Résumé
Background: Data on determinants of long-term disease progression in HIV-infected patients on antiretroviral therapy (ART) are limited in low and middle-income settings.
Methods: Effects of current CD4 count, viral load and haemoglobin and diagnosis of AIDS-defining events (ADEs) after start of combination ART (cART) on death and new ADEs were assessed using Poisson regression, in patient aged >= 18 years within a multi-centre cohort in Thailand.
Results: Among 1,572 patients, median follow-up from cART initiation was 4.4 (IQR 3.6-6.3) years. The analysis of death was based on 60 events during 6,573 person-years; 30/50 (60%) deaths with underlying cause ascertained were attributable to infections. Analysis of new ADE included 192 events during 5,865 person-years; TB and Pneumocystis jiroveci pneumonia were the most commonly presented first new ADE (35% and 20% of cases, respectively). In multivariable analyses, low current CD4 count after starting cART was the strongest predictor of death and of new ADE. Even at CD4 above 200 cells/mm 3, survival improved steadily with CD4, with mortality rare at >= 500 cells/mm 3 (rate 1.1 per 1,000 person-years). Haemoglobin had a strong independent effect, while viral load was weakly predictive with poorer prognosis only observed at >= 100,000 copies/ml. Mortality risk increased following diagnosis of ADEs during cART. The decline in mortality rate with duration on cART (from 21.3 per 1,000 person-years within first 6 months to 4.7 per 1,000 person-years at >= 36 months) was accounted for by current CD4 count.
Conclusions: Patients with low CD4 count or haemoglobin require more intensive diagnostic and treatment of underlying causes. Maintaining CD4 >= 500 cells/mm(3) minimizes mortality. However, patient monitoring could potentially be relaxed at high CD4 count if resources are limited. Optimal ART monitoring strategies in low-income settings remain a research priority. Better understanding of the aetiology of anaemia in patients on ART could guide prevention and treatment.