Pre- and early postpartum psychosocial stress trajectories in mothers and child body mass index at 3 years: a birth cohort study

Data were derived from the Ulm SPATZ Health Study, a birth cohort study, conducted in Ulm, Germany. Families in whom the mothers gave birth to a child at Ulm University Medical Centre from 04/2012 to 05/2013 were asked to participate. Participation was 49%, n = 970 mothers, 1006 children − among them 934 mothers with singleton children. The Department of Gynaecology and Obstetrics at Ulm Medical Centre is the only maternity clinic in Ulm with a good representation of the source population. Details of the baseline examination were described elsewhere [17]. The study population comprised all mothers with their singleton children for whom data on at least one exposure variable and on the outcome (n = 596) were available (for further information on the recruiting process and missing values see Fig. 1). The study was approved by the ethics board of Ulm University (No. 311/11). All participants gave written informed consent. Informed consent was also obtained from the parent and/or legal guardian of minors. The study was performed in accordance with the Declaration of Helsinki.

At delivery (0 M), at 6 (6 M) and 12 months (12 M) postpartum, chronic maternal stress was measured using the Screening Scale of the Trier Inventory of Chronic Stress (SSCS-TICS) [18] included in a self-administered standardized questionnaire. SSCS-TICS was developed to assess chronic concerns, lacking social recognition, work overload, excessive demands, and social stress three months prior to assessment. Furthermore, the German version of Hospital Anxiety and Depression Scale (HADS), referring to symptoms of anxiety and depression in the last week prior to completion of the questionnaire was used [19]. Summary scores of eleven or more in each subscale indicate moderate to severe, scores between 8 and 10 mild symptoms, and scores < 8 are considered normal. Hair samples were collected from the posterior vertex region on the maternal head by trained study staff (at 0 M) or by the families themselves (at 6 M and 12 M). HCC were determined in the scalp-near 3 cm hair segment, assuming to reflect cumulative steroid secretion over the period of the preceding 3 months. Further details including the analytical process were described in details elsewhere [17, 20]. HCC values of < 0.09 and > 90 were excluded, [20] so were values > 3 standard deviations from the mean [21].

Child body height and weight measured at the 3-year-screening examination (child age 36.0 months, standard deviation: 1.1) was transcribed by the mothers from the child’s paper-based medical record (“Kinderuntersuchungsheft”), which is started at birth (day 3 to 10) and includes up to 9 screening exams up to age 9 years. Height and weight were measured by the paediatrician, normally including clothing but not shoes. However, the information whether the child was dressed during the measurement was not given in detail. The outcome was the body-mass index (BMI (height/weight²)) and the age- and sex-standardized BMI according to a German population [22]. For this purpose, missing age at BMI measurement was set to 3 years (36 months) in n = 3 participants corresponding to the mean of the remaining study population. For descriptive findings BMI was classified into underweight or normal, overweight, and obese based on < 90^th percentile (P), ≥ P90 – < P97, ≥ P97 of a German population. The classification according to the International Obesity Task Force (IOTF) [23], and the World Health Organization (WHO) [24] was used additionally. All three classifications are age- and sex-specific.

Variables describing the study population such as maternal age (≤ 25, 26–35, ≥ 36 years), number of persons in the household at delivery (2–3, 4, > 4 persons), nationality (derived from nationality and place of birth), maternal education (< 12, ≥ 12 years), smoking during pregnancy (yes, no), and smoking in the year before pregnancy (yes, no) were collected following delivery using a self-administered standardized questionnaire. We used smoking before pregnancy as covariate due to a possible underreporting of smoking during pregnancy. Maternal BMI was calculated based on mother-reported height and physician-recorded measured maternal body weight. The latter was obtained from paper documentation of routine preventive examination that obstetricians issue to their patients when pregnancy is clinically established. However, if first examination took place at > 90 days of gestation, we used self-reported maternal body weight assessed when pregnancy was established (n = 96). Maternal BMI was classified into underweight or normal, overweight, and obese corresponding to the related BMI < 25.00, 25.00–30.00, > 30.00 kg/m² for purpose of description whereas for regression models, maternal BMI was used continuously. Child birth weight was drawn from the clinic documentation system. For regression models, child birth weight was used continuously. The following categories were used for description < 2500, 2500– < 3000, 3000– < 3500, 3500– < 4000, ≥ 4000 g. Breastfeeding duration (days) was mother-reported via self-administered standardized questionnaire at a 6-week-, 6 M-, 12 M-, and the 24 M-examination and considered until the mother had reported that she had completely stopped breastfeeding. Further paternal covariates were considered as follows: Paternal BMI was calculated based on father-reported height and weight 6 weeks after child birth. At this time point, information on paternal education (< 12, ≥ 12 years) was further collected in a self-administered questionnaire.

We describe study characteristics and compare them to the baseline population by indicating 95% CI of the binomial distributions. Measures of central tendency are given for maternal stress as well as depression and anxiety symptoms, and HCC at 0 M, 6 M, and 12 M, including Spearman correlation coefficients. Sex-stratified analyses were performed. We further show depression and anxiety symptoms categorized into relevant groups as suggested [25]. We identified latent class mixed model trajectory groups for chronic stress, depression and anxiety symptoms as well as HCC separately. The appropriate number of trajectory groups was determined by increasing it from 1 to 4 and subsequently staying with the number with the lowest Bayesian Information Criterion (BIC), given that the group size was ≥ 10. For HCC for example, BIC was in favor of 3 groups, but one consisted of only one person. The procedure above was also applied for sex-stratified analyses to examine if a sex-stratified analysis would fit better.

Missing values for the exposure variables were considered as missing at random and were estimated using the R lcmm function. Missing values of the confounders were imputed using PROC MI. All variables were considered as either continuous or were coded as dummy variables. For numbers of the respective missing values see Fig. 1. For analyses, 10 completed data sets were generated and combined in a final step. We included all confounders in our imputation model, all exposures, and the outcome variables as potential predictors of missing values. Two separate imputation models were calculated, one for BMI and one for BMI-SDS. We compared the results between multiple imputation data and the complete case approach.

A null-effect of maternal stress on child BMI/BMI-SDS is partly in line with two recent reviews, [10, 11] as the authors argued that the two main types of stressors that were indeed associated to subsequent child overweight or obesity were environmental stressors [10, 11] and mother–child dysfunctional relationship [10]. Both types of stress are not specifically targeted by the SSCS-TICS questionnaire. In one more recently conducted longitudinal study [26] however, including n = 9206 mother–child dyads, maternal psychosocial stress at 5 years of age predicted child overweight and obesity at 11 years. Yet, accounting for early factors as well as sociodemographic, dietary or physical activity attenuated this association to statistical non-significance. Furthermore, postnatal maternal stress during the first year after birth had a positive longitudinal relationship with children’s BMI z-scores up to the age of five years [27] but sex-stratified analyses revealed that only girls showed this positive association while boy’s BMI z-scores were unaffected by maternal stress. It should be noted that our study might be underpowered to show statistically significant differences as it comprises a high proportion of mothers with relatively low stress levels. Interestingly, findings for the relationship between maternal stress and children's weight-related behaviours showed no evidence for an association with children's unhealthy dietary intake, but for an association of maternal stress with children's lower physical activity and higher sedentary behaviour [28]. Further studies including these variables allowing sex-specific analyses would be desirable.

According to a review, [29] a stable moderate-high or high symptom trajectory of depression was reported in most studies on perinatal depression trajectories, suggesting a high-risk group with persistent depression symptoms. Accordingly, chronic but not episodic maternal depression is associated with a greater risk for child overweight [12]. We did not find such a group but a low-stable and a low-increasing depression symptom group. However, our results concerning the association with the outcome variable are in line with further studies [13, 14] which did not find consistent associations between (multiply assessed) depression symptoms and BMI or BMI z-scores. Instead, antenatal and postpartum depressive symptoms were partly differentially associated with the outcome.

Few studies have examined maternal anxiety trajectories [30] and data on anxiety symptoms and child weight are scarce. We found higher child BMI with mothers indicating low-increasing compared to low-stable anxiety symptoms in girls which may suggest a higher impact of more recent as opposed to earlier experiences of anxiety symptoms on child BMI. The positive association between high-increasing anxiety trajectory and child BMI would correspond to this assumption. However, also moderate-decreasing maternal anxiety symptoms are positively associated with higher child BMI in girls (see Table 4). These results are contradictory and need further explanation. Due to the relatively small sample size, our results do certainly need confirmation in large scaled studies. Especially since Shay et al. 2020 [31] observed no direct relationship between maternal anxiety and child overweight but both variables were associated with decreased weeks of breastfeeding, where we did not find a respective association. Notably, the authors analyzed clinically significant anxiety symptoms which were not highly prevalent in our study.

After excluding extremely high cortisol concentrations as suggested by Marceau et al. 2020 [21], we found no association between maternal HCC and child weight. It is well known that during pregnancy, cortisol concentrations are about 3-fold higher compared to a non-pregnant state. Little is known however about the post-pregnancy period and HCC “return to normal” or whether the different trajectories of peripartum HCC reflect different stress or anxiety or depression trajectories. In their review Meyer et al. 2021 [32] concluded that the current literature lacks consistent evidence that HCC are related to chronic stress, anxiety or depression symptoms. This finding is also supported by our group [33, 34] and by Spearman correlation coefficients ≤ 0.09 when analyzing the associations between chronic psychosocial stress, depression or anxiety symptoms and HCC at 6 M and 12 M.

Our study design allowed to identify trajectories of maternal psychosocial stress and related symptoms following delivery and thus may address chronicity of stress exposure or anxiety/depression symptoms. Yet, a possible selection bias due to loss to follow-up of non-German, less educated or larger families might impact the results as experiencing stress may cluster with social and economic adversities. Thus, the real magnitude of the association might be underestimated in our study. Further variables such as birth mode, seasonality, batch of hair cortisol sample (see e.g., [17, 21]) might influence trajectories of hair cortisol concentrations. Larger studies are yet needed to analyze the respective subsamples and their trajectories as well as their associations with child weight. Interestingly, the prevalence of children who suffer from overweight or obesity is slightly lower in our study compared to the general population in Germany, where 10.8% (95% CI: 7.0–16.5) of girls and 7.3% (4.7–11.1) of boys aged 3 to 6 years were either overweight or obese [35]. Unfortunately, our study provides no data to elucidate a mechanism between depression/anxiety symptoms or maternal stress and child BMI.