The median age of the 777 HIV-positive eligible women was 35 years. Black, non-Hispanic was the largest racial group (61.5%), followed by white, non-Hispanic (20.2%) and Hispanic/Latino (18.3%). Most women were single (43.0%), with a median reported age at first intercourse of 15 years. Among those women sexually active at baseline, almost half (46.6%) reported that they did not always use a condom in the 6 months prior to the baseline visit. HSV-2-positive women were more likely to be black, non-Hispanic (P = 0.0003) and were more likely to be older (P = 0.0009), although the difference in median age between HSV-2-positive and HSV-2-negative women was small ().
Baseline Characteristics of 777 HIV-Positive US Women From the HIV Epidemiology Research Study (1993–2000), Stratified by HSV-2 Serostatus
Approximately two-thirds of the women (67.8%) were seropositive for HSV-2 at baseline. Almost three-quarters of the women (74.1%) were seropositive for HSV-1 at baseline, 38% of the women were culture-positive for Candida (38.0%), and almost 13% were positive for T. vaginalis. Prevalences of chlamydia (1.2%) and gonorrhea (0.5%) were relatively low at baseline. Prevalence of syphilis was higher, at 7.7%. HSV-1-seronegative women were more likely to be HSV-2 seropositive, as were women with syphilis. Among HSV-2-seronegative women, the prevalence of hepatitis C virus was 60.4%, while the prevalence of hepatitis C virus among HSV-2-seropositive women was 60.8%. Since hepatitis C virus positivity was not associated with HSV-2 serostatus, it was not considered a confounder in the mixed models.
The median baseline HIV viral load was similar between HSV-2-seropositive (3.1 log copies/mL, 95% confidence interval: 1.4, 5.7) and HSV-2-seronegative (3.2 log copies/mL, 95% confidence interval: 1.4, 5.3) women (P = 0.44) (). Baseline CD4 count was slightly higher (P = 0.06) for those who were HSV-2 seronegative at baseline (median CD4 count 412 cells/mm3 compared with 362 cells/mm3 for HSV-2-seropositive women) ().
After censoring, women completed between 2 and 10 study visits, with a median of 7. There were 278 visits (5.2%) where a woman was currently receiving HAART after censoring. Women were not using any antiretrovirals at 3,074 visits (57.5%) and were on sub-HAART regimens at 1,992 visits (37.3%). Data on HAART use were missing for 2 women at one visit. There were 193 visits (3.6%) for which a value for viral load was missing and 175 visits (3.3%) for which a CD4 count was missing.
The final HIV viral load model was adjusted for baseline CD4 count and its associated spline terms, presence of genital infections, and race (). Age was not retained in the final model since removing it changed the HSV-2 coefficient estimate by less than 10%. The HSV-2 coefficient and the coefficients representing the interaction between HSV-2 and time were nonsignificant in the viral load model. We also assessed whether the effect of HSV-2 on viral load over time was nonlinear by including an interaction term between HSV and the polynomial term for time. The likelihood ratio test comparing the model without the interaction term with the model with it was nonsignificant (P = 0.10), so the term was not retained in the final model.
Table 2. Summary of Estimated Regression Coefficients From Final Mixed Models for Associations Between Baseline HSV-2 Seropositivity and 1) HIV Viral Load and 2) CD4 Count/100 Over Study Follow-up, HIV Epidemiology Research Study, United States, 1993–2000 (more ...)
The final CD4 model was adjusted for baseline viral load and race. Neither age, presence of genital infections, nor the spline terms for baseline viral load were retained in the final model since their removal changed the HSV-2 coefficient estimate by less than 10%. The interaction term between HSV-2 and time was statistically significant (P = 0.006), indicating that HSV-2-seropositive women, compared with HSV-2-seronegative women, had an increase of 8 cells/mm3 per year over the study period. This result, while statistically significant, was in the opposite direction from that expected and is likely not clinically meaningful because of the small change in CD4 count over time. As in the viral load model, we assessed whether the effect of HSV-2 on CD4 over time was nonlinear by including an interaction term between HSV and the polynomial term for time. The likelihood ratio test comparing the model without the interaction term with the model with it was nonsignificant (P = 0.32), so the term was not retained in the final model.
HSV-2 therefore did not notably influence the trajectories of HIV viral load and CD4 counts over time. The mean predicted HIV plasma viral load values and CD4 counts from the mixed models are shown graphically in , respectively, stratified by HSV-2 serostatus. The graphs presented are the predicted lines for those individuals in the referent group of each category (median baseline viral load, median CD4 count, absence of genital infections, and nonblack race). Plots for other groups looked similar but are not shown.
Figure 1. Mean predicted values from the mixed model by herpes simplex virus type 2 (HSV-2) status with individual visit data for A) human immunodeficiency virus (HIV) viral load (log copies/mL) and B) CD4 counts (cells/100), United States, 1993–2000. Solid (more ...)
When data were restricted to the 57.5% of study visits where women reported no current use of antiretrovirals, results from the viral load model were similar. In the CD4 model, the interaction between HSV and time was smaller (HSV-2-seropositive women, compared with HSV-2-seronegative women, had an increase in CD4 of 3 cells/mm3 per year over the study period) and was nonsignificant (P = 0.45). When study visits where women reported current use of the antivirals Cytovene (n = 12 visits), Foscavir (n = 328 visits), and amantadine (n = 2 visits) were excluded, results were similar for both the viral load and CD4 models.
These results were consistent with the Kaplan-Meier curves examining the time to AIDS diagnosis by baseline HSV-2 serostatus (data not shown). The hazard ratio comparing HSV-2-seropositive women with those who were HSV-2 seronegative at baseline was 0.99 (95% confidence interval: 0.77, 1.23).