Background A model that quantifies the risk of peritoneal recurrence would be a useful tool for improving decision-making in patients undergoing curative-aim gastrectomy for gastric cancer (GC). Methods Five Italian centers participated in this study. Two risk scores were created according to the two most widely used pathologic classifications of GC (the Lauren classification and the presence of signet-ring-cell features). The risk scores (the PERI-Gastric 1 and 2) were based on the results of multivariable logistic regressions and presented as nomograms (the PERI-Gram 1 and 2). Discrimination was assessed with the area under the curve (AUC) of receiver operating curves. Calibration graphs were constructed by plotting the actual versus the predicted rate of peritoneal recurrence. Internal validation was performed with a bootstrap resampling method (1000 iterations).
ResultsThe models were developed based on a population of 645 patients (selected from 1580 patients treated from 1998 to 2018). In the PERI-Gastric 1, significant variables were linitis plastica, stump GC, pT3-4, pN2-3 and the Lauren diffuse histotype, while in the PERI-Gastric 2, significant variables were linitis plastica, stump GC, pT3-4, pN2-3 and the presence of signet-ring cells. The AUC was 0,828 (0.778-0.877) for the PERI-Gastric 1 and 0,805 (0.755-0.855) for the PERI-Gastric 2. After bootstrap resampling, the PERI-Gastric 1 had a mean AUC of 0.775 (0.721-0.830) and a 95%CI estimate for the calibration slope of 0.852-1.505 and the PERI-Gastric 2 a mean AUC of 0.749 (0.693-0.805) and a 95%CI estimate for the slope of 0.777-1.351. The models are available at www. perig astric. org. Conclusions We developed the PERI-Gastric and the PERI-Gram as instruments to determine the risk of peritoneal recurrence after curative-aim gastrectomy. These models could direct the administration of prophylactic intraperitoneal treatments.