Although the importance of the initial regimen in highly active antiretroviral therapy (HAART) has been emphasized, the effect on clinical outcome of early modification of the initial HAART regimen has not been well-defined. The analysis included antiretroviral-naive HIV patients who started HAART between 1999 and 2005 at a university hospital. Early modification was defined as any drug change within 2 months of starting HAART. The effect of early regimen modification on the occurrence of new AIDS-defining illness or death was assessed using Kaplan-Meier survival estimates, and hazard ratios were estimated using Cox's proportional hazards regression model. Of 398 patients beginning HAART, 21% experienced early modification of their regimen, the most common reason being gastrointestinal toxicity (49%). After adjusting for risk factors for occurrence of a new AIDS event or death, identified by univariate analysis, the hazards ratio contrasting early modification with maintenance of initial HAART regimen was 3.06 (95% confidence interval, 1.36-6.90; p = 0.007). There were significant differences between the AIDS event-free survival curves of patients in clinical categories B or C with or without early modification of initial HAART regimen (p = 0.0002), but no significant difference in patients in clinical category A (p = 0.706). A considerable proportion of patients who started HAART changed treatment regimen mainly due to intolerance early after start of HAART. Early modification of an initial HAART regimen was associated with poor clinical outcome in HIV patients in the advanced clinical categories.