A nomogram to predict distant metastases after multimodality therapy for patients with localized esophageal cancer

Background: Among patients with localized esophageal cancer (LEC), 35% or more develop distant metastases (DM) as first relapse, most in the first 24 months after local therapy. Implementation of novel strategies may be possible if DM can be predicted reliably. We hypothesized that clinical variables could help generate a DM nomogram. Patients and Methods: Patients with LEC who completed multimodality therapy were analyzed. Various statistical methods were used, including multivariate analysis to generate a nomogram. A concordance index (c-index) was established and validated using the bootstrap method. Results: Among 629 patients analyzed (356 trimodality/ 273 bimodality), 36% patients developed DM as first relapse. The median overall survival from DM was only 8.6 months (95% CI, 7.0-10.2). In a multivariate analysis, the variables associated with a higher risk for developing DM were poorly differentiated histology (hazard ratio [HR], 1.76; P<.0001), baseline T3/T4 primary (HR, 3.07; P=.0006), and baseline N+ LEC (HR, 2.01; P<.0001). Although variables associated with a lower risk for DM were age of 60 years or older (HR, 0.75; P=.04), squamous cell carcinoma (HR, 0.54; P=.013), and trimodality therapy (HR, 0.58; P=.0001), the bias-corrected c-index was 0.67 after 250 bootstrap resamples. Conclusions: Our nomogram identified patients with LEC who developed DM with a high probability. The model needs to be refined (tumor and blood biomarkers) and validated. This type of model will allow implementation of novel strategies in patients with LEC.