We aimed to examine whether the duration of respiratory events and related oximetric values are associated with outcomes of multilevel upper airway surgery in patients with moderate-severe obstructive sleep apnea (OSA). The records of patients with a preoperative apnea-hypopnea index (AHI) >15 events/h, who underwent uvulopharyngopalatoplasty plus tongue base suspension with or without septoplasty between 2012 and 2014, were reviewed retrospectively. If the postoperative 6th month AHI was <20 events/h with at least a 50 % reduction from preoperative rates, the outcome of surgery was regarded as successful, otherwise, as failure. To calculate factors predictive of surgical outcomes, a receiver operating characteristic (ROC) analysis was performed. Logistic regression analyses were utilized to obtain the Odds ratio (OR) and 95 % confidential interval (CI). In total, 82 patients were enrolled in the study. Sixty-one patients (74.4 %) met the success criteria. The mean obstructive apnea duration (OAD) was the sole variable with a significant and satisfactory area under the curve (AUC) value [AUC (95 % CI) = 0.719 (0.597-0.842), p = 0.003]. The cutoff value was found to be 26.75 s with 71.4 % sensitivity, 72.1 % specificity, 88.0 % positive predictive value, and 46.9 % negative predictive value. Univariate analysis revealed an association between surgical failure and mean OAD > 26.75 s, total apnea duration, lowest SaO2, mean SaO2, mean O2 desaturation, and oxygen desaturation index, although only mean OAD > 26.75 s remained to be an independent predictor for unfavorable outcome after adjustment for other confounders in multivariate analysis [OR (95 % CI) = 3.92 (1.08-14.17), p = 0.041]. The current study suggests that OSA patients having longer OAD are in the risk of having surgical failure.