Background:  We asked whether “non-HIV” biomarkers of anemia, liver injury, renal injury, and chronic viral hepatitis improve differentiation of mortality risk and whether these biomarkers are associated with HIV markers.

Methods:  We identified veterans with HIV infection initiating combination ART (cART) within the Veterans Health Administration between January 1, 1997 and August 1, 2002. Association between HIV and “non-HIV” biomarkers was tested using nonparametric tests. After splitting the data into development and validation sets, nested multivariable Poisson models were fitted to HIV markers (CD4 cell count, HIV RNA, AIDS-defining conditions) and to “non-HIV” biomarkers (hemoglobin, transaminases, platelets, creatinine, and hepatitis B and C serology). All models were also adjusted for age and substance abuse or dependence. C statistics and quintiles of risk were compared. Sensitivity analyses employed inverse weighting for the propensity to be missing and included adjustment for year of cART initiation.

Results:  Of 13,586 veterans initiating cART during this interval, 9789 (72%) had complete data and 2566 of these died. Subjects were predominantly black (51%), male (98%), with a median age of 45 years. HIV biomarkers were associated with “non-HIV” biomarkers (p <0.0001). For example, the correlation with CD4 cell count was 0.39 for hemoglobin and –0.21 for an index of liver injury (FIB 4). In development and validation sets HIV biomarkers (C statistics in both, 0.69) and “non-HIV” biomarkers (C statistics, 0.72, 0.71) discriminated mortality. When models were combined, discrimination improved (C statistic in both 0.74, p <0.0001) resulting in better differentiation of risk. For example, among those at highest risk (fifth quintile) mortality rates increased from 13.9, 95%CI 12.7 to 15.1 using the HIV biomarkers only to 17.1 95%CI 15.6 to 18.7 deaths/100 person-years when “non-HIV” markers were also included. Findings were robust after adjusting missing data and year of cART initiation.

Conclusions:  “Non-HIV” biomarkers improve differentiation of mortality risk achieved by HIV markers, are strongly associated with HIV biomarkers, and therefore likely reflect HIV pathology. Combined with those of the SMART study, our findings underscore the overlapping nature of pathologic injury in HIV infection. After further validation, a combined index of HIV and “non-HIV” biomarkers may prove a superior management tool and surrogate endpoint for clinical trials.