Objective: To assess whether different forms of family violence against women were associated with impaired size at birth and early childhood growth.
Methods: A substudy embedded into a community-based food and micronutrient supplementation trial (MINIMat) of pregnant women in rural Bangladesh included a 2-year follow-up of the 3164 live-born children of participating women. Anthropometric data were collected from birth up to 24 months of age, and converted to WHO growth standard SD scores. Size at birth and early childhood growth were assessed in relation to women’s exposure to physical, sexual and emotional violence and the level of controlling behaviour in the family.
Results: Fifty per cent of all women reported a lifetime experience of some form of family violence. The mean birth weight was 2701 g, 30% were low birth weight (<2500 g), mean birth length was 47.8 cm (17.5%, ⩽2 SD) and at 24 months of age 37% were underweight and 50% of the children were stunted. Exposure to any form of violence was negatively associated with weight and length at birth and weight-for-age and height-for-age SD scores at 24 months of age, as well as a change in weight and height SD score from birth to 24 months of age (p<0.05, adjusted for potential confounders).
Conclusions: Violence against women was associated with an increased risk of fetal and early childhood growth impairment, adding to the multitude of confirmed and plausible health consequences caused by this problem.
Statistics from Altmetric.com
Violence against women causes serious long-term physical and mental health consequences for women all over the world.1 2 3 Physically and sexually abused reproductive age women have an increased risk of pregnancy complications and their offspring more frequently have a low birth weight.4 5 6 Furthermore, two studies from rural India have indicated an association between women’s experience of physical abuse and perinatal and infant death.7 8 In a Nicaraguan study we found that partner violence was an important risk factor for infant and child mortality. The adjusted risk of death before the age of 5 years was six times higher if the mother had been exposed to both physical and sexual violence by a partner at any point in her life.9 Moreover, among daughters (but not sons) of educated mothers from rural Bangladesh we found an association between severe physical violence as well as a high level of controlling behaviour in marriage and increased under-five mortality.10
Research on the consequences of violence against women on small children’s physical health is scarce. However, studies have revealed associations between physical partner violence against women and lower immunisation coverage of their children,11 and recent data from a hospital-based case–control study in Brazil indicated a threefold increase in the risk of severe acute malnutrition in young children in families with severe physical female partner abuse.12 Furthermore, in a cross-sectional survey in Uganda lifetime intimate partner violence was suggested to be associated with overall infant illness and diarrhoea.13
In south Asia gender inequality is prominent14 and partner violence is commonly reported; in a WHO-coordinated study of 1500 ever-married rural Bangladeshi women, 69% reported a lifetime experience of intimate partner violence.3 15 South Asia is also the region with the highest prevalence (30–35%) of low birth weight in the world.16 17 Low birth weight strongly contributes to pervasive early childhood malnutrition,17 18 which is a main cause of under-five mortality in the region.17 18
We hypothesise that violence against women increases the risk of fetal and early childhood malnutrition, thus contributing to a possible association between violence against women and an increased risk of infant and under-five mortality. This study is embedded into a large prenatal food and micronutrient trial in Bangladesh, with careful monitoring during pregnancy and follow-up of children after birth. The aim of this paper is to analyse whether different forms of family violence (of which intimate partner violence represents >75%)3 against women is associated with impaired size at birth and early childhood growth. The longitudinal study design, the community-based sample, the assessment of potential confounding factors, and the sample size of more than 3000 live births followed for 2 years provide unique possibilities for studying the potential effects of violence against women on the growth of their offspring.
The study site was a rural area in the Bangladesh delta region where a well-established health and demographic surveillance system enables pregnancy identification and longitudinal follow-up. Data come from a prenatal food and micronutrient supplementation trial that included 4436 pregnant women and in this analysis includes a 2-year follow-up of live-born children of participating women. Enrolment of women took place from November 2001 to October 2003, and the children were subsequently born from April 2002 to June 2004.
At enrolment information was collected on women’s age, parity, marital status, educational level, occupation and religion. The participating women were randomly assigned into six equally sized micronutrient and food supplementation groups (with early or “usual” start during pregnancy) and either to counselling for exclusive breast feeding or a different health education message of equivalent intensity (L Persson, S Arifeen, E Ekström, et al, personal communication). Household economic status was estimated by constructing a wealth index through asset ownership.19
In the third trimester women were interviewed by paramedics regarding experiences of violence, the interviews were performed in private in the clinic in a non-judgemental manner. These paramedics had been trained in collecting this type of information. A short modified version of the WHO collaborative study questionnaire was used that was based on the conflict tactic scale.3 15 20 The different forms of violence covered were physical, sexual and emotional violence and controlling behaviour, and as possible perpetrators intimate male partners and/or other family members were considered. Data on both lifetime and current exposure were collected for physical, sexual and emotional violence. For controlling behaviour only lifetime data were collected.
A birth notification system had been established in order to measure birth anthropometry within 72 h. Data on newborn’s sex, birth weight, birth length and breast-feeding practices were collected. In a minority of cases the home visit was performed later than 72 h after birth. To adjust for changes in anthropometry occurring from birth to time at first anthropometry the obtained results were adjusted to the day of birth by a SD score transformation based on the growth pattern observed during that period for the entire group of children. During the following 2-year study period the mother–child pairs were visited monthly in their homes during the first year and every third month during the second year. On each occasion, anthropometric and breast-feeding data were collected. At 2, 6 and 12 months of age weight and length were also measured at a clinic visit (on average 2 days after the home visit). Data were controlled for consistency between home visits and clinic visits, and a few per cent had only clinic visit data available.
Measurements of weight were done by electronic and beam scales, with a maximum measurement error of 10 g (baby scales; SECA Ltd, Birmingham,UK). Locally manufactured two-track wooden length boards were used for supine length measurements, with a maximum measurement error of 1 mm. Two length measurements were recorded on each occasion and means were calculated. All equipment was calibrated and standardised on a daily basis. An independent team of data collectors repeated a 5% random sample of the interviews and measurements to ensure data quality.
Study sample and exclusions
Out of a total of 3558 registered singleton live births, 158 were excluded from follow-up due to incomplete or missing anthropometric measurements at birth, incomplete data on maternal violence exposure or severe medical conditions. A further 233 children were lost to follow-up because of outmigration or being absent from home during three subsequent home visits and three mothers refused to participate. Finally, a cohort of 3164 live-born singletons with complete maternal violence data and anthropometry registered at birth was followed for 24 months, on average 14.4 of the scheduled 16 anthropometric measurements per child were completed. At 12 of the monthly visits over 80% of the children included in the study were measured. There was no difference in violence exposure between mothers of children having anthropometric data collected at a certain age and mothers of children with missing data at the same age. Seventy-nine children in the cohort died during the follow-up period, 43% of them were neonatal deaths.
If children’s age at examination differed from the scheduled monthly control age the anthropometric data were extrapolated and adjusted to each child’s growth pattern, to match within age limits; ±7 days for the one month home visit, ±14 days for 2–11 months and ±28 days for 12–24 months home visits.
Mothers of children lost to follow-up did not differ in socioeconomic status or in partner violence exposure in comparison with study participants.
Anthropometric data were converted into Z scores using ANTHRO 2005 software and the scores were calculated for the exact age in days of every child (WHO, 2006). Growth velocity was calculated both as absolute values in centimetres per year and as yearly changes in height/length-for-age Z scores (HAZ).
The individual child was the unit of analysis and a model for evaluating the association between mother’s exposure to intimate partner violence and child growth was constructed using the general linear model technique. Analyses were carried out for different forms of violence against women and the association with height-for-age Z score (HAZ), weight-for-height Z score (WHZ), weight-for-age Z score (WAZ) and growth velocity (ΔHAZ, ΔWHZ and ΔWAZ) in their daughters and sons from birth to the age of 2 years. The following factors were considered as possible confounders: maternal intervention groups; maternal education; maternal age and parity (birth order of child), which were closely interrelated; household asset score; duration of exclusive breast feeding and religion. Any co-factor with a p<0.20 for any linear or non-linear association with exposure to violence and growth was included in the model. Adjustment for confounding factors caused only small changes in the effect estimators, thus potential residual confounding is unlikely in this model. Data were analysed using the statistical package for social sciences (version 12.01).
Informed consent was obtained at a community level and from the participating women, and data were handled with strict confidentiality. The paramedics performing the interviews were trained in mental health counselling, and women reporting exposure to violence were invited for counselling. The study was reviewed and approved by the Ethical Review Committee at ICDDR, B, Dhaka, Bangladesh.
The pregnant women were on average 26 years old (SD 5.8, range 14–44) at enrolment and 32% were expecting their first child (range 1–11). Thirty-four per cent had less than 3 years of formal schooling. All mothers were currently married and almost all (92%) did not participate in any income-earning activity. Initiation of breast feeding was universal and of long duration (>12 months), but the mean duration of exclusive breast feeding was only 3 months.
Half (50%) of all mothers had lifetime experience of some form of family violence, see table 1.
Experience of any form of family violence, as well as lifetime physical, severe physical, sexual and emotional violence was more common among mothers being Muslims, having low asset scores (poor), low educational level, older age and being multipara. A high level of controlling behaviour was more frequently reported by women being Muslims, having low asset scores and low educational level. Women’s violence exposure was not associated with the duration of exclusive breast feeding, apart from women exposed to severe physical violence breast feeding longer than others. There was no association between any form of family violence and gestational age at birth. Furthermore, women’s violence exposure did neither differ between the food interventions groups, nor between the two breast-feeding counselling groups in the MINIMat trial.
The average birth weight of the 3164 children was 2701 g (SD 403, range 1180–4250 g) and proportion low birth weight (<2500 g) was 33%. The mean birth length was 47.8 cm (SD 2.2, range 38–55 cm). Almost 9% of the children were born preterm (<37 gestational weeks, according to ultrasound). The children’s average growth velocity was 16.3 cm/year. Changes in mean WAZ was −0.15 SD/year and in mean HAZ −0.54 SD/year over the 2 years of follow-up.
Mother’s educational level and household asset scores were positively associated with higher WAZ, WHZ and HAZ for both daughters and sons, at all monthly visits. Older maternal age and multiparity were associated with overall lower anthropometric Z scores for children. Newborn children of Hindu mothers had lower weights and were shorter than children of Muslim mothers; however, after the neonatal period there was no association between mother’s religion and child weight or height.
Family violence and child growth
Children of mothers with lifetime experience of any form of family violence were smaller in body size at birth and at every monthly home visit up to 24 months of age, in comparison with children of non-abused mothers, fig 1. At 24 months the proportion of underweight was 41.9%, wasting 13.3% and stunting 55.5% among children of mothers exposed to any form of lifetime family violence, as compared with 37.0% underweight, 11.3% wasting and 49.8% stunting among children of non-abused mothers.
Also after adjusting for asset scores, maternal educational level, parity (or maternal age) and religion there was a statistically significant (p<0.05) association between mothers’ exposure to any form (whatsoever) of violence and lower WAZ and HAZ scores at birth as well as at 24 months of age for both girls and boys, table 2.
Moreover, lifetime physical, sexual and emotional violence and a high level of controlling behaviour in marriage, respectively, were associated with low birthweight (p<0.05), and also independently associated with (p<0.05) lower WAZ and HAZ at 24 months of age (not shown in table). Length at birth was shorter (p<0.05) among infants of women with any history of physical and emotional violence and a high level of controlling behaviour. All analyses were adjusted for possible confounding factors. Physical violence during the actual pregnancy was numerically associated with lower body size at birth of the offspring, although not statistically significant after adjusting for confounders.
Impaired growth velocity, ie, changes in mean height/length-for-age (HAZ) Z scores from birth to 24 months of age, was significantly more pronounced (p<0.05) for children of mothers exposed to any violence (table 2), but also for children of mothers exposed to lifetime physical and sexual violence or a high level of controlling behaviour in marriage (not shown in table).
In this cohort of 3164 rural Bangladeshi children, mothers’ exposure to family violence was significantly associated with lower weight and length-for-age SD scores, at birth and during the first 2 years of life. Furthermore, children of mothers with lifetime violence experience also had slower growth velocity during this period. These associations persisted when controlling for potential confounders. Being exposed to violence per se seemed to be more important for growth retardation than the exact form or timing of the violent acts.
This study was population based and of longitudinal design, minimising the risk of selection bias and making it possible to assess the relationship between exposure and outcome. Great efforts were made to maintain the high quality of anthropometric data. However, underreporting of exposure to violence is possible, as the results showed a lower proportion of women reporting experience of partner violence than in a WHO-coordinated study previously performed in the area.3 15 Comparing our data with only the subgroup of pregnant Bangladeshi women in the WHO study still points towards an underreporting of violence in the present study (RT Naved, unpublished data). The interviewers collected information on experiences of violence in privacy, after receiving careful training in these procedures. However, the protocol included several other time-consuming study procedures, which might have resulted in underreporting experiences of violence.21
What is already known on this topic
Violence against women causes serious long-term physical and mental health consequences for exposed women.
Physical and sexual violence against women is described to be associated with low birth weight of the offspring, as well as with an increased risk of neonatal, infant and under-five mortality.
What this study adds
This is the first study on the topic with a longitudinal follow-up of a community-based sample.
Any lifetime family violence against women, as well as physical, sexual and emotional violence and a high level of controlling behaviour in marriage were independently associated with smaller size at birth and growth impairment in early childhood.
Our finding of an increased risk of low birth weight in newborns of women exposed to violence confirms earlier reports of an association between physical or sexual partner violence and low birth weight of the offspring.5 22 In our study, retarded fetal growth rather than preterm delivery explained that association. Former reports are contradictory regarding a possible relation between violence against women and preterm births.6 23 The association between violence and fetal growth restriction could have several explanations; ie, all different form of violence might cause depression and stress1 2 3 among exposed pregnant women, leading to increased levels of stress hormones, which reduce placental circulation24 25 and also to an alteration in immunological factors, which increases the risk of severe infections.26 27 Furthermore, deficient rest, malnutrition and insufficient medical care during pregnancy are more common among abused women than others.6 28
The most important new finding in this study was an increased risk of growth impairment also after the fetal period for the offspring of mothers with experience of family violence. This implies a continued increased risk of chronic undernutrition among children of abused mothers. Even though most of the differences in size between children of abused mothers and others are already present at birth, the violence-related growth retardation became more pronounced during the 2 years of follow-up.
An association between family violence and undernutrition in early childhood could be explained by abused mothers being depressed, emotionally stressed and socially isolated,1 3 23 which reduces their ability to cope with the needs of a small child, and reduces quality of feeding interactions and other care-giving behaviour.29 30 31 Former studies have indicated a shorter duration of exclusive breast feeding among abused women.11 However, in this study the duration of exclusive breast feeding did not differ in relation to mothers’ violence exposure, apart from a small group of severely abused women breast feeding longer than others. In addition, abused women are less likely to seek external support when needed.32 We found an overall tendency of violence against women being most harmful for the growth of boys, which may reflect the increased biological vulnerability of male infants in this young age group.33 34
We conclude that family violence against women is associated with an increased risk of fetal and early childhood growth impairment and undernutrition of their offspring. These data contribute to the understanding of a possible association between gender-based violence against women and an increased risk of infant and under-five mortality.
The authors gratefully acknowledge the participation of all pregnant women and families in Matlab.
Funding The MINIMat research study was funded by the United Nations Children’s Fund (UNICEF), Swedish International Development Cooperation Agency (Sida), UK Medical Research Council, Swedish Research Council, Department for International Development (DfID), International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDR,B), Global Health, Research Fund-Japan, Child Health and Nutrition Research Initiative (CHNRI), Uppsala University and United States Agency for International Development (USAID).
Competing interests None.
Ethics approval The study was reviewed and approved by the Ethical Review Committee at ICDDR, B, Dhaka, Bangladesh.
Patient consent Obtained.
Contributors LÅP initiated and was one of the principal investigators of the main MINIMat intervention study in which this substudy was embedded. RTN was responsible for the design and supervision of the violence interviews. KÅ-M performed the data analyses and all authors participated in the interpretation of the results and writing up the manuscript. All three authors have contributed to and reviewed the final version of the manuscript. All three authors guarantee the integrity of the research.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.