Seasonal trends and maternal characteristics as predictors of maternal undernutrition and low birthweight in Eastern Maharashtra, India

Abstract Few studies have assessed whether women and infants in rural and peri‐urban communities in South Asia experience seasonal fluctuations in nutritional status; however, a handful of studies have documented seasonal variability in risk factors for undernutrition including food availability, physical activity and infections. We used data from the Maternal and Newborn Health (MNH) registry, a population‐based pregnancy and birth registry in Eastern Maharashtra, India, to analyse seasonal trends in birthweight and maternal nutritional status—body mass index (BMI) and haemoglobin—in the first trimester of pregnancy. We plotted monthly and seasonal trends in birthweight, and maternal BMI and haemoglobin, and used multivariable regression models to identify seasonal and maternal characteristics that predicted each outcome. Between October 2014 and January 2018, MNH included 29,253 livebirths with recorded birthweight. BMI was assessed in 15,252 women less than 12 weeks of gestation and haemoglobin in 18,278 women less than 13 weeks of gestation. Maternal characteristics (age, education, parity and height) were significantly associated with nutritional status; however, there were minimal seasonal fluctuations in birthweight or maternal nutrition. There were significant secular trends in maternal haemoglobin; between 2014 and 2018, the prevalence of maternal anaemia decreased from 91% to 79% and moderate or severe anaemia from 53% to 37%. The prevalence of maternal underweight (45.3%) and overweight (9.8%) and low birthweight (19.1%) remained relatively constant over the study period. Our findings highlight that in some rural and peri‐urban areas in South Asia, tackling systemic drivers of malnutrition may be more effective than targeted interventions based on season.


| INTRODUCTION
Globally, 15%-20% of births, or 20 million infants, are born with a low birthweight (LBW; <2500 g) (WHO, 2014). Southern Asia, with the highest prevalence of LBW worldwide, is home to an estimated 9.8 million LBW infants born each year (Blencowe et al., 2019).
LBW infants have an increased risk of mortality, developmental delays and chronic diseases (Black et al., 2013). Undernutrition during pregnancy from conception through delivery-marked by low prepregnancy body mass index (BMI), inadequate gestational weight gain, micronutrient deficiencies and anaemia-are all major contributors to maternal and infant mortality and morbidity, including LBW (Black et al., 2013;Patel et al., 2018;Scholl, 2011;Victora et al., 2008).
In South Asia, the majority of undernourished women and children reside in rural areas where subsistence and semisubsistence agriculture serve as the primary livelihoods (Gillespie, van den Bold, Hodge, & Herforth, 2015;Kadiyala, Harris, Headey, Yosef, & Gillespie, 2014). Notably, agriculture is driven by annual cycles in weather, which affect labour demands, food supplies and disease vectors in the environment (Devereux, Sabates-Wheeler, & Longhurst, 2013). In South Asia, the weather patterns are dominated by the monsoon season (which begins in June) followed by several months of cool weather (winter) before peak heat in April and May prior to the next monsoon cycle. For households living close to the poverty line, seasonal variations can have drastic negative health and economic consequences for individuals and communities (Devereux et al., 2013). Recently, seasonal variations have been proposed as an 'indirect determinant' of undernutrition in the first 1000 days in South Asia, particularly because of the effect of seasonality on agriculture and food systems (Madan, Haas, Menon, & Gillespie, 2018). If seasonal stress is a driver of malnutrition, this would warrant (1) further research into the specific drivers of seasonal variability and (2) an investment in programmes designed to address seasonal trends such as the use of supplementary feeding or social safety net programmes during lean seasons or programmes designed to target seasonal fluctuations in food systems, dietary intake, physical activity, infection and other individual drivers of malnutrition (Madan et al., 2018).
To date, season of birth has been shown to be associated with nutritional status in pregnancy, birthweight and offspring health in several studies in sub-Saharan Africa (Anya, 2004;Moore et al., 1997;Prentice, Cole, Foord, Lamb, & Whitehead, 1987;Rayco-Solon, Fulford, & Prentice, 2005;Rickard et al., 2012;Roba, O'Connor, Belachew, & O'Brien, 2015;Verhoeff, Brabin, Chimsuku, Kazembe, & Broadhead, 1999;Waterland et al., 2010). Fewer studies have assessed the effects of seasonality on the nutrition and health of women and children in South Asia, and no consistent seasonal trend has been identified (Bondevik, Lie, Ulstein, & Kvåle, 2000;Madan et al., 2018;Rao, Kanade, Yajnik, & Fall, 2009;Shaheen et al., 2006;Stevens et al., 2016). Research on the local impact of seasonality on nutritional status is essential for collecting high-quality surveillance data as well as the development of context-appropriate nutrition interventions. The discovery of wide seasonal variability in nutrition indicators indicates a need to carefully time surveillance, research or programme-related surveys to account for seasonal trends; it would also warrant the need to conduct studies to further understand the specific local drivers of seasonal variability to strengthen local programmes and policies. By contrast, if there is no seasonal variability in nutritional status in a given area, this could highlight the importance of tackling long-term, underlying drivers of poverty and malnutrition.
In this study, we utilize data from the ongoing Maternal and Newborn Health (MNH) registry in Nagpur, Eastern Maharashtra, India, to examine whether there were seasonal trends in birthweight and maternal nutritional status (BMI and haemoglobin concentration in the first trimester of pregnancy) in pregnancies from 2014 to 2018.
Given the importance of poverty-related risk factors for malnutrition, we also assess long-term, nonseasonal, demographic characteristics as predictors of birthweight and maternal nutrition in this Central Indian population.
Mothers enrolled in MNH come from rural and peri-urban households in the districts surrounding Nagpur City (Nagpur, Bhandara, Wardha and Chandrapur), where the majority of households work in agriculture. We thus hypothesized that there would be seasonal trends in maternal and child undernutrition given the seasonal fluctuation in important risk factors for undernutrition including food availability (Hillbruner & Egan, 2008;C Panter-Brick, 1997), infectious diseases (Bondevik et al., 2000;Madan et al., 2018), agricultural workloads and energy expenditure (Panter-Brick, 1993;Rao et al., 2009;Stevens et al., 2016). For women, we hypothesized that there would be an increased risk of malnutrition

Key messages
• Seasonal variability has been proposed as an important driver of undernutrition in the first 1000 days because of studies documenting seasonal fluctuations in risk factors including food availability, physical activity and infections.
• South Asia is considered particularly sensitive to seasonal trends because of the monsoon season.
• We did not find seasonal variability in the nutritional status of pregnant women or in birthweight in Eastern Maharashtra, India. Low birthweight and maternal undernutrition remained consistently high and were strongly correlated with maternal characteristics (age, education, parity and height).
• In Eastern Maharashtra, tackling underlying drivers of malnutrition may be more effective than targeting interventions based on seasonal risk factors.
in women during the monsoon season and the months immediately after the monsoon, as has been documented in other studies in South Asia in different ecological zones (Bondevik et al., 2000;Hillbruner & Egan, 2008;Panter-Brick, 1993). In most of South Asia, the monsoon season coincides with an increase in infectious diseases including malaria, influenza and dengue (Bondevik et al., 2000;Madan et al., 2018), increased agricultural workloads and energy expenditure (Panter-Brick, 1993;Rao et al., 2009;Stevens et al., 2016) and lower dietary diversity (Hillbruner & Egan, 2008;Panter-Brick, 1997). The one existing study of seasonality on birthweight in Maharashtra found that birthweights were highest in the summer (February-May), followed by the rainy season (June-September), and lowest in the winter (October-January) (Rao et al., 2009). Consistent with these findings, and given the importance of women's nutritional status in the second and third trimesters for birthweight (Hasan et al., 2018;Hickey, Cliver, McNeal, Hoffman, & Goldenberg, 1996;Siega-Riz et al., 2009;Strauss & Dietz, 1999;Young et al., 2019), we hypothesized that children born in the winter (the season following the monsoon) would have the lowest birthweight as has been documented in two studies in South Asia (Rao et al., 2009;Shaheen et al., 2006). Although seasonality is likely to affect acute conditions that affect maternal and child nutrition, we also hypothesized that underlying maternal characteristics (age, education, parity and height) would influence our outcomes of interest and sought to quantify the relationship between these underlying characteristics and maternal and infant nutrition outcomes. At the first antenatal visit, information regarding the enrolled women's characteristics including age, education, parity, gestational age, height, weight and haemoglobin were recorded by a trained research administrator (a medical officer or auxiliary nurse-midwives employed at PHCs and subcentres). Weight was measured to nearest 100 g using a spring balance, and height to nearest centimetre was measured using a nonflexible measuring tape fixed to a wall. Haemoglobin was estimated within 2 weeks of enrolment using Sahli's method (Path, 1997). Birthweight was measured within 24 h of birth by using either a pan spring or a pan digital weighing scale. Data collection and management for the Nagpur site of MNHR have also been published previously (Patel et al., 2018).

| Study population
Briefly, trained registry administrators collected data on standardized paper forms. Each form was then manually reviewed prior to data entry at Lata Medical Research Foundation, Nagpur. After entry, data were transmitted to the GN data management centre (Research Triangle Institute, Durham, North Carolina, USA) where digital data checks were performed on a monthly basis. The data management centre forwarded edit, monitoring and performance reports to the Nagpur site team for data cleaning and management.
The data were continually reviewed for quality by NICHD-appointed data monitoring committee and by the Nagpur site data collection team.

| Study sample and key variables
Our study sample is composed of all women enrolled in the MNH registry between October 1, 2014, and January 31, 2018. Mothers could enrol in MNH at any point in their pregnancy including before or immediately after labour and delivery. For our birthweight sample, we included all infants born between October 1, 2014, and January 31, 2018, with a recorded birthweight regardless of when the mother enrolled in the registry. Maternal nutritional status was assessed at the enrolment visit only, preventing an assessment of change in nutritional status (i.e. gestational weight gain) throughout pregnancy. Given that maternal BMI changes substantially over the course of pregnancy, to enhance comparability across mothers, we limited our maternal BMI sample to only mothers with BMI assessed in the first 12 weeks of pregnancy, which best approximates prepregnancy BMI. Maternal blood volume also changes dramatically over the course of pregnancy, resulting in different recommendations for anaemia cut-offs based on pregnancy trimester (Centers for Disease Control and Prevention, 1989;WHO, 2011WHO, , 2016. We thus limited our maternal haemoglobin sample to only women who had their haemoglobin assessed in the first trimester of pregnancy (<13 weeks). All mothers who had their nutritional status assessed in the first trimester were thus eligible for inclusion in the BMI and haemoglobin samples. For the haemoglobin sample, haemoglobin assessment dates were confirmed to be within 13 weeks of the date of the reported last menstrual period. For the BMI sample, an additional survey question confirming that mothers had BMI assessed before 12 weeks of gestation was also used for inclusion LOCKS ET AL. criteria. Because the first trimester analyses aim to assess whether there are seasonal trends in maternal nutritional status in the first trimester, there were no exclusions for birth outcomes or subsequent maternal or neonatal morbidity or mortality. If a mother enrolled prior to January 31, 2018, and had her nutritional status assessed in the first trimester of pregnancy, she was eligible for inclusion in the BMI and haemoglobin samples, even if her child was born after January 31, 2018, and thus did not contribute to the birthweight sample.
Maternal first-trimester BMI was calculated by dividing weight in kilograms (kg) by squared height in meters. Categorical variables were then created for the outcomes of interest. BMI classifications were based on the WHO expert consultation on BMI in Asian populations, which highlighted that although the international cutoff for overweight is set at 25 kg/m 2 , in Asian populations, cardiometabolic risk increases with BMI > 23 ( Barba et al., 2004). We defined underweight as BMI < 18.5 and overweight as BMI > 23 in accordance with the WHO consultation. Anaemia is defined as haemoglobin < 11 g/dl, moderate anaemia as 7-9.9 g/dl and severe anaemia as <7 g/dl in accordance with the WHO cut-offs for women in their first trimester of pregnancy (WHO, 2011(WHO, , 2016. Because severe anaemia was very rare in this population, moderate and severe anaemia were collapsed into a single category for analyses. LBW was defined as a birthweight < 2500 g. Cut-offs for height categories (≤145, 145-149.9, 150-154.9 and ≥155 cm) were determined based on a meta-analysis on the relationship of maternal short stature and offspring size for gestational age (Kozuki et al., 2015).
The Nagpur climate is classified as tropical wet and dry with a pronounced dry winter and heavy rain in the monsoon season (Geiger, 1954). Nagpur district experiences average annual rainfall of Eastern Maharashtra (Author, Tiple, & Khurad, 2009;Kakde, Kakde, & Saoji, 2001;Rao et al., 2009) combined with information from the U.S. national weather service. We defined summer as February-May, monsoon season as June-September and winter as October-January.

| Data analysis
Descriptive statistics-frequencies with percentages and means with standard deviations-were used to present the socio-demographic characteristics of the study samples. Data are presented for the birthweight, BMI and haemoglobin samples separately. Plots were created of the mean BMI, haemoglobin and birthweight as well as the prevalence of maternal underweight, overweight, anaemia, moderate or severe anaemia and LBW by month and season.
Monthly and seasonal means and prevalences and their corresponding standard errors were obtained using survey procedures accounting for clustering. The seasonal data are presented in three ways: (1) by calendar month (averaged across years), (2) by calendar month and year and (3) by season (winter, monsoon or summer) and year.
To assess whether maternal characteristics and/or seasonal variables predicted birthweight, maternal BMI or maternal haemoglobin concentration, generalized estimating equations (GEEs), with an exchangeable correlation structure to account for clustering, was used. Underlying maternal characteristics of interest were selected based on a review of the literature and what was available in mothers in the MNHR. This included mother's age, education, parity and height. Although these characteristics are modifiable through long-term poverty reduction and family planning interventions, they all precede the current pregnancy by years and are nonmodifiable in the context of ante-natal care interventions aimed at improving nutrition during or immediately before pregnancy. Categories for predictors were used because the exposureoutcome relationships were assumed to be non-linear. Univariable, reduced multivariable and full multivariable GEE models are presented. Reduced models were used to assess the relationship between maternal characteristics and each outcome of interest. All reduced models adjust for maternal age and education, and for the maternal BMI and maternal haemoglobin outcomes, the reduced models also adjust for gestational age at outcome assessment. Full multivariable models were used to assess the relationship between seasonality and each outcome for interest. The full multivariable models for birthweight adjust for maternal age, education, parity and height. For maternal BMI, the full multivariable models adjust for gestational age when weight and height were assessed as well as maternal age, education and parity, and for maternal haemoglobin, the full multivariable models adjust for gestational age when haemoglobin was assessed as well as maternal age, education, height and parity. The association between seasonality and each outcome was assessed with independent models defining seasonality in three different ways: (1) based on month of outcome assessment (January-December), (2) based on season of outcome assessment (winter, summer or monsoon) and (3) based on season with an interaction term for the year. Independent full multivariable models adjusting for underlying maternal characteristics (age, education, parity and height) were also used to assess maternal BMI and maternal haemoglobin in the first trimester as predictors of birthweight. Maternal nutritional status in the first trimester is excluded as a covariate in models assessing the association between seasonality and birthweight because these variables would be on the causal pathway through which season could affect birthweight.

| Ethical approvals
The Maternal and Newborn Health Registry was approved by the Boston University Medical Campus and Lata Medical Research Foundation Institutional Review Boards. All mothers provided informed consent for themselves and their infants to participate in the study.
The study was monitored by a Data and Safety Monitoring Committee appointed by NICHD. The committee reviewed the study annually.

| RESULTS
Between October 1, 2014, and January 31, 2018, 35,387 women enrolled in the MNH registry in Nagpur, and there were 29,253 births with a recorded birthweight (our final sample for birthweight analyses). Infants whose mothers enrolled during this period but were born after January 31, 2018, were not included in birthweight analyses because their data were not available at the time of analysis, but their mothers were eligible for inclusion in the first-trimester BMI and haemoglobin analyses; 19,092 women, 54% of all women who enrolled during this period, enrolled during their first trimester of pregnancy (<13 weeks of gestation). Among the women who enrolled in their first trimester, 18,278 women had a recorded value for haemoglobin before 13 weeks of gestation, and 15,252 has their weight and height assessed before 12 weeks of gestation.
Mothers who enrolled in the first trimester and comprised the BMI and haemoglobin samples had similar characteristics and infant birthweights to mothers in the birthweight sample (Table 1). The one exception was maternal BMI in the haemoglobin sample: Underweight women (BMI < 18.5 kg/m 2 ) were more likely to have their haemoglobin assessed prior to 13 weeks gestation than women with a BMI ≥ 18.5 kg/m 2 . Fifty-five percent of mothers in the haemoglobin sample were underweight compared with 46.2% and 45.4% in the birthweight and BMI samples. In all three groups, approximately half of women had a secondary education or less.
Over three quarters of mothers were between 21 and 29 years of age, and half were pregnant for the first time. Maternal undernutrition-both chronic and acute-was highly prevalent in this population. Approximately two-thirds of mothers were below 155 cm in height, and a quarter were below 150 cm. Over a third of mothers who had their BMI assessed in the first 12 weeks of pregnancy were underweight, and over eight-tenths of those who had their haemoglobin assessed in the first 13 weeks of pregnancy were anaemic at enrolment.

| Birthweight
Mean birthweight in our sample was 2681 g ± 446 (SD), with monthly means ranging from 2659 to 2706 g across the 12 calendar months (Figure 1). The overall prevalence of LBW was 19.1%, and it remained between 16.8% and 20.6% across the 12 calendar months.
On average, birthweight was highest in the summer months (with a peak in April of 2706 g) and lowest during the monsoon and the months immediately following the monsoon (with a nadir in October of 2659 g). There was a small secular decline in mean birthweight from October 2014 to January 2018. The small differences in mean birthweight across calendar months and the secular decline were statistically significant in multivariable models adjusting for underlying maternal characteristics; however, the difference in means never exceeded 50 g (Table 2). There were, however, statistically and clinically significant differences in birthweight based on maternal characteristics. Maternal age, education, parity, height, BMI in the first trimester and anaemia status in the first trimester were all significantly associated with birthweight. Of particular note, in multivariable models adjusted for maternal age and education, short stature remained a significant predictor of birthweight: Mothers with statures below 145 cm gave birth to infants who were on average 206.2 g lighter than mothers with statures above 155 cm. In multivariable models adjusting for all underlying maternal characteristics (age, education, height and parity), maternal BMI in the first trimester of pregnancy was significantly correlated with birthweight: Mothers who were extremely underweight in the first trimester (BMI < 15 kg/m 2 ) gave birth to infants who were 150.3 g lighter than mothers with BMIs in the 'normal' range (18.5-23.0 kg/m 2 ).

| Maternal BMI in the first trimester
Mean BMI remained between 19.1 and 19. 4 kg/m 2 across all 12 calendar months, with small month-to-month variability ( Figure 2 in the multivariable models adjusting for maternal characteristics, but the differences were small. Mothers who enrolled in the summer months (February-May) had a mean BMI that was 0.2 (0.0, 0.3) kg/m 2 higher (P = 0.01) than mothers who enrolled during the monsoon and its proceeding months (June-September). from 91% to 79% during this same time period and the prevalence of moderate or severe anaemia, which ranged from 50% to 55% in the early seasons of the study, decreased to 37% by winter 2017-2018.

| Maternal haemoglobin in the first trimester
Maternal characteristics were significantly associated with haemoglobin concentration (Table S2). In multivariable models, maternal age, parity and height were all significantly associated with haemoglobin concentration. Notably, mothers with more than a secondary education had an average mean haemoglobin concentration All women who enrolled in MNH between October 2014 and January 2018 were eligible for inclusion. All infants with a recorded birthweight during this time period are included in the birthweight sample. Because weight and haemoglobin change substantially over the course of pregnancy, maternal BMI and haemoglobin analyses are restricted to women who completed assessments in the first trimester of pregnancy in order to ensure comparability of nutritional indicators. All women with haemoglobin assessed <13 weeks are included in the haemoglobin sample. A specific screening question with a dichotomous cut-off for weight assessment at 12 weeks was used for BMI; thus, the maternal BMI sample includes only mothers whose weight was assessed before 12 weeks of gestation. b Proportion of women in each BMI category only provided among women who had BMI assessed in their first trimester of pregnancy. c Anaemia categories based on the WHO guidelines for defining anaemia in the first trimester of pregnancy (WHO, 2011(WHO, , 2016. Severe anaemia is defined as hemogloin <7 g/dl, moderate anaemia as 7 g/dl ≤ haemoglobin <8 g/dl, mild anaemia as 8 g/dl ≤ haemoglobin <11 g/dl and no anaemia (haemoglobin > 11 g/dl). d Dataset does not include birthweight for births that occurred after January 2018. Adjusted for maternal age, education, parity and height.
c Difference in means and corresponding 95% confidence intervals and P values obtained with generalized linear models accounting for cluster correlations using an exchangeable correlation structure.

LOCKS ET AL.
F I G U R E 2 Monthly, annual and seasonal trends in maternal body mass index (BMI) among women who enrolled in the registry from October 2014 to January 2018 and were less than 12 weeks of gestation at the time of weight assessment (n = 15,252) F I G U R E 3 Monthly, annual and seasonal trends in maternal haemoglobin concentration among women who enrolled in the registry from October 2014 to January 2018 and were less than 13 weeks of gestation at the time of haemoglobin assessment (n = 18,278) that was 0.32 (0.23, 0.42) g/dl higher than mothers who had less than a secondary education adjusting for maternal age and gestational age at time of haemoglobin assessment. After adjustment for gestational age at haemoglobin assessment, maternal age and maternal education, mothers with two or more previous pregnancies had haemoglobin concentrations that were 0.24 (0.18, 0.29) g/dl lower than primigravid mothers, and mothers with a very short stature (height < 145 cm) had concentrations that were on average 0.20 (0.13, 0.28) g/dl lower than mothers who were 155 cm or taller.

| DISCUSSION
Our findings from this population-based birth registry in Eastern Maharashtra indicate that there were no substantial seasonal fluctuations in birthweight or maternal nutritional status in the first trimester of pregnancy in this population. We did find, however, that underlying maternal characteristics (age, education, height and parity) were associated with significant differences in birthweight and Africa (Anya, 2004;Bouvier et al., 1997;Chodick, Shira Flash, & Shalev, 2017;Hlimi, 2015;Hoestermann, Ogbaselassie, Wacker, & Bastert, 1996;Hounton et al., 2008;Moore et al., 1997;Prentice et al., 1987;Rayco-Solon et al., 2005;Rickard et al., 2012;Roba et al., 2015;Verhoeff et al., 1999;Waterland et al., 2010). Research on the seasonality of maternal and child nutrition in South Asia is more limited. A recent review of 14 studies on seasonal variability and maternal and child nutrition in South Asia concluded that there is a compelling case that seasonal variation affects several of the proximal causes malnutrition in the first 1000 days-food access and availability, dietary intake, health and infection, maternal energy expenditure and child care practices (Madan et al., 2018). Notably, the review only identified four studies in pregnant and lactating women (Panter-Brick, 1993;Rao et al., 2009;Schmid, Salomeyesudas, Satheesh, Hanley, & Kuhnlein, 2007;Stevens et al., 2016), with 10 additional studies in children, the majority of which assessed diarrhoea and infectious morbidity as their outcomes of interest. There may be greater seasonal variability in the risk factors for malnutrition in children than in pregnant women; however, this analysis does not include data on childhood malnutrition beyond birthweight. Individual studies in South Asia have found seasonal variability in determinants of nutritional status such as household food insecurity (Hillbruner & Egan, 2008;Stevens et al., 2016), dietary intake (Rao et al., 2009;Schmid et al., 2007;Stevens et al., 2016) and energy expenditure (Panter-Brick, 1993;Rao et al., 2009); however, few studies include data on women's nutritional status or offspring birthweight as an outcome. One study from Nepal reported lower mean haematocrit and a higher prevalence of moderate or severe anaemia in the monsoon season (June through September) (Bondevik et al., 2000), and two separate studies (Rao et al., 2009;Shaheen et al., 2006), including one that was also conducted in Maharashtra (Rao et al., 2009 ally, approximately half of anaemia is estimated to be due to iron deficiency (Stoltzfus, 2003), and the 2016-2018 Comprehensive National Nutrition Survey indicated that among adolescents in India (aged 10-19 years), 41.5% of anaemic adolescents (aged 10-19 years) also had iron deficiency. The attribution of anaemia due to iron deficiency may also be higher in Maharashtra where the prevalence of iron deficiency among 10-19 years old is 31.2%, higher than the national average of 21.5%. Despite the significant decline in the prevalence of anaemia in pregnant women in our study from 2014 to 2018, it should be noted that the prevalence of anaemia in pregnant women in Eastern Maharashtra remained above 75% throughout the study period, indicating that anaemia among pregnant women continues to be a severe public health problem in this population (WHO, 2011).
Our study has several limitations. Some mothers did not enrol in the MNH registry in their first trimester of pregnancy; thus, we do not have comparable maternal BMI or haemoglobin on all mothers; however, notably, there were no differences in key variables (birthweight or other maternal characteristics) between mothers in the birthweight sample and mothers with nutritional information from the first trimester of pregnancy. We also did not collect data on the change in nutritional status over the course of pregnancy, nor do we have data on dietary intake, energy expenditure or infectious diseases, so we are limited in our ability to assess seasonal variability in the determinants of nutritional status. The majority of households in Eastern Maharashtra are engaged in agriculture as their primary livelihood (Directorate of Economics and Statistics, Planning Department, G. of M., 2019); however, we did not collect individual-level information on livelihoods and cannot identify which specific households engage in subsistence agriculture and thus could be more vulnerable to seasonal changes. Our study also has several strengths including the collection of birthweight data from a population-based registry of 30,000 pregnancies over a continuous 40-month period. Over half of pregnancies also included anthropometric and haemoglobin assessments in the first trimester of pregnancy. We also collected sociodemographic information about mothers, allowing us to adjust for important confounders. Our findings highlight that rates of maternal and child undernutrition in Eastern Maharashtra remain alarmingly high and that only maternal anaemia has begun to decline in recent years.

| CONCLUSION
Research on the local impact of seasonality on nutritional status is essential for collecting high-quality surveillance data as well as the development of context-appropriate nutrition interventions. The discovery of wide seasonal variability in nutrition indicators indicates a need to (1) carefully time national or programme-related survey assessments to account for seasonal trends and (2)