Predictive value of CD4 cell count nadir on long-term mortality in HIV-positive patients in Uganda.

OBJECTIVE: Although international guidelines recommend initiating antiretroviral therapy (ART) when a patient's CD4 cell count is ≤350 cells/μL, most patients in resource-limited settings present with much lower CD4 cell counts. The lowest level that their CD4 cell count reaches, the nadir, may have long-term consequences in terms of mortality. We examined this health state in a large cohort of HIV+ patients in Uganda.

DESIGN: This was an observational study of HIV patients in Uganda aged 14 years or older, who were enrolled in 10 major clinics across Uganda.

METHODS: We assessed the CD4 nadir of patients, using their CD4 cell count at initiation of ART, stratified into categories (,50, 50-99, 100-149, 150-249, 250+ cells/μL). We constructed Kaplan-Meier curves to assess the differences in survivorship for patients left-censored at 1 year and 2 years after treatment initiation. We used Cox proportional hazards regression to model the associations between CD4 nadir and mortality. We adjusted mortality for loss-to-follow-up.

RESULTS: Of 22,315 patients, 20,129 patients had greater than 1 year of treatment follow-up. Among these patients, 327 (1.6%) died and 444 (2.2%) were lost to follow-up. After left-censoring at one year, relative to lowest CD4 strata, patients with higher CD4 counts had significantly lower rates of mortality (CD4 150-249, hazard ratio [HR] 0.60, 95% confidence interval [CI]: 0.45-0.82, P = 0.001; 250+, HR 0.66, 95% CI, 0.44-1.00, P = -0.05). Male sex, older age, and duration of time on ART were independently associated with mortality. When left-censoring at 2 years, CD4 nadir was no longer statistically significantly associated with mortality.

CONCLUSION: After surviving for 1 year on ART, a CD4 nadir was strongly predictive of longer-term mortality among patients in Uganda. This should argue for efforts to increase engagement with patients to ensure a higher CD4 nadir at initiation of treatment.