This study aimed to evaluate the safety of antiretroviral treatment interruption (TI) in HIV‐infected patients who started treatment based on earlier guidelines, and to identify baseline factors predictive of the time to reach fixed criteria for treatment resumption. Prospective, open‐label, multicenter trial. Patients were eligible if they had a CD4 cell count >350/mm3 and plasma HIV RNA <50,000 copies/ml when they first started antiretroviral therapy (ART); and if they had a CD4 count >450/mm3 and stable plasma HIV RNA <5,000 copies/ml for at least 6 months prior to enrolment. The criteria for ART resumption were a CD4 cell count <300/mm3 and/or a CDC stage B or C event. 116 patients had received ART for a median of 5.3 years. The median CD4 cell count and plasma HIV RNA values at inclusion were 809/mm3 and 2.6 log copies/ml, respectively. Median HIV DNA load at inclusion was 2.3 log copies/106 peripheral blood mononuclear cells (PBMCs). Thirty‐six months after TI, 63.9% of the patients had not yet reached the criteria for ART resumption, and 55.9% of patients had not resumed ART. In Cox multivariable analysis, a high HIV DNA level at TI, a low CD4 nadir, and pre‐existing AIDS status were the only significant risk factors for reaching the criteria for ART resumption (hazards ratio: 2.15 (1.02–4.53), 4.59 (1.22–17.24), and 5.74 (1.60–20.56), respectively). Patients who started ART with a CD4 cell count above 350/mm3 were able to interrupt treatment for long periods without a high absolute risk of either AIDS or severe non‐AIDS morbidity/mortality. A high PBMC HIV DNA level at TI was a strong predictor for more rapid treatment resumption. J. Med. Virol. 82:1819–1828, 2010. © 2010 Wiley‐Liss, Inc.