Grandmaternal smoking during pregnancy and grandchild health: A systematic review and meta-analysis

doi:10.21203/rs.3.rs-1625650/v1

Background

Maternal smoking during pregnancy is associated with a host of detrimental effects on the exposed child. However, emerging evidence suggests that the adverse effects of smoking during pregnancy may be transmitted across generations. We conducted a systematic review to comprehensively summarize evidence on the association between grandmaternal smoking during pregnancy and health outcomes in the grandchild.

Methods

We searched MEDLINE, EMBASE, and 10 other databases from inception to March 02, 2021, updated on October 18, 2021, to identify analytic epidemiologic studies (prospective or retrospective cohort and case-control design) investigating the association between grandmaternal smoking during pregnancy and multiple health outcomes in the grandchild in all countries and settings. Three investigators independently screened records, extracted data, and assessed study quality using the Effective Public Health Practice Project (EPHPP) tool. Random-effects robust variance estimation was used to combine effect estimates whenever possible, separately for maternal and paternal line.

Results

Twenty-four reports from 15 unique studies with 23 health outcomes and 231,478 grandchildren were included. The overall study quality was moderate. In maternal line, grandmaternal smoking during pregnancy was associated with an increased risk of asthma (seven studies, 148,527 grandchildren, risk ratio (RR) 1.10, 95% CI 1.02 to 1.18) and higher birth weight (four studies, 19,478 grandchildren, mean difference (g) 40.04, 95% CI 22.60 to 57.49) in the grandchild. There was suggestive evidence that maternal grandmaternal smoking was associated with increased risks of acute lymphoblastic leukaemia, any cancer, autism, and attention deficit hyperactivity disorder, decreased risks of early-onset myopia and small birth weight for gestational age, and higher birth length and body mass index at birth. In paternal line, grandmaternal smoking during pregnancy was not associated with asthma (four studies, 41,069 grandchildren, RR 1.01, 95% CI 0.85 to 1.19), but there was suggestive evidence for lower intelligence quotient scores and decreased risks of early-onset myopia and intolerance to loud sound.

Conclusions

The effects of maternal smoking during pregnancy may not be limited to the exposed child, but be transmitted to the grandchild, possibly through germline epigenetic inheritance. Further public health efforts are warranted to strengthen smoking cessation programs for pregnant women.

acute lymphoblastic leukaemia

asthma

attention deficit hyperactivity disorder

autism

cancer

grandchild

grandmother

intelligence quotient

smoking during pregnancy

systematic review

Globally, the prevalence of maternal smoking during pregnancy is estimated to be around 2% [1]. However, in many countries, smoking during pregnancy is still common. More than 10% of pregnant women smoked during pregnancy in 29 (17%) of 174 countries and more than 20% in 12 (7%) countries [1]. Specifically, the prevalence of smoking during pregnancy was on average 8% in Europe (ranging from 0.5–38%) and 17% in the USA [1]. Notably, more than 70% of pregnant women who smoked during pregnancy smoked daily rather than occasionally [1]. With decades of research, smoking during pregnancy has been linked to a host of detrimental effects on the exposed child. For example, smoking during pregnancy is associated perinatally with low birth weight [2], preterm birth [3, 4], various birth defects [5], still birth [6], and neonatal death [7]; and postnatally with sudden infant death syndrome [8], asthma [9], obesity [10], high blood pressure [11], low intelligence quotient (IQ) scores [12], conduct problems [13], and attention deficit hyperactivity disorder (ADHD) [14]. These findings suggest that smoking during pregnancy may affect offspring’s health already in utero, with potentially life-long negative health consequences.

The biologic mechanisms by which maternal smoking may lead to diverse adverse health in the offspring are complex. Extensive human and laboratory studies have shown that cigarette smoke contains more than 7,000 chemicals, many of which (e.g., nicotine, carbon monoxide) can cross the placental barrier to influence fetal programming or have a direct toxicologic effect on the fetus [15, 16]. In recent years, it is increasingly recognized that many of the mechanisms are mediated epigenetically [17, 18]. The epigenome refers to the biochemical information that lies ‘over and above’ the DNA sequence and regulates the expression of genes through mechanisms such as DNA methylation (DNAm), histone modifications and non-coding RNAs [19, 20]. There is a growing body of literature that smoking during pregnancy is associated with numerous epigenetic alterations in the exposed child, which may in turn affect health outcomes [17, 18]. For example, many studies consistently found that DNAm at specific gene loci mediated in part the relationship between maternal smoking and offspring low birth weight [21–26]. An increasing number of human epigenome-wide association studies showed that DNAm marks at some specific gene regions were associated with ADHD and asthma, suggesting that epigenetic mechanisms may play an important role in their etiology [27–39]. In rat models, maternal nicotine exposure during pregnancy (F0) induced epigenetic changes not only in the first generation (F1), but also in the germ cells of F1 generation that form the second generation (F2) (termed as intergenerational epigenetic inheritance) and of F2 generation that form the third generation (F3) (as transgenerational epigenetic inheritance) [40–42]. Furthermore, recent human epigenome-wide association studies found that grandmaternal smoking during pregnancy was associated with differential DNAm in the grandchild [43, 44].

While the detrimental health effects of smoking during pregnancy on the exposed child are well established, it is unclear whether these effects are transmitted to the subsequent generation(s) through epigenetic inheritance. Whilst the plausibility of epigenetic inheritance is being elucidated [20, 45, 46], a number of epidemiologic intergenerational studies have investigated the impact of grandmaternal smoking during pregnancy on various health outcomes in the grandchild, but results are inconsistent for some outcomes (e.g., birth weight, asthma) [47–50]. To elucidate the underlying evidence, we undertook a systematic review of observational epidemiologic studies investigating the association between grandmaternal smoking during pregnancy and multiple health outcomes in the grandchild. We sought to comprehensively summarize the available evidence on the potential intergenerational transmission of the effects of maternal smoking during pregnancy.

The protocol for this systematic review was registered in the International Prospective Register of Systematic Reviews (PROSPERO) (CRD42017084039). This study was reported as per the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guideline [51] (Additional file 1). Ethical approval was not required for this study.

Literature search and eligibility criteria

We searched MEDLINE, EMBASE, Cochrane Library, ISI Web of Science, CINAHL, Google Scholar, Scopus, Global Health, PsycINFO, Zetoc, CAB International and WHO Global Health Library from inception to March 02, 2021. The database search was updated on October 18, 2021. No language restriction was applied. The search strategies are available in the PROSPERO protocol. Three investigators (GQZ, YL and HHZ) independently screened the titles and/or abstracts and reviewed full-text articles for eligibility. Any discrepancies were resolved by discussion or by a third investigator (BIN). We also manually checked the references of included studies and relevant reviews to identify additional eligible studies.

Studies were selected for inclusion if they met the following Population, Exposure, Comparator, Outcome, Study design (PECOS) criteria: 1) Population: grandmother-mother/father-child triads of any ethnicity in any country or setting; 2) Exposure: maternal grandmaternal smoking during pregnancy (i.e., grandmother smoked any amount of cigarettes when pregnant with the mother) or paternal grandmaternal smoking during pregnancy (i.e., grandmother smoked any amount of cigarettes when pregnant with the father); 3) Comparator: no maternal or paternal grandmaternal smoking during pregnancy; 4) Outcome: any health outcome in the grandchild of any age; 5) Study design: analytic epidemiologic studies (prospective or retrospective cohort and case-control design). We also included studies investigating the effects of both grandmaternal and maternal smoking during pregnancy relative to neither of them smoked during pregnancy. We excluded reviews, commentaries, studies of grandmaternal smoking not during pregnancy, and studies of indirect outcome measures (e.g., DNAm, dietary inflammatory index, lung function parameters). For conference abstracts, corresponding authors were contacted to request additional data. If no information was provided, they were excluded from our review (eTable 1 in Additional file 2).

Data extraction and quality assessment

Three investigators (GQZ, YL and HHZ) independently extracted data by outcome using a customized data extraction form. For each article, we extracted first author, year of publication, country, study design, exposure group, reference group, sex and age of the grandchild, outcome examined, number of cases and controls in case-control studies or total population in cohort studies for binary outcomes, number of total population in cohort studies for continuous outcomes, type of effect measure (mean difference (MD), risk ratio (RR), odds ratio or hazard ratio), and maximally-adjusted effect estimate with 95% confidence interval (CI). When studies reported multiple effect estimates for different exposure levels relative to non-smoking [47, 49, 52, 53], or estimates relative to a reference group other than non-smoking [54], corresponding authors were contacted to obtain the estimates comparing smoking versus non-smoking. We received the requested data from three studies [47, 49, 54]. We categorized outcomes according to the International Classification of Diseases 11th Revision (ICD-11) [55].

Three investigators (GQZ, YL and HHZ) independently assessed the risk of bias of included studies by outcome using the Effective Public Health Practice Project (EPHPP) tool [56]. The EPHPP tool consists of six domains: study design, selection bias, confounders, exposure measurement, outcome measurement and statistical analysis. Each domain is rated as strong, moderate, or weak. Disagreements during data extraction and quality assessment were resolved by discussion or by a third investigator (BIN).

Data synthesis and analysis

We developed a priori an analysis protocol for this review (Additional file 3), which was adapted from our earlier work [57]. We used the random-effects robust variance estimation (RVE) method to calculate the summary average effect and its 95% CI [58]. The RVE method does not require any distributional assumptions on the population effects, can accommodate dependence among effect estimates (e.g., multiple estimates based on the same individuals at different time points, or computed using a common control group), and can provide correct inference even in meta-analyses with a small number of studies [58, 59]. We quantified the extent of heterogeneity by estimating the standard deviation of population effects [60, 61]. To evaluate the robustness of meta-analysis results to potential publication bias, we calculated the S-value [62], which represents the severity of publication bias that would hypothetically be required to shift the summary estimate to the null. For binary outcomes, we also calculated the E-value [63–65], which assesses how strong residual confounding would have to be to “explain away” an observed exposure-outcome association.

The primary analyses focused on the average effects of grandmaternal smoking during pregnancy. These analyses were conducted separately for maternal and paternal line. Subgroup analyses were conducted according to maternal smoking status during pregnancy (smoking versus non-smoking) and sex of the grandchild (boys versus girls). We also conducted analyses for the effects of grandmaternal and maternal smoking during pregnancy in combination compared to neither of them smoked during pregnancy. For binary outcomes, all estimates were converted to the RR scale [63, 66], and results were presented on that scale. All statistical analyses were performed using R software (version 4.0.4) [67]. The R scripts and datasets are respectively available in Additional file 4 and at the Open Science Framework.

Characteristics of included studies

Overall, we identified 1,152 records, scrutinized 49 full-text articles, and ultimately included 24 reports from 15 unique studies, including 18 reports from nine prospective cohort studies [48–50, 53, 54, 68–80], three reports from three retrospective cohort studies [47, 52, 81], one report from one nested case-control study [82], and two reports from two case-control studies [83, 84]. These reports were published between 2003 and 2021. Figure 1 shows the study selection process. Studies excluded at the full-text screening stage with reasons for their exclusion are available in eTable 1 in Additional file 2. Characteristics of the included studies are summarized in Table 1. Figure 2 presents a summary of quality assessment by outcome across all included studies. Each EPHPP domain judgment for each outcome is available in eTable 2 in Additional file 2.

Table 1. Characteristics of observational epidemiologic studies investigating grandmaternal smoking during pregnancy and health outcomes in the grandchild

Study	Country	Study design	Participants		Exposure assessment	Outcome assessment	Confounding factors adjusted
Study	Country	Study design	Sampling method, source	No. recruited/ analyzed	Exposure assessment	Outcome assessment	Confounding factors adjusted
Diseases of the respiratory system
Accordini 2018 [74]	Europe^a, Australia	Prospective cohort	Sampling method: all children Children born from parents who participated in the ECRHS I and III	Maternal line: 5,002/4,666 Paternal line: 4,563/4,192	Parent report at ECRHS I	Parent-reported ever asthma with or without nasal allergies in children aged 0–51 years	Grandmother’s and grandfather’s ever asthma and education level, maternal and paternal ever asthma, age, smoking and education level, and offspring’s gender and age
Bråbäck 2018 [47]	Sweden	Retrospective cohort	Sampling method: general population Children born after June 31, 2005, from all mothers born between 1982-1986 in Sweden	Maternal line: 15,265/10,329 Paternal line: 15,265/10,329	Grandmother report at gestational week 8-12	Purchase of any asthma medication (inhaled steroid and/or leukotriene antagonist) by age 6 years	Maternal smoking habits, maternal and paternal grandmother’s age, level of education, social welfare, BMI and asthma medication, maternal and paternal grandfather’s asthma medication, maternal and paternal age, sex of the child, and county of residence at mother’s and father’s birth
Li 2005 [82]	USA	Nested case-control	Sampling method: controls were countermatched on in utero exposure to maternal smoking within grade, sex, and community of residence Cases: asthmatic children from public schools (grades 4, 7, 10) in 12 southern California communities in 1993; Controls: non-asthmatic children from the same cohort	Maternal line: Cases: 338/235 Controls: 570/335	Mother report	Parent-reported doctor-diagnosed asthma in the first 5 years of life	Grade, sex, community of residence, race/ethnicity, gestational age, and secondhand smoke exposure