297 Background: The peritoneum is a common and devastating site of metastases for several types of gastrointestinal cancers (GI). Patients with PM and its manifestations, such as ascites, are frequently excluded from enrolling on clinical trials of GI cancers. While exclusion criteria of clinical trials are designed to minimize the likelihood of harm for patients considering enrollment, routine exclusion criteria without a strong rationale may present downstream harm as novel therapies are used in the clinic without specific data on the PM population. We sought to characterize the frequency of PM as an exclusion criterion from trials registered on clinicaltrials.gov. Methods: A natural language processing (NLP) algorithm was designed and validated to identify all interventional clinical trials of studies of gastrointestinal cancers registered on clinicaltrials.gov from 1/1/2014-12/31/2024. Validation was performed by expert review of a random sample of 100 selected studies to determine the receiver operator characteristics of the algorithm. We analyzed data by the proportion of studies with PM exclusion criteria. Linear regression and spearman modeling were used to identify trends over time by date of clinical trial registration. Subset analyses were conducted by upper (esophageal, gastric) and lower (colon, rectal, and anal) GI cancers. Results: Data from clinicaltrials.gov was obtained on 5/21/2025. A total of 88761 studies were identified; these were sub-stratified to 8265 Lower GI and 5230 Upper GI studies, of which 3453 and 2773, respectively, utilized interventional drugs and were included in the analysis. During validation of 100 randomly selected studies, we found 2 false positives and 4 false negatives (yielding a sensitivity and specificity of 88% and 95%, respectively). Our rate of inappropriate inclusion was 2%. Of the included studies, we found that over a 15-year period, the rate of Peritoneal metastases exclusion was 18.9% for lower GI studies and 22.9% for upper GI studies. An overall positive linear association was identified over time for both lower GI studies (slope 1.08%/year, R2=0.32, p=0.05) and upper GI studies (2.51 %/year, R2=0.89, p < 05). Spearman correlation was significant for the trend in both lower and upper GI studies (rs = .55 P < .05 e-02 and rs= .98 p < 0.05 e-08, respectively). Conclusions: PM is a common exclusion criteria in clinical trials of GI cancers and appears to be increasing over time. Without a compelling rationale for exclusion, the routine exclusion of PM from patients with GI cancers may limit trial candidacy and potentially harm external validity as therapies become available in the clinic. We urge thoughtful consideration regarding PM as routine exclusion criterion for trials of GI cancers.
Brown et al. (Sat,) studied this question.