Data from 1686 women were analyzed: 1203 (71.4%) were categorized as nonusers, 429 (25.4%) as intermittent users, and 54 (3.2%) as persistent users of crack. Their characteristics are presented in .
Characteristics of 1686 HIV+ women in a multi-site cohort according to longitudinal patterns of crack use, 1996-20041
There were 419 deaths during the follow-up period: 197 (47.0%) were AIDS-related, 138 (33.0%) were non-AIDS related, and 84 (20.0%) were indeterminate. Time to death was assessed with a Kaplan-Meier function (). The estimated survival rates at 8.2 years (3000 days) were 89% for nonusers, 90% for intermittent users, and 65% for persistent users (log-rank test=6.6, p<.05). In a Cox proportional hazards model () adjusting for age, race, income, education, problem drinking, adherent HAART use, CD4 count < 200 cells/mm3 at baseline, HIV-1 RNA level > 100,000 copies/ml at baseline, illness duration, and study site, compared with that for nonusers, the risk of AIDS-related death was significantly higher for persistent users (hazard ratio=3.61, p<.001), but not for intermittent users.
Figure 1 Survival by patterns of crack use in a cohort of HIV-1 infected women. Compared with nonusers (dashed line), and intermittent users (dotted line), days to death for persistent crack users (solid line) was significantly shorter and survival rates significantly (more ...)
Cox proportional hazards models of effects of patterns of crack use on AIDS-related mortality and AIDS-defining illnesses, N=1686: Models control for study site.
Of the total group of 1686 women, 543 (32.2%) were found to have a newly-acquired AIDS-defining illness during the follow-up period. Significantly higher proportions of intermittent users (42.0%, n=180) and persistent users (38.9%, n=21) reported a new illness during this time period than did nonusers (28.4%, n=342) (chi square=27.6, p<.001). The most frequently reported AIDS-defining illnesses were bacterial pneumonia (n=98, 18% of all cases), pneumocystis carinii pneumonia (n=52, 10%), herpes simplex virus-non-pulmonary (n=49, 9%), esophogeal candidiasis (n=48, 9%), cryptosporidiasis (n=30, 6%), dementia/encephalopathy (n=27, 5%), wasting syndrome (n=27, 5%), and tuberculosis (n=20, 4%). Among these, persistent and/or intermittent users were significantly more likely than nonusers to report bacterial pneumonia (chi square=18.8, p<.001), tuberculosis (chi square=16.6, p<.01), and esophogeal candidiasis (chi square=6.4, p<.05). Time to new AIDS-defining illness was assessed in the three groups with a Kaplan-Meier function (). The average days to illness or censoring was 2592 days for nonusers, 2305 days for intermittent users, and 2211 days for persistent users (log-rank test=27.5, p<.001). In a Cox proportional hazards model () the risk of AIDS-defining illness was significantly higher for intermittent crack users (hazard ratio=1.57, p<.001) and consistent users (hazard ratio=1.65, p<.05) than for nonusers, adjusting for all covariates.
Figure 2 Time to newly acquired AIDS-defining illness by patterns of crack use in a cohort of HIV-1 infected women. Compared with nonusers (dashed line), days to illness for intermittent users (dotted line) and persistent crack users (solid line) were significantly (more ...)
and present the unadjusted proportions over time by pattern of crack cocaine use of women with CD4 count<200 cells/mm3, and HIV-1 RNA>100,000 copies/ml. Throughout most of the study period, those reporting persistent crack use had higher viral load concentrations and poorer immune function, while those reporting no use had the lowest HIV-1 RNA levels and best immune health, with intermittent crack users falling in between.
Figure 3 Unadjusted proportions of women with CD4 lymphocyte count <200 copies/mm3 over 18 semi-annual study visits. Nonusers (dashed line) had generally lower proportions, while persistent users (solid line) typically had the highest proportions, with (more ...)
Figure 4 Unadjusted proportions of women with HIV-1 RNA viral load >100,000 copies over 18 semi-annual study visits. Nonusers (dashed line) had consistently lower proportions, while persistent users (solid line) generally had the highest proportions, with (more ...)
presents the results of a time-varying random regression analysis of the effects of persistent and intermittent crack cocaine use on CD4 < 200 and HIV-1 RNA > 100,000. Across both models, persistent crack use, intermittent-active, and intermittent-abstinent crack use were significantly associated with HIV disease progression, controlling for adherent HAART use, problem drinking, women’s socio-demographic characteristics, study site, illness duration, baseline viral load (in the CD4 model), and baseline CD4 (in the viral load model). Persistent problem drinking was positively associated with disease progression defined by high viral load but not low CD4. In both models, adherent HAART use was protective against disease progression.
Random regression analysis of effects of time-varying patterns of crack use on markers of HIV disease progression, N=1686: Models control for study site.
We tested five additional covariates that could account for the relationship between crack use and disease progression, with the same models used in the Cox proportional hazards and random regression analyses. Results (not shown) remained highly similar controlling, separately, for heroin use, intravenous drug use, tobacco smoking, Hepatitis C virus co-infection, and depressive symptoms (using the Center for Epidemiologic Studies-Depression Scale clinical cutoff of 16).26
The only exceptions were for intermittent-abstinent crack use, which became non-significant in the viral load models when controlling for smoking and for depression.
Finally, to explore the impact of crack use on immune reconstitution, we conducted a supplementary analysis of associations between patterns of use (nonuse, inactive use, and active use) and immunologic response. Following Lucas and colleagues,9
for all women remaining in the cohort at the end of the study period (n=1,053), we defined change in HIV-1 RNA (log10
copies/ml) as the difference between the most recent viral load and peak HIV-1 RNA level, and change in CD4 as the difference between the most recent and nadir CD4 lymphocyte counts. We found that the median reduction in HIV-1 RNA level was highest in nonusers, at 1.7 log10
copies/ml, compared to 1.4 log10
copies/ml in inactive crack users and 1.0 log10
copies/ml in active users (F=4.94, df=2/1035, p<.01). The median CD4 increase was highest in nonusers, at 161 cells/mm3
, compared with 123 cells/mm3
in inactive users, and 100 cells/mm3
in active users (F=6.99, df=2/1035, p<.01). In multivariate linear regression models (not shown), active and inactive crack use remained significant after adjustment for race/ethnicity, use of HAART, HAART adherence 95% of time, prior HAART exposure, nadir CD4 count, and peak HIV-1 RNA level. Here, compared with nonusers, active and inactive crack users had smaller median reductions in HIV-1 RNA from baseline, and smaller median increases in CD4. These results confirm that inferior virologic and immunologic responses are associated with both active and inactive use of crack.