The association of maternal postpartum depression with infant weight growth: a prospective birth cohort study

Untreated maternal postpartum depression has consequences for infant weight, which may vary with infant growth time and postpartum depression duration. Dynamic assessment of the association between maternal postpartum depression and infant weight growth is crucial for early detection of the suspicious abnormal effects of maternal postpartum depression on infant weight growth and taking corresponding intervention measures. But, none of published studies continuously and dynamically evaluated these effect changes on infant weight growth. This study was aimed to evaluate the dynamic effects of maternal postpartum depression on infant weight growth at a prospective birth cohort. new 1-month on the negative of changing, and regular Larger-scale and longer-term studies are recommended to evaluate the long-term impacts of depression from pre-pregnancy to postpartum even longer on infant and children’s psychological and physical development based on these combination models. our underscore perinatal BMI: body mass index, CHMIS: community health management information system, EPDS: Edinburgh postpartum depression scale, GEE: generalized estimating equation, LMM: linear mixed models, MDD: major depressive disorder, PPD: postpartum depression, WAZ: weight-for-age z score.


Abstract
Background Untreated maternal postpartum depression has consequences for infant weight, which may vary with infant growth time and postpartum depression duration. Dynamic assessment of the association between maternal postpartum depression and infant weight growth is crucial for early detection of the suspicious abnormal effects of maternal postpartum depression on infant weight growth and taking corresponding intervention measures. But, none of published studies continuously and dynamically evaluated these effect changes on infant weight growth. This study was aimed to evaluate the dynamic effects of maternal postpartum depression on infant weight growth at a prospective birth cohort.
Methods 960 mother-infant pairs between 2015 to 2018 in Changsha, China were followed up at ages of 1, 3, 6, 8, and 12 months. Data were obtained through household surveys. Depression of mothers was assessed at 1 month postpartum. Linear mixed models and generalized estimating equation models were used to test the connection and its changes between maternal postpartum depression and infant weight growth at ve different periods of 1-12 months.

Conclusions
Maternal postpartum depression was continuously associated with a lighter weight of ifants from 1 to 12 months and higher risk of being underweight in infants from 1 to 3 months. It seems important to put early prevention, screening, diagnosis, and treatment of maternal depression into practice as soon as possible to avoid adverse consequences.

Background
Maternal postpartum depression (PPD) is considered a major depressive disorder (MDD) with a global prevalence of 0.5%-68.8% depending on used diagnostic standards [1,2]. Generally, PPD begins within 4 to 6 weeks after delivery, and clinical symptoms similar to MDD may enlist depressed mood, loss of interest or pleasure in activities, sleep disturbance, appetite disturbance, loss of energy, feelings of worthlessness or guilt, diminished concentration, irritability, anxiety, and thoughts of suicide [3,4]. Evidence suggested that mothers in postpartum have a greater risk of depression than other periods even up to 20 times, and untreated PPD has consequences for infant weight which were not limited to the time of depression screened [1,[5][6][7][8][9][10][11][12][13].
At present, explanatory mechanisms of how maternal PPD affected infant weight growth were based on the following hypotheses [6-17]: (1) Maternal PPD may affect maternal rearing behavior and quality, especially feeding behavior, as well as the quantity and quality of interaction between mother and infant, thus resulted in a change in infant weight growth. (2) The effects of maternal PPD on infant weight growth may vary with infant growth time and duration of maternal PPD, and these effects were not limited to infancy, but also extended to early childhood, preschool, school-age, adolescence, and even adulthood.
To date, many epidemiologic studies examined the association between maternal PPD and infant weight growth which may vary with the income level of population [1,15,[18][19][20][21][22][23][24][25][26][27]. Maternal PPD seemed to stunt infant weight growth in India [19][20], Bangladesh [21], Zambia [22], Nigeria [23], Latin America [24] and other low-income countries. However, in United States [25], Europe [26], and other high-income countries, maternal PPD seemed to have no or few associations with infant weight gain. But none of previous research reported dynamic associations between maternal PPD and infant weight, because most of their data used a single time-point infant-weight measurement from cross-sectional or cohort survey [15,[18][19][20][21][22][23][25][26][27][28], which ignoring effect changes of maternal PPD on infant weight after1-month postpartum and longer. Dynamic evaluation of the association between maternal PPD and infant weight can not only reveal the duration of the impact of maternal PPD on infant weight growth, but also clearly re ect the change characteristics of the impact of maternal PPD on infant weight growth at different stages, so as to nd out the suspected abnormal growth impact of maternal PPD on infant weight as soon as possible and take intervention measures. Moreover, the majority of published studies did not fully control the confounding factors at the individual and family level of infants and mainly used t-test, chi-square test, analysis of variance, correlation analysis and other basic statistical methods in the statistical analysis, which may be unable to clearly estimate the effect size of maternal PPD on infant weight growth [15,[19][20][21][22][23][25][26][27][28].
The weight growth of infants is crucial to early childhood even the entire life. Simultaneously, the effects of maternal PPD on infant weight growth were not limited to its screened time. Therefore, it is urgent to do some prospective panel data studies in China, which has 500,000 to 8.6 million new mothers with postpartum depression every year [29][30], to characterize effect changes of maternal PPD on infant weight growth [31][32][33][34][35][36], and then reveal the character and duration of maternal PPD on infant weight to take timely intervention measures to improve the growth level in infancy even childhood. Thus, this study was designed, to evaluate the dynamic effects of maternal PPD on the growth of infant weight by linear mixed models (LMM) and generalized estimating equation (GEE) models in a prospective birth cohort.

Study Design and Subjects
This study was designed as a prospective ongoing birth cohort study, which was initiated in 2015 and conducted in three communities in the Kaifu District of Changsha City, Hunan Province, China. Subjects were recruited from the mothers and their infants after birth in Changsha City of Hunan province from January to December 2015, and then received face-to-face questionnaires at 1, 3, 6, 8, and 12 months. The inclusion criteria included: (1) mothers and infants were inhabitants in the Kaifu District of Changsha City; (2) mothers and infants had complete health care records in the Community Health Management Information System (CHMIS); (3) mothers agreed to participate in the investigation and signed written informed consents. Exclusion criteria included: (1) mothers had severe mental illnesses or brain diseases; (2) mothers were unable to give informed consent.

Depressive symptoms
Depressive symptoms of mothers were assessed at 1 month postpartum with a self-report ten-item depression subscale of Edinburgh Postpartum Depression Scale (EPDS) [37]. The EPDS scale has 10 items, including mood, fun, self-blame, anxiety, fear, insomnia, coping ability, sadness, crying and self-injury, and each item has 0 to 3 points with a total of 30 points. We calculated depression subscale scores and used a cut-off value of 8 points to screen postpartum depression which was recommended by Lee et al [38].

Infant weight and covariates
Outcomes of interest were infant weights at 1, 3, 6, 8, and 12 months, obtained by self-designed questionnaires through household surveys. And the weight was then converted into weight-for-age z score (WAZ) by the child growth standards of Department of maternal and child health and community health, Ministry of health of China in 2009 [39], which WAZ less than − 2 was de ned as underweight of infants.
(2) Infant covariates: infant gender (male or female), gestational week (< 37, or ≥ 37weeks), birth weight (< 2500, 2500-3999, or ≥ 4000g). Gestational age less than 37 full weeks was de ned as a premature infant. Birth weight less than 2500g was de ned as low birth weight and greater than 3999g was de ned as macrosomia according to Obstetrics and gynecology (9th ed) [44].

Sample size
Given the probable persistence of the effect of maternal PPD on infant weight growth and the regular pattern of infant weight growth over time, we hypothesized that the association between maternal PPD and infant weight was negative and varied with weight growth, and tested this hypothesis by carrying out a prospective birth cohort study. Thus, the sample size of this study was estimated based on the calculation formula of a cohort study [45]. Previous literature reported that the rate of underweight in infants with non-depressed mothers at 1year-old was 3.3% [24]. We assumed that the relative risk (RR) of the maternal PPD group was 4.0 and did with a two-sided test by α = 0.05 and β = 0.10, then the total sample size required was 713.

Statistical analysis
Frequency, rate, mean and standard deviation were used to describe group information of subjects. A Chi-square test was used to compare the differences in maternal-infant characteristics between maternal PPD and non-PPD infants. Repeated measures ANOVA was used to test the growth pattern of infant weight in different months between depressed and non-depressed mothers and identify the interaction of maternal PPD and potential confounding factors with time on infant weight.
Linear mixed model (LMM) was used to examine the association between maternal PPD and infant weight at ve periods of 1-12 months, including 1 month, 1 to 3 months, 1 to 6 months, 1 to 8 months,1 to 12 months, which maternal PPD as the independent variable, infant weight as the dependent variable, potential confounding variables maternal age, education, pre-pregnancy BMI, history of gestation, pregnancy disease, daily use of mobile phones during pregnancy, average monthly household income, and feeding pattern of infants from birth to 12 months, infant gender gestational week, birth weight as covariates. And the interaction between maternal-infant variables and time found in repeated measures ANOVA would be introduced in the model as well. Unstructured (UN) structure of G (variance/covariance of random effect) matrix was selected in LMM model which considered in most cases 46 and unrestricted structure of R (residual variance/covariance) matrix was assumed.
In addition, when the LMM model tting results showed an association between maternal PPD and infant weight, the generalized estimation equation (GEE) model was used to test the connection between maternal PPD and infant underweight at different periods of 1-12 months, which maternal PPD as the independent variable, infant underweight as the dependent variable, maternal confounding variables found to be related to infant weight by LMM tting as the covariates. We used binomial distribution and log link function in our GEE model analyses.
2lnL, AIC, BIC of information criteria were used to evaluate the LMM models tting effect, and QIC of information criteria was used to evaluate the GEEs models tting effect. Estimates of regression coe cients (β) and corresponding 95% con dence intervals (CI), risk ratio (RR) and corresponding 95% con dence intervals (CI) were used to quantify the effects of maternal PPD in LMM and GEE models respectively. A two-tailed P-value less than 0.05 was considered to be statistically signi cant. All statistical analyses were completed in statistical software IBM SPSS 22.

Results
The ow chart of participants in this birth cohort study was shown in Fig. 1. After excluding the mothers who were none inhabitant and had no health care records (n=265), refused to investigate (n=45), did not complete the PPD scale (n=16), 960 mother-infant pairs were included in our study. Besides, there were 8 (0.8%), 23 (2.5%), 30 (3.2%), 43 (4.7%) infants who were lost to follow-up at 3 months, 6 months, 8 months, and 12months, respectively.
Background data of the involved 960 maternal-infant pairs between maternal PPD and the non-PPD group were listed in Table 1. In total, 77 mothers reported PPD 1 month after delivery, and the prevalence of maternal PPD was 8.0%. Mothers who had pregnancy disease, more daily use to mobile phones during pregnancy, lower average monthly household income, premature and low birth weight infants were more likely to reported PPD, while mothers who were obese before pregnancy reported lower rates. Further, depressed mothers were more towards mix and arti cial feeding in the rst month of their infants (see Table 1).  The measurement results of weight in infants at different months were listed in Table 2. The result of repeated measures ANOVA showed infant weight was linearly gained by month with the largest type III square and mean square (F=22232.89 P<0.05) and there was a difference between depressed and non-depressed mothers over the 12-months (F=6.76 P<0.01). Thus, we considered using the linear model as the basic mixed model to t the growth trajectory of infant weight, and added the interaction covariates of infant gender (F=14.62, P<0.001) and birth weight (F=14.62, P<0.001) with month-old to model to examine the effect of maternal PPD on weight growth of infant during different periods. Ultimately, estimated xed-effect parameters of maternal PPD in the nal LMM models at ve different periods of 1-12 months were shown in Table 3. Furthermore, effects parameter estimations of maternal PPD on underweight of infants in nal GEEs models were shown in Table 4.
LMM analyses indicated that maternal PPD was negatively associated with infant weight from 1 to 12 months (P<0.05) (see Table 3). GEEs analyses found that maternal PPD was positively related to infant underweight from 1 to 3 months (P<0.05) (see Table 4). On closer analysis, the average weights of infants with maternal PPD were 0.14kg, 0.13kg, 0.13kg, 0.13kg, and 0.16kg lighter than that without PPD at 1 month, 1-3 months, 1-6 months, 1-8 months, and 1-12 months respectively (see Table 3). Then, the risks of being underweight in maternal PPD infants were 3.19 times and 3.19 times higher than that none of PPD at 1 month and 1-3 months accordingly (see Table 4).

Discussion
This birth cohort study reveals two majors ndings. First, infants of postpartum depressed mothers were continuously lighter than infants of non-depressed mothers from 1 to 12 months,even after adjustment for confounding factors, and these weight differences in infants were declined at the 3rd month and peaked at the 12th month. Second, maternal depression at 1 month postpartum was associated with underweight infants from 1 to 3 months, but not signi cant at 6-12 months. Thus, both the results were found to support the study hypotheses.
The negative association between maternal postpartum depression and infant weight growth found in this study was consistent with those of developing countries [19][20][21][22][23][24], but this study added some new ndings that this negative association changed dynamically and lasted throughout infancy. Various factors contribute to these new ndings, but likely at least two factors.
First, postpartum depression could impair mothers to feed their infant healthily and properly, infants with high dependence and nutritional needs on the depression mothers were more likely to have lightweight or low weight. For one thing, exclusive breastfeeding for infants in the rst 6 months recommended by the World Health Organization was critical for infants to obtain adequate nutrition [47]. However, the majority of published studies reported depressed mothers were more commonly to cease exclusive breastfeeding in the rst and second months [1,[48][49]. In this study, we also found that depressed mothers were more likely to discontinue exclusive breastfeeding in the rst month after comparing the differences of feeding patterns in infants between depressed and non-depressed mothers in ve time periods, which there was no difference in the other four periods. In addition, it has been proven that early discontinue of exclusive breastfeeding was a key factor in infant under-nutrition. Thus, early cessation of exclusive breastfeeding by depressed mothers may lead to their infants being lighter and more prone to low weight in the rst 3 months when infants were highly dependent on breastfeeding. For another, depression mothers were more intrusive in mother-infant interaction of feeding practice such as more stress, restrictive and emotional feeding to infants, and more breastfeeding problems (e.g. increased breastfeeding di culties, lack of breastfeeding con dence, and decreased levels of breastfeeding self-e cacy) due to their depression symptoms [48, [50][51][52][53][54][55][56], and this could make it di cult or impossible for infants to establish a sense of security for mothers, which made infants more likely to unhealthy eating behavior, and then delay weight gain according to the attachment theory [54,57]. A cohort study in Austria reported that maternal depression was associated with less responsive feeding practices at 2 years postpartum [52]. Although we did not nd any difference in feeding patterns of infants between depressed and non-depressed mothers in the other four-time periods except the rst month, the difference in feeding quality in these time periods may also lead to the slowdown of infant weight growth. In future research, we will further clarify the association between feeding quality and infant weight growth.
Second, the duration of maternal depression and the catch-up growth of low birth weight infants and may play a mediating role in explaining the continuous and dynamic association between maternal postpartum depression and infant weight growth. Firstly, although it is unclear whether maternal postpartum depression can account for infant weight growth beyond the current period of depression [58], several studies have reported a long-term association extending to preschool stage [24,59], which re ected postpartum depression may become chronicity. Besides, there was some evidence that mothers were chronic exposure to depression from 2 weeks to 6 months postpartum with various incidence [14,[60][61]], but we did not examine the association of the duration of maternal PPD and infant weight growth because of the evaluation of maternal PPD by single time in our population. Secondly, studies have con rmed that mothers of low birth weight infants were more likely to report postpartum depression [42] which also found in this study, and low birth weight infants were usually catch-up growth in infancy when given by appropriate nutrition after birth but still lighter than normal birth weight infants [62-64]. Thus, to some extent, the catch-up growth of low birth weight infants may explain that maternal postpartum depression was not associated with low-weight infants after 3 months when there was no difference in feeding patterns of infants between depressed and non-depressed mothers.
Our ndings have important implications. First, our results add new growing evidence that the effect of maternal depression at 1-month postpartum on the negative growth of infant weight is continuous, changing, and regular in the whole of infancy. Larger-scale and longer-term studies are recommended to evaluate the long-term impacts of maternal depression from pre-pregnancy to postpartum even longer on infant and children's psychological and physical development based on these combination models. Second, our results underscore the urgency of screening and treatment of postpartum depression and even perinatal depression in China

Limitations
This study was primarily limited by two aspects as follows. First, we collected our data from three streets of Kaifu District, Changsha City, China, limiting generalizability to some extent. However, our data of infant development, growth, and health were from a whole population birth cohort of the targeted community, which will be traced to the pre-school age from the maternal pregnancy, and collected as comprehensively as possible through household surveys, health information system and extraction of children's health manual. Therefore, to some extent, the infants' demographic, growth, and development characteristics of this birth cohort and large sample population are basically consistent. Second, the collection of maternal PPD data requires mothers to recall experiences from delivery to 1 month postpartum, which may have a recall bias. However, strict investigation training, proper investigation skills, standard content lling, and timely quality control of on-thespot investigation are also adopted to reduce this bias in this study.

Conclusions
This study found that the association between maternal postpartum depression and infant weight was negative, dynamic, and continuous in the whole infant period, but the association with the low weight of infants was only signi cant in the rst 3 months. Future research is recommended to focus on the longer-term effects of postpartum depression on children's physical and psychological development. And future policymaking is recommended to pay attention to prevent, screen, and treat depression of pregnant women to avoid poor growth of children. Availability of data and materials The datasets analyzed speci cally for use in this study are not publicly available due to on-going research, but reasonable requests for data can be made to corresponding author at the end of the research.
Author' contributions YY and QH designed the study; QH performed data pre-processing and data analysis, and drafted the manuscript; GC, SMH, GT, XWX, NJ, XYM, CL, RL, YS, and TZ interpreted the ndings and critically reviewed and edited the manuscript. YY nalized the manuscript. All authors reviewed and approved the nal manuscript.
Ethics approval and consent to participate All methods performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee (The Independent Ethics Committee of Clinical Pharmacology Institute, Central South University, Changsha, China (CTXY-130041-3-2) and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Written informed consent was obtained from all individual participants include in the study.

Consent for publication
Not applicable.
Competing interests Figure 1 Flow chart of participants in this birth cohort study