Maternal birth weight has been considered an anthropometric indicator for predicting the birth weight of children(Reference Ounsted, Scott and Ounsted1,Reference Hypponen, Power and Smith2) . Studies that assessed the intergenerational transmission of birth weight identified relationships between low birth weight (LBW) in the mother and LBW in the child. At the same time, the relationships of higher birth weights between mothers and their children have also been evidenced in some studies(Reference Klebanoff, Graubard and Kessel3–Reference Cnattingius, Villamor and Lagerros6). In addition, studies also evaluated the association between paternal birth weight and offspring birth weight(Reference Veena, Kumaran and Swarnagowri7,Reference Coutinho, David and Collins8) .
A systematic review of the intergenerational transmission of birth weight suggests that a 100 g increase in the mother’s birth weight leads to a 10–20 g gain in the child’s birth weight. Paternal birth weight was also associated with child birth weight, but this association was not as strong as maternal birth weight(Reference Ramakrishnan, Martorell and Schroeder9). Thus, this difference in the strength of association is possibly due to the fact that birth weight is related to maternal anthropometric factors, such as height and pre-pregnancy BMI, in addition to maternal weight gain during pregnancy. The influence of the maternal lineage on the birth weight of children, which possibly indicates an additional effect represented by intra-uterine influences on birth weight, resulting from maternal health conditions, behaviour and socio-economic status(Reference Lawlor and Mishra10), what would explain this difference in the intergenerational relationship among mothers/fathers and their children.
The relationship between the birth weight of both parents and children has been studied previously(Reference Ramakrishnan, Martorell and Schroeder9), and a meta-analysis has examined intergenerational differences in birth weight(Reference Ramraj, Pulver and Siddiqi11). In this context, the purposes of this study were (1) to systematically review the literature on the evidence of the intergenerational transmission of birth weight from parents to their children and (2) to quantify this association by performing a meta-analysis.
Methods
Protocol and registration
The review was carried out following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses(Reference Page, Moher and Bossuyt12). The study protocol was registered in the International Prospective Register of Systematic Reviews – (registration number: CRD42021230962).
Selection of studies
Potentially relevant papers were identified by searching the electronic databases PubMed, Science Direct, Web of Science, Embase, Scopus, CINAHL and LILACS completed on 21st of January 2021 (Fig. 1). The literature search used the following terms: ‘birth weight’ OR ‘birthweight’ OR ‘birth-weight’ OR ‘size at birth’ AND ‘family’ OR ‘parents’ OR ‘mother*’ OR ‘father*’ OR ‘offspring’ AND ‘intergenerational’ OR ‘generation’.
Fig. 1. Flow chart of the study selection.
After excluding the duplicates, two independent reviewers (RRO and EPS) screened the titles to remove irrelevant studies. The full texts of the remaining studies were retrieved and those studies that were eligible for this review were identified. In addition to the electronic search, reference lists of the selected studies were examined to identify manuscripts that had not been captured by the database search. Disagreements were solved by a third reviewer. Additional search was made by scanning the reference lists of the identified studies and the previously published systematic reviews.
Selection criteria
We included original studies, performed in humans, that evaluated the intergenerational transmission of birth weight. Thus, studies presenting birth weight data from two generations were selected, including mothers, fathers or parents in the first generation, and daughters, sons or children in the second generation. We excluded those studies that included review studies, editorials, comments and studies conducted with animals. Studies were excluded if they did not provide birth weight data; studies with a different design than the longitudinal; studies focused on fetal growth or prematurity and studies focused on specific samples, such as studies performed with twins.
Exposure and outcome
Study exposure was first-generation birth weight, including studies conducted with fathers and mothers. Outcome was second-generation birth weight for both offsprings, sons or daughters. In some studies, when available, birth weight in both generations was considered a continuous variable (measured in grams), and the combined mean difference (MD) between the first and second generations was analysed. Other studies, with available data on OR and other measures of effect, were included for the LBW analysis.
Extraction and quality assessment
The extraction of data and assessment of quality were performed separately and blindly by two reviewers (RRO and EPS) using a structured form generated in Microsoft Excel 2016 (Microsoft). Differences were resolved by consensus and discussion with a third reviewer (DPG). We extracted the following information from each manuscript: publication year; country of data collection; data source; sample size; exposure; outcome; control for confounding and main results.
When reported, mean birth weight and standard deviation or OR and 95 % CI were extracted. If these data were not informed or could not be calculated, the first author of the study was contacted by email.
Study quality
Methodological quality assessment was based on the Newcastle–Ottawa scale(Reference Wells, Shea and O’Connell13), a quality assessment scale for cohort studies. For each study, a maximum of nine points could be achieved. The Grading of Recommendations Assessment, Development and Evaluation approach was used to assess the overall quality and strength of evidence. By this approach, the quality of the totality of evidence can be graded as ‘very low’, ‘low’, ‘moderate’ or ‘high’. Evidence derived from observational studies receives an initial grade of ‘low’.
Data analysis
Two meta-analyses were conducted. The first one used mean birth weights of both generations and their respective standard deviations, obtaining the MD and 95 % CI which were calculated for each study, as well as a pooled estimate. The second study outcome, an intergenerational assessment of LBW, was assessed by OR, and 95 % CI for LBW was generated for each study as well as pooled estimate in the second meta-analysis. For the mean birth weight meta-analysis, when the study contained information on the mean birth weight of both parents or children of both sexes, the weighted mean and standard deviation were calculated for each study. In the OR meta-analysis, we followed the birth weight classifications in groups established by the studies, with more than one estimate being performed in each study. Pooled summary statistics were calculated using a random-effects model. Forest plots were generated to explore heterogeneity, graphically.
To evaluate the pooled effect size, we used the random-effects models and evaluated the heterogeneity among studies using the I 2 statistics. To explore the heterogeneity sources of this association, the variables year of publication (before 2010 and after 2010), study design (retrospective cohort and prospective cohort), sample (<1000, 1000–5000 and >5000), setting (high-income country, middle/low-income country), relationship (parents, mothers and fathers) and adjustment for confounding variables confounding (no and yes). Meta-regression was performed to evaluate the pooled effect according to the characteristics of the studies. Funnel plots and the Egger’s test were used to evaluate publication bias under variables (year, study design, sample, setting, relationship, control for confounding). Analysis was performed using Stata 16.
A sensitivity analysis was performed to assess the robustness of the observed results. Therefore, according to the Newcastle–Ottawa scale, studies of low quality, less than or equal to five points, were excluded in the sensitivity analysis.
Results
Study characteristics
Fig. 1 shows the study selection flow chart. The search identified 9878 studies. After excluding duplicates (n 1873), 8005 titles were read and 154 abstracts were selected. Of these, eighty-four abstracts were excluded mainly because they did not assess birth weight and did not include two generations or related birth weight of children with socio-demographic, behavioural and health characteristics of the parents. After excluding the abstracts, seventy manuscripts read in full were selected.
Of these, fifty-three were excluded, and the reasons for this exclusion were presented in the flow chart (Fig. 1). Other three studies were identified through a search in the references of the manuscripts selected. For studies in which we were unable to extract mean birth weight and standard deviation or OR and 95 % CI, we sent an email to the authors. In case they did not respond to the email or did not provide the necessary data, we excluded the meta-analysis study (n 12).
Thus, twenty studies were included in the meta-analysis, fourteen in the MD analysis(Reference Little14–Reference Liu, Lin and Su27) and eleven in the OR(Reference Klebanoff, Graubard and Kessel3,Reference Coutinho, David and Collins8,Reference Winkvist, Mogren and Högberg15,Reference Agnihotri, Antonisamy and Priya18,Reference Costa, Silva and Da Silva Gomes23,Reference Sherf, Sheiner and Shoham Vardi26–Reference Sepulveda-Martinez, Rodriguez-Lopez and y Mino31) .Table 1 shows a description of the studies included in the meta-analysis.
Table 1. Description of studies included in meta-analysis (n 20)
BW, birth weight; LBW, low birth weight; SES, socio-economic status.
All studies were cohort studies, of which ten were prospective and ten retrospective cohorts. The studies were conducted in the USA (n 8), Israel (n 2), Sweden (n 2), England (n 1), Malta (n 1), Spain (n 1), Argentina (n 1), India (n 1), Brazil (n 1), Norway (n 1) and a study by the Consortium on Health Oriented Research in Transitional Societies with a sample of the countries Brazil, Guatemala, India and the Philippines.
Mean birth weight
Fig. 2 shows the results of the overall meta-analysis of mean birth weight. The pooled MD in birth weight (measured in grams) between the first generation and second generation, across all studies, was not statistically significant (fourteen studies; MD 19·26, 95 % CI 28·85, 67·36; P = 0·43). Using a random-effects model, these results were found to be highly heterogeneous (I 2 = 99·96 %).
Fig. 2. Meta-analysis of mean birth weight. The pooled mean difference (MD) in birth weight (measured in grams) between the first generation and second generation (fourteen studies).
Low birth weight
The pooled association between LBW in the first generation and LBW in the second generation is shown in Fig. 3. Offspring of LBW parents were 69 % more likely to have LBW (effect size 1·69, 95 % CI 1·46, 1·95; I 2 :85·8 %). These results were found to be statistically significant and of high heterogeneity.
Fig. 3. Meta-analysis on the association of the low birth weight between the first generation and the second generation (random effect). ES effect size (eighteen estimates from eleven studies). aMother born with 2,7-3,6kg bMother born with 1,8-2,7kg cAfrican americans mother dWhite mother eAfrican americans father fWhite father gMother small for gestational age hMother with LBW IFather with LBW jNon-hispanic black LBW kNon-hispanic white LBW lParents small for gestational age
The subgroup analysis (Fig. 4) was carried out on studies performed with mothers (eight studies, ten estimates; OR 1·80, 95 % CI 1·59, 2·03), with parents (three studies, seven estimates; OR 1·55, 95 % CI 1·22, 1·97) and with fathers (one study, two estimates; OR 2·19, 95 % CI 1·00, 4·80).
Fig. 4. Meta-analysis of low birth weight between the first generation and second generation by subgroup. aMother born with 2,7-3,6kg bMother born with 1,8-2,7kg cAfrican americans mother dWhite mother eAfrican americans father fWhite father gMother small for gestational age hMother with LBW IFather with LBW jNon-hispanic black LBW kNon-hispanic white LBW lParents small for gestational age
When performing meta-regression, no significant differences were observed between year of publication (P = 0·70), type of study (P = 0·47), sample (P = 0·34), setting (P = 0·17), relationship (P = 0·49) and control by confounding (P = 0·29) (Table 2). Despite the funnel plot showing evidence of publication bias (Fig. 5), the Egger’s test was not significant (P = 0·67).
Table 2. Meta-analysis showing heterogeneity and meta-regression of the associations between low birth weight (LBW) in the first generation and LBW in the second generation (eighteen estimates from eleven studies) (Odds ratios and 95 % confidence intervals)
Fig. 5. Funnel plot of the effects measured by the studies of low birth weight included in meta-analysis (eighteen estimates from eleven studies).
Quality assessment
The results of the literature quality evaluation are shown in Table 3. Of the twenty studies included in this review, seventeen studies met more than half of the methodological quality criteria score(Reference Klebanoff, Graubard and Kessel3,Reference Little14–Reference Hyppönen and Power16,Reference Agnihotri, Antonisamy and Priya18–Reference Sepulveda-Martinez, Rodriguez-Lopez and y Mino31) . Furthermore, three studies had four points in methodological quality(Reference Coutinho, David and Collins8,Reference Cuestas, Darauich and Corredera17,Reference Costa, Silva and Da Silva Gomes23) . The overall strength and quality of the evidence were assessed by Grading of Recommendations Assessment, Development and Evaluation, the default level for observational studies (Table 4).
Table 3. Example of Newcastle–Ottawa scale for assessment of quality of cohort studies
*Represents a point on the scale score.
Table 4. Grading of Recommendations Assessment, Development and Evaluation assessment
* Evidence of significant inter-study heterogeneity (I 2 = 99·96 %).
† Evidence of significant inter-study heterogeneity (I 2 = 82·3 %) that cannot be explained by meta-regression.
In the sensitivity analysis, after excluding low-quality studies(Reference Winkvist, Mogren and Högberg15,Reference Cuestas, Darauich and Corredera17,Reference Costa, Silva and Da Silva Gomes23,Reference Sherf, Sheiner and Shoham Vardi26) , we found that there were no significant changes in mean birth weight (ten studies; MD, 17·33, 95 % CI 30·32, 64·97; P = 0·48). In the sensitivity analysis of LBW, after excluding low-quality articles(Reference Klebanoff, Graubard and Kessel3,Reference Coutinho, David and Collins8,Reference Winkvist, Mogren and Högberg15,Reference Costa, Silva and Da Silva Gomes23,Reference Sherf, Sheiner and Shoham Vardi26) , the result of our study was more evident (six studies, thirteen estimates; OR 1·76, 95 % CI 1·58, 1·95; P = 0·04).
Discussion
In this systematic review and meta-analysis, we assessed the association between intergenerational birth weight. The overall mean birth weight was slightly higher among offspring, compared with parents’ birth weight; however, this association was not statistically significant. We also assessed the association between LBW over the generations. Children of parents with LBW at birth had a higher risk of being born with LBW. In this sense, we can observe the roles that intergenerational factors can play on the birth weight of the next generation.
Intergenerational factors are characteristics of pregnancy, childbirth, exposure to events, situations and/or substances that affect the health status of one generation and can affect the growth and development of the next generation(Reference Masho and Archer32). Furthermore, maternal social environment, socio-economic status at birth and the child growth pattern are important factors in predicting the weight of children at birth(Reference Morton, De Stavola and Leon33,Reference Spencer34) .
The associations between the birth weight of the parents and children are well known, with most of these studies reporting a stronger relationship with the birth weight of the mother rather than with the birth weight of the father(Reference Ramakrishnan, Martorell and Schroeder9). This difference in the strength of the association between mothers and fathers is possibly due to birth weight related to maternal anthropometric factors, such as height and pre-pregnancy BMI, and maternal weight gain during pregnancy. Moreover, birth weight results from maternal factors such as smoking, diabetes and hypertension during the gestational period. Some studies suggest the effects of paternal smoking during pregnancy, that is, passive smoking, can influence the reduction of birth weight of children(Reference Ion, Wills and Bernal35,Reference Andriani and Kuo36) .
Our meta-analysis found that the LBW of the parents increases the chance of the child having LBW. The WHO defines LBW as a birth weight of less than 2500 g and remains a significant public health problem worldwide. According to data from UNICEF and from WHO(37), almost 15 % of all children in the world are born with LBW, undermining their survival, health and development. Hence, reducing LBW is one of the global nutritional targets – WHO intends to reduce LBW by 30 % worldwide by 2025(38,39) .
As we have seen, LBW has an intergenerational transmission. The consequences of LBW are both short and long term, including neonatal mortality and morbidity, and an increased probability of stunted growth, poor cognitive development(Reference Risnes, Vatten and Baker40) and lower(Reference Gu, Wang and Liu41). In adulthood, the risk of chronic diseases such as obesity, diabetes and CVD increases(Reference Risnes, Vatten and Baker40,Reference Jornayvaz, Vollenweider and Bochud42) .
Most studies were carried out in high-income countries. It is known that there is considerable variation in the prevalence of LBW among regions worldwide and within each country. Nevertheless, the vast majority of people with LBW occur in low- and middle-income countries and especially in the most vulnerable populations(Reference Kim and Saada43,Reference Muglia and Katz44) . Between 2000 and 2015, almost 95 % of LBW children were found in less developed region(37). In less developed regions, LBW is mainly caused by low fetal growth associated with maternal malnutrition before and during pregnancy. In more developed regions, LBW is associated with prematurity (defined as a baby born before 37 weeks of pregnancy) due to high maternal age, smoking, multiparity and caesarean delivery(Reference Kramer45). Most studies investigating the intergenerational transmission of birth weight have been based on American or European populations.
This is the first meta-analysis on the intergenerational transmission of birth weight, with the inclusion of longitudinal studies as a strong point. In relation to the publication bias, although visual inspection of the funnel plot showed asymmetry, the Egger’s test did not confirm publication bias. However, this study presents some limitations that should be considered. Initially, there was a lack of information in some studies, such as mean birth weight and standard deviation, making it impossible to include it in the meta-analysis. Second, the high heterogeneity was observed both in the mean birth weight and in the LBW analysis through the OR. We used meta-regression to investigate the source of heterogeneity, which is used to explore associations between study characteristics and the effect found, but we were unable to obtain an explanation with the variables included in our analyses. Furthermore, regarding the assumption of independence of the data that regular meta-analysis packages assume, we can consider these data as paired and not independent. However, in the meta-analysis we have the variability between studies, but we would not be able to do a meta-analysis with the variability within the studies.
Therefore, some methodological differences identified in the studies must be taken into consideration. Although the variables did not explain the high heterogeneity between studies, the possible explanation could be the source of information on birth weight, which varied between studies. Some of them used measurements from population records and hospital records; others collected information through a parent questionnaire.
In addition, some studies have been adjusted for few variables. For instance, important socio-economic variables that influence birth weight have not been adjusted, so the pooled estimates of associations may be affected by residual confounding. However, some studies that controlled for confusion seem to have included possible mediators in the model. Adjusting for a mediator may underestimate the magnitude of the association. Accordingly, it is clear that further studies should use an adequate conceptual framework when analysing the association of birth weight of parents and children.
Conclusions
This meta-analysis did not find an effect on mean birth weight between parents and offspring. However, we have found that having a LBW parent increases the odds of their child being born with LBW. Thus, more studies are needed, especially to assess the intergenerational transmission of birth weight in low- and middle-income countries. We also need more studies in order to understand the potential determinants, confounding factors and possible mediators of the association between birth weight of parents and children.
Acknowledgements
This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) -Finance Code 001.
The authors declare that they have no conflict of interest.
RRO participated in all stages of the manuscript (definition and search in databases, selecting, reading articles, extracting data, and analyzing), interpreted the results, and wrote down the text. EPS participated in the selection, reading of articles, and extracting data. TRF collaborated with data analyses, and did a critical review of the manuscript. DPG guided and critically reviewed the manuscript.
