Prenatal maternal stress and risk of neurodevelopmental disorders in the offspring: A systematic review and meta-analysis protocol

Description of the exposure

A growing body of evidence increasingly supports the hypothesis that indicates that early adverse experiences, including maternal stress during pregnancy, may have a significant effect on the fetal and postnatal brain development and subsequent organization of behaviour¹. Prenatal maternal stress (PNMS) refers to a status of low or negative well-being experienced during pregnancy. The term “stress” in this context often covers negative life events, anxiety and depressive feelings, and the prevalence of depression, anxiety, and stress in pregnant women has been estimated at 12%, 28% and 31%, respectively². Stressful life events may include significant daily stressors and high stress perception at individual level, but may also include massive stress events such as natural disasters at the population level. A number of studies have reported an association between PNMS and increased risk of psychopathology in children^3–6. Specifically a number of studies have reported that maternal exposure to significant life stressors (such as bereavement, job loss, divorce, trauma) at critical periods during pregnancy may change neurobehavioral outcomes in exposed offspring^7–11.

Description of the outcomes

Attention-deficit/hyperactivity disorder (ADHD) is a neurodevelopmental disorder that affects up to 1 in 20 children in the USA¹². It is a developmental condition characterized by inattention, hyperactivity and impulsivity behaviour. These symptoms often continue into adulthood and can significantly impact quality of life¹³. Autism spectrum disorder (ASD) is another lifelong neurodevelopmental disorder characterized by three behaviourally altered domains including social deficits, impaired communication, and stereotyped and repetitive behaviours¹⁴, and these symptoms begin in early childhood^15,16. The aetiological basis of these disorders is complex and multifactorial, and while twin studies show that genetic factors play a key role^17,18, environmental factors, such as PNMS, have also been proposed to play a role in their aetiology¹⁹.

In support of this, there is a growing literature reporting an association between PNMS and increased risk of these neurodevelopmental disorders in exposed offspring^4–6,20. For example, Class et al. reported an increased risk of ADHD in offspring of mothers exposed to the death of a first-degree relative during pregnancy²⁰. Van Batenburg et al. found that maternal antenatal depressive symptoms were related to attention problems in affected children at 3–4 years of age²¹. Moreover, exposure to prenatal maternal anxiety has also been reported to be significantly associated with ADHD symptoms in early childhood²². Accumulating evidence suggests that the impact of PNMS and other environmental factors on the development of the fetal brain is dependent on the gestational stage at which the exposure occurs^23–25. Accordingly, evaluating the timing of the exposure to the stressor is particularly important, because associations during a specific time increase the probability of a causal association²⁶. For example, Li et al. found an association between preconception stress (experienced 0–6 months before pregnancy) and increased risk of ADHD, but only in male offspring⁶, while Class et al. found an increased risk of ASD and ADHD in children whose mothers were exposed to bereavement during the third trimester of pregnancy²⁰.

However, there are inconsistencies in the literature regarding the association between PNMS and neurobehavioral outcomes in affected offspring, with some studies finding no association between PNMS and the risk of neurodevelopmental disorders in the offspring^27–29. Moreover for those studies that have reported an association between PNMS and adverse neurodevelopmental outcomes in the offspring, there is no clear consensus on the period of gestation that is most critical^27,28,30.

Why it is important to perform this review

Developing a clear consensus on this issue is important as identifying children at increased risk of ASD and ADHD may allow for greater paediatric surveillance of those most at risk to facilitate early intensive behavioural interventions, which can improve behavioural outcomes^31–34. For example, the Early Start Denver Model is a behavioral intervention aimed at improving cognitive, social and behavioral outcomes of toddlers diagnosed with ASD. This intervention can improve cognitive and adaptive behaviour with reduction of the severity of the diagnosis³¹. Moreover, a consensus on this issue may underscore the need for alternative interventions during pregnancy aimed at reducing stress, anxiety and depression^35,36. Currently there is only one intervention that starts during pregnancy, involving multiple home visits by qualified nurses, known as the Nurse Family Partnership³⁶. This intervention has been reported to lead to long-term improvements in parenting behaviour, as well as in child outcome, (including but not limited to the reduction of anger and criminal behaviours)³⁷. Accordingly, a better knowledge of risk factors for neurodevelopmental disorders may allow increased screening and early intervention for children exposed to PNMS, with the aim of improving their long-term neurodevelopmental outcome.

Rationale for current systematic review

Evidence suggests that PNMS can have a role in increasing the risk of ASD and ADHD^9–11,20,38 and that antenatal depression symptoms could be related to child behavioural problems^10,21. However, a number of studies reported no association between PNMS and adverse neurodevelopmental outcome in children^27,28,30. Accordingly, the aim of this systematic review and meta-analysis is to summarise the existing literature on the effects of PNMS, anxiety and depression feelings on the risk of ASD and ADHD, and if possible, to provide a quantitative estimate of this relationship.

Research questions

The primary objective of this study will be to systematically report the evidence of an increased risk of ASD and ADHD in offspring exposed to PNMS.

The specific review questions to be addressed are:

Systematic review to identify eligible papers

Study registration. This study has been registered with the international Prospective Register of Systematic Reviews (PROSPERO) under the registration number CRD42018084222.

Study design. According to the Preferred Reporting Items for Systematic review and Meta-analysis protocol (PRISMA-P) guidelines³⁹, we will use a flow diagram to show the step by step study selection process and we will provide a rationale for those studies that are excluded studies. In addition we will report the descriptive characteristic of all studies included in the systematic review and we will report the meta-analysis results using a forest plot.

Ethics and dissemination. Ethical approval is not required for this systematic review and meta-analysis given that we will not include any experimental research on humans or include any identifying personal information which requires informed consent. The results of this systematic review and meta-analysis will be disseminated in peer-reviewed journals.

Types of studies. Cohort studies and case-control studies that assess maternal exposure to stressful events and the risk of ASD and/or ADHD diagnosis in the offspring, will be included in this review. The search will not be restricted by language, although we can only include studies in English and Italian because of limited resources. Studies in other languages with an English abstract will be described and reported in the review, and when possible, the full text of the article will be translated into English. Only studies published in peer-reviewed journals will be considered for the systematic review. Studies published every year from database inception until March 2018 will be considered. The following inclusion criteria in accordance with the PICOS principle (population, intervention, comparison, and outcome) will be considered.

Population. Studies evaluating women who have experienced stress or/and anxiety or depression symptoms at any time during pregnancy or in a particular trimester will be included. The search will not be restricted to humans, although only human studies will be included in the review.

Exposure. PNMS will be defined as any type of stress such as stressful life events, natural disasters, significant life events, maternal bereavement, anxiety feeling or traumatic stress measured by:

a) Validated self-report questionnaire

b) Interview

c) Physiological parameters, such as hormone levels (e.g. prolactin, corticosteroids etc.).

d) Objective measures of stress and population level events

Prenatal maternal depression will be defined according to screening scales, such as the Hamilton depression rating scale, the Beck Depression Inventory, the Patient Health Questionnaire, the Centre for Epidemiologic Studies Depression Scale. If possible, we will perform a subgroup analysis including studies on prenatal maternal depression and ASD or ADHD. This analysis will be clearly highlighted as post-hoc depending on the available data.

Outcomes. A diagnosis of ASD or ADHD.

Method for identifying studies for inclusion

Electronic searches. Two authors (NM and GWOK) will independently perform the search for relevant articles using PUBMED, PsycINFO, Web of Science, EMBASE and SCOPUS databases. In addition we will hand-check the bibliographies of studies selected for inclusion to identify if there are any other potentially eligible studies for inclusion.

Search strategy. We will use the following searching strategy:

(Prenatal OR Antenatal OR Pregnant) AND (Stress) OR (Distress) OR (Anxiety) OR (Bereavement) AND (Offspring OR Child) AND (Autism spectrum disorder) OR (Attention deficit/hyperactivity disorder).

The detailed search strategy is presented in Supplementary File 1.

Data collection, extraction and assessment

Study selection. Two authors (NM and GWOK) will independently screen the titles and the abstracts for all articles, in order to identify studies that potentially meet the inclusion criteria outlined above. The titles and abstracts obtained from each database will be collected and processed in Endnote X8®. Duplicates will be excluded using the Endnote function “remove duplicates”. The full texts of potentially relevant articles will be examined independently by both authors and it will be a requirement that all eligibility criteria be met in order to select each study for review.

Data extraction. Two authors (NM and FB) will independently extract data using a standardised data extraction form (see Supplementary File 2). Data extraction will include author and year of publication, study outcome, exposure (type, timing and duration), study design, offspring gender, data sources, cohort size, definition of the outcome used, exclusion criteria, crude and adjusted estimates [if reported, including the relative risk (RR), odds ratio (OR), hazard ratio (HR) and 95% confidence interval (CI)]. If disagreements occurs a third reviewer (ASK) will be included to resolve them where necessary. If the situation arises where data reported in the manuscripts selected for inclusion are insufficient or unclear for the purposes of analysis, NM will contact the authors of these studies by e-mail to request clarification or the missing data.

Assessment of risk of bias. A quality assessment tool will be used to critically appraise all included studies by two independent review authors (NM and KMS) (see Supplementary File 3). Risk of bias will be assessed with the Cochrane collaboration’s tool for observational studies⁴⁰, considering six domains: selection bias, exposure, outcome measurement, statistical analysis, study attrition and confounding. Each domain will be assessed and study bias will be classified as having a high, moderate, low, minimal or unreported risk of bias for each domain. Each study will then be rated as having high, moderate or low risk of bias. For example, those with all six domains rated as minimal or low will be classified studies having a low-risk of bias.

Quantitative data synthesis. If possible, a meta-analysis will be performed in order to calculate a quantitative estimate of the relationship between PNMS and the risk of ASD and ADHD. The meta-analysis will be performed on all studies that report crude and/or adjusted estimates in the form of an OR, HR or RR, with 95% CIs. The generic inverse variance method will be used for crude and adjusted results. In particular, studies reporting adjusted estimates will undergo a separate in unadjusted and adjusted models in order to evaluate potential study confounders. The analysis will be performed for ASD and ADHD separately.

The Review Manager (Cochrane Collaboration Software, RevMan) 5.3 software will be used to perform statistical analysis and to combine results in a forest plot, using random-effect models. The pooled OR and standard error (SE) will be used to generate a funnel plot in order to estimate the likelihood of publication bias if 10 or more studies are included in the meta-analysis.

Statistical methods

Assessment of heterogeneity. Statistical heterogeneity will be assessed using the I² statistic with an alpha of 0.05 for statistical significance, as per the Cochrane Handbook for Systematic Reviews threshold recommendations⁴⁰. I² reflects the percentage of total variation across studies that is attributable to heterogeneity rather than chance. An I² negative or up to 0% suggests no observed heterogeneity. I² between 0% and 25% represents low heterogeneity; I² from 25% to 50% represents moderate heterogeneity; and 50% to 75% represents substantially high heterogeneity⁴¹.

Analysis of subgroups or subsets. If there are sufficient numbers of studies and data available, we will perform the following subgroup analysis:

1)   For timing of stress exposure (e.g. First trimester vs Second trimester vs. Third trimester);

2)   According to gender (male vs female)

3)   According to different type of stress (e.g. objective vs subjective stress);

4)   According to the study design setup (e.g. cohort studies vs case-control studies);

5)   According to study quality (e.g. those with minimal or low quality vs moderate to high quality).

Any additional analyses will be described as post-hoc. Additional subgroup and sensitivity analyses will be performed as appropriate to assess the causes of heterogeneity as appropriate.

Assessment of reporting biases. If a sufficient number of studies is available (at least 10 studies), we will assess publication bias using a funnel plot and Egger test. A funnel plot will be produced to estimate the likelihood of publication bias using the pooled OR and SE.

Potential limitations

This study may have several limitations. First, the search will be restricted to English and Italian language literature only. Second, we will not include unpublished literature, possibly resulting in an increased risk of publication bias for the included studies, given that reports that describe a significant finding or a positive outcome have a greater probability of being published when compared to those that do not. However, it should be considered that unpublished studies (and studies written in languages other than English) are more likely to report non-significant or small effect sizes, thereby leading to a weakening of the overall estimate. Where possible, we will use a funnel plot for assessing the presence of publication bias. In addition, another limitation may be the presence of confounding, such as maternal age, maternal smoking or alcohol abuse during pregnancy, paternal age, family history of mental illness and family’s socio-economic status. We will make separate analyses for adjusted and unadjusted results and subgroup analyses in order to evaluate the potential role of these confounders.