STUDY LIMITATIONS
The data of this study extracted from the normal medical record which can veritably reflect the outcome of the current second-line ART management. As a retrospective study, selection bias was possibly introduced because of the fact that patients with incomplete records would be excluded. Some important variables associated with mortality like the viral load and body mass index when switch to second-line ART which were not recorded for most of the patients. Adherence of ART was confirmed variable don’t record in the normal medical record.