Background:  It is controversial whether use of HAART in primary HIV infection has long-term benefit.

Methods Since 1992, individuals with primary HIV infection were enrolled into an observational cohort study. After 1996, some subjects received HAART at the discretion of their primary care providers. We evaluated the effect of HAART on clinical outcomes, including AIDS-defining opportunistic infections, a pre-defined group of HIV-related diagnoses, and time to HAART-related adverse events. Treatment groups began HAART ≤30 days (group 1), 31 to 180 days (group 2), or >180 days (group 3) from the estimated HIV infection date and were compared to historical and contemporary controls. We analyzed time to first diagnosis >4 weeks after HIV infection. Cox proportional hazard models were adjusted for age, gender, race/ethnicity, acute retroviral symptom severity, and baseline HIV levels.

Results:  A total of 243 men and 7 women contributed 925 person-years. Groups 1, 2, and 3 had 41, 81, and 34 subjects, respectively, and there were 29 historical and 65 contemporary control subjects. Group 3 subjects began HAART a median of 652 days (IQR 351 to 2553) from HIV infection. Of 17 subjects, 16 experienced opportunistic infections and 6 died. The 161 HIV-related diagnoses included herpesvirus infections (n = 38), warts (n = 26), sinusitis (n = 23), dermatitis (n = 19), candidiasis (n = 18), and bronchitis (n = 11). Median time in years (IQR) to diagnosis was 3.8 (2.9 to 7.8) for group 1, 5.6 (3.9 to 6.6) for group 2, 7.3 (4.4 to 10.2) for group 3, 4.7 (3.4 to 6.7) for contemporary controls, and 1.8 (1.1 to 2.9) for historical controls. Compared to contemporary controls, adjusted hazard ratios were 0.62 (95%CI 0.29 to 1.3, p = 0.2) for group 1, 0.58 (0.33 to 1.0, p = 0.06) for group 2, 0.23 (0.11 to 0.45, p <0.001) for group 3, and 2.8 (1.4 to 5.6, p = 0.003) for historical controls. Groups 2 and 3 had a slight decrease in risk of grade 2-4 adverse events than group 1.

Conclusions:  The differing outcomes in historical versus contemporary controls underscores the need for understanding limitations of control groups. Subjects treated acutely did not have significantly less risk of HIV-related diagnoses than untreated contemporary controls, but this result may be biased by association of lower pre-treatment CD4 counts and higher HIV RNA levels with both use of HAART and poorer prognosis. Historical controls may be a more representative population for comparison, but differences in risk could be attributed to factors other than HAART.