Background.African Americans with human immunodeficiency virus type 1 (HIV-1) infection and kidney disease are at increased risk of end-stage renal disease requiring renal replacement therapy (RRT), particularly in urban areas with high rates of poverty and injection drug use. It is unknown how the widespread use of highly active antiretroviral therapy (HAART) has affected survival during RRT in this vulnerable population.
Methods.African American patients infected with HIV-1 who required RRT were identified from 2 cohorts that included 4509 Africans Americans infected with HIV-1 who were recruited during the period 1988–2004 in Baltimore, Maryland. Survival after initiation of RRT was compared for those who initiated treatment in the pre-HAART and the HAART eras using Kaplan-Meier curves. Cox proportional hazards regression analysis was used to adjust for potential confounders.
Results.RRT was initiated in 162 patients (3.6%) during 10.6 years of follow-up (119 during the HAART era). Compared with patients who started RRT in the pre-HAART era, those in the HAART era were older (P < .001) and more likely to have CD4 cell counts of ⩽200 cells/mm3 (P = .01). A total of 126 patients (78%) died during follow-up; among those who initiated RRT during the HAART era, 87 deaths occurred (73%). Median survival time in the pre-HAART era was 22.4 months (95% confidence interval [CI], 9.3–30.7); during the HAART era, it was 19.9 months (95% CI, 14.7–26.5; P = .94). In the multiple Cox regression model, factors independently associated with increased mortality included age (hazard ratio [HR], 1.30; 95% CI, 1.06–1.60; P = .01), lower serum albumin level (HR, 0.72; 95% CI, 0.57–0.91; P < .007), lower CD4 cell count (HR, 0.90; 95% CI, 0.82–0.99; P < .03), and the lack of HAART (HR, 0.52; 95% CI, 0.33–0.82; P = .005).
Conclusions.Older age, lower serum albumin level, lower CD4 cell count, and the lack of HAART are independent predictors of poor survival among African Americans infected with HIV-1 undergoing RRT in a resource-limited urban area. RRT survival was similar in the pre-HAART and HAART eras, likely reflecting inadequate HIV treatment in this population.
In both the general and HIV-1–infected populations, African Americans are at higher risk for kidney disease than are other race groups [1–3]. In 2005, African Americans accounted for an estimated 49% of new HIV-1 infections in the United States  but only 13% of the general population. African American men aged 25–44 years account for 40% of HIV-1–infected patients receiving renal replacement therapy (RRT), although they account for just 2% of the US population and only ∼1% of the prevalent population with end-stage renal disease in the United States . The disproportionately high rates of HIV-1 infection among African Americans who have higher kidney disease burden may result in increases in HIV-infected patients needing RRT. Initial reports suggested a dismal prognosis for HIV-infected patients undergoing maintenance dialysis [6–8]. However, subsequent data suggested better survival for those patients, particularly with the improved longevity afforded by the introduction of HAART [9–11].
There are few data from well-characterized cohorts that compare survival during RRT in HIV-1–infected African Americans prior to and after the widespread use of HAART. In this study we compare survival among HIV-1–infected African Americans requiring RRT observed in 2 Baltimore-based cohorts between 1988 and 2004. In addition, factors associated with survival were ascertained.
Study participants.The Johns Hopkins HIV-1 Cohort (JHHC) and the AIDS Linked to the Intravenous Experience (ALIVE) study are longitudinal cohorts based in Baltimore, Maryland. The JHHC includes data from HIV-1–infected participants who received primary care in the Johns Hopkins HIV-1 Clinic . Participants have been enrolled on a continuing basis since 1990. Demographic, clinical, laboratory, and pharmaceutical data were collected at times corresponding to enrollment in the clinic and at 6-month intervals thereafter. The ALIVE study is a longitudinal cohort of HIV-1–infected and HIV-1–seronegative injection drug users that enrolled participants in 1988, 1994, 1998, and 2000 . Participants were followed at 6-month intervals, at which time clinical, behavioral, and laboratory data were obtained according to a standardized protocol . In both studies, dates of death were confirmed by matches with the National Death Index. Study participants who were enrolled in both cohorts were considered to be in the JHHC for all analyses. The JHHC and ALIVE were approved by the Johns Hopkins University Medicine Institutional Review Board and Committee on Human Research, respectively. All subjects provided written informed consent.
Definitions.Incident RRT cases were identified by matching participant identifiers with the United States Renal Data System (USRDS). The pre-HAART and HAART eras were defined as the calendar periods before and after 1 January 1996, respectively. Use of antiretroviral therapy was defined in the window of time from 3 months before to 6 months after inception of RRT. HAART was defined as use of ⩾3 drugs in a regimen that included a nonnucleoside reverse-transcriptase inhibitor, a protease inhibitor, or abacavir. Injection drug use was defined as a history of this HIV-1 transmission risk factor identified at enrollment to the cohort. Of note, all participants from ALIVE had a history of injection drug use as stipulated by enrollment criteria. Laboratory values, including albumin and hemoglobin levels and CD4 cell counts, were those obtained closest in time to initiation of RRT.
Statistical analyses.Statistical analyses were performed using Stata Software, version 8 (Stata Corporation). Demographic and clinical characteristics were compared in participants who initiated RRT in the pre-HAART and HAART eras using medians with interquartile ranges or proportions, as appropriate. The Wilcoxon rank sum test and Fisher's exact test were used to compare continuous and categorical variables, respectively.
Survival after the initiation of RRT was estimated using the Kaplan-Meier method and the log rank test. Cox proportional hazards models with 95% CIs were used to estimate the relative hazards of mortality according to baseline covariates. The proportional hazards assumption was examined by graphical methods and Schoenfeld residuals . Factors that were associated with mortality (P < .1) were considered for inclusion in multivariate models. Potential interactions between pairs of variables included in multivariate models were assessed by combining terms in regression analysis. To assess whether survival among subjects who initiated RRT in the pre-HAART era was biased by survival into the HAART era, we conducted a sensitivity analysis in which follow-up time was censored at 1 January 1997 in those who initiated RRT in the pre-HAART era.
Over the course of follow-up, RRT was initiated in 162 (3.6%) of 4509 HIV-1–infected African American subjects from the 2 cohort studies. Hemodialysis therapy was the modality used for the majority of patients (94%). Participants' demographic and clinical characteristics are shown in table 1. Compared with subjects who initiated RRT in the pre-HAART era (n = 43), those who started RRT in the HAART era (n = 119) were significantly older, had significantly lower nadir CD4 cell counts, and were significantly more likely to have had prior exposure to antiretroviral therapy. Additionally, subjects who initiated RRT in the HAART era were significantly more likely to have a CD4 cell count of ⩽200 cells/mm3 at the time of RRT inception than were those in the pre-HAART era (77% vs. 56%; P = .01), although the difference in median CD4 cell count did not achieve statistical significance (table 1). Subjects initiating RRT in the HAART era were slightly less likely to have a history of injection drug use and to be coinfected with hepatitis C virus than were those starting RRT in the pre-HAART era; these differences were not statistically significant. The pre-HAART and HAART era groups were similar by sex, serum albumin concentration, and serum hemoglobin concentration at RRT inception (P > .05).
Participants were significantly more likely to be observed in the JHHC cohort in the HAART era than in the pre-HAART era (90% vs. 60%; P < .001; table 1). This difference likely results from different temporal patterns of subject enrollment in the 2 cohorts. ALIVE conducted its primary enrollment in 1988 and 1989, whereas the JHHC, which was initiated in 1990, enrolled subjects on a rolling basis, with the largest number being enrolled in 1997. As shown in table 2, compared with subjects followed in ALIVE, those in JHHC had a lower median nadir CD4 cell count (109 vs. 139 cells/mm3; P = .04) and a lower median hemoglobin concentration among men (10.0 vs. 11.0 mg/dL; P = .04). In the HAART era, subjects in the JHHC were more likely to be receiving HAART at the time of RRT inception than were those in ALIVE (53 [50%] of 107 vs. 3 [25%] of 12; P = .007).
The median duration of survival after RRT inception was 22.4 months (95% CI, 9.3–30.7 months) in the pre-HAART era and 19.9 months (95% CI, 14.7–26.5 months) in the HAART era (P = .94, by log rank test) (figure 1). The rate of survival at 1 year and 2 years was 56% (95% CI, 40%–69%) and 46% (95% CI, 31%–60%), respectively, in the pre-HAART era and 63% (95% CI, 53%–71%) and 43% (95% CI, 34%–53%), respectively, in the HAART era. As expected, patients who did not receive HAART or antiretroviral therapy in the HAART era (n = 54) had higher median CD4 cell counts at initiation of dialysis than did those who received HAART (n = 56; 222 vs. 119 cells/mm3; P = .001). The median survival time of the HAART recipients was 26 months, compared with 15 months in those who were untreated (adjusted hazard ratio [HR], 0.54; 95% CI, 0.34–0.88; P = .01). The median survival among patients who received HAART (26 months) was slightly longer than that of patients who received antiretroviral therapy (24 months). However, the relative hazards were significantly lower in HAART than antiretroviral therapy recipients after adjustment for baseline differences in CD4 cell count, age, and albumin concentration (HR, 0.54; 95% CI, 0.33–0.91; P = .21). A serum albumin concentration of >3.0 g/dL but not a CD4 cell count of >200 cells/mm3 was associated with better survival during RRT (figure 2).
In the multiple Cox regression model (table 3), factors independently associated with mortality included age (HR, 1.30; 95% CI, 1.06–1.60; P = .01), serum albumin level (HR, 0.72; 95% CI, 0.57–0.91; P < .007), CD4 cell count (HR, 0.90; 95% CI, 0.82–0.99; P < .03), and the lack of HAART (HR, 0.52; 95% CI, 0.33–0.82; P = .005). Initiation of RRT in the HAART era was not significantly associated with mortality (HR, 1.01; 95% CI, 0.69–1.49). To assess whether transition of pre-HAART subjects into the era when HAART became widely used affected this comparison, we conducted a sensitivity analysis in which follow-up time in subjects who initiated RRT in the pre-HAART era was censored at 1 January 1997. This maneuver did not substantively affect the estimated relative risk of death in the HAART era compared with the pre-HAART era (HR, 1.18; 95% CI, 0.70–1.84).
We also considered the possibility that duration of HIV-1 infection may have differed in the 2 treatment eras and affected our point estimates for treatment era or type of antiretroviral use. Ideally this concern would be addressed by incorporating time since HIV-1 seroconversion into the models. However, because this information was unknown, we used the date of enrollment into the cohort as a surrogate. We then conducted a sensitivity Cox proportional hazards analysis using staggered entry, in which enrollment to the cohort was taken as the time origin, but entry into the risk set occurred on the date of RRT initiation. With this method, HR estimates are adjusted for the duration of time under observation before end-stage renal disease. Univariate and multivariate estimates for both HIV-1 treatment era and type of therapy used were not meaningfully different from the model in which staggered entry was ignored (data not shown). No statistically significant interactions among variables included in multivariate models were identified.
This study highlights the relatively poor survival after initiating RRT among HIV-1–infected African Americans with end-stage renal disease in a resource-limited urban area. The median survival time was 22.4 months in the pre-HAART era and just 19.9 months during the HAART era. After adjustment for selected prognostic variables, which were more favorable in the earlier era, the relative hazards of death were nonsignificantly lower in the HAART era than in the pre-HAART era. Although the overall survival was similar during the pre-HAART and HAART eras, we found that older age, lower serum albumin level, CD4 cell count, and the lack of HAART were independent predictors of poor survival.
HIV-1–related renal disease is the third-leading cause of end-stage renal disease among African Americans aged 20–64 years . As a result of African American overrepresentation in the HIV-1 epidemic in the United States and longer survival with HAART [16–18], the number of HIV-1–infected patients reaching end-stage renal disease is projected to increase significantly over the next decade .
Using data from the USRDS, Ahuja et al.  evaluated 6166 HIV-infected patients with end-stage renal disease who received dialysis in the United States and found that during 1990–1999, 1-year survival of HIV-infected patients undergoing dialysis improved from 56% to 74%, and the annual death rates decreased from 458 deaths to 240 deaths per 1000 patient-years. They concluded that survival of HIV-infected patients undergoing dialysis in the United States has remarkably improved because of the introduction of HAART. The authors pointed out that the data were limited by the lack of information regarding HAART use and stage of HIV disease. Similarly, Macrae et al. , using USRDS data, found that the odds of surviving to 1 year had increased >2-fold among African Americans starting RRT in the HAART era compared with those beginning RRT in the pre-HAART era. In contrast, we found only a marginal increase in 1-year survival after beginning RRT in the HAART era compared with the pre-HAART era (63% vs. 56%).
The similar RRT survival rates that we observed in the pre-HAART and HAART eras may have several explanations. First, our patient sample was drawn from 2 inner city cohorts with high rates of poverty and substance abuse, factors that have been linked to worse-than-expected outcomes during RRT . Although we did not find injection drug use to be significantly associated with survival, this risk factor was over-represented in our sample (67% of subjects), limiting our ability to meaningfully compare with non–drug users. Lucas et al.  demonstrated that active drug use is temporally linked to HIV disease progression and mortality. Second, although we did not find an improvement in RRT survival in the HAART era, we did find that individual use of HAART in the period around RRT initiation was associated with significantly longer survival. Strikingly, in our study, nearly half of patients who initiated RRT in the HAART era were not receiving HAART. This implies that the absence of a temporal improvement in RRT survival over this period was in part due to underuse of antiretroviral therapy.
Third, our data also indicate that patients initiating RRT in the pre-HAART era had substantially better prognostic factors than did those in the HAART era, including younger age and higher current and nadir CD4 cell counts. Although we attempted to adjust for some of these differences in prognostic multivariate models, residual confounding is likely. For example, compared with HAART-era patients, those who initiated RRT in the pre-HAART era may have had other favorable prognostic characteristics, such as a higher level of adherence to medical care. The reason for the differences in prognostic factors in the 2 HIV-1 treatment eras is not immediately clear. Many participants in both cohorts were followed from the pre-HAART period into the HAART era. Subsequently, the HAART-era participants may represent those with onset of renal failure among subjects with a longer duration (and more advanced progression) of HIV infection. HIV-infected subjects surviving into the HAART era therefore were older and had lower CD4 cell counts. However, patients with advanced HIV-1 disease and chronic kidney disease may have been less likely to survive until end-stage renal disease in the earlier era. Alternatively, RRT may have been offered to patients with worse prognostic factors in the later era under the assumption that HIV disease should be reversible with HAART. Patients in the pre-HAART era with kidney dysfunction may have not been offered RRT because of their expected higher mortality rate. Patients who survived to RRT in the pre-HAART era were healthier, with less advanced HIV-1 disease, and therefore their mortality may have been related to their kidney disease rather than HIV-1 infection. If the pre-RRT temporal differences that we observed in HIV-1–related prognostic factors are nationally representative, it implies that analyses of USRDS data [11, 20], which lack HIV-1 prognostic factors, might actually be underestimating improvement in survival that has occurred in the HAART era relative to the pre-HAART era.
It is also important to note that, compared with the dismal prognosis for HIV-infected patients undergoing RRT in the late 1980s , overall survival has significantly improved in recent years. However, survival for those undergoing RRT remains significantly lower than for patients not infected with HIV undergoing RRT . In our study, the 1-year and 2-year survival during the HAART era were 63% and 43%, respectively, which is significantly lower than the reported 80% and 68%, respectively, in non—HIV-1–infected African Americans undergoing dialysis . Clearly, a more effective strategy is needed to offer a meaningful improvement in the health of this vulnerable population.
Serum albumin level, a measure of nutritional status, is a well recognized predictor of survival in patients undergoing dialysis [24–27], and in accord with previous reports, we observed a strong association between serum albumin concentration and survival during dialysis. Most notably, the impact of serum albumin on survival is rapid (within months) and is magnified with time (figure 2B). The effect of dialysis prescription and dose on nutrition and survival could not be assessed in this cohort. In addition, survival may have been influenced by other factors related to undergoing dialysis, such as modality, type of dialysis access (catheter, graft, or fistula), infections, level of anemia, and calcium phosphate balance.
In the present study, higher CD4 cell count was a significant predictor of survival during RRT. This is not unexpected, in light of the many epidemiological studies in the general HIV-infected population, with or without kidney disease, that show the relationship of CD4 cell counts and survival.
Prevention of kidney disease should be a primary goal in HIV-infected African American patients. Identifying patients at higher risk and treating aggressively with HAART are likely to provide positive impact on outcome. We have previously shown a correlation between increasing HIV load, decreasing CD4 cell count, and the occurrence of proteinuria and renal failure . Previous case reports have demonstrated a reversal of acute renal failure and histological changes on biopsy after treatment with antiretroviral therapy [29, 30]. In addition, several observational cohort studies have demonstrated improved renal survival in patients with HIV-associated nephropathy and a decreased need for renal replacement therapy for patients treated with HAART, compared with those not so treated .
There are several limitations to our study. The study was based at a single academic center in a predominantly poor urban area, which may not be representative of experiences in other settings. However, the growth of the HIV epidemic is most rapid in poor urban areas in the United States. A second limitation of our analysis was that we could not control for different reasons HAART was not being used, such as nonadherence, critically ill status, or poor access to care. Finally, we did not have information on dialysis prescription, dose, or adherence or other end-stage renal disease–related data, such as bone disease and anemia management, that may affect outcomes.
Our study has several implications for treatment of HIV-1–infected patients undergoing RRT. First, examining the barriers to use of health services and HAART in this vulnerable group of patients may provide effective means of improving clinical outcome. Second, comanagement of injection drug use in this population should be an integrative part of RRT. Third, nephrologists, HIV care providers, and dietitians caring for those patients should focus on nutritional status, identify causes of inadequate nutrition, and use corrective measures. In addition, we recommend adequate dialysis prescription, which is a key to improved nutrition in patients undergoing RRT.
We conclude that an HIV-infected person's use of HAART is associated with an ∼50% reduction in mortality during RRT. However, at a population level, the experience of our inner-city African American cohort demonstrates that HAART was dramatically underused, resulting in overall survival rates similar in the pre-HAART and HAART eras. Failure to use HAART likely results from a constellation of issues, including patient health knowledge, difficulty taking complicated HAART regimens during dialysis, poverty, active substance abuse, and clinician factors. Further investigation of barriers to HAART use among RRT patients should be aggressively pursued to maximize the survival benefit from both.
Financial support.National Institutes of Health (R01DA11602, R21AA15032, and K24DA00432 to R.D.M.; R01DA04334 and K23DA15616 to G.M.L.; and R01DA012568 to S.H.M.).
Potential conflicts of interest.All authors: no conflicts.
- kidney diseases
- survival analysis
- kidney failure, chronic
- antiretroviral therapy, highly active
- cd4 count determination procedure
- renal replacement therapy
- serum albumin
- intravenous drug user
- hiv infection
- cox proportional hazards models
- african american
- median survival time