Currently, there exists well-established scientific evidence supporting the practice of exclusive breast-feeding (EBF)(Reference Kramer and Kakuma1, Reference Horta, Bahl and Martinés2). WHO recommends EBF up to 6 months and thereafter its complementation with safe foods until the child reaches ≥2 years of age(Reference Kramer and Kakuma1, 3). Despite the benefits afforded by EBF, the global rates of breast-feeding still fall short of acceptable levels(Reference Yngve and Tseng4). According to WHO, only 35 % of children under 4 months of age are exclusively breast-fed(Reference Kramer and Kakuma1).
The practice of breast-feeding is a result of biological, innate and behavioural impulses and moderated through certain attributes of the mother and the newborn, as well as through the context in which they live. Among maternal factors, the repercussions of intimate partner violence (IPV) on the onset and duration of EBF have been increasingly debated in the literature(Reference Scott, Landers and Hughes5, Reference Taveras, Capra and Braveman6).
Several hypotheses justify adding IPV to this debate. According to the Pan-American Health Organization, women who suffer from IPV during pregnancy may have reduced self-esteem and thus a reduced ability to care for the newborn and adopt health-promoting measures such as sustaining EBF(7). Experiencing situations of intimate violence before, during and after pregnancy may also interfere directly in the production of milk by changing the pattern of adrenaline and noradrenaline release acting on the hypothalamo-hypophyseal axis, which is largely responsible for milk production and the let-down reflex(Reference Boutet, Vercueil and Schelstraete8). The impact of IPV on the self-esteem and self-confidence of a pregnant woman may also hamper her capacity to perceive herself as the child's exclusive source of food(Reference Mezzacappa and Katlin9).
Published studies on the repercussions of IPV on the initiation, duration and management of breast-feeding are still scarce. In a recent MEDLINE literature review, only five studies were identified with this focus and the results were rather conflicting. Acheson(Reference Acheson10) and Lourenço and Deslandes(Reference Lourenço and Deslandes11) showed that IPV reduced the time of maternal breast-feeding, whereas Bullock et al.(Reference Bullock, Libbus and Sable12), Silverman et al.(Reference Silverman, Decker and Reed13) and Lau and Chan(Reference Lau and Chan14) found no link. Although some evidence on an IPV–EBF relationship does exist, available studies fall short of providing a comprehensive account. An issue that requires investigation is with regard to the role that severe physical violence during pregnancy (SPVP) may play on EBF. In order to address this gap, the present study aims to investigate whether SPVP is an independent risk factor for early termination of EBF.
Methods
Setting and participants
The sample comprised randomly selected mothers of children under 5 months of age waiting to be consulted in five large public primary health-care (PHC) facilities of Rio de Janeiro, Brazil. Data collection took place from January to July 2007. A woman was considered ineligible if she did not experience at least one month of an intimate relationship with a partner during pregnancy or postpartum; gave birth to twins; or if there was an absolute counter-indication for breast-feeding. Out of the 853 women invited for the study, eighteen (2·1 %) were not eligible; among the remaining 835, twenty-four (2·9 %) refused to participate. Thus, 811 women were effectively interviewed in a reserved area without the presence of anyone but the interviewer, once anonymity and confidentiality of information had been guaranteed.
Conceptual model, variables and measurements
Figure 1 shows the theoretical–conceptual model of the study, which covers the dimensions most often studied in IPV during pregnancy and early weaning. The first hierarchical dimension relates to the socio-economic situation, here represented by women's educational status measured through ‘schooling achievement’, ‘work status during postpartum’ and the overall ‘economic situation of the family’. The latter variable uses the Brazilian Criterion of Economic Classification (BCEC), which encompasses information on the level of education of the household's main breadwinner, the possession of selected appliances and durable assets, and the presence or absence of a domestic employee at home(15). This variable involves five decreasing economic strata/levels (A–E).
Theoretical–conceptual model (SPVP, severe physical violence during pregnancy)
The intermediate level of the model encompasses the main variable of interest – ‘SPVP’ – as well as variables on women's lifestyles and demographic and reproductive characteristics. SPVP was assessed through the revised Conflict Tactics Scale (CTS2), which has been formally adapted for use in Brazil(Reference Straus, Hamby and Boney-McCoy16–Reference Reichenheim, Klein and Moraes20). A case was considered positive when the women reported having experienced at least one of the acts comprising the severe physical assault subscale during pregnancy, either as a victim or as a perpetrator.
The demographic and reproductive dimensions comprised ‘age’ (mothers and children), ‘sex’ (child), ‘number of children in the household’ (per mother) and the ‘mother's desire to get pregnant’. The lifestyle dimension was represented by the current tobacco smoking habits (yes/no) and alcohol and illicit drug use during gestation. Alcohol (mis)use among pregnant women was evaluated through TWEAK (Tolerance; Worry; Eye-opened, Amnesia; C/Kut-down), a screening tool previously adapted for use in Brazil(Reference Moraes, Viellas and Reichenheim21, Reference Russell, Martier and Sokol22). A Brazilian version of the NSDUQ (Non-Student Drug Use Questionnaire) was used to identify illicit drug use. The consumption of at least one illicit substance (marijuana, cocaine or solvents) defined a positive case(Reference Smart, Arif and Hughes23).
Three dimensions were postulated at the proximal level and envisaged as possible mediators of the effect of SPVP on breast-feeding: child health, utilization of health services and maternal mental health. The first dimension was represented by the ‘mother's perception regarding the child's health status’ and ‘gestational age’ at birth. The use of health services was characterized by the ‘gestational age when the mother started her prenatal follow-up’, ‘number of prenatal health-care visits’ and ‘quality of maternity care’, the latter typified by whether the hospital in which the child was born adhered to a state-sponsored pro-breast-feeding programme (Baby-Friendly Hospital Initiative (BFHI))(24). Depicting the third mediating dimension, ‘maternal self-esteem’ was assessed through a Portuguese version of the Rosenberg scale(Reference Rosenberg25, Reference Assis, Avanci and Silva26).
A child was considered to be in EBF regime when receiving only breast milk, either directly from the breast or mechanically extracted, and no other liquid or solid except for supplementary vitamins and minerals, oral rehydration solution and/or medicines(27). Since most of the mothers went to a health centre to give birth to the child seen in the paediatric sector, the assessment of EBF was based on maternal recall of infant feeding during the preceding 7 d. This strategy attempted to recover the child's established feeding routine outside the period of acute illness or convalescence. Infants are more inclined to accept only breast milk on such occasions, which increases the likelihood of false-positive information on the EBF regime. Categorizations (levels) of the other variables used in the analysis are self-evident and are displayed in Table 1 (see Results section).
Statistical analysis
The data analysis was based on the ‘current status approach’(Reference Grummer-Strawn28, Reference Aarts, Kylberg and Hornell29). Among several suggested models, preference was given to the complementary log–log model(Reference Grummer-Strawn28). This model estimates the cumulative hazard function equivalent to that estimated by the Weibull survival model(Reference Selvin30).
The theoretical model presented in Fig. 1 guided the statistical modelling process. First, an analysis was carried out to scrutinize the variables unconditionally associated with outcome. A P value ≤0·20 was considered as the cut-off point. A multivariate analysis followed, on the basis of two models. The first (Model I) assumed that the dimensions related to the child's health, use of health services and maternal mental health were mediators in the causal pathway between exposure (SPVP) and outcome (early cessation of EBF) and were thus excluded. To test this hypothesis, Model II included all three dimensions. One may assume that this hypothesis is corroborated if the effect of SPVP on early cessation of EBF declines or even disappears with the introduction of the three dimensions. Variables that modified the SPVP coefficient by at least 10 % and/or those that presented a P value ≤0·05 were retained in the models.
Ethical considerations
The present study was approved by the Research Ethics Committee of the Rio de Janeiro Municipal Health Department in conformity with the Declaration of Helsinki. All participants gave their written informed consent after anonymity and confidentiality of information had been guaranteed. The women also received advice about public facilities that help families affected by violence in Rio de Janeiro. Further contacts were encouraged if perceived to be necessary.
Results
According to Table 1, the prevalence of severe physical abuse among partners was 18·9 %. Over half of the children (53·9 %) were no longer exclusively breast-fed at the time of the interview. The median time spent in EBF was 30 d. Most of the children were aged <90 d (73·6 %), the average being 59 (sd 41·7) d. The initial univariate exploratory analysis indicated that most of the variables, such as maternal sociodemographic, lifestyle, reproductive, use of health services and child and mental health, were associated with early cessation of EBF.
Study population profile and univariate analysis
HR, hazard ratio; BFHI, Baby-Friendly Hospital Initiative.
*For complementary log–log model.
†Variable operationalized on five decreasing levels of economic stratification (from A to E).
‡Variable with 810 observations.
§Variable with 800 observations.
∥Variable represented by BFHI – the variable did not present a linear effect.
¶Reference category.
Table 2 shows the results of the multivariate analysis. The association between SPVP and early cessation of EBF remained statistically significant even after allowing for confounders (Model I). Yet, when the variables representing ‘child health’, ‘utilization of health services’ and ‘degree of maternal self-esteem’ were introduced as possibly intervening in the process, the hazard ratio reduced from 1·30 to 1·17 and lost significance. The inclusion of the first and second variables in the multivariate model reduced the SPVP point estimate by only 0·5 % and 3·3 %, respectively, but adding ‘maternal self-esteem’ led to an almost 9 % reduction and to loss of statistical significance (data not shown).
Multivariate analysis by means of Models I and II: relationship between SPVP and EBF
SPVP, severe physical violence during pregnancy; EBF, exclusive breast-feeding; HR, hazard ratio; BCEC, Brazilian Criterion of Economic Classification; BFHI, Baby-Friendly Hospital Initiative.
*For complementary log–log model.
†Variable operationalized on five decreasing levels of economic stratification (from A to E).
‡Reference category.
§Variable with 800 observations.
Whereas all of the variables fitted in Model I stayed statistically significant, the variables depicting the economic situation (BCEC), maternal age and gestational age at the beginning of the prenatal period ceased to be significant at the 0·05 level in Model II.
The probability of early cessation of EBF was also higher among women who were suspected of alcohol misuse during pregnancy and/or postpartum. Early cessation also increased: according to the mother's perception of her baby's declining health; among those who did not deliver their children in maternity homes with baby-friendly initiatives or delivered in those pending accreditation; and among women with low self-esteem.
Figure 2 shows the relationship between SPVP and the duration of EBF. Bearing a multivariate scenario described by Model I, the likelihood of children remaining in EBF among women not exposed to SPVP is consistently higher than among women exposed to SPVP throughout the postnatal period.
Probability of remaining in exclusive breast-feeding according to the occurrence or not of severe physical violence during pregnancy (
, no; 
, yes; curves fitted according to Model I, with covariates set at their mean values)
Discussion
The present study adds to the knowledge on IPV and maternal breast-feeding by providing evidence of the relationship between the physical component of this type of violence and early cessation of EBF. The results corroborate the hypothesis that SPVP is an independent risk factor for early cessation of EBF since SPVP still increases the risk of outcome by >30 %, even after controlling for confounding variables.
Women who are subjected to chronically stressful situations, such as situations of severe physical violence, tend to develop mental disorders ranging from minor ones such as anxiety and depressive symptoms to more serious forms with risk of suicide(Reference Ludermir, Schraiber and D'Oliveira31–Reference Tiwari, Chan and Fong33). Many women in this situation suffer from psychological conditions. They are unable to perform routine activities of daily living, much less care for the newborn. Thus, it seems unlikely that a mother who lives under conditions of severe violence would find the right environment and, given her own priorities, perhaps the willingness to perform such a demanding task as breast-feeding(Reference Boutet, Vercueil and Schelstraete8, Reference Rondo and Souza34–Reference Hasselmann, Werneck and Silva35).
Interestingly, the independent effect of SPVP could not be ratified once the variables related to child health, use of health services during the prenatal period and the degree of maternal self-esteem were introduced. These results suggest that the effect of SPVP in reducing EBF is partly due to the diminished self-care and care of the child as captured by a poorer health-care service attendance during pregnancy and delivery, which is a practice that usually goes in tandem with low self-esteem following repeated and degrading experiences of IPV, and by a worsening of the child's health.
Keeping a pregnant woman away from adequate assistance on account of late entry, small numbers of prenatal care appointments or because the mother gave birth in a maternity home that does not explicitly encourage breast-feeding (e.g. BFHI) makes professional guidance on the practice and importance of EBF fairly difficult. This also reduces the chances of sharing experiences with other pregnant women, especially in joining organized breast-feeding groups held in health centres and maternity homes, which is recognizably an effective incentive to initiating and prolonging breast-feeding(Reference de Oliveira, Camacho and Tedstone36, Reference Abrahams and Labbok37). Moreover, poor-quality health care during the prenatal, delivery and/or postpartum periods may be associated with the worsening health status of children and, in turn, may negatively influence the establishment of long-term EBF.
Likewise, the proposition that reduced maternal self-confidence and self-esteem is sometimes in the path between SPVP and EBF seems quite tenable. This hypothesis is strengthened by connecting previous findings pointing out that low confidence and feelings of worthlessness are consistent with the long-term psychological consequences of SPVP victimization(Reference Zlotnick, Johnson and Kohn38, Reference Rokach39), and with other studies indicating that the psychosocial well-being of women during pregnancy is an important protective factor for early weaning(Reference Scott, Landers and Hughes5, Reference Rondó and Souza40–Reference Bonet, Blondel and Khoshnood47). The importance of maternal self-esteem as a mediator of the relationship between SPVP and EBF is also reasonable when comparing the gradual effects of introducing the dimensions in the multivariate model. On a practical level, acting upon improving a woman's self-esteem and confidence during the prenatal and postpartum period may have beneficial offshoots at several levels. This would not only be a means to break the cycle of violence but would also be of help in improving the woman's attitude towards herself and her offspring, and would ultimately have a positive effect on breast-feeding.
The results of the present study must be seen in the light of its strengths and limitations. On the positive side stands the quality of information on the exposure (SPVP) and outcome (EBF) of interest. The CTS2 intersperses items depicting overt situations of violence for settling conflicts between intimate partners with socially acceptable ones, thus increasing its acceptability(Reference Straus, Hamby and Boney-McCoy16). The instrument's adequacy has been ratified by the auspicious psychometric evaluations carried out in different contexts, be it in Brazil(Reference Moraes, Hasselmann and Reichenheim18–Reference Reichenheim, Klein and Moraes20) or elsewhere(Reference Calvete, Corral and Estevez48–Reference Jones, Peter and Beck50). The option for collecting data after childbirth was another positive point, since applying the instrument during the postpartum period may have helped to detect situations of violence occurring or intensifying right at the end of pregnancy(Reference Cokkinides, Coker and Sanderson51). The multivariate approach used to analyse the data, based on a comprehensive theoretical model as a guide to control important potential confounders and test for possible mediators, may also strengthen the validity of the results(Reference Krug, Mercy and Dahlberg52).
The approach used to characterize EBF may also be commended. The literature suggests that the type of current status approach based on a single recall may lead to an overestimation of the prevalence of children being breast-fed(Reference Aarts, Kylberg and Hornell29, Reference Piwoz, Creed de Kanashiro and Lopez de Romana53). Accordingly, the ideal procedure would be to collect information on breast-feeding of children under the age of 2 years by direct observation in their homes through daily visits. This proved to be unfeasible in the present study, given the cross-sectional approach to collecting information. This option took into account the fact that intimate violence may not be studied through follow-up designs in most situations. For ethical reasons, the detection of violence would forcibly initiate a series of measures aimed at mitigating, reducing or even interrupting/ending the exposure of interest, which would inevitably alter the informative ability of the study. However, the slightly longer recall period (7 d) attempted to avoid overestimating the outcome, to the extent that it would enable the recovery of the child's usual feeding pattern beyond the periods of acute illness or recovery. Both periods are traditionally associated with increased breast-feeding(Reference Piwoz, Creed de Kanashiro and Lopez de Romana53, Reference Sharps, El-Mohandes and Nabil El-Khorazaty54).
A drawback of the present study is the impracticality of separately studying the role that different types of violence (psychological, minor physical and severe physical) play in the duration of breast-feeding on account of the reduced number of women subjected exclusively to each type. Since many cases of SPVP also involved situations of psychological violence and milder forms of physical violence, doubt exists as to whether it is the occurrence of severe forms that actually leads to increasing the likelihood of weaning or if the effect results from simultaneous victimization to different facets of IPV. Future research investigating the specific role that each type of violence has in the initiation and duration of EBF would help in setting up more targeted strategies to be used in health services.
Despite these limitations, results suggest that it would be unwise to disregard more extreme physical types of IPV in the management of breast-feeding. Sensitizing and expanding the awareness of health personnel with regard to investigating the occurrence of family conflicts and other psychosocial issues while attending to women's and children's health before and after delivery is critical. Detecting SPVP early on, thus enabling immediate referral along with effective counselling on the practice and advantages of EBF, would certainly increase the chances of delaying breast-feeding. Reciprocally, early cessation of EBF could serve as a sentinel event of situations of conjugal conflict. The knowledge that a puerpera is having a hard time breast-feeding in the absence of any other clinical and social factors as an explanation may be a useful lead for suspecting cases of IPV and a first step towards an early detection of the problem.
The findings were consistent with the basic hypothesis holding SPVP as a risk factor for early cessation of EBF, thus corroborating some previous studies(Reference Acheson10, Reference Lourenço and Deslandes11) and refuting others(Reference Bullock, Libbus and Sable12–Reference Lau and Chan14). Nevertheless, there is a long way ahead before reaching a more complete understanding of the processes underlying this relationship. In this sense, the present study should be seen as contributing towards understanding early weaning; however, new research is still necessary to explore in greater depth the role of the various forms of IPV in the initiation and duration of maternal breast-feeding. An agenda comprising studies carried out in broader contexts would be welcome, since the present findings are limited to families using public PHC services. Studying specific high-risk groups that are usually followed up in more complex health-care services (e.g. premature babies or infants with neurological problems) may shed some light on the intricacies of IPV on EBF. Similarly, much would be gained from focusing on families of higher socio-economic status who tend to be seen more often in private health-care services.
Acknowledgements
The present study was financially supported by the Fundação Carlos Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro (FAPERJ), Grant no. E-26/110.365/2007. A.S.D.O. was partially supported by the Brazilian National Research Council (CNPq), Grant no. 150426/2009-7. C.L.M. was partially supported by the CNPq, Grant no. 302851/2008-9. M.E.R. was partially supported by the CNPq, Grant no. 306909/2006-5 and 301221/2009-0. The authors have no conflict of interest to declare. A.S.D.O. collaborated in designing the study and writing the protocol, coordinated the data collection process, undertook the literature review, assisted in analysing the data and wrote the first draft of the manuscript; G.L. collaborated in designing the study and writing the protocol, coordinated the data collection process and assisted in writing the final draft of the manuscript; C.L.M. and M.E.R. managed funds for the project, designed the study, wrote the protocol, supervised the data collection process, undertook the statistical analysis and collaborated in writing the final draft of the manuscript. The authors are thankful to the field interviewers and the employees of the Silveira Martins, Marcolino Candau, Heitor Beltrão, Ariadne Lopes de Menezes and Jorge Saldanha Bandeira de Mello Municipal Health Centers (CMS) for their collaboration in collecting data.
