NSCLC are seldom reported. Method: Consecutive patients with curative resected NSCLC and confirmed EGFR mutations at Fudan University Shanghai Cancer Center from January 2007 to December 2017, were retrospectively enrolled. The initial recurrence sites were recorded and categorized into five groups: thoracic recurrence, brain recurrence, neck recurrence, abdominal recurrence, and bone recurrence. The indicators of overall and site-specific recurrence were identified using the Cox regression model, where a panel of routinely used IHC markers (including Her2, Ki67, TTF-1, CK20, CK7, CK5/6, p53, RRM1, NapsinA, p40, syn, Bcl-2, CDX2, ERCC1 and p63) were incorporated. A nomogram was developed based on variables selected in multivariate analysis. The bootstrapping method (1000 repetitions) was applied to internally validate the nomogram Result: After a median follow-up of 32 (range, 5-122) months, disease recurrence was observed in 197(37.1%) out of the 531 patients, with a median recurrence-free survival (RFS) of 19 (95% CI, 16.63-21.37) months. Most patients (n¼136; 69.0%) had thoracic recurrence, followed by brain recurrence (n¼41; 20.8%), bone recurrence (n¼41; 20.8%), abdominal recurrence (n¼14; 7.1%), and neck recurrence (n¼13; 6.6%). Sex, tumor size, Ki67, and N stage were independent indicators of thoracic recurrence. Tumor size, N stage, CK20, and Syn were independent indicators of brain recurrence. N stage and Ki67 were independent indicators of bone recurrence. N stage was the independent indicator of abdominal recurrence and neck recurrence. Tumor size, Ki67, CK20, and N stage were independently associated with overall recurrence, and thus a nomogram predicting the 1-, 2-, and 3-year RFS probability was developed based on these four factors. The concordance index (C-index) was 0.723 (95% confidence interval, 0.675 to 0.771) and the calibration curves displayed good agreement between the predicted RFS and the actual observation.