The association between prepregnancy obesity and screening positive for postpartum depression

Authors


Dr DY LaCoursiere, Department of Reproductive Medicine, Division of General Obstetrics and Gynecology, University of California, San Diego, 200 West Arbor, #8433, San Diego CA 92103, USA. Email ylacoursiere@ucsd.edu

Abstract

Please cite this paper as: LaCoursiere D, Barrett-Connor E, O’Hara M, Hutton A, Varner M. The association between prepregnancy obesity and screening positive for postpartum depression. BJOG 2010;117:1011–1018.

Objective  To describe the association between reported prepregnancy body mass index (BMI) and screening positive for depression.

Design  Cohort study.

Setting  Four urban hospitals in Utah, USA.

Population  Women delivering a term, singleton, live-born infant at one of four urban hospitals in Utah in the period 2005–2007.

Methods  Women were enrolled immediately postpartum. Demographic, anthropometric, stressors, psychiatric, and medical/obstetric and family-history data were obtained. Prepregnancy height, weight, and pregnancy weight gain were self-reported. The primary exposure variable, prepregnancy BMI, was calculated. Women were stratified into the six World Health Organization BMI categories (underweight, normal weight, pre-obese, or obese class 1–3).

Main outcome measure  At 6–8 weeks postpartum, women were screened for depression using the Edinburgh Postnatal Depression Scale (EPDS). The primary outcome measure was a prespecified EPDS score of ≥12.

Results  Among the 1053 women studied, 14.4% of normal weight women screened positive for postpartum depression. This proportion was greater in women classed as underweight (18.0%, n = 11), pre-obese (18.5%, n = 38), obese class 1 (18.8%, n = 16), obese class 2 (32.4%, n = 11), and obese class 3 (40.0%, n = 8) (P < 0.01). Controlling for demographic, psychological, and medical/obstetric factors, prepregnancy class-2 (aOR 2.87, 95% CI 1.21–6.81) and class-3 (aOR 3.94, 95% CI 1.38–11.23) obesity remained strongly associated with screening positive for postpartum depression, compared with women of normal weight.

Conclusions  Self-reported prepregnancy obesity may be associated with screening positive for depression when measured postpartum.

Introduction

Although the data have been inconsistent, some studies suggest that obesity is a risk factor for depression in women.1,2 Women with a history of depression are at an increased risk of developing postpartum depression (PPD).3 However, the possible association of pregravid obesity with subsequent PPD has not been sufficiently studied. In the USA, the significant increase in the prevalence of women being overweight and obese prepregnancy makes this a particularly timely question.4–6 Although the prevalence of prepregnancy overweight and obesity presses this issue, ultimately it is the untoward effects of undiagnosed and untreated depression on the mother (severity and recurrence of depression) and her child (behavioral issues and cognitive delays) that make identification of at-risk women important.7,8

We previously reported results from a cross-sectional survey showing that women with extremes of prepregnancy body mass indices (BMIs) were at increased risk of self-reported postpartum depressive symptoms.9 However, a crude measure of depressive symptoms was used in that preliminary analysis. One of the only other published studies supporting the association of obesity with PPD was limited by a small sample size and by the use of a dichotomous BMI exposure.10 A secondary analysis of overweight and obese women who enrolled postpartum in a randomised controlled behavioral intervention trial did not find an association of BMI group and screening positive for depression. However, these women were all self-selected participants in a 9-month study that included 18 classes and six counselling sessions, and thus may be less likely to include depressed women.11 In a recently reported study of women during pregnancy, depression was one of six negative psychological states found to be associated with self-reported pregravid BMI. This study found trends among all raised weight categories [overweight (BMI > 26.0–29.0), obese (BMI > 29.0–34.9), and severely obese (BMI > 35.0)], and strong associations between BMI > 35 and antepartum perceived stress, trait anxiety, depressive symptoms, powerful others, locus of control, restrained eating, and dieting and weight cycling.12

In the present study, women were recruited immediately postpartum and were followed for 6–8 weeks to test the hypothesis that prepregnancy BMI is associated with screening positive for depression measured postpartum, after controlling for other demographic, psychological, and medical and obstetrical risk factors.

Methods

Study population

The study population was recruited from English-speaking women delivering using the obstetric services of four large urban hospitals in Utah in the period 2005–2007. Women not on medication for depression were invited to participate while hospitalised within 24–48 hours after the delivery of a singleton, term (≥37 weeks of gestation) live-born infant.

Study design

After obtaining Institutional Review Board approval at each participating institution, study coordinators obtained written informed consent, and used an interview and self-administered questionnaire to ask about demographic and anthropometric data, pregnancy stressors, and psychiatric, medical, obstetric, and family history. Additional information was obtained by review of the in-patient medical record.

Demographic data included maternal age, education, marital status, and insurance status. Insurance status was defined as insured if women reported private insurance, underinsured if women reported publically funded prenatal or medical insurance, and uninsured if women did not have health insurance during the antepartum period. Self-reported maternal race/ethnicity was collected using Utah state birth record categories. Pregnancy history and current pregnancy outcome, provider type, and whether the provider discussed weight or mood were recorded. Abuse, social, personal and family psychiatric history data were collected. Pregnancy stressors were identified via self-administration of the 13 ‘stressors’ questions from the Centers for Disease Control and Prevention Pregnancy Risk Assessment Monitoring System (PRAMS). For the present analyses, the responses to questions were combined and used to determine the presence or absence of emotional, partner-associated, financial, and traumatic stressors, as defined by the PRAMS protocol.13 Stressors were included in the analysis as it is unclear where stressful events lie along the causal pathway between obesity and depressed mood.

Patients reported their height, prepregnancy weight, and pregnancy weight gain immediately postpartum; their body mass index (BMI) before pregnancy was also calculated. This approach was used because most women are not weighed while hospitalised for delivery. Thus, the primary exposure variable was self-reported prepregnancy BMI, and weight gain was used as a covariate. Prepregnancy BMI was stratified by the World Health Organization’s classification system: underweight, <18.5; normal weight, 18.5–24.9; pre-obese, 25.0–29.9; obese class 1, 30.0–34.9; obese class 2, 35.0–39.9; and obese class 3 (>40.0).14

At 6–8 weeks after delivery, patients received a self-administered questionnaire by mail that contained the Edinburgh Postnatal Depression Scale (EPDS),15,16 and additional questions regarding their postpartum weight and the practices of their pregnancy care provider. The criterion used to define screening positive for depression at 6–8 weeks postpartum was a prespecified score of ≥12 on the EPDS. The EPDS is well validated: a score of ≥12 has been shown to have a sensitivity, specificity, and positive predictive value of PPD of 83, 78, and 73%, respectively.16 Women who screened positive on the EPDS were contacted and referred to support services. Information regarding whether they were already being evaluated and treated for PPD was collected. The pregnancy care provider of all women screening positive for depression was notified via mail that their patient had screened positive as part of the study. Women screening positive who were unable to be contacted via telephone were sent a letter informing them of their depression screening results.

Statistical analyses

Univariate analyses of categorical data were performed using a chi-square test (when cell counts were ≥5) or Fisher’s exact test (when cells counts were <5), and analyses of continuous data were performed using nonparametric tests. Statistical hypotheses were tested using two-tailed 95% confidence intervals for the primary hypothesis. Additional hypotheses were tested using P < 0.05, with one and five degrees of freedom. Regression assumptions were met and data were analysed in a cohort design using multiple logistic regression modelling with demographic data, stressors, abuse history, depression history, and prepregnancy BMI as explanatory variables with EPDS ≥ 12 as the dichotomous outcome variable for logistic modelling. Given the small sample of Black, Asian/Pacific Islanders, and other ethnicity categories, these women were excluded from the multivariate model. Maternal age, gravidity, and education were included as continuous variables. Ordinal scaled data were coded using dummy variables. Forward elimination was used to develop a parsimonious model. The analyses were performed with the Statistical Package for the Social Sciences, spss v16.

Results

Of the 1543 women recruited after delivery, 1053 women completed both the immediate postpartum and the follow-up (6–8 weeks postpartum) portions of the study (68.2%). Women with incomplete height, prepregnancy weight, and EPDS data were excluded (n = 26). Compared with women who enrolled but did not complete the postpartum questionnaire, women completing both portions of the study were more likely to be older (27.6 ± 5.3 years versus 26.1 ± 5.3 years; P < 0.01), more educated (14.7 ± 2.7 years versus 13.3 ± 2.4 years; P < 0.01), married (91.1% versus 74.3%; P < 0.01), insured (69.0% versus 50.3%; P < 0.001), and non-Hispanic white (86.9% versus 73.4%; P < 0.01). Also, those completing both portions were less likely to report a history of abuse (11.5% versus 16.9%; P < 0.01) and had a lower prepregnancy BMI (24.4 ± 5.5 versus 25.5 ± 6.5; P < 0.01). Women in higher BMI categories were less likely to complete the follow-up questionnaire (normal prepregnancy BMI strata 70.2%, underweight 66.3%, pre-obese 67.9%, class 1 60.7%, class 2 57.6%, and class 3 47.6%). There was no statistically significant difference between women completing both portions of the study and those completing only the baseline portion in terms of mean gravidity (2.5 ± 1.6 versus 2.6 ± 1.7; P = 0.37) or history of depression (16.9% versus 19.4%; P = 0.24).

Overall, obese women were older, less educated, and more frequently parous and non-Hispanic whites, compared with women of normal weight (Table 1). They also gained and retained less weight, and reported more partner-associated stressors than their normal weight counterparts, as shown in Table 2. There was a trend for women with an increased prepregnancy BMI to more frequently report a history of depression and PPD (Table 2). Seventeen percent of all women reported a history of depression. The least frequent report of depression history was seen in underweight women (n = 6, 9.8%), compared with 15.5% of normal weight women (n = 97), 18.7% of pre-obese women (n = 38), 22.4% of class-1 obese women (n = 19), 29.4% of class-2 obese women (n = 10), and 25.0% of class-3 obese women (n = 5).

Table 1.   Baseline demographic data of women stratified by baseline prepregnancy BMI category
 Total
n = 1027
Underweight
n = 61
Normal
n = 624
Pre-obese
n = 203
Class-1
obese
n = 85
Class-2
obese
n = 34
Class-3
obese
n = 20
P*
  1. *Chi-square (when cell counts were ≥5)/Fisher’s exact (when cells counts were <5) and Kruskal–Wallis tests were used for significance testing. Sum of categories may equal less than listed total because of invalid or incomplete data.

 mean ± standard deviation
Age, years27.7 ± 4.926.5 ± 4.927.4 ± 4.728.4 ± 5.127.9 ± 5.030.0 ± 4.927.2 ± 5.50.012
Gravidity2.5 ± 1.62.3 ± 1.12.4 ± 1.62.7 ± 1.62.6 ± 1.82.9 ± 1.62.7 ± 1.80.043
Education, years14.7 ± 2.715.1 ± 2.914.9 ± 2.814.7 ± 2.614.1 ± 2.514.3 ± 2.013.7 ± 2.40.010
 n (percentage)
Race/ethnicity
White (non-Hispanic)903 (87.9)50 (82.0)555 (88.9)180 (88.7)75 (88.2)28 (82.4)15 (75.0) 
White (Hispanic)68 (6.6)6 (9.8)38 (6.1)10 (4.9)6 (7.1)4 (11.8)4 (20.0) 
Black5 (0.5)0 (0)2 (0.3)1 (0.5)1 (1.2)0 (0)1 (5.0) 
Asian/Pacific22 (2.1)2 (3.3)16 (2.6)2 (1.0)2 (2.4)0 (0)0 (0) 
Islander
Other29 (2.9)3 (4.9)10 (4.9)10 (5.0)1 (1.2)2 (5.9)0 (0)0.028
Marital Status
Single82 (8.1)3 (5.0)49 (8.0)15 (7.5)6 (7.1)4 (12.9)5 (25.0) 
Married919 (91.2)56 (91.5)561 (91.5)182 (91.5)78 (91.8)27 (87.1)15 (75.0) 
Divorced/Separated7 (0.7)1 (1.7)3 (0.5)2 (1.0)1 (1.2)0 (0)0 (0)0.314
Insurance Status
Insured716 (69.7)40 (65.6)451 (72.3)136 (67.0)57 (67.1)24 (70.6)8 (40.0) 
Underinsured207 (20.2)14 (23.0)114 (18.3)42 (20.7)22 (25.9)6 (17.6)9 (45.0) 
Uninsured104 (10.1)7 (11.5)59 (9.5)25 (12.3)6 (7.1)4 (11.8)3 (15.0)0.123
Table 2.   Obstetric history, pregnancy stressors, risk factors, personal and family psychiatric history, total and stratified by prepregnancy BMI category
 Total
n = 1027
Underweight
n = 61
Normal
n = 624
Pre-obese
n = 203
Class-1
obese
n = 85
Class-2
obese
n = 34
Class-3
obese
n = 20
P*
  1. *Chi-square (when cell counts were ≥5)/Fisher’s exact (when cells counts were <5) and Kruskal–Wallis tests were used for significance testing.

 mean ± standard deviation
Gestational age (weeks) at first prenatal visit9.3 ± 4.09.8 ± 2.99.3 ± 3.99.4 ± 3.89.5 ± 4.68.1 ± 3.79.2 ± 3.40.08
Gestational age (weeks) at delivery39.2 ± 1.139.2 ± 1.139.2 ± 1.139.1 ± 1.139.0 ± 1.439.0 ± 0.839.2 ± 1.80.44
Prepregnancy BMI24.4 ± 5.517.6 ± 0.821.7 ± 1.727.0 ± 1.432.1 ± 1.537.3 ± 1.444.8 ± 3.9<0.01
Weight gain (pounds)33.3 ± 13.532.8 ± 12.134.7 ± 11.433.2 ± 15.230.8 ± 13.723.1 ± 17.218.3 ± 26.8<0.01
 n (percentage)
Weight retained
Less than 5 lbs retained236 (28.8)11 (20.8)107 (21.3)59 (38.5)33 (48.5)14 (63.6)11 (61.1)<0.01
More than 10 lbs retained405 (49.5)27 (50.1)270 (53.8)72 (46.2)25 (36.8)5 (22.7)6 (33.3)0.04
History of physical or sexual abuse119 (11.7)6 (9.8)65 (10.5)24 (12.1)14 (16.5)6 (17.6)4 (20.0)0.36
Financial stressors497 (49.3)30 (50.0)317 (51.7)93 (46.7)32 (38.6)16 (48.5)9 (45.0)0.31
Partner-associated stressors195 (19.3)10 (16.7)103 (16.8)45 (22.6)22 (26.5)10 (30.3)5 (25.0)0.08
Emotional stressors352 (34.9)15 (25.0)215 (35.0)71 (35.7)29 (34.9)15 (45.5)7 (35.0)0.51
Traumatic stressors103 (10.2)6 (9.8)63 (10.2)20 (10.0)9 (10.8)3 (9.1)2 (10.0)0.99
Smoked tobacco during pregnancy22 (2.1)2 (3.3)13 (2.1)4 (2.0)2 (2.4)0 (0)1 (5)0.86
Drank alcohol during pregnancy132 (12.9)7 (11.5)81 (13.0)29 (14.3)9 (10.6)5 (14.7)1 (5)0.84
Used elicit drugs during pregnancy7 (0.7)1 (1.6)3 (0.5)2 (1.0)0 (0)0 (0)1 (5.0)0.17
History of depression175 (17.0)6 (9.8)97 (15.5)38 (18.7)19 (22.4)10 (29.4)5 (25.0)0.08
History of PPD; multiparas only, n = 61487 (14.3)8 (21.2)45 (12.5)18 (14.0)13 (25.5)1 (4.2)2 (16.7)0.08
History of other psychiatric disease12 (1.2)0 (0.0)9 (1.4)1 (0.5)0 (0)0 (0)1 (5.0)0.28
Family history psychiatric disease157 (15.3)9 (14.8)106 (17.0)26 (12.8)8 (9.4)3 (8.8)5 (25.0)0.20

The proportion of women screening positive for depression measured 6–8 weeks postpartum, as defined by an EPDS ≥ 12 varied by prepregnancy BMI, as depicted in Figure 1. The lowest depression rate was in women with a normal BMI (BMI = 18.5–24.9). In total 14.4% of all 624 women with a normal BMI screened positive for PPD, whereas the proportion of the 61 underweight women screening positive was greater (18.0%), and was similar to the proportions of the 205 pre-obese women (18.5%) and the 85 class-1 obese women (18.8%). These differences were not statistically significant. Rates were significantly higher in the 34 class-2 obese (32.4%) and the 20 class-3 obese (40.0%) women (P = 0.03). Overall, only 54% of all women recalled having their mood addressed at their postpartum visit, and 4.3% of all women studied reported being diagnosed with PPD by their provider. The percentage of women who recalled being diagnosed with PPD was higher in the group who recalled that their mood was addressed by their provider at their postpartum visit (7.0% versus 1.5%; P < 0.05). Two of the six women who reported being diagnosed with PPD, but not at their postpartum visit, had a history of PPD, and received the diagnosis from a psychiatric provider.

Figure 1.

 The percentage of women screening positive for postpartum depression (EPDS ≥ 12) by prepregnancy BMI category.

After controlling for maternal age, race/ethnicity, education, history of depression, pregnancy stressors, and history of abuse, prepregnancy obesity was an independent risk factor for screening positive for depression postpartum. The adjusted odds ratio of screening positive for PPD in class-2 obese women was 2.87 (95% CI 1.21–6.81), and was 3.94 (95% CI 1.38–11.22) in class-3 obese women. The adjusted odds ratios of screening positive for depression among prepregnancy underweight, pre-obese and class-1 obese women were not significantly increased compared with women of normal weight (Table 3). The variables listed in Table 3 represent all variables used in the model. To address the smaller sample sizes in the class-2 and -3 groups, the adjusted odds ratio of all classes of obesity were combined, and revealed an adjusted odds ratio of 1.74 (95% CI 1.04–2.87).

Table 3.   Multiple logistic regression model of prepregnancy BMI and screening positive for PPD
 Adjusted OR95% CI
  1. Model predicts 83.4% of cases.

Age0.990.95–1.03
Race/ethnicity
White, non-HispanicReference 
White, Hispanic1.160.59–2.29
Education0.950.87–1.03
History of depression
NoReference 
Yes2.901.94–4.33
Stressors
NoneReference 
Partner-associated2.611.63–4.18
Traumatic1.660.87–3.70
Both8.484.25–16.91
Prepregnancy BMI
NormalReference 
Underweight1.220.54–2.73
Pre-obese1.090.66–1.80
Obese – class 11.060.54–2.09
Obese – class 22.871.21–6.81
Obese – class 33.941.38–11.22
History of abuse
NoReference 
Yes1.160.69–1.96
Constant0.26 

As expected, a history of depression was positively associated with screening positive for depression measured postpartum (aOR 2.90; 95% C.I 1.94–4.33). Partner-associated stress during the pregnancy was also associated with screening positive for depression, and a multiplicative association was found when women endorsed concurrent partner-associated and traumatic stress (aOR 8.48; 95% CI 4.25–16.91). When stressors were excluded from the model the association persisted and the adjusted odds ratios were as follows: obese class 1, 1.24 (95% CI 0.66–2.34); class 2, 2.96 (95% CI 1.28–6.84); and class 3, 3.20 (95% CI 1.14–9.03). Pregnancy complications in this cohort of term women were uncommon, and were not associated with PPD; therefore, they were not included in the model. When included as covariates, gravidity and weight gain did not impact the prediction of PPD, and were subsequently excluded. Likewise when weight gain data were categorised to inadequate, adequate, and excess weight gain according to the new Institute of Medicine recommendations (with adequate as the reference) there was no significant change in the model.17 Whereas 46.7% of underweight and 39.5% of normal weight women gained the recommended weight during pregnancy, only 20.3 and 12.4% of preobese and obese women gained the recommended weight. In the latter two groups over two-thirds gained excess weight.

Seventy-six percent of women reported recalling that weight was reviewed at some time during their prenatal care, but only 24% reported that this was discussed postpartum. Approximately half recalled that their provider discussed their mood during pregnancy (44%) and postpartum (54%).

Discussion and conclusion

In this study, we found an association between prepregnancy BMI and screening positive for depression postpartum using a validated measure of postpartum mood. Whereas 14.4% of normal weight women screened positive for depression measured postpartum, the PPD test positive rate increased progressively as BMI increased, to 18.8% of class-1 obese, 32.4% of class-2 obese, and 40.0% of class-3 obese women. The increased odds of screening positive for depression conferred by prepregnancy obesity persisted after controlling for multiple covariates, including maternal age, race/ethnicity, education, personal history of depression, pregnancy stressors, and history of abuse. The overall proportion of women screening positive for PPD in this study is slightly higher than estimates of PPD in the literature. Dietz et al.18 identified a PPD rate of 10% in a group of predominantly White women. Dietz et al.’s study used the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis, and treatment with medication, to define their cases. Thus, it is likely that the method of ascertaining the outcome, specifically by using a screening test, raises the estimate of women with PPD. Additionally, using a higher cut-off point (EPDS score ≥ 13), which has been shown to improve specificity,19 did not substantially change the association between screening positive for PPD and pregravid BMI. The higher cut-off point simply decreased the proportions to 14.8% (underweight), 11.2% (normal), 14.9% (pre-obese), 15.3% (obese class 1), 26.5% (obese class 2), and 30.0% (obese class 3).

Despite the exclusion of women being treated for depression on enrollment, some of the women screening positive for depression at the postpartum assessment may have had antecedent depression that was unrecognised during pregnancy, thus increasing the PPD percentage. However, many published rates of PPD also include antepartum depression that is identified in the postpartum period, and thus may not entirely explain the increased percentage of women screening positive for PPD. Although screening cut-off points and antenatal depression continuing into the postpartum period may account for some of the increased rate, it is possible that the overall rates are greater in the study population. The study participants were recruited from the state with the highest major depression rate in the nation.20

This is the first study of adequate sample size to allow the stratification of the extremes of prepregnancy BMI while using a well-validated measure for the primary outcome variable (EPDS) to describe the association of obesity and screening positive for depression in the postpartum period. These findings mirror the well-known association between obesity and depression in non-pregnant women, and support the recent work of Laraia et al. showing that severe obesity during pregnancy is associated with depressed mood.1,2,12 These findings contrast with the lack of association seen in the randomised controlled study by Krause et al.11 The exposure variable and study population varied between the current study and the results from the active mothers postpartum (‘AMP’) study. The ‘AMP’ study used BMI at 6 weeks postpartum as the exposure variable, and did not include women of normal weight. Additionally, the women participating in the study were less likely to report a history of depression, only 1.4%, and 41% of women reported an income of greater than $60 000 annual income, and thus may not represent the ‘at-risk’ population.11 Unlike our previous cross-sectional analysis, being underweight was not associated with screening positive for PPD.9 In the current study obesity was the independent variable of interest: the study was not powered to detect a significant difference between underweight and normal weight women.

Although all women should be screened for PPD, identifying women at increased risk of PPD affords the opportunity for early diagnosis as well as research into targeted early interventions, which is of utmost importance because unrecognised or untreated depression can have a profound effect on both mother and child. Untreated women are more likely to have recurrent or persistent depression,7 and their infants are more likely to exhibit behavioural problems and cognitive delays.8 The association of obesity and depression during and after pregnancy should be considered when managing the obese parturient, not only for the issues pertaining to mood, but also for how mood might impact on other adverse outcomes that co-occur with obesity.

There is evidence that the association between obesity and depression in women is bidirectional. Plausible physiological and behavioral mechanisms exist in this pathway including the hypothalamic–pituitary–adrenal axis and immunologic dysregulation, functional impairment, body-image dissatisfaction, and binge eating.21 Although this study excluded women taking medication for depression at enrollment, women could have been untreated or undiagnosed during pregnancy, and less frequently before pregnancy, and remain in the study cohort. Thus some women may have antecedent depression and not PPD per se, which is a limitation of this study. However, screening for antenatal depression is not typically included in prenatal care, and as such this cohort, which excludes women under treatment for depression, is likely to represent the population who should be screened in clinical practice. In retrospect, including these women and performing a subgroup analysis would have strengthened the study. Thus, from these data it is not possible to determine the causal pathway. Regardless of the directionality and precise mechanism of the obesity–depression association, acknowledgement of this association affords targeted surveillance and intervention. Additional studies to identify the trajectory of mood experienced by these women is necessary to better describe the association as primarily antenatal or postnatal, and to guide the timing of potential interventions in obese women.

The difference between women who were initially recruited and those who completed both the immediate and later postpartum evaluation may limit the generalisability of the results, and the accuracy of the point estimate for increased risk. Because women who did not complete the postpartum part of the study were more likely to be obese before pregnancy, non-White, of lower socioeconomic status, and to report a history of abuse, inclusion of these women would be expected to increase the strength of the association, given the generally accepted associations of these variables with both obesity and depression.

Some limitations of this study should be noted. Although the population studied reflects the overall largely White race/ethnicity of the state of Utah, these results may not be generalisable to more ethnically diverse populations. Although the EPDS has been translated and validated in many different languages, logistical issues resulted in the exclusion of non-English speaking women. These results cannot necessarily be applied to non-English speakers.

Optimally, anthropometric data would be collected before pregnancy by trained personnel or by review of the patient’s prenatal record. The study design did not permit measurement of antepartum BMI, and prenatal records were incomplete and not uniformly available in the in-patient hospital record. The self-reported height and prepregnancy weight used to calculate prepregnancy BMI could have been over- or under-reported. Nevertheless, data regarding the validity of self-reported height and weight in reproductive-age women, and estimates of prepregnancy weight by women during pregnancy, suggest acceptable levels of correlation with measured self-reported weight, and the latter has been used successfully to characterise associations in other epidemiologic investigations.12,22–25 Pre-obese women may underestimate their prepregnancy weight, and lean women may overestimate their weight. In the present study, this misclassification bias would be expected to reduce the strength of the associations observed, i.e. a conservative bias. Some studies show that race/ethnicity influences the association between self-reported and measured height and weight.23 The homogeneity of the Utah study population decreases this potential misclassification bias.

Although we were unable to perform a comprehensive psychiatric evaluation to diagnose depression in all women screening positive, the EPDS has been shown to be a good surrogate measure of depression in circumstances in which it is not feasible to do personal clinical assessments.16 Only 4.3% of all women in the study reported being diagnosed with PPD, despite the fact that three-fold more women screened positive using the EPDS instrument. Although some of this discrepancy may result from error in diagnosis with the screening test, the disparate rates of PPD (1.5% versus 7.0%), based on whether women reported that mood was addressed by their postpartum care provider, lends support to the fact that some women were not evaluated for depression outside of their participation in this study. Women recalled that only 54% of providers addressed postpartum mood, despite the publication of the August 2006 Committee Opinion, by the American College of Obstetrics and Gynecology, which advocated screening for depression and for psychosocial risk factors as part of comprehensive perinatal care for women.26

Despite the limitations of this study, this is a large and comprehensive study of prepregnancy obesity and screening positive for depression in the postpartum period, showing an association independent of multiple other potential risk factors for depression, and highlighting the limited evaluation for depression during or after pregnancy. These findings call for further research identifying the timing of the onset of depression, mechanisms, and the need for heightened attention to weight, mood, and their interaction in pregnant women.

Disclosure of interests

None.

Contribution to authorship

DYL was involved in all phases of this study. She had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. EB-C contributed to the data analysis and interpretation. MO’H provided aid in data interpretation and consultation during the design of the study. AH provided design assistance and provision of the patient safety net. MWV made substantial contributions to the conception, design, and interpretation of the data. All authors participated in the drafting and revising of the manuscript.

Details of ethics approval

This study was conducted after the approval of the University of Utah and Intermountain Health Care Institutional Review Boards, approval #12618 and 05.2062, respectively.

Funding

This work was supported by a National Institutes of Health grant RO3 HD048865.

Acknowledgements

None.

Ancillary