Adherence of 95% or more to unboosted protease regimens is required for optimal virologic suppression in HIV-1-infected patients. Whether the same is true for nonnucleoside reverse transcriptase inhibitor (NNRTI)-based therapy is unclear. To assess the relationship between adherence to NNRTI-based therapy and viral load in treatment-naive patients. Observational cohort study. Private-sector HIV and AIDS disease management program in South Africa. 2821 adults infected with HIV who began NNRTI-based therapy between January 1998 and March 2003 (2764 patients [98%] were enrolled after December 2000). Adherence was assessed by monthly pharmacy claims. The primary end point was sustained viral load suppression (<400 copies/mL) in 100% of recorded viral load measurements throughout follow-up. Secondary end points included time to initial viral load suppression and time to subsequent virologic failure (>400 copies/mL). The median follow-up period was 2.2 years (interquartile range, 1.7 to 2.7 years). The proportion of patients with sustained viral load suppression ranged from 13% (41 of 325 patients) in patients who filled less than 50% of antiretroviral drug prescriptions to 73% (725 of 997 patients) in those who filled 100% of antiretroviral drug prescriptions. Each 10% increase in pharmacy claim adherence greater than 50% was associated with a mean absolute increase of 0.10 in the proportion of patients with sustained virologic suppression (P < 0.001). Predictors for shorter time to virologic failure after initial suppression in multivariable Cox regression included CD4+ T-cell counts of 0.50 x 10(9) cells/L or less (hazard ratio, 1.60 [95% CI, 1.22 to 2.10] vs. CD4+ T-cell counts >0.20 x 10(9) cells/L), baseline viral load greater than 10(5) copies/mL (hazard ratio, 1.39 [CI, 1.14 to 1.70]), nevirapine-based regimen (hazard ratio, 1.43 [CI, 1.16 to 1.75]), and low pharmacy claim adherence (hazard ratio, 1.58 [CI, 1.48 to 1.69], per 10% decrease in adherence to 50%). Observational study with adherence stratification at study end and lack of standardized timing for outcome measurement. Virologic outcomes improve in a linear dose-response manner as adherence to NNRTI-based regimens increases beyond 50%.