Systematic review
DOI: 10.1111/1471-0528.13928 www.bjog.org
Intimate partner violence during pregnancy and the risk for adverse infant outcomes: a systematic review and meta-analysis BM Donovan,a CN Spracklen,b ML Schweizer,c,d KK Ryckman,a,e AF Saftlasa a
Department of Epidemiology, College of Public Health, University of Iowa, Iowa City, IA, USA b Department of Genetics, University of North Carolina, Chapel Hill, NC, USA c Department of Internal Medicine, Carver College of Medicine, University of Iowa, Iowa City, IA, USA d Iowa City VA Health Care System, Iowa City, IA, USA e Department of Pediatrics, Carver College of Medicine, University of Iowa, Iowa City, IA, USA Correspondence: Dr AF Saftlas, University of Iowa College of Public Health, 145 North Riverside Drive, S427 Epidemiology Department, Iowa City, IA 52240, USA. Email audrey-saftlas@uiowa.edu Accepted 6 January 2016. Published Online 9 March 2016.
Background Intimate partner violence (IPV) is of particular
concern during pregnancy when not one, but two lives are at risk. Previous meta-analyses have suggested an association between IPV and adverse birth outcomes; however, many large studies have since been published, illustrating the need for updated pooled effect estimates. Objectives To evaluate the relationship between IPV during
pregnancy and the risk of preterm birth (PTB), low-birthweight (LBW), and small-for-gestational-age (SGA) infants. Search strategy We searched PubMed and SCOPUS (from
inception until May 2015), and the reference lists of the relevant studies. Selection criteria Observational studies comparing the rates of at
Main results Intimate partner violence (IPV) was significantly associated with PTB (OR 1.91, 95% CI 1.60–2.29) and LBW (OR 2.11, 95% CI 1.68–2.65), although a large level of heterogeneity was present for both (I2 = 84 and 91%, respectively). The association with SGA was less pronounced and marginally significant (OR 1.37, 95% CI 1.02–1.84), although fewer studies were available for meta-analysis (n = 7). Conclusions Our meta-analysis indicates that women who
experienced IPV during pregnancy are at increased risk of having a PTB, and an LBW or an SGA infant. More studies examining the association between IPV and SGA are needed. Keywords Domestic violence, low birthweight, partner abuse, premature, small for gestational age.
least one adverse birth outcome (SGA, LBW, or PTB) in women who experienced IPV during pregnancy and those who did not.
Tweetable Abstract Meta-analysis of IPV during pregnancy finds
Data collection and analysis Data extracted from 50 studies were pooled and pooled odds ratios were calculated using randomeffects models.
Linked article: The article has journal club questions by EYL Leung, p. 1300 in this issue. To view these visit http://dx.doi.org/ 10.1111/1471-0528.13926.
increased risk for preterm birth, LBW and SGA infants.
Please cite this paper as: Donovan BM, Spracklen CN, Schweizer ML, Ryckman KK, Saftlas AF. Intimate partner violence during pregnancy and the risk for adverse infant outcomes: a systematic review and meta-analysis. BJOG 2016;123:1289–1299.
Introduction Intimate partner violence (IPV) is a major health concern both nationally and globally. It is a violation of human rights, and can lead to both physical and mental ailments.1 The US Centers for Disease Control and Prevention (CDC) defines IPV as physical, sexual, or psychological harm by a current or former partner or spouse.2 Its frequency and severity can range from one hit or emotional put-down to
ª 2016 Royal College of Obstetricians and Gynaecologists
severe physical injury and emotional humiliation.2 Each year, an estimated 1.5–4.0 million US women are victims of IPV.3–5 IPV is of particular concern during pregnancy when not one, but two lives are at risk. Indirect or direct exposure to IPV during pregnancy has been shown to increase a woman’s risk of having an adverse birth outcome.6,7 Physical assault to the abdomen or sexual trauma experienced during pregnancy may increase the risk of spontaneous
1289
Donovan et al.
abortion, preterm delivery, low birthweight (LBW), or neonatal death.7 Adverse birth outcomes may also be indirectly facilitated through negative maternal behaviours, inadequate nutrition or prenatal care, and increased stress levels.7 Two previous meta-analyses examined IPV as a risk factor for adverse birth outcomes. One meta-analysis, published in 2001, included only studies that assessed the relationship between IPV and LBW.8 The other meta-analysis assessed LBW, preterm birth (PTB), and small for gestational age (SGA) among English-language studies only, and did not attempt to contact authors to obtain sufficient data for analyses.9 Both meta-analyses found significantly increased unadjusted odds of LBW among infants of women who were exposed to IPV compared with women who were not exposed. Shah and Shah also showed a significant increase in the adjusted odds of LBW, and both the unadjusted and adjusted odds of PTB; the unadjusted odds of SGA were increased, but not statistically significantly.9 Several large-scale studies have been published since 2010, which may add new insights to the existing body of knowledge.10–17 Previous meta-analyses on this topic did not analyse the effects of specific types of violence on adverse birth outcomes. The objective of this study was to systematically review and statistically summarize observational studies examining the risk of PTB, LBW, and SGA births among women who experienced IPV during pregnancy, compared with those who did not. We hypothesised that women who experienced IPV during pregnancy will be at higher risk of delivering PTB, LBW, and SGA infants than women who did not experience IPV during pregnancy.
Methods Search strategy This meta-analysis follows the Meta-analysis of Observational Studies in Epidemiology (MOOSE) criteria.18 A systematic literature search of eligible studies was conducted using the PubMed/MEDLINE and Scopus databases, from their dates of inception through May 2015, for studies evaluating the association between IPV during pregnancy and PTB, LBW, and SGA births. No specific search software was used. The following medical subject headings (MeSH) and keywords were used to find relevant articles in both databases: (psychological abuse OR controlling behavior OR humiliation OR emotional abuse OR non-physical violence OR threats of violence OR battering OR domestic violence OR spouse abuse OR sex offenses OR coercion OR social isolation OR family violence OR abuse assessment screen OR conflict tactics scale OR intimate partner violence) AND (premature birth OR infant, premature OR infant, small for gestational age OR infant, low birthweight
1290
OR fetal growth retardation OR pregnancy outcome OR birth outcome).
Study selection Publications identified from the literature search were screened for duplicates. Two investigators screened titles and abstracts (BMD and CNS), and potentially relevant articles were selected for full-text review. Studies were considered for inclusion in our meta-analysis if they met the following criteria: (1) an original, observational study that examined the association between IPV during pregnancy and SGA, LBW, and/or PTB (outcomes had to be mutually exclusive); (2) included a comparison group of women who did not experience IPV during pregnancy; (3) presented the raw data necessary to construct an unadjusted 2 9 2 contingency table; and (4) was published in English or another language that uses the Latin alphabet (for translation purposes). For the purpose of this study, PTB, LBW, and SGA were defined as infants born before 37 weeks of gestation, weighing <2500 g, and with a birthweight below the tenth percentile for a given gestational age, respectively.19 Reference lists of abstracted articles were hand-searched for additional relevant articles. Only published studies were included in our analysis. The grey literature was not examined due to the large number of studies that met our inclusion criteria. Google TRANSLATE was used to translate articles into English.
Data abstraction Full-text articles were obtained for all studies that initially met the inclusion criteria. Two independent reviewers (BMD and CNS) abstracted all studies for potential inclusion and quality using a piloted, customised data abstraction form, resulting in a concordance rate of 81%. Inconsistencies between the two reviewers were adjudicated by a third, independent reviewer (AFS). Information collected included study characteristics (author, year of publication, study location, study dates, and study design), subject recruitment procedures, population characteristics, definition of IPV-exposed and control groups, screening tool used, outcome definition(s), study exclusion criteria, ascertainment method for exposure and outcomes, information on potential confounding factors, data on the association between IPV and the outcomes (2 9 2 contingency table information), and crude and adjusted odds ratios (when available). We were unable to abstract the information needed for analysis (i.e. complete the 2 9 2 contingency table) for 36 studies. In an attempt to keep these studies in our final analysis, we contacted 32 study authors who had corresponding email addresses listed in their publication. Twelve responded with the requested data on either the first or
ÂŞ 2016 Royal College of Obstetricians and Gynaecologists
IPV During Pregnancy and Poor Infant Outcomes
second contact attempt, and data received from two of the 12 studies provided all the information necessary for inclusion in the meta-analysis.
Quality assessment The quality assessment was performed by applying the Newcastle–Ottawa Scale.20 In customising the scale to fit this study, we took into account the study sampling methods and similarities between the study groups regarding adjustment for confounding factors, the ascertainment of exposure and outcomes, and study design. Our abstracting instrument included a total of eight questions worth a maximum of 9 points possible, with higher scores reflecting studies of higher quality. BMD and CNS independently performed the quality assessment while abstracting the data for the meta-analysis. The quality scores of the two abstractors were averaged.
Data synthesis and analysis Pooled odds ratios (ORs) and 95% confidence intervals (95% CIs) were calculated to estimate the association between PTB, LBW, and SGA among women who experienced IPV during pregnancy relative to unexposed women. Random-effects models were used to allow for the significant heterogeneity found between studies.21 The degree of heterogeneity present between the studies was expressed using the I2 statistic. Potential publication bias was evaluated using funnel plots. All statistical analyses were twosided and performed using REVMAN 5.3.22 A priori decisions were made to analyse unadjusted and adjusted data separately for each outcome, and to stratify the studies by study design, high and low–medium income settings, type of violence experienced, and average quality score. All included studies that adjusted for potential confounding variables, regardless of the specific variables that were adjusted for, were pooled to produce adjusted estimates. High-income countries were defined as those with a gross national income (GNI) of $12 736 or more per capita, as defined by The World Bank. Low–middle income countries were defined as those with a GNI <$12 736 per capita.23 Low quality scores were listed as those receiving five or fewer points on the Newcastle–Ottawa Scale.20 High-quality scores received six or more points.
Results Literature search Results from our search strategy are summarized in Figure 1. We identified 1575 publications. Of these, 361 were duplicates between the two databases, and an additional 1135 were excluded based on reviews of the titles and abstracts. Hand-searching of the reference lists yielded an additional 23 studies for inclusion. The remaining 102
ª 2016 Royal College of Obstetricians and Gynaecologists
articles were eligible for full-text review and abstraction. After reviewing the full-text articles, 52 additional articles were excluded, with the most common reason for exclusion being that the data presented were not sufficient for metaanalysis of our specific research question. Ultimately, 50 unique studies met the inclusion criteria: 30 studies assessed PTB, 41 assessed LBW, and seven assessed SGA.
Study characteristics The 50 included studies examined a total of 5 087 388 participants: 14 906 women who experienced IPV during pregnancy and 5 072 482 women who did not. This included 490 405 women with a PTB infant, 298 284 women with an LBW infant, 6916 women with an SGA infant, and 4 786 957 women with normally-grown infants at term. Characteristics of the included studies are shown in Table S1. Included studies arose from populations in 17 countries spanning six continents, with most studies originating from North America (46%), Asia (16%), and Africa (12%). Sample sizes ranged from 130 to 4 833 286 participants. In women who experienced IPV during pregnancy, rates of PTB, LBW, and SGA were 2.5–70.4%, 3.3–68.8%, and 6.3–27.2%, respectively. In women who did not experience IPV during pregnancy, rates of PTB, LBW, and SGA were 2.0–57.0%, 1.2–52.1%, and 2.7–22.3%, respectively.
Preterm birth Thirty studies reported on the association between IPV and PTB (Table 1). The unadjusted pooled analysis showed a nearly two-fold increase in the odds of delivering a preterm infant among victims of IPV during pregnancy, compared with women who were not exposed (OR 1.91, 95% CI 1.60–2.29; Figure 2). We detected a high level of heterogeneity among the unadjusted studies (I2 = 84%). Symmetry of the studies within the funnel plot for all unadjusted PTB estimates suggested that publication bias was not an issue (Figure S1). Thirteen of the 30 studies reported adjusted ORs for the association between IPV and PTB (Table 1). The adjusted pooled OR was statistically significant and differed little from the unadjusted estimate (OR 1.89, 95% CI 1.43–2.48; Figure S2). A large level of heterogeneity was also present in the adjusted analysis (I2 = 83%). Symmetry of the studies within the funnel plot for all adjusted PTB estimates suggested that publication bias was not an issue (Figure S3). Several unadjusted subanalyses were performed in an effort to identify the sources of heterogeneity between the studies (Table 1). The first subanalysis stratified the included studies by income status (i.e. countries with high income and countries with low–middle income) and study design (cohort, case–control, and cross-sectional). Among studies performed in high-income countries (n = 21), the
1291
Donovan et al.
Search yielded 1575 total cita ons from PubMed (n = 738) and Scopus (n = 837) 361 duplicates
1214 tles reviewed 1074 tles excluded because they did not fit inclusion criteria 140 abstracts reviewed 61 excluded: 1 interven on study 11 not IPV 23 ar cles included from hand searching references
19 not PTB/LBW/SGA 3 pre-pregnancy 20 reviews 102 full- text ar cles abstracted 7 not a study 52 excluded: 4 Not IPV 10 Not PTB/LBW/SGA 32 Data was not in correct form for meta-analysis
50 ar cles included in Final Analysis * 1. 2. 3.
Low birth weight (n = 41) Preterm birth (n = 30) Small for gesta onal age (n = 7)
3 Unable to translate 3 Duplicate study
*Ar cles may include more than one outcome
Figure 1. Flow diagram of search strategy.
association between IPV and PTB was highest among the eight cross-sectional studies (Figure S4). The effect estimate was slightly lower in the ten cohort studies, whereas no effect was found among the three case–control studies. Among studies performed in countries with low–middle income (n = 9), the pooled odds ratio for PTB was highest among the three cross-sectional studies; however, the effect estimate was still significantly increased among the four case–control studies (Figure S5). The pooled effect estimate for the two cohort studies was not statistically significant. Although the subanalyses among the studies performed in high-income countries remained highly heterogeneous, heterogeneity was much lower among studies performed in countries with low–middle income. Another subanalysis was performed by stratifying PTB studies by violence type: physical, emotional, and sexual violence (Figure 3). Studies that combined at least two of the violence types were pooled together (‘combined violence’). The pooled ORs were highest among women who experienced at least two types of IPV during pregnancy (OR 2.33, 95% CI 1.88–2.88; Figures S6 and
1292
S7). Heterogeneity was moderate in this subgroup (I2 = 54%). For the final subset analysis, the studies were split into groups according to study quality (Table 1). High- and low-quality studies (n = 19 and 11, respectively) had the same pooled effect estimate, showing nearly a two-fold increase in the odds of delivering a preterm infant for women exposed to IPV during pregnancy (Figure S8). Studies with a lower quality score had a moderate level of heterogeneity, whereas those with a higher quality score were highly heterogeneous (I2 = 64 and 84%).
Low birthweight The association between IPV during pregnancy and delivery of an LBW infant was estimated using data from 41 studies (Table 1). The unadjusted pooled OR show more than a two-fold increase in odds of LBW among women who experienced IPV during pregnancy compared with those who did not (OR 2.11, 95% CI 1.68–2.65; Figure S9). Fourteen studies provided adjusted ORs; after pooling the data, the association was somewhat attenuated compared
ª 2016 Royal College of Obstetricians and Gynaecologists
ª 2016 Royal College of Obstetricians and Gynaecologists
OR (95% CI)
5.9 6.7
6.7 6.4 6.3 4.3 6.0 6.0 5.3 4.4 6.8
83 59 90 52 71 3 0 64 88
Average quality score
84 87
I2 (%)
15 26
6 6 2
17 7 3
14
41 14
No. of studies
NA, not enough studies to subanalyse (n = 10). *OR is significant. **Quality score is based on Newcastle–Ottawa scale and has a maximum of 9 points.
Overall unadjusted estimates 30 1.91 (1.60–2.29)* Overall unadjusted estimates 13 2.26 (1.64–3.12)* (for studies with adjusted estimates) Overall adjusted estimates 13 1.89 (1.43–2.48)* Unadjusted estimates for high-income countries Cohort studies 10 1.63 (1.31–2.03)* Cross-sectional studies 8 1.91 (1.35–2.71)* Case–control studies 3 0.95 (0.47–1.92) Unadjusted estimates for low/middle-income countries Cohort studies 2 3.11 (0.68–14.25) Cross-sectional studies 3 3.81 (2.96–5.08)* Case–control studies 4 2.20 (1.75–2.78)* Unadjusted estimates for average quality score** Low quality (≤5 points) 11 1.91 (1.39–2.62)* High quality (>5 points) 19 1.91 (1.53–2.39)*
No. of studies
Preterm birth
1.80 (1.24–2.62)* 2.29 (1.71–3.06)*
3.87 (0.98–15.29) 2.70 (1.85–3.94)* 2.98 (0.90–9.89)
1.82 (1.33–2.48)* 1.66 (1.17–2.36)* 1.06 (0.57–1.99)
1.92 (1.34–2.73)*
2.11 (1.68–2.65)* 2.84 (1.70–4.73)*
OR (95% CI)
79 94
96 52 74
85 89 38
90
91 95
I2 (%)
Low birthweight
4.7 6.9
6.3 6.2 5.0
6.3 5.9 4.3
6.9
6.1 6.9
Average quality score
3 4
2 NA NA
3 2 NA
2
7 2
No. of studies
Table 1. Summary of pooled effect estimates from meta-analyses of the association between IPV and adverse birth outcomes, 1950–May 2015
1.29 (0.92–1.81) 1.66 (0.82–3.33)
1.32 (0.87–2.02) NA NA
1.10 (0.85–1.43) 2.22 (0.68–7.25) NA
1.58 (0.61–4.10)
1.37 (1.02–1.84)* 2.03 (0.54–7.67)
OR (95% CI)
0 66
0 NA NA
0 82 NA
78
32 89
I2 (%)
5.0 8.3
6.5 NA NA
7.3 6.5 NA
8.5
6.9 8.5
Average quality score
Small for gestational age
IPV During Pregnancy and Poor Infant Outcomes
1293
Donovan et al.
Figure 2. Forest plot of the unadjusted effect estimates for all PTB studies.
7.29
Odds RaƟos With 95% Confidence Intervals
n= 2
Physical Violence Only EmoƟonal Violence Only Sexual Violence Only Combined Violence Types n = 26
n= 4
n = 17
2.7
n = 17
n = 15
n= 3
n= 3
n= 2 n= 2
1
n= 0
0.3703704
Preterm Birth
n=0
Low Birth Weight
Small for GestaƟonal Age
Figure 3. Graphical representation of violence type effect estimates.
with the unadjusted estimate, although still statistically significant (OR 1.92, 95% CI 1.34–2.73; Figure S10). Both pooled analyses had a considerable level of heterogeneity,
1294
necessitating further subanalyses. Funnel-plot symmetry of the unadjusted and adjusted effect estimates suggested that publication bias was not an issue (Figures S11 and S12).
ª 2016 Royal College of Obstetricians and Gynaecologists
IPV During Pregnancy and Poor Infant Outcomes
Subanalyses that stratified by study design and economic status were also performed for the LBW outcome (Table 1). For high-income countries (n = 27), the association between IPV and delivery of an LBW infant was higher in cohort studies (n = 17) than the associations estimated from cross-sectional (n = 7) and case–control studies (n = 3) (Figure S13). Cohort and cross-sectional studies had a large level of heterogeneity, whereas case–control studies had a moderate level of heterogeneity. Among countries with low–middle incomes (n = 14), the odds of delivering an LBW infant associated with IPV was significantly elevated in cross-sectional studies (n = 6), but not in cohort (n = 6) and case–control (n = 2) studies (Figure S14). Although all three study designs showed a moderate to high degree of heterogeneity, the I2 values for the cross-sectional and case–control studies were lower than the I2 values for the overall analysis, suggesting that study design and/or income level may explain a portion of the heterogeneity among the pooled studies. When stratified by violence type, women who experienced at least two types of IPV during pregnancy had a 2.5–fold increased odds of LBW compared with women who were not abused during pregnancy (OR 2.46, 95% CI 1.73–3.51; Figures 3 and S15). Heterogeneity was high in this subgroup (I2 = 89%). Women who experienced either physical or emotional abuse during pregnancy also had increased odds for delivering an LBW infant, compared with unexposed women. Although the association of sexual violence with LBW was greatly increased, the results were not statistically significant (Figure S16). The association between IPV and LBW was smaller in studies of lower quality (n = 15) compared with the association among studies of higher quality (n = 26) (Figure S17; Table 1). A substantial level of heterogeneity was present in both subsets (79 and 94%).
associated with IPV during pregnancy (Figure S20). The three cohort studies also showed a slight non-significant increase in the odds for delivering an SGA infant. Two studies examining the association between IPV and SGA were performed in countries with low–middle incomes, both of which were cohort studies (Figure S21). The pooled effect estimate showed slightly increased odds of SGA associated with IPV during pregnancy that was not statistically significant. Heterogeneity remained high in the cross-sectional studies, and was not present within either set of cohort studies, suggesting that study design and/or income level is a major source of heterogeneity among the SGA studies. When analyses of each birth outcome were stratified by violence type, the pooled effect estimates for delivering an SGA infant were slightly increased among women who experienced more than one type of IPV (OR 1.33, 95% CI 0.88– 1.99), and among women who experienced emotional IPV exclusively (OR 1.16, 95% CI 0.94–1.44) (Figures 3, S22, and S23). Women who experienced physical IPV had increased odds of having an SGA infant compared with women who did not experience any IPV during pregnancy (OR 1.79, 95% CI 0.92–3.49); however, the results for this outcome were not statistically significant. Heterogeneity was moderate among the physical violence studies, and was not present in studies of combined IPV and emotional IPV only. Studies that were of lower quality (n = 3) had lower odds of SGA than studies of higher quality (n = 4) (Figure S24; Table 1). The pooled effect estimates for both subsets were not statistically significant. Heterogeneity was not present in studies with a low average quality score, whereas studies with a high average quality score were moderately heterogeneous (I2 = 0 and 66%, respectively).
Discussion
Small for gestational age
Main findings
Seven studies assessed the association between IPV experienced during pregnancy and delivering an SGA infant (Table 1). The unadjusted pooled OR for SGA outcomes was significantly increased among women who experienced IPV during pregnancy, compared with women who did not (OR 1.37, 95% CI 1.02–1.84; Figure S18). Although not statistically significant, the adjusted pooled effect estimate based on two studies was slightly higher (OR 1.58, 95% CI 0.61–4.10) than the comparable unadjusted pooled estimate (Figure S19). Heterogeneity was moderate in the unadjusted analysis and high in the adjusted analysis. Similar subanalyses were performed for the SGA outcome as those previously described for PTB and LBW. Five of the seven studies that assessed SGA were conducted in highincome countries (Table 1). The two cross-sectional studies showed a non-significant, two-fold increase in odds of SGA
Our meta-analysis of 50 studies indicates that women who experience IPV during pregnancy are at increased risk of having an adverse birth outcome compared with women who don’t experience IPV during pregnancy. The pooled odds ratios for PTB and LBW were significantly increased, even after adjusting for potential confounding factors. Although a limited number of studies examined the association of IPV with SGA, ours is the first meta-analysis to report an adjusted pooled effect estimate for this outcome. Both the pooled effect estimate for SGA were increased, and although the adjusted analysis was not statistically significant, it included only two studies. We detected high levels of heterogeneity, particularly among studies examining LBW and PTB. A sensitivity analysis was subsequently run for each outcome to determine if the study sample sizes contributed to the large level of heterogeneity. Hetero-
ª 2016 Royal College of Obstetricians and Gynaecologists
1295
Donovan et al.
geneity estimates remained high even after studies with n < 500 participants were removed from the analyses, necessitating further subanalyses. To identify possible sources of heterogeneity and describe differences in effect estimates, we stratified our results by national income level of the study populations, study design, IPV type, and study quality. For all three outcomes, the stratified results suggested that national income level and study design accounted for some of the heterogeneity. This may be explained by the frequency and severity of violence, the availability of prenatal care, and the range of socio-economic status for populations living in countries of high and middle–low incomes. For many, but not all, of our outcomes, the cross-sectional studies saw a stronger association between IPV and the outcome than cohort or case–control studies. This may be because of confounding (cohort studies and case–control studies are stronger study designs) or recall bias. Violence type and study quality were also identified as possible sources of heterogeneity. The pooled effect estimates were greatest among women who experienced a combination of violence types. Multiple types of IPV exposure appear to yield a larger influence on the effect estimate than just one type alone. Few studies examined the sole effects of sexual violence on adverse birth outcomes, making it difficult to assess its impact in particular. The studies included in this meta-analysis were generally of higher quality, contributing to a small risk of bias. Lower quality studies didn’t include adjusted analyses and had poor ascertainment of the outcome, which may have biased the pooled effect estimates towards the null.
Strengths and limitations A great strength of our meta-analysis was the use of a thorough and sensitive search strategy aimed at identifying a large number of studies at the beginning of the search. By hand-searching the references of all identified studies, we were able to locate an additional set of studies, and two studies were included after obtaining the information from the authors. Furthermore, we were able to translate all but three studies that were written using a non-Latin alphabet. As a result of the large number of included studies examining PTB and LBW, it is unlikely that these three studies would have altered our conclusions. In addition, we provide the first pooled effect estimates for the association of isolated violence types with adverse birth outcomes. Our findings are limited by the high degree of heterogeneity in the pooled analyses; however, our subanalyses identify national income level and type of IPV as likely sources of heterogeneity in addition to study design and quality, which were described as sources of heterogeneity in a previous meta-analysis.9 Unfortunately, because of limited data in the
1296
original studies, we were unable to account for other possible sources of heterogeneity such as race/ethnicity, maternal smoking, and pre-pregnancy body mass index.
Interpretation Although our findings are consistent with those of the previous meta-analyses, we found substantially stronger associations between IPV and PTB or LBW, largely because of the inclusion of several recently published, high-quality studies of over 500 subjects that reported large increases in the odds of PTB, LBW, and SGA outcomes.10–15 Murphy et al.8 found that women who experienced physical, sexual, or emotional abuse during pregnancy were more likely than non-abused women to deliver an LBW infant (OR 1.4, 95% CI 1.1–1.8). Shah and Shah examined LBW, PTB, and SGA outcomes, and found the odds of delivering an infant with any of these adverse birth outcomes was 50–80% higher in women who experienced IPV during pregnancy or immediately before conception (LBW – OR 1.61, 95% CI 1.28–2.02; PTB – OR 1.81, 95% CI 1.31–2.26; SGA – OR 1.57, 95% CI 0.90–2.77), compared with those who did not.9 Compared with the most recent meta-analysis, ours includes an additional 19 LBW, 16 PTB, and three SGA studies in its unadjusted pooled effect estimates, and an additional two LBW, seven PTB, and two SGA studies in the adjusted pooled estimates.9 Our meta-analysis appears to be the first to analyse the effects of specific violence types on adverse birth outcomes. Exposure to IPV during pregnancy can facilitate pathways to adverse birth outcomes, both directly and indirectly.6,7 Direct physical assault to the abdomen or sexual abuse has been associated with pregnancy complications such as placental damage, uterine contractions, premature rupture of membranes, and genitourinary infections.6,24 Abuse may also lead to an increase in behavioural risk factors associated with adverse birth outcomes, such as maternal smoking, alcohol or drug use, inadequate prenatal care, or insufficient prenatal weight gain.6,15,25–35 Many studies have shown that women who are abused tend to have higher levels of stress, less support from their partners, and lower self-esteem.6,15,24,32,36 These factors may indirectly facilitate biological mechanisms that contribute to adverse birth outcomes.6,15,26
Conclusion The relationship between IPV during pregnancy and the risk for adverse birth outcomes has been assessed among women from a wide distribution of different geographical areas and socio-economic backgrounds. Whereas the association of physical IPV with adverse birth outcomes, particularly PTB and LBW, has been examined in numerous
ª 2016 Royal College of Obstetricians and Gynaecologists
IPV During Pregnancy and Poor Infant Outcomes
studies, few studies have focused on the exclusive effects of emotional or sexual violence. Additional studies examining the biological mechanism are needed to more clearly understand the association between IPV and SGA. Although IPV prevalence rates vary across the globe, the detrimental effects of IPV on health are conclusive.37,38 To address this problem, numerous health professional associations now have clinical guidelines on the detection and subsequent care for abused women.37,39 In addition, many nations have mandated abuse screening protocols that all healthcare professionals must follow.37,40 Interventions effective in reducing IPV incorporated home visitation programmes and multifaceted counselling interventions; however, the evidence for the effect of these interventions on secondary outcomes, such as postnatal depression, quality of life, and delivery of an LBW or premature infant, was less conclusive.41 A shift in the focus of IPV research towards the effectiveness of interventions to prevent violence during pregnancy is strongly needed. Providing an effective therapeutic intervention for women who disclose abuse could aid in the prevention of adverse birth outcomes.37 In addition, the evaluation of advocacy and mentor support programmes is necessary to better support women post-disclosure.37,42,43 Primary prevention in the form of prenatal education to inform couples about the risk of adverse birth outcomes in relation to IPV exposure could aid in the recognition and determent of these behaviours.
Disclosure of interests None declared. Completed disclosure of interests form available to view online as supporting information.
Contribution to authorship AFS was responsible for the concept of the study. AFS, BMD, KKR, MLS, and CNS are responsible for the design of the study. BMD and CNS performed the search, selected abstracts, obtained the full manuscripts, and performed the initial data extraction. AFS was responsible for reconciling differences in the data extraction. BMD and AFS attempted to contact authors of the original studies in order to obtain the data needed to include the study in this meta-analysis. BMD performed the meta-analysis. AFS, BMD, CNS, MLS, and KKR contributed to the interpretation of the data, and the drafting and approval of the final article.
Details of ethics approval Not applicable.
Funding No funding was sought for this study.
ÂŞ 2016 Royal College of Obstetricians and Gynaecologists
Acknowledgements We would like to thank Caitlin J. Smith for her helpful comments and contribution to manuscript preparation and data analysis. We thank Mr Chris Childs for his contribution to the development of the search strategy. Special thanks go to Dr Jeanne Alhusen, Dr Alissa Huth-Bocks, and Dr Nastassja Koen for contributing additional data for these analyses.
Supporting Information Additional Supporting Information may be found in the online version of this article: Figure S1. Funnel plot of the unadjusted effect estimates for all PTB studies. Figure S2. Forest plot of the adjusted effect estimates for all PTB studies. Figure S3. Funnel plot of the adjusted effect estimates for all PTB studies. Figure S4. Forest plot of the unadjusted effect estimates for PTB studies performed in high income countries stratified by study design. Figure S5. Forest plot of the unadjusted effect estimates for PTB studies performed in low/middle income countries stratified by study design. Figure S6. Forest plot of the unadjusted effect estimates for PTB studies that analyzed combined violence exposure. Figure S7. Forest plot of the unadjusted effect estimates for PTB studies that analyzed independent violence types. Figure S8. Forest plot of the unadjusted effect estimates for PTB studies stratified by average quality score. Figure S9. Forest plot of the unadjusted effect estimates for all LBW studies. Figure S10. Forest plot of the adjusted effect estimates for all LBW studies. Figure S11. Funnel plot of the unadjusted effect estimates for all LBW studies. Figure S12. Funnel plot of the adjusted effect estimates for all LBW studies. Figure S13. Forest plot of the unadjusted effect estimates for LBW studies performed in high income countries stratified by study design. Figure S14. Forest plot of the unadjusted effect estimates for LBW studies performed in low/middle income countries stratified by study design. Figure S15. Forest plot of the unadjusted effect estimates for LBW studies that analyzed combined violence exposure. Figure S16. Forest plot of the unadjusted effect estimates for LBW studies that analyzed independent violence types. Figure S17. Forest plot of the unadjusted effect estimates for LBW studies stratified by average quality score.
1297
Donovan et al.
Figure S18. Forest plot of the unadjusted effect estimates for all SGA studies. Figure S19. Forest plot of the adjusted effect estimates for all SGA studies. Figure S20. Forest plot of the unadjusted effect estimates for SGA studies performed in high income countries stratified by study design. Figure S21. Forest plot of the unadjusted effect estimates for SGA studies performed in low/middle income countries stratified by study design. Figure S22. Forest plot of the unadjusted effect estimates for SGA studies that analyzed combined violence exposure. Figure S23. Forest plot of the unadjusted effect estimates for SGA studies that analyzed independent violence types. Figure S24. Forest plot of the unadjusted effect estimates for SGA studies stratified by average quality score. Table S1. Characteristics of Studies Included in the Meta-Analysis of IPV Exposure during Pregnancy and Risk of PTB, LBW, and SGA, 1950-May 2015 Data S1. Powerpoint slides summarising the study. &
References 1 Violence Against Women. 2014; Available from: http:// www.who.int/mediacentre/factsheets/fs239/en/. 2 Intimate Partner Violence: Definitions. November 25, 2014 [cited 2015 April 7]; Available from: http://www.cdc.gov/violencepreven tion/intimatepartnerviolence/definitions.html. 3 Bailey BA. Partner violence during pregnancy: prevalence, effects, screening, and management. Int J Womens Health 2010; 2:183–97. 4 Campbell JC, Lewandowski LA. Mental and physical health effects of intimate partner violence on women and children. Psychiatr Clin North Am 1997;20:353–74. 5 Tjaden P, Thoennes n. Prevalence, Incidence, and Consequences of Violence Against Women: Findings from the National Violence Against Women Survey. Washington, DC: National Institute of Justice, Centers for Disease Control and Prevention; 1998. 6 Heaman MI. Relationships between physical abuse during pregnancy and risk factors for preterm birth among women in Manitoba. J Obstet Gynecol Neonatal Nurs 2005;34:721–31. 7 Coker AL, Sanderson M, Dong B. Partner violence during pregnancy and risk of adverse pregnancy outcomes. Paediatr Perinat Epidemiol 2004;18:260–9. 8 Murphy CC, Schei B, Myhr TL, Du Mont J. Abuse: a risk factor for low birth weight? A systematic review and meta-analysis. CMAJ 2001;164:1567–72. 9 Shah PS, Shah J. Maternal exposure to domestic violence and pregnancy and birth outcomes: a systematic review and metaanalyses. J Womens Health (Larchmt) 2010;19:2017–31. 10 Abdollahi F, Abhari FR, Delavar MA, Charati JY. Physical violence against pregnant women by an intimate partner, and adverse pregnancy outcomes in Mazandaran Province, Iran. J Family Community Med 2015;22:13–8. 11 Alijahan R, Hazrati S, Mirzarahimi M, Pourfarzi F, Ahmadi Hadi P. Prevalence and risk factors associated with preterm birth in Ardabil, Iran. Iran J Reprod Med 2014;12:47–56.
1298
12 Henriksen L, Schei B, Vangen S, Lukasse M. Sexual violence and neonatal outcomes: a Norwegian population-based cohort study. BMJ Open 2014;4:e005935. 13 Meuleners LB, Lee AH, Janssen PA, Fraser ML. Maternal and foetal outcomes among pregnant women hospitalised due to interpersonal violence: a population based study in Western Australia, 2002-2008. BMC Pregnancy Childbirth 2011;11:70. 14 Nunes MA, Camey S, Ferri CP, Manzolli P, Manenti CN, Schmidt MI. Violence during pregnancy and newborn outcomes: a cohort study in a disadvantaged population in Brazil. Eur J Public Health 2011;21: 92–7. 15 Sanchez SE, Alva AV, Diez Chang G, Qiu C, Yanez D, Gelaye B, et al. Risk of spontaneous preterm birth in relation to maternal exposure to intimate partner violence during pregnancy in Peru. Matern Child Health J, 2013;17:485–92. 16 Scribano PV, Stevens J, Kaizar E. The effects of intimate partner violence before, during, and after pregnancy in nurse visited first time mothers. Matern Child Health J 2013;17:307–18. 17 Watson LF, Taft AJ. Intimate partner violence and the association with very preterm birth. Birth 2013;40:17–23. 18 Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson GD, Rennie D, et al. Meta-analysis of observational studies in epidemiology: a proposal for reporting. Meta-analysis of observational studies in epidemiology (MOOSE) group. JAMA 2000;283:2008–12. 19 Birthweight and Gestation. January 22, 2015 [cited 2015 April 8]; Available from: http://www.cdc.gov/nchs/fastats/birthweight.htm. 20 Wells G, Shea B, O’Connell D, Peterson J, Welch V, Losos M, et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomized studies in meta-analyses. [cited 2015 February 13]; Available from: http://www.ohri.ca/programs/clinical_epidemiology/ oxford.htm. 21 DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7:177–87. 22 Review Manager (RevMan) [Computer Program]. The Cochrane Collaboration: Version 5.3. The Nordic Cochrane Centre: Copenhagen; 2014. 23 Group TWB Country and Lending Groups. 2015 June 12, 2015]; Available from: http://data.worldbank.org/about/country-and-lendinggroups. 24 Curry MA Stress Related to Domestic Violence during Pregnancy and Infant Birth Weight, in Empowering Survivors of Abuse: Health Care for Battered Women and Their Children. Thousand Oaks, CA: Sage Publications; 1998: 98–108. 25 Cokkinides VE, Coker AL, Sanderson M, Addy C, Bethea L. Physical violence during pregnancy: maternal complications and birth outcomes. Obstet Gynecol 1999;93:661–6. 26 Curry MA, Doyle BA, Gilhooley J. Abuse among pregnant adolescents: differences by developmental age. MCN Am J Matern Child Nurs 1998;23:144–50. 27 Goodwin MM, Gazmararian JA, Johnson CH, Gilbert BC, Saltzman LE. Pregnancy intendedness and physical abuse around the time of pregnancy: findings from the pregnancy risk assessment monitoring system, 1996-1997. PRAMS Working Group. Pregnancy Risk Assessment Monitoring System. Matern Child Health J 2000;4:85–92. 28 Huth-Bocks AC, Levendosky AA, Bogat GA. The effects of domestic violence during pregnancy on maternal and infant health. Violence Vict 2002;17:169–85. 29 McFarlane J, Parker B, Soeken K. Physical abuse, smoking, and substance use during pregnancy: prevalence, interrelationships, and effects on birth weight. J Obstet Gynecol Neonatal Nurs 1996;25: 313–20.
ª 2016 Royal College of Obstetricians and Gynaecologists
IPV During Pregnancy and Poor Infant Outcomes
30 Parker B, McFarlane J, Soeken K. Abuse during pregnancy: effects on maternal complications and birth weight in adult and teenage women. Obstet Gynecol 1994;84:323–8. 31 Stewart DE, Cecutti A. Physical abuse in pregnancy. CMAJ 1993;149:1257–63. 32 Wiemann CM, Agurcia CA, Berenson AB, Volk RJ, Rickert VI. Pregnant adolescents: experiences and behaviors associated with physical assault by an intimate partner. Matern Child Health J 2000;4:93–101. 33 Campbell J, Torres S, Ryan J, King C, Campbell DW, Stallings RY, et al. Physical and nonphysical partner abuse and other risk factors for low birth weight among full term and preterm babies: a multiethnic case-control study. Am J Epidemiol 1999;150:714–26. 34 McFarlane J, Parker B, Soeken K. Abuse during pregnancy: associations with maternal health and infant birth weight. Nurs Res 1996;45:37–42. 35 Bohn DK. Lifetime and current abuse, pregnancy risks, and outcomes among Native American women. J Health Care Poor Underserved 2002;13:184–98. 36 Muhajarine N, D’Arcy C. Physical abuse during pregnancy: prevalence and risk factors. CMAJ 1999;160:1007–11. 37 O’Doherty LJ, Taft A, Hegarty K, Ramsay J, Davidson LL, Feder G. Screening women for intimate partner violence in healthcare settings: abridged Cochrane systematic review and meta-analysis. BMJ 2014;348:g2913.
ª 2016 Royal College of Obstetricians and Gynaecologists
38 Global and regional estimates of violence against women: Prevalence and health effects of intimate partner violence and nonpartner sexual violence. WHO, London School of Hygiene and Tropical Medicine, South African Medical Research Council; 2013. 39 Responding to Intimate Partner Violence and Sexual Violence Against Women: WHO Clinical and Policy Guidelines. Geneva: World Health Organization; 2013. 40 Moyer VA. Screening for intimate partner violence and abuse of elderly and vulnerable adults: U.S. preventive services task force recommendation statement. Ann Intern Med 2013;158:478– 86. 41 Van Parys AS, Verhamme A, Temmerman M, Verstraelen H. Intimate partner violence and pregnancy: a systematic review of interventions. PLoS ONE 2014;9:e85084. 42 Ramsay J, Carter Y, Davidson L, Dunne D, Eldridge S, Feder G, et al. Advocacy interventions to reduce or eliminate violence and promote the physical and psychosocial well-being of women who experience intimate partner abuse. Cochrane Database Syst Rev, 2009; Cd005043. 43 Taft AJ, Small R, Hegarty KL, Watson LF, Gold L, Lumley JA. Mothers’ AdvocateS In the Community (MOSAIC)–non-professional mentor support to reduce intimate partner violence and depression in mothers: a cluster randomised trial in primary care. BMC Public Health 2011;11:178.
1299