Incidence and Risk Factors of First-Line HAART Discontinuation: Is it Worth Choosing Competing Risk or Standard Survival Approaches?

International audienceObjectives: To estimate the incidences of first-line HAART discontinuation (for intolerance, treatment failure or treatment simplification) and their risk factors by standard survival (1-KM, Cox model) or competing risk approach (CIF, Fine-Gray model) in HIV infected patients. Methods: We studied 1136 patients receiving first-line Highly Active Antiretroviral Therapies (HAART), aged over 18 years, from the Dat'AIDS cohort, Toulouse, France, between January 2000 and June 2008. Cumulative incidence was estimated with 1-KM and CIF estimators and risk factors with Cox and Fine-Gray models. Results: There were 265 discontinuations for intolerance, 136 simplifications, 101 treatment failure and 274 other reasons. One year incidences were 19.0% versus 16.8%, 8.0% versus 6.0%, 6.3% versus 4.8% and 20.0% versus 17.3%, with the estimators 1-KM and CIF, respectively. For intolerance, both models identified similar risk factors. For risk factors of simplification or treatment failure, results differed by the model. Conclusions: As expected, the 1-KM overestimates the incidence of treatment discontinuation. For early and frequent events such as intolerance, the Cox and the Fine-Gray models appear to give similar results. For late and rare events, potentially exposed to competing risk, results differed. The common or specific nature of a factor may also play a role