Validation of Prognostic Indices in Locally Advanced and Metastatic Oesophago-gastric Cancer
Michael J Devlin1, Laura Feeney1, Peter Gallagher1, Jirhe Okugbeni2, Kyle Crawford1, Rachel Campbell1, Claire Harrison1, Colin Purcell1, Martin Eatock1, Richard C Turkington1,2
1Northern Ireland Cancer Centre, Belfast City Hospital, Belfast, Northern Ireland.
Survival outcomes in advanced oesophago-gastric adenocarcinoma are poor and there is a need to improve patient stratification to aid clinical decision-making.
Prognostic indices have been developed by the Royal Marsden Hospital (RMH)1, Japanese Clinical Oncology Group (JCOG)2 and Korea Cancer Center (KCC)3. We sought to validate each index in a western, non-trial oesophago-gastric adenocarcinoma population.
183 patients with locally advanced or metastatic oesophago-gastric adenocarcinoma were treated with Epirubicin, Cisplatin and 5-Fluorouracil/Capecitabine chemotherapy at the Northern Ireland Cancer Centre between 2007 and 2012. Survival analysis was performed using the Kaplan-Meier method with Hazard ratios (HR) calculated using the log-rank test.
For the RMH index the median survival rates for the good, moderate and prognostic groups were 11.5, 9.6 and 8.7 months respectively. Compared to the good risk group, the moderate risk group had a 1.4-fold (HR1.42; 95% CI 1.04-1.95, p=0.03), and the poor risk group an over 2-fold (HR 2.25; 95% CI 0.9-5.59, p=0.08) increased risk of death. For the JCOG index the median survival rates for the good, moderate and prognostic groups were 11.5, 9.2 and 5.0 months respectively. Compared to the good risk group, the moderate risk group had a 1.3-fold (HR1.32; 95% CI 0.96-1.8, p=0.09), and the poor risk group an almost 4-fold (HR 3.89; 95% CI 1.3-11.5, p=0.01) increased risk of death. The KCC index was unable to separate the patients into prognostic groups. For the RMH and JCOG indices the range of survival and risk of death between the prognostic groups was
We successfully validated the RMH and JCOG prognostic indices in a non-trial population. Each score did not differentiate between good, moderate and poor prognostic patients as effectively as previously published, indicating a need to adapt these indices for a non-trial population.