A 4-gene signature predicts survival of patients with resected adenocarcinoma of the esophagus, junction, and gastric cardia

The incidence of esophageal and junctional adenocarcinoma has increased 6-fold in the past 30 years and 5-year survival remains approximately 20%. Current staging is limited in its ability to predict survival which has ramifications for treatment choices. The aim of this study was to generate and validate a molecular prognostic signature for esophageal adenocarcinoma.

The incidence of esophageal and junctional adenocarcinoma has increased 6-fold in the past 30 years and 5-year survival remains approximately 20%. Current staging is limited in its ability to predict survival which has ramifications for treatment choices. The aim of this study was to generate and validate a molecular prognostic signature for esophageal adenocarcinoma.

This study has generated a clinically applicable prognostic gene signature that independently predicts survival in an external validation cohort and may inform management decisions.

Gene expression profiling was performed and the resulting 42,000 gene signatures correlated with clinical and pathologic features for 75 snap-frozen esophageal and junctional resection specimens. External validation of selected targets was performed on 371 independent cases using immunohistochemistry to maximize clinical applicability.

A total of 119 genes were associated significantly with survival and 270 genes with the number of involved lymph nodes. Filtering of these lists resulted in a shortlist of 10 genes taken forward to validation. Four genes proved to be prognostic at the protein level (deoxycytidine kinase [DCK], 3'-phosphoadenosine 5'-phosphosulfate synthase 2 [PAPSS2], sirtuin 2 [SIRT2], and tripartite motif-containing 44 [TRIM44]) and were combined to create a molecular prognostic signature. This 4-gene signature was highly predictive of survival in the independent external validation cohort (0/4 genes dysregulated 5-year survival, 58%; 95% confidence interval [CI], 36%-80%; 1-2/4 genes dysregulated 5-year survival, 26%; 95% CI, 20%-32%; and 3-4/4 genes dysregulated 5-year survival, 14%; 95% CI, 4%-24% (P = .001). Furthermore, this 4-gene signature was independently prognostic in a multivariable model together with the existing clinical TNM staging system (P = .013). Conclusions: This study has generated a clinically applicable prognostic gene signature that independently predicts survival in an external validation cohort and may inform management decisions.