Mortality, CD4 Cell Count Decline, and Depressive Symptoms Among HIV-Seropositive Women: Longitudinal Analysis From the HIV Epidemiology Research Study

Context The impact of depression on morbidity and mortality among women with human immunodeficiency virus (HIV) has not been examined despite the fact that women with HIV have substantially higher rates of depression than their male counterparts.

Objective To determine the association of depressive symptoms with HIV-related mortality and decline in CD4 lymphocyte counts among women with HIV.

Design The HIV Epidemiologic Research Study, a prospective, longitudinal cohort study conducted from April 1993 through January 1995, with follow-up through March 2000.

Setting Four academic medical centers in Baltimore, Md; Bronx, NY; Providence, RI; and Detroit, Mich.

Participants A total of 765 HIV-seropositive women aged 16 to 55 years.

Main Outcome Measures HIV-related mortality and CD4 cell count slope decline over a maximum of 7 years, compared among women with limited or no depressive symptoms, intermittent depressive symptoms, or chronic depressive symptoms, as measured using the self-report Center for Epidemiologic Studies Depression Scale.

Results In multivariate analyses controlling for clinical, treatment, and other factors, women with chronic depressive symptoms were 2 times more likely to die than women with limited or no depressive symptoms (relative risk [RR], 2.0; 95% confidence interval [CI], 1.0-3.8). Among women with CD4 cell counts of less than 200 × 10⁶/L, HIV-related mortality rates were 54% for those with chronic depressive symptoms (RR, 4.3; 95% CI, 1.6-11.6) and 48% for those with intermittent depressive symptoms (RR, 3.5; 95% CI, 1.1-10.5) compared with 21% for those with limited or no depressive symptoms. Chronic depressive symptoms were also associated with significantly greater decline in CD4 cell counts after controlling for other variables in the model, especially among women with baseline CD4 cell counts of less than 500 × 10⁶/L and baseline viral load greater than 10 000 copies/µL.

Conclusions Our results indicate that depressive symptoms among women with HIV are associated with HIV disease progression, controlling for clinical, substance use, and sociodemographic characteristics. These results highlight the importance of adequate diagnosis and treatment of depression among women with HIV. Further research is needed to determine if treatment of depression can not only enhance the mental health of women with HIV but also impede disease progression and mortality.

Depression has been associated with immune suppression and other health outcomes in studies of individuals with and without chronic disease.^1-5 Although there has been variability in results, the association between depression and immune outcomes are strongest and most consistent in both older and hospitalized patients.⁵ In the past decade, there has been particular interest in the impact of depression on the health of persons already immunocompromised because of infection with human immunodeficiency virus (HIV).

Investigators have examined the relationship of depression with morbidity and mortality among gay and bisexual men with HIV. Studies have documented associations between depression and disease progression,^6-8 HIV-related symptoms,^9-11 and mortality.¹² Some studies found that HIV-seropositive gay men who reported depressive symptoms demonstrated immunological changes associated with HIV activity and progression (eg, in CD4, CD8, and CD56 cell counts and proliferative response).^13-15 In contrast, other studies found no association between depression, immune function, and disease progression among gay men with HIV.^9,10,16-20 Possible reasons for these discrepancies include differences between studies in research design, length of follow-up, number and quality of control variables, and measurement of depression or study outcome.

To our knowledge, no studies have reported an association of depression with morbidity and mortality for women with HIV, despite depression being more prevalent among women than men. Clinical levels of depression have been reported by 30% to 60% of women with HIV in community and clinic samples,^21-24 whereas among men with HIV, prevalence of depression is generally documented as 20% or less.^9,20,25-28 In studies examining sex differences, women with HIV reported significantly more depressive symptoms than men with HIV.^11,18,24,29 This may be due to the more impoverished social status and drug use history of women compared with men with HIV.³⁰ For example, in a study of demographically comparable female and male injection drug users with HIV, there was no sex difference in prevalence of depressive disorders, though women had more severe depressive symptoms.¹⁸

The primary objective of this study was to determine whether depressive symptoms were associated with mortality among women with HIV. We also examined the link between depression and decline in CD4 lymphocyte count. To address these questions, we used data from the HIV Epidemiologic Research Study (HERS), a large, prospective cohort study of the biological and psychosocial manifestations of HIV in women followed up for 7 years.³¹ HERS data enabled us to examine the influence of depressive symptoms on clinically relevant outcomes, while controlling for potential confounding factors for which information is often not available. Additionally, because of the longitudinal design of this study, depressive symptoms were defined using repeated assessments, thus providing more reliable estimates proximal to changes associated with physical illness.

HIV disease progression has been linked to clinical, substance use, and sociodemographic factors. We hypothesized that after controlling for these factors, chronic (ie, persistent) depressive symptoms would have a more deleterious impact on health outcomes than intermittent episodes and that intermittent depressive episodes would be more deleterious than limited or no depressive symptoms. We also hypothesized that the effects of depression would be strongest for those with the most advanced disease progression (ie, lower CD4 cell count and higher viral load). If depression is associated with mortality, CD4 cell count decline, or both among women with HIV, early pharmacologic and psychotherapeutic interventions could reduce this risk.

The HERS study cohort included 871 women infected with HIV recruited from 4 study sites: Johns Hopkins School of Hygiene and Public Health, Baltimore, Md; Montefiore Medical Center, Bronx, NY; Miriam Hospital, Brown University, Providence, RI; and Wayne State University School of Medicine, Detroit, Mich. Inclusion criteria were age 16 to 55 years, English or Spanish fluency, documented HIV status within 60 days or consent to HIV testing, and reporting of 1 or more HIV-related risk behaviors. Women with an AIDS (acquired immunodeficiency syndrome) diagnosis (1987 US Centers for Disease Control and Prevention [CDC] surveillance definition) or any AIDS-defining opportunistic infections were considered late in disease progression and were ineligible; 14.5% of participants developed a clinical AIDS diagnosis during the study period. For these analyses, 12% of participants (n = 106) were excluded because they made fewer than 3 study visits, the minimum required to examine reliably the presence of depressive symptoms and disease progression over time. This resulted in a study sample of 765 women.

Staff members were centrally trained and used common data collection procedures.³¹ Recruitment was conducted from April 1993 through January 1995; follow-up visits were completed by March 2000. At 6-month intervals, women were interviewed, underwent physical examination, medical record abstraction, and specimen collection. Specimens were shipped to the CDC for centralized processing and storage (at −70°C). Procedures were in place at all sites for referral to mental health services for diagnostic evaluation and treatment, if needed. The average number of study visits per participant was 10 (SD, 3.5; range, 3-15); few participants were lost to follow-up (n = 91; 11.9%) or voluntarily withdrew from the study (n = 21; 2.7%). Participation was completely voluntary, confidential, and did not influence provision of health care in any way. Study procedures were approved by an institutional review board at each site and at the CDC.

HIV-Related Mortality. HIV-related mortality was determined from standardized review of medical records and confirmed using the National Death Index and individual death certificates. Individuals who died of causes other than HIV were included in analyses and censored at time of death (n = 121).

CD4 Lymphocyte Count and Slope Decline. Flow cytometry from whole blood was used to determine CD4 cell counts at each study visit. All sites passed periodic proficiency panels.³²

Change in CD4 cell count was expressed by regressing CD4 cell count on time (months in study). The resulting standardized β weights represented the slope of CD4 cell count change. Because change may be nonlinear, rate of change was also calculated using the natural logarithm and square root of CD4 cell count and CD4 cell count cubed. Slopes based on transformations were not significantly different from slopes based on untransformed CD4 cell count; thus, only untransformed CD4 cell counts were used in subsequent analyses. To ensure reliability, slopes were not calculated for respondents with fewer than 3 valid CD4 cell count measurements (n = 4).

The Center for Epidemiologic Studies Depression Scale (CES-D) contains 20 items asking about depressive symptoms in the past 7 days, with responses from 0 (rarely/none of the time) to 3 (most of the time).³³ This scale has been used extensively in population-based studies,^33-35 studies of clinical samples,^33,34,36 and of persons with HIV.^10,13 In general population studies, approximately 20% of people have scores greater than 16 and are classified as "probable cases of depression."^33,37,38 The CES-D has good sensitivity (88%), but somewhat lower specificity (73%).^39,40 Additionally, we computed a subscale of "depressed affect," dropping 5 somatic complaints that also reflect symptoms of HIV (ie, fatigue, poor appetite, lack of energy, restless sleep, and poor concentration).^10,13,21

Respondents were classified as having chronic depressive symptoms if CES-D scores were 16 or higher for at least 75% of their study visits (eg, 9 of 12 study visits). Respondents were classified as having intermittent depressive symptoms if their CES-D score was 16 or higher at 26% to 74% of study visits. Respondents were classified as having limited or no depressive symptoms if their CES-D scores were 16 or higher for no more than 25% of study visits. Conservatively, we retained the cut-off score of 16 and used the same classification scheme for the affect-only subscale.

Viral Load.

Quantification of HIV RNA viral load was performed using a branched-DNA (B-DNA) signal amplification assay (Chiron Corporation, Emeryville, Calif). Viral load was categorized as less than 500, 500-9999, or 10 000 or more copies/µL.

Antiretroviral Therapy Use. Participants were hierarchically classified regarding use of antiretroviral medications during the study period. If participants reported taking highly active antiretroviral therapy (HAART) at 2 consecutive study visits, they were classified as receiving HAART for 12 months or more. Of the remaining participants, if women ever received HAART during the study, they were classified as receiving HAART for less than 12 months. If participants never received HAART and reported simultaneous use of 2 or more other antiretroviral medications, they were classified as receiving combination therapy. If women reported taking a single antiretroviral medication, they were classified as receiving monotherapy. The remaining participants were classified as receiving no antiretroviral medication.

HIV-Related Symptoms. Presence of the following HIV-related symptoms was documented: oral thrush; diarrhea; elevated temperature greater than 38°C (100°F); problems concentrating or remembering; numbness, tingling, or burning sensations in the arms, legs, hands, or feet; unexpected weight loss of more than 4.5 kg (10 lb) that lasted longer than 1 month. Unless specified, symptoms lasted longer than 2 weeks and were not due to another, identifiable cause.

Participants reported use of crack or cocaine in noninjection form (yes/no) and use of injection drugs (yes/no).

Age, race/ethnicity, education, employment status, income, and whether receiving public assistance were collected via structured interview.

The main objective of this study was to determine the association of depressive symptoms with mortality and slope of CD4 lymphocyte decline among women with HIV, after controlling for clinical features (eg, CD4 cell count, viral load, antiretroviral medication use, and HIV-related symptoms), substance use, and sociodemographic characteristics. Kaplan-Meier survival analysis examined the extent to which having depressive symptoms predicted time to death. We used Cox proportional hazards modeling to examine this association, controlling for the clinical features, substance use, and sociodemographic characteristics significantly associated with mortality in univariate analyses. Hierarchical linear regression was used to examine the association between depressive symptoms and CD4 cell count slope over time, controlling for clinical features, substance use, and sociodemographic characteristics. By using a hierarchical approach, it is possible to examine change in variance accounted for by each step of the regression equation, in this case, to examine impact of depressive symptoms after controlling for all other variables.⁴¹ Logistic regression (for mortality) and analysis of variance (for CD4 cell count slope decline) were used to test our hypothesis that effects of depression would be strongest for those with most advanced disease progression (ie, interaction between level of depressive symptoms with CD4 cell count and viral load). Statistical adjustments were made for all post hoc analyses to control for type I error rate inflation.

Participants ranged in age from 19 to 55 (mean [SD], 35.5 [6.7]) years; 62% were black, 21% white, and 18% Hispanic or other race/ethnicity (Table 1). At baseline, average social class status for study participants was low: 45% of respondents had not completed high school, 84% were unemployed, 74% reported annual household income of $12 000 or less, and 65% received public assistance. Fifty-four percent of participants reported no illicit drug use during the study period, the remainder reported noninjection crack or cocaine use and/or injection drug use. At baseline, 33% of participants had CD4 cell counts of >500 × 10⁶/L, 26% had undetectable viral load (<500 copies/µL), and 46% were asymptomatic. Nearly half of participants received HAART during the study period. Of those who never received antiretroviral therapy, the majority may have had no clinical indication for treatment (84% had CD4 cell counts >200 × 10⁶/L and 57% had viral loads <10 000 copies/µL throughout the study).

Forty-two percent of participants were categorized as having chronic depressive symptoms, 35% with intermittent symptoms, and 23% with limited or no depressive symptoms. Mean (SD) baseline CES-D scores for those women categorized as having chronic, intermittent, and limited or no depressive symptoms were 30.4 (10.9), 19.2 (11.3), and 8.5 (6.5), respectively. Means were relatively stable across all study visits, thus indicating that depression severity did not increase over time on average.

Participants included in the analyses compared with the 106 women excluded (for having <3 study visits) differed on several characteristics (P<.05): they were less likely to be Hispanic or of other ethnicity (18% vs 28%), have baseline CD4 cell counts lower than 200 ×10⁶/L (15% vs 33%), baseline viral loads of 10 000 copies/µL or greater (18% vs 38%), and received no ART (18% vs 57%) or monotherapy (12% vs 41%). Groups did not differ on other sociodemographic characteristics, illicit drug use, or baseline symptoms, nor did they differ with regard to depressive symptoms.

During the 7 years of follow-up, 106 (14%) women died of HIV-related causes. Figure 1A shows Kaplan-Meier survival curves for HIV-related mortality for participants with chronic, intermittent, and limited or no depressive symptoms. The difference between these curves is significant (log-rank test [2 df], 15.11; P<.001). Whereas 8% of women with limited or no depressive symptoms died, 16% of women with intermittent and 23% of women with chronic depressive symptoms died. Similar survival curves were obtained using the affect-only depressive symptoms subscale (log-rank test [2 df], 19.3; P<.001, data not shown); HIV-related mortality rates were 12%, 17%, and 25% for those with limited or no, intermittent, and chronic affect-only depressive symptoms, respectively.

Univariate Association Between HIV-Related Mortality With Depressive Symptoms and Baseline Characteristics. Recognizing that other factors influence HIV-related mortality, we investigated the effects of confounding variables (Table 2, column labeled "Bivariate Odds Ratio"). Women with chronic and intermittent depressive symptoms, respectively, were nearly 3 and 2 times more likely to die compared with those with limited or no depressive symptoms. Respondents who died differed from those who survived on most clinical features: they had lower baseline CD4 cell counts, higher viral load, a greater number of HIV-related symptoms, and did not receive HAART (bivariate odds ratios, 2.0-26.7; Table 2). Study participants aged 35 years or older and those unemployed at baseline were 1.5 and 2.0 times more likely to die compared with those younger and employed. HIV-related mortality was unrelated to all other sociodemographic features (race/ethnicity, education, income, receipt of public assistance) and illicit drug use (crack/cocaine and injection drug use) (results not shown).

Depressive Symptoms as a Predictor of Time to Death. Cox proportional hazards modeling was used to evaluate hierarchically the effect of depressive symptoms on time to HIV-related death, controlling for potential confounding variables associated with mortality at the P<.10 level (ie, baseline CD4 cell count, viral load, HIV-related symptoms, ART use, age, and employment). Women with chronic depressive symptoms were still 2 times more likely to die than women with limited or no depressive symptoms, as depicted in Figure 1B (overall χ²₁₄ = 336.2; P<.001). Table 2 contains RRs from the final Cox regression model (column labeled "Multivariate Relative Risk").

Finally, analyses were conducted to better understand how depressive symptoms interact with disease progression, measured by baseline CD4 cell count and viral load, to influence HIV-related mortality. For women with baseline CD4 cell counts lower than 200 × 10⁶/L, 54% of those with chronic depressive symptoms died during the 7-year study period, 48% with intermittent depressive symptoms died, and only 21% with limited or no depressive symptoms died (RR, 4.3; 95% confidence interval [CI], 1.6-11.6 and RR, 3.5; 95% CI, 1.1-10.5, for chronic and intermittent compared with limited or no depressive symptoms, respectively) (Figure 2A). For women with baseline CD4 cell counts of 200 to 500 × 10⁶/L, there was a trend for higher mortality among women with chronic and intermittent depressive symptoms compared with those with limited or no depressive symptoms (16%, 11%, 7% of women, respectively). For women with CD4 cell counts higher than 500 × 10⁶/L, there was no difference in HIV-related mortality by depressive symptoms. Similar patterns were observed stratifying for viral load (Figure 2B).

Association of CD4 Lymphocyte Decline Over Time With Depressive Symptoms and Baseline Characteristics. Participants with chronic or intermittent depressive symptoms had significantly greater CD4 cell count slope decline than those with limited or no depressive symptoms (−0.35 and −0.27 vs 0.13) (β = −.14; P<.001). Women with baseline CD4 cell count of >500 × 10⁶/L, those not receiving HAART or receiving HAART for less than 12 months, blacks, and those of lower social class (ie, low income, less formal education, not employed, and receiving public assistance) had significantly greater slope decline (P<.05) (Table 3 ).

Depressive Symptoms as a Predictor of CD4 Lymphocyte Decline Over Time. Hierarchical linear regression was used to estimate the association between depressive symptoms and CD4 lymphocyte decline, controlling for other factors in the model (Table 4). In step 1, clinical features, substance use, and sociodemographic characteristics associated with CD4 lymphocyte decline at the P<.10 level in the univariate analyses were entered. Step 1 was statistically significant, accounting for 10.3% of the variance in CD4 cell count decline (P<.001). In step 2, depressive symptoms were added, thus increasing amount of variance explained to 11.7% (P = .001).

Finally, analyses were conducted to determine how depressive symptoms interact with disease progression (measured by baseline CD4 cell count and viral load) to influence CD4 cell count slope decline. Significant differences were obtained for participants with baseline CD4 cell counts less than 200 × 10⁶/L (F_2,111 = 5.0; P = .009) and CD4 cell counts 200 to 500 × 10⁶/L (F_2,385 = 6.4; P = .002). Post hoc analyses (using Tukey honestly significant difference procedure to adjust for type I error–rate inflation) indicated that among women with baseline CD4 cell count less than 200 × 10⁶/L, those with chronic depressive symptoms had a CD4 cell count slope decline (−0.40) statistically different from the relatively unchanged slope of women with limited or no depressive symptoms (0.06). Among women with baseline CD4 cell counts of 200 to 500 × 10⁶/L, those with chronic depressive symptoms had a statistically larger CD4 cell count decline ( −0.32) than both those with intermittent or limited or no depressive symptoms (−0.15 and −0.04, respectively). Similar patterns were observed for viral load. Significant differences in slope decline across levels of depression were obtained for women with baseline viral loads of ≥10 000 copies/µL (F_2,134 = 5.4; P = .006) and 500 to 9999 copies/µL (F_2,411 = 5.3, P = .005). Post hoc analyses indicated that among participants with a baseline viral load of 10 000 copies/µL or greater, women with either chronic or intermittent depressive symptoms had statistically larger slope declines (−0.48 and −0.39) than women with limited or no depressive symptoms (−0.05). Among participants with baseline viral load of 500 to 9999 copies/µL, women with chronic depressive symptoms had a statistically larger slope decline (−0.36) than women with limited or no depressive symptoms (−0.12).

HIV-related mortality among women was associated with baseline CD4 cell count, viral load, and antiretroviral medication use during the study period. After controlling for these clinical features known to be associated with morbidity and mortality, HIV-seropositive women with chronic depressive symptoms were 2 times more likely to die compared with those with limited or no depressive symptoms. As predicted, those with the lowest CD4 lymphocyte counts and highest viral load were most vulnerable to adverse consequences of depressive symptoms. For example, HIV-related mortality for those with chronic or intermittent depressive symptoms was 54% and 48%, compared with 21% for those with limited or no depressive symptoms, when CD4 lymphocyte count was less than 200 × 10⁶/L. Additionally, after controlling for potential confounding variables, women with chronic depressive symptoms had a more rapid decline in CD4 lymphocyte count compared with women with limited or no depressive symptoms. Rates of depressive symptoms among women in this study were high: 42% reported chronic, persistent depressive symptoms during the 7 years of follow-up.

Determining a clear cause-and-effect relationship between depression and HIV disease progression is complex because symptoms for both conditions overlap.^42,43 Indeed, the association between depressive symptoms and mortality could be attributable to disease progression. However, survival analyses revealed a significant relation between depressive symptoms and mortality while controlling for clinical indicators of disease status. Furthermore, we conducted a supplemental time-dependent survival analysis in which depression, as well as the clinical indicators associated with disease status (CD4 cell count, viral load, HIV-related symptoms, and ART use), were allowed to vary across time; this enabled us to investigate temporal trends between these variables and mortality. Results indicated that participants with depressive symptoms were nearly twice as likely to die than those without depressive symptoms, after controlling for clinical evidence of declining health over time. Moreover, post hoc analyses indicated that among respondents with no reported HIV-related symptoms at baseline, those with chronic depressive symptoms were 3.6 times more likely to die than those with limited or no depressive symptoms. Finally, depression remained significantly associated with mortality even after somatic symptoms of depression were eliminated. Therefore, although not conclusive, our results suggest that depressive symptoms contribute uniquely to mortality and are not simply an artifact of declining health.

The association between depression and mortality has been documented in other studies^3,44-49; however, the mechanisms are not well understood. Depression has been associated with neuroendocrine and immunological changes that influence morbidity and mortality.^5,50,51 Thus, depression may interact with biological vulnerabilities caused by medical illness. The effects of chronic, as opposed to intermittent, depressive symptoms may be a function of increased "allostatic load"; that is, the pathophysiological costs to the individual resulting from persistent depression.⁵² Depression may also have indirect effects via its influence on damaging behaviors (eg, smoking and alcohol abuse) and health care use.^49,53 Perhaps most importantly, depression has been associated with poorer adherence to medication regimens among patients with HIV^54-56 and so has critical implications for viral replication, resistance, and treatment failure.⁵⁷ Moreover, like other studies indicating suboptimal treatment for women with HIV,⁵⁸ only 49% of women in this study reported HAART use. Protease inhibitor use has been associated with a significant reduction in depressive symptoms.⁵⁹ It may be that many women are unable to take advantage of this "secondary gain" of using protease inhbitors, that is, improvements in mental as well as physical health. Further research is needed to elucidate the mechanisms by which depression adversely influences disease progression among women and men with HIV as well as other chronic illnesses.

This study was limited by use of self-reported depressive symptoms: the CES-D is a screening instrument, not a psychiatric diagnostic tool. Additionally, there were several factors not measured that may be mediators between depression and HIV-related mortality (eg, treatment adherence, health care utilization, and psychiatric treatment). Depressive symptoms accounted for only a small proportion of variance in CD4 cell count slope decline. This is similar to the explanatory power in prior studies¹³; however, it raises questions with regard to clinical significance. This may be due to the fact that disease progression, severity of physical symptoms, onset of opportunistic infections, and survival itself varies among individuals with the same CD4 cell count.⁶⁰

In contrast, there were several notable strengths of this study. The cohort was large and recruited from diverse geographic areas. Longitudinal data were available to track prevalence and impact of depressive symptoms during 7 years. This was one of the first analyses to focus on the impact of depressive symptoms among women with HIV. Finally, the study was strengthened by the integration of a broad measurement framework, including objective clinical indicators known to influence disease progression as well as psychosocial factors.⁴⁹

Results of this study may have implications for treatment of depression among women with HIV. Health care providers should recognize the risks of depression for disease progression and mortality among women with HIV. Identifying these at-risk patients early so they can be targeted for closer monitoring and treatment of depression may be critical. As such, the integration of primary clinical services and mental health services for women with HIV should be considered. The combination of depression with substance use, poverty, interpersonal violence, and the burden of caring for themselves and others creates numerous social challenges for women with HIV.^21,61,62

Several common classes of antidepressants are effective and well-tolerated among persons with HIV,^63-66 and psychotherapy interventions also appear promising.⁶⁶ When depressed patients with HIV are treated with antidepressants, medical care costs are significantly reduced.⁶⁷ We must recognize, however, that barriers to successful treatment may include refusal or nonadherence to antidepressants because of the stigma associated with mental illness, adverse effects, and adding another medication to complex regimens.^63,68,69 Outreach, education, and attention to the complex psychosocial needs of women with HIV are essential to treat depression effectively. Further research is needed to determine if pharmacologic and psychotherapeutic treatments for depression can be effective to enhance not only the mental health of women with HIV, but also to enhance their physical health and impede disease progression.