Upper GI
Kyle J. Hitscherich, DO
Clinical Research Fellow
National Institutes of Health
Gaithersburg, Maryland, United States
Disclosure: Disclosure information not submitted.
Malignant potential of Gastrointestinal Stromal Tumors (GISTs) is based on characteristics (size, location, etc.) which have been used to create risk calculators to predict recurrence. The most utilized nomogram (Gold 2009) was applied to a patient population from a statewide academic community-based cancer center to evaluate performance compared to the Memorial Sloan Kettering (MSK), Mayo Clinic and National Spanish Registry (GEIS) populations.
Methods:
A retrospective review of operative pathology reports for the past 14 years (January 2005 – May 2019) identified all patients diagnosed with GISTs. Patient demographics, pathology characteristics and recurrence free survival (RFS) were compiled. Those without recurrence or death were censored at the date of their last follow-up within hospital system. The nomogram prediction of RFS was estimated using methods described in the original publication. Concordance probabilities for risk stratified quartiles were calculated based on nomogram scores to test performance of the nomogram against our Kaplan-Meier curve.
Results:
A total of 195 patients were treated for GISTs at our institution within the study timeline. Average age at presentation was 64 years (21-90). One patient was lost to follow up. The most prevalent tumor location was in the stomach (68%), and overall mitotic index was preferentially less than 5 mitosis per HPF (77%). Microscopically negative surgical margins were obtained in 93% of resections. Average tumor size on final pathology was 3.4cm (0.2-29).
Within our patient population, 16 developed recurrence with 1 mortality. Kaplan-Meier estimated RFS rates were 91% at 2 years and 88% at 5 years with a mean disease-free survival of 8.8 years. Concordance probability of the nomogram with our data was 0.82 (95% CI: 0.71-0.92), compared to the MSK (0.78), GEIS (0.76) and the Mayo (0.80) populations. In this study, the nomogram predicted a worse 2- and 5-year RFS particularly for the highest risk malignant GISTs (0.82 v 0.38 and 0.76 v 0.21).
Conclusions:
The concordance of the predictive nomogram was similar in our population to other published series. However, the nomogram overpredicts recurrence rates for the highest risk quartile of GIST patients.