Type and prevalence of zidovudine (ZDV) resistance mutations in HIV-1-infected children in clinically stable condition and on ZDV monotherapy were analyzed to evaluate the effect of switching to didanosine (ddI) monotherapy or to ZDV plus ddI on the pattern of mutations and on the clinical outcome. Monthly clinical and laboratory controls for HIV-1 infection status were performed; at enrollment and every 4 to 6 months after treatment randomization mutant proviral sequences were evaluated in all the children, whereas viral burden was performed only in a small subgroup of patients randomly selected in each of the three treatment groups. ZDV resistance-associated proviral DNA mutations were defined as low-level resistance (LLR) mutations or medium/high-level resistance (MHLR) mutations; clinical outcome was considered as stable or deteriorating. Results showed that at entry into the study the duration of ZDV therapy was significantly correlated with the presence of mutations, and that the level of resistance given by mutations was associated with the severity both of symptoms and immunodeficiency. After randomization to treatment, in patients with mutations that confer LLR a better clinical outcome with ddI monotherapy than with ZDV plus ddI and ZDV alone was observed in the subsequent 6 months, whereas in patients with mutations that confer MHLR no significant difference among the three treatment groups was found. Data showed also that levels of viral burden at the time of changing therapy are related to clinical outcome if measured by plasma viral load. These results suggest that genotypic resistance assays, together with viral load, may prove useful for rational treatment decisions both at the start of therapy and with failure.