesity corresponded to BMI values of <20, 25.0-29.9and ≥30 kg/m 2 unless otherwise stated.Adjusted risk estimate is used where review authors combined adjusted odds ratios, hazard ratios and rate ratios without converting them to a single measure of association.

| Health behaviours and wider determinants of health
Figure 2 shows there is high-certainty evidence that maternal folate supplementation (<0.Low-certainty evidence was found that maternal periconception folate supplementation is associated with a reduced risk of preterm birth 54 and that there is limited evidence of associations between this exposure and the risk of twinning, miscarriage, low birthweight, congenital heart defects and birth defects other than neural tube and congenital heart defects and cleft lip and palate. 41Lowcertainty evidence was also found that maternal vitamin supplementation (any) beginning preconceptionally and terminating in the first trimester does not affect the risk of either miscarriage or stillbirth, 12 and that paternal consumption of any alcohol in the 3 months before conception is associated with an increased risk of congenital heart defects. 76Further low-certainty evidence was found for associations between maternal preconception iron intake and an increased risk of gestational diabetes mellitus, 59 adherence to a Mediterranean or High Alternate Healthy Eating Index diet and a reduced risk of gestational diabetes mellitus, 59 and maternal experience of abuse (physical, emotional or sexual) at any time before pregnancy and an increased risk of low birthweight. 61

| Demographic and reproductive exposures
Figure 3 shows moderate-certainty evidence that paternal age of ≥40 years is associated with a greater risk of miscarriage, and lowcertainty evidence that paternal age of 35-39 years may not be associated with greater or reduced risk of this outcome. 42Low-certainty evidence was found that paternal age of <20 years is associated with an increased risk of spina bifida, 51 that maternal age of ≥35 years is associated with a greater risk of urinary incontinence, 36 and that maternal age of ≥45 years is associated with increased risk of: an abnormal five-minute Apgar score, foetal loss, pregnancy complications and caesarean delivery, relative to maternal age of <45 years. 52w-certainty evidence was also found that interpregnancy intervals   shorter than 6 months are associated with an increased risk of low weight, small-for-gestational-age and preterm births, and that interpregnancy intervals of ≥60 months are also associated with an increased risk of these outcomes. 38Further low-quality evidence was found that interpregnancy or birth intervals of >48 months are associated with a greater risk of pre-eclampsia 39 and that multiparity is associated with an increased risk of urinary incontinence, relative to nulliparity but not primiparity. 36

| Pre-pregnancy body mass index (BMI)
Figure 4 shows moderate-certainty evidence that maternal BMI in the overweight (BMI 25-29) and obese (BMI >30) ranges is associated with a greater risk of gestational diabetes mellitus, 74 pregnancy-induced antenatal hypertension 65 and pre-eclampsia 81 and a reduced risk of placental abruption. 31Moderate-certainty evidence was also found that maternal obesity is associated with an increased risk of foetal distress, 82 shoulder dystocia (vs BMI <30), 75 foetal macrosomia (vs BMI 18.5-24.9) 56and large for gestational age births (vs BMI <25). 44Further moderate-certainty evidence was found that, relative to the lowest category of BMI, increasing maternal BMI is associated with a greater risk of both miscarriage and neonatal death, 77 and that maternal underweight (BMI <18.5) is associated with a reduced risk of pregnancy-induced hypertension. 65gure 5 shows low-certainty evidence that maternal underweight is associated with a reduced risk of pre-eclampsia, 81 gestational diabetes mellitus, 74 large-for-gestational-age and macrosomic births 56 and an increased risk of small-for-gestational age birth 55 and placental abruption. 31Regarding maternal overweight, lowcertainty evidence suggested associations between this exposure and an increased risk of foetal distress, 82 large-for-gestational-age and macrosomic births 56 and a reduced risk of small-for-gestationalage birth. 56Further low-certainty evidence was found that maternal overweight and obesity are both associated with an increased risk of miscarriage, 35 post-term pregnancy and a longer hospital stay 82 and that maternal obesity, specifically, is associated with an increased risk of maternal depression, 60 failure to progress in labour, caesarean delivery, a low Apgar score at five minutes postpartum 82 and neonatal intensive care unit admission. 56

| Interpregnancy weight change
Figure 6 shows moderate-certainty evidence that maternal interpregnancy weight gain of >2 BMI units, relative to BMI maintenance (change of ≤2 units), is associated with a greater risk of gestational hypertension, 57 and that interpregnancy weight gains of both 1 to <3 and ≥3 BMI units, relative to BMI maintenance (≤1 unit change), are associated with greater risk of caesarean delivery, 63,73 large-for-gestational-age birth 73 and gestational diabetes mellitus. 72Low-certainty evidence was found that: interpregnancy BMI losses and gains of >1 unit are associated with an increased and reduced risk of small-for-gestational-age birth, respectively 63 ; that interpregnancy BMI reduction of >1 unit is associated with an increased risk of preterm birth and a reduced risk of large-for-gestational-age birth 73 ; and that interpregnancy weight gain of >2 units is associated with an increased risk of pre-eclampsia. 57I G U R E 4 Pre-prepregnancy weight/BMI exposures and their associations with adverse pregnancy, birth and postpartum outcomes (moderate GRADE ratings only).Legend: Pre, Preconception; Post, Postconception; Obs, Observational; Int, Interventional; BMI, Body mass index; OR, Odds ratio; aRR, Adjusted risk ratio.a Author defined; b Pregnancy-induced BMI 20-24.9kg/m² BMI 20-24.9kg/m² BMI 18.5-24.9kg/m² BMI 18.5-24.9kg/m² BMI 18.5-24.9kg/m² BMI 18.5-24.9kg/m²   exposures were not pre-specified in the search strategy, and no language, publication year or study design search restrictions were applied, resulting in the inclusion of a diverse range of exposureoutcome associations.We used the GRADE approach, which critically evaluates evidence certainty; this type of summary, which incorporates both an assessment of methodological quality and strength of the evidence base, is useful to clinical practice and policymakers. 83

| Limitations of the data
Our review has some noted limitations.We did not search for grey literature; unpublished systematic reviews may contain further associations or cover reported associations more comprehensively.
Moreover, where a Cochrane Review either did not exist or was not all-inclusive for a particular exposure-outcome association, our selection of the overlapping review containing the greatest number of studies for the association may not have reflected the most methodologically sound review and the quality of the data we have reported is ultimately constrained by the quality of the systematic reviews from which they are derived.However, as an inevitable limitation of umbrella reviews is that the included systematic reviews may not have included relevant primary studies, this approach allowed us to maximise the number of studies covered while still avoiding the issue of review overlap.Nonetheless, whether any 'missed' primary studies would have been sufficient to alter our findings remains unclear.
There also remains the possibility that some reviews have not described relevant exposures as pre-, peri-or inter-conception factors in their titles or abstracts and were therefore not identified by our search.While our use of citation tracking helped to identify some of these reviews, it is possible others may have been missed.Moreover, while the GRADE approach has been argued to reflect 'most' of the Bradford Hill criteria for establishing causality, it was not explicitly developed for this purpose. 84Our GRADE ratings should therefore only be interpreted as reflecting the certainty of the evidence for reported exposure-outcome associations and not as evidence that reported exposures cause their associated outcome.Lastly, while we took care to report exposures occurring only within the periconceptional period, it is possible that some observed associations-particularly those derived from observational studies relating to late pregnancy or postpartum outcomes-may be at least partially attributed to an exposure's continued presence beyond the periconceptional period.

| Interpretation
Fewer than half of women practice preconception folate supplementation in countries such as England, Scotland and the United States, 85 mean paternal age at conception continues to rise in highincome countries, 86,87 and globally, approximately 55% of women are overweight or obese 88 and 31.7% are not sufficiently physically active. 89The main implication of our findings for public health policy is that there needs to be greater investment in policies and interventions to support preconception supplementation, physical activity, maintenance of a healthy weight and reproductive life planning.Improvement in these exposures is likely to reap long-term health and economic dividends. 90Moreover, while the review by Schoenaker et al. 8 found that maternal folic acid supplementation and weight were listed in the preconception guidelines, statements and reports of five of the health organisations considered, maternal physical activity was mentioned in the guidelines, statements and reports of just three organisations, paternal age in just one and interpregnancy weight change was not mentioned in any.This policy divergence suggests that greater collaboration may be required to achieve consensus in messaging around preconception health.
The findings of this review also highlight several research gaps.
Few reviews or primary studies considered the preconception health of men, with only 6.8% of reported associations relating to paternal exposures.There was also a gap in the examination of the wider determinants of physical and mental health, such as pollution, poverty and abusive relationships, relative to individual-level exposures like maternal weight and folate supplementation, where multiple systematic reviews presented overlapping data for the same outcomes.There is a need for higher-quality systematic reviews in this area, with a priori protocol registration, comprehensive literature searches, justified study exclusions and appropriate statistical methods.There is also a need for lower-and middle-income countries to be better represented in this research and, where possible, for more randomised controlled trials and other research seeking to establish epidemiologic evidence of causality to be conducted.

| Conclusions
In conclusion, we found high-and moderate-certainty evidence that maternal preconception folate supplementation, BMI, interpregnancy weight change, physical inactivity and advanced paternal age are associated with adverse pregnancy, birth and postpartum outcomes.

ACK N OWLED G EM ENTS
The authors thank the authors of included articles who supplied additional data and information and Sarah Herring for her assistance in developing the review's search strategy.

Michael
Included reviews were published between 2005 and 2020, with the publication dates for their included studies ranging from 1973 to 2019.Two were published in Portuguese40,80 and therefore translated to English, with the remainder published in English.There were 205 exposure-outcome associations reported across 1,383 studies.Twelve of these associations were based on clinical and/or randomised trials, and five included evidence from a combination of trials and observational studies.The remainder relied on observational studies only.A total of 994 (71.9%) primary studies were conducted in high-income countries and 355 in low-and middle-income countries.

F I G U R E 5
Principal findingsOur review identified 53 systematic literature reviews reporting 205 unique associations for exposures in the preconception period.Only 17 of these associations used evidence gathered from randomised trials.The methodological quality of reviews was generally poor; only two reviews were assessed as high-and two as moderate quality.The only high-certainty findings were that maternal folate supplementation commencing preconceptionally reduces the risk of neural tube defectaffected births and pregnancy terminations for a foetal anomaly.4.2 | Strengths of the studyStrengths of this review include its comprehensive search, where both observational and interventional studies were included, Pre-prepregnancy weight/BMI exposures and their associations with adverse pregnancy, birth and postpartum outcomes (low GRADE ratings only).Legend: Pre, Preconception; Post, Postconception; Obs, Observational; Int, Interventional; BMI, Body mass index; OR, Odds ratio; (a)RR, (adjusted) Risk ratio; NICU, Neonatal Intensive Care Unit.a Author defined; b At five minutes postpartum 29.9 kg/m² BMI 25-29.9kg/m² BMI ~25-29.9kg/m² BMI ≥30 kg/m² BMI ≥30 kg/m² BMI 25-29.9kg/m² Review (AMSTAR 2 rating) Torloni et al. [critical] Wang et al. [critical] Liu et al.b [critical] Liu et al.b [critical] Adane et al. [critical] Liu et al.a [critical] Molyneaux et al. [low] Heslehurst et al. [critical] Heslehurst et al. [critical] Heslehurst et al. [critical] Heslehurst et al. [critical] Heslehurst et al. [critical] Heslehurst et al. [critical] Heslehurst et al. [critical] Heslehurst et al. [critical] Liu et al.b [critical] Liu et al.b [critical] Balsells et al. [critical] Balsells et al. [critical] Liu et al.b [critical] Liu et al.b [critical] . Quenby S, Gallos ID, Dhillon-Smith RK, et al.Miscarriage matters: the epidemiological, physical, psychological, and economic costs of early pregnancy loss.The Lancet.2021;397:1658-1667.2. Chawanpaiboon S, Vogel JP, Moller A-B, et al.Global, regional, and national estimates of levels of preterm birth in 2014: a systematic review and modelling analysis.Lancet Global Health. 12019;7:e37-e46.