Due to the preferential selection of the fittest HIV mutants, drug-resistant variants are often overgrown by wild-type virus after treatment interruption. Our objective was to investigate the dynamics of the 103N mutation (which usually does not reduce HIV fitness) following the withdrawal of non-nucleoside reverse transcriptase inhibitors (NNRTIs). Patients who were found to have the 103N mutation at or after failure of a NNRTI were selected from an observational database. Two groups of patients were identified: one which continued antiretroviral treatment without NNRTIs (group A) and one which discontinued all anti-retrovirals after failure of an NNRTI (group B). Genotype was obtained by direct sequencing of the replicating plasma virus. Sixty-two subjects tested between July 1998 and December 2002 were included in the analysis, 39 in group A and 23 in group B. At the time of the first resistance test, median (IQR) CD4+ T-lymphocytes and HIV-RNA were 269 (150-449) cells/microL and 25,000 (9,600-83,300) copies/mL. In 31 (50%), 30 (48%), and one case (2%), the 103N mutation was selected by nevirapine, efavirenz, and by delavirdine, respectively. A total of 149 tests were analyzed, with a median (IQR) of 2 (2-3) tests/patient. The median (IQR) interval between failure of NNRTIs and the last resistance test was 11 (5-22) months. Overall, a reversion to wild-type at position 103 was observed in 23/62 (37%) subjects, 14/39 (36%) in group A and 9/23 (39%) in group B. In group A, 14/23 (61%) patients tested within 12 months, 10/16 (63%) of those tested between 12 and 24 months, and 12/14 (86%) of those tested beyond 24 months from NNRTI discontinuation had the 103N mutation. In group B, 14/20 (70%) patients tested within 12 months, 3/4 (75%) of those tested between 12 and 24 months, and none out of two tested beyond 24 months from NNRTI discontinuation had the 103N mutation. In conclusion, following NNRTI discontinuation, in the majority of patients HIV variants carrying the 103N mutation are not overgrown for long by wild-type quasispecies at this position. This suggests that the 103N mutation per se influences minimally the viral fitness in vivo.