The objective of this study was to evaluate how much limb fat is needed to be lost for lipoatrophy to become clinically evident. Antiretroviral drug-naive patients from a randomized trial comparing stavudine or abacavir plus lamivudine and efavirenz, who had subjective assessment to detect clinically evident lipoatrophy (standardized questionnaire) and objective measurements of limb fat (dual X-ray absorptiometry) at baseline, 48 weeks, and 96 weeks were included. ROC curves were used to assess the sensitivity and specificity of several cut-off values of absolute and percent limb fat loss for diagnosing lipoatrophy. Of 54 patients included, 13 (24%) had subjective lipoatrophy at 96 weeks. After 96 weeks, median limb fat change was -2.3 kg (interquartile range: -5.2, +0.2) and 0.4 kg (interquartile range: -7.2, +3.4) in patients with and without lipoatrophy, respectively. Median percent limb fat change was -45.5% (interquartile range: -78.0, +3.7) and 5.5% (interquartile range: -62.8, +95.6), respectively. The cut-off values of absolute and percent limb fat loss showing the best sensitivity and specificity values were -1.5 kg (sensitivity, 77%; specificity, 76%) and -30% (sensitivity, 85%; specificity, 73%). At least 30% limb fat is needed to be lost in HIV-infected patients for lipoatrophy to become clinically evident.