In this study of 2398 women, there was an increased risk of any antenatal depression associated with preexisting diabetes in the unadjusted analysis; however, this association was attenuated and no longer statistically significant after adjusting for demographic, clinical, and pregnancy characteristics. There was no independent association between GDM and either measure of antenatal depression in the unadjusted or adjusted analyses.
Our findings differ from some of those reported in prior studies. Kozhimannil et al.,11
in a study of 11,024 pregnant women on Medicaid, found a 2-fold increased odds of an ICD-9 code for perinatal depression (depression 6 months before 1 year after delivery) or a prescription for antidepressants among women with preexisting diabetes and women with GDM. Although in the unadjusted analysis, we found a similar association of preexisting diabetes and any antenatal depression, this association was largely accounted for by the greater percentage of women with preexisting diabetes who also had other chronic medical conditions. Type 2 diabetes is often associated with comorbid illnesses, such as hypertension. Thus, the burden of chronic illness, and not diabetes alone, may be more likely to be associated with depression. The study by Kozhimannil et al.11
did not adjust for the presence of other chronic medical conditions. Additionally, this study used a Medicaid sample and given the higher prevalence of risk factors for diabetes among low-income women,25
it is possible that a greater proportion of women diagnosed with GDM in this study had previously unrecognized type 2 diabetes, which manifested as more severe GDM.
Another key difference between the two studies is that we relied on antenatal diagnostic screening for depression that was applied to the entire study population independent of diabetes status. Kozhimannil et al.11
defined depression based on ICD-9 codes and use of antidepressants in the 6 months before and 1 year after delivery, thus focusing on women with recognized depression and including PPD in their study outcome. Only 20%–50% of women meeting the criteria for major depression in pregnancy are accurately diagnosed, and physicians are likely to recognize women with more severe and persistent symptoms of depression.2,4
Moreover, if women with preexisting diabetes or GDM are more likely to be diagnosed with perinatal depression because of a greater number of healthcare visits, this might explain the observed increased odds of depression among women with preexisting diabetes and GDM in the study by Kozhimannil et al.11
With respect to the lack of association between GDM and antenatal depression, our findings were similar to those of three prior smaller studies.12–14
However, evidence from the Australian Carbohydrate Intolerance Study in Pregnant Women, a randomized trial of treatment of hyperglycemia in pregnancy, suggests that treatment of hyperglycemia in pregnancy may improve mood profile,26
and if women with GDM in our study were screened for depression after beginning treatment for GDM, this could explain our null findings. Therefore, we conducted a sensitivity analysis excluding women screened for depression after 4 months of gestation. The results (not shown) did not differ from the results of our primary analysis.
The current study has several important strengths. We identified antenatal depression using a diagnostic instrument (PHQ-9) that was administered independent of a woman's diabetes status; therefore, we were able to identify clinically relevant depression in an unbiased manner. Further, use of the PHQ-9 also enabled us to distinguish between probable minor and major depression to further refine our outcome definition and analysis. We were also able to collect information about current use of antidepressants in addition to detailed demographic, pregnancy, and clinical information. Finally, this study included a large and diverse population, making our findings more generalizable to the broader population of pregnant women than previous studies, which had small sample sizes or studied a Medicaid population.
There are also several limitations to consider. Despite the diversity of the study population, our study was from a single large obstetrics clinic in one geographic region of the United States, thus limiting the generalizability of our findings. This study examined prevalent antenatal depression and did not distinguish between preexisting and incident depression or include information on past history of depression. Depression is thought to have a bidirectional effect on chronic diseases, such as diabetes,9
and thus is both a risk factor for and a consequence of diabetes. Therefore, by focusing on prevalent antenatal depression, this study could not determine the direction of causality. Depression status was determined by the PHQ-9, not a structured psychiatric interview. Thus, we used the term “probable” major and minor depression, as clinical interviews were not done to confirm questionnaire diagnosis. Although we were able to adjust for a large number of confounders, information on prepregnancy BMI was not available. However, two prior studies that examined the association of depression and diabetes found that adjusting for BMI did not appreciably change the measured association.27,28
Additionally, prepregnancy BMI may have been a consequence of preexisting depression or management of preexisting diabetes through weight loss, and, therefore, adjustment for prepregnancy BMI would have been inappropriate. Finally, a portion of data on important covariates was missing. Therefore, we used multiple imputation, a technique shown to decrease bias and increase efficiency when data are missing at random.23
Results using complete-case analysis were similar to the multiple imputation results.