To evaluate whether fat-free mass index (FFMI) could be predictive of prolonged air leak (PAL) complicating video-assisted thoracoscopic (VATS) lobectomy for non-small-cell lung cancer (NSCLC). A retrospective study was conducted on the prospectively-maintained database in our institution between January 2015 and July 2017. The gender-specific median values of FFMI for males and females were applied as their respective cutoffs to stratify patients into low-FFMI group and high-FFMI group in initial univariable analyses. An effective multivariable logistic-regression analysis was then performed to demonstrate the predictive value of dichotomized FFMI. There were 1,091 surgical patients with NSCLC included (616 males and 475 females), with a PAL incidence of 14.6%. The median FFMI values among males and females were 17.3 and 14.6 kg/m2, respectively. PAL cases in both male (16.9±1.5 vs. 17.4±1.5 kg/m2; P=0.002) and female (14.0±0.9 vs. 14.6±1.1 kg/m2; P<0.001) groups had a significantly lower mean FFMI than that of non-PAL cases. The incidence of PAL was significantly increased in male patients with FFMI <17.3 kg/m2 (23.7% vs. 14.3%; P=0.003) and female patients with FFMI <14.6 kg/m2 (12.7% vs. 5.0%; P=0.003). Lower dichotomized FFMI was also significantly associated with prolonged time to air leak cessation and length of stay (LOS). Finally, multivariable logistic-regression analysis indicated that lower dichotomized FFMI [odds ratio (OR) =1.98; 95% confidence interval (CI): 1.33-2.96; P=0.001] could independently predict the occurrence of PAL. FFMI acts as an excellent categorical risk factor for PAL complicating VATS lobectomy and shows a much superior significance than body mass index (BMI) in terms of the prediction of PAL.