Six RCTs were included in the review (n=550). Only one of the six studies described randomisation method and allocation concealment.
There was no significant difference between treatment groups for viral suppression using the ITT analysis (pooled RD 3%, 95% CI -4 to 9%) or the on-treatment analysis (RR 3%, 95% CI -2 to 7%). No heterogeneity was present in either analysis (I2=0 and 7%).
There was no significant difference between treatment groups for mean CD4 count, cholesterol or triglycerides. Significant heterogeneity was found between trials that evaluated cholesterol (I2=98.4%, p<0.00001) and triglycerides (I2=92.7%, p<0.00001). None of the study aspects investigated in subgroup analyses were associated with heterogeneity. Amelioration of lipodystrophy was described in the nevirapine group of two out of six RCTs; four out of six did not detect any difference between treatments. Nevirapine treatment was associated with significantly better QoL scores than continuation with protease inhibitor (WMD 1.40, 95% CI 1.22 to 1.58, p<0.00001, I2=0; two RCTs).
There was no significant difference in overall withdrawal between treatment groups, however, nevirapine therapy was associated with grade 3 or 4 liver toxicity that led to drug discontinuation (RD 7%, 95% CI 3 to 12%, p=0.0009; three RCTs). Liver toxicity was found in 7% (13 out of 175) of patients who took nevirapine. There were no deaths due to adverse events.
The funnel plot suggested that publication bias may have been present for the outcome virological suppression by intention-to-treat analysis (p=0.046), but not by on-treatment analysis or outcome withdrawals due to adverse events.