Recently, d4T/ddI combination has been associated with the selection of thymidine analogue mutations (TAMs) in 50% of patients with a detectable viral load after 6 to 12 months of this bi-therapy (ALBI, STADI and BMS A1460 tests). We evaluated the rate of selection of the TAMs in naive patients with a viral load of > 200 copies/mL after: 6 months to 1 year of d4T/3TC bi-therapy (group 1); 1 year or more of a treatment including d4T/3TC (group 2); and more than 6 months of tri-therapy including d4T/ddI (group 3). The reverse transcriptase gene has ben studied in 33 patients in group 1, 17 patients in group 2 and ten patients in group 3. For the latter patients, the tri-therapies were as follows: d4T/ddI/PI (n = 5), d4T/ddI/NNRTI (n = 4), d4T/ddI/NRTI (n = 1). For the group 1 patients, at baseline, two patients already had TAMs. At M6, all the patients acquired the 3TC-associated mutations, M184V. Only one patient selected a MDR mutation profile (F116Y, Q151M). At M12, 26 of 33 plasmas were analysed. Only one patient selected one TAM (T215Y). For the group 2 patients, only three patients selected TAMs after more than 30 months of treatment. For the group 3 patients, at baseline, only one patient already harbored TAMs. None of the other patients had selected TAMs. In patients who received d4T/ddI/3TC, only the M184V, the 3TC-associated mutation, was selected. In conclusion, stavudine in association with 3TC selected a low rate of TAMs; in patients receiving a treatment including d4T/3TC, time of exposure to stavudine seemed to be an important parameter for the selection of TAMs; and in contrast to results obtained on d4T/ddI, tri-therapies including d4T/ddI did not select any TAMs, whatever the combination (NRTI, NNRTI, PI).