Abstract
Purpose :
Local recurrence remains a major problem for the patients with eyelid sebaceous gland carcinoma (SGC). To perform Mohs micrographic surgery (MMS) in eyelid SGC is controversial. To develop and validate a nomogram for individualized recurrence prediction in patients with eyelid SGC. Furthermore, to assess the benefit from MMS in preventing recurrence.
Methods :
A retrospective cohort study. The study included 238 consecutive patients (training cohort: n=167; validation cohort: n=71) with eyelid SGC. Least absolute shrinkage and selection operator (LASSO) regression was applied to select the features for nomogram. The predictive accuracy, discrimination ability and clinical usefulness of this model were determined by concordance index (C-index), calibration curve and decision curve analysis (DCA), respectively. And these characteristics were compared with TNM staging system. The results were validated with bootstrap resampling and an independent cohort. Multivariate logistic regression was used to confirm the efficacy of MMS in preventing relapsed SGC.
Results :
This nomogram exhibited satisfactory discrimination ability and good calibration for both training (C-index:0.94) and validation (C-index:0.89) cohorts. The discrimination ability compared significantly favorable than that of TNM staging (C-index: training cohort:0.68, validation cohort:0.59, all p<0.001). DCA demonstrated that this nomogram was clinically useful. The exclusion of MMS from the nomogram sharply reduced prognostic value (C-index: training cohort:0.80, validation cohort:0.73, all p<0.001). Multivariate logistic regression consistently proved that initial treatment with MMS was the only independent predictor for recurrence, in both training (OR:0.02, p<0.001) and validation (OR:0.03, p<0.001) sets.
Conclusions :
This nomogram provides improved individualized estimates of recurrence after primary excision for Chinese patients of eyelid SGC. This model may promote precision medicine in clinical application and affect therapeutic decisions.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.