Currently submitted to: Journal of Medical Internet Research
Date Submitted: Oct 18, 2026
Open Peer Review Period: Oct 19, 2026 - Dec 14, 2026
(currently open for review)
Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
Prediction Models for Immune Checkpoint Inhibitor Efficacy in Gastric Cancer: A Systematic Review and Meta-analysis
ABSTRACT
Background:
Treatment efficacy varies substantially among patients with gastric cancer receiving immune checkpoint inhibitors (ICIs). Although prediction models are increasingly being developed, their performance, methodological quality, and clinical applicability have not been systematically evaluated.
Objective:
We aimed to comprehensively assess existing models for predicting the efficacy of ICI-based therapy in gastric cancer.
Methods:
PubMed, Embase, the Cochrane Library, Web of Science, MEDLINE via EBSCOhost, and Ovid were searched from inception to June 25, 2026. Reporting followed PRISMA 2020, extraction was guided by CHARMS, and risk of bias was assessed using PROBAST. AUCs were synthesized in an exploratory record-level Sidik–Jonkman random-effects analysis with Knapp–Hartung adjustment. Standard errors were approximated from AUCs and extracted analysis sample sizes. Subgroup and leave-one-record-out analyses were conducted.
Results:
Twelve reports contributed 80 model-performance records. Seven records with a reported AUC of 1.000 were summarized descriptively; 73 records were pooled. The pooled AUC was 0.80 (95% CI, 0.77–0.82; I²=88.29%). Pooled AUCs were 0.83 (95% CI, 0.79–0.86) for machine-learning models, 0.82 (95% CI, 0.78–0.87) for logistic regression models, 0.80 (95% CI, 0.66–0.94) for deep-learning models, and 0.68 (95% CI, 0.65–0.71) for single-biomarker ROC analyses. The seven-category comparison was significant (P<0.001), but categories were uneven and performance records were correlated. Leave-one-record-out estimates ranged from 0.79 to 0.80. All 12 reports had high overall risk of bias. Calibration was explicitly reported for 2 records, and complete matched contingency tables were available or reconstructable for 19.
Conclusions:
Published models provide preliminary evidence on discrimination among patients receiving ICI-based therapy, but high risk of bias, heterogeneous outcomes, correlated performance records, approximate standard errors, and limited external validation constrain interpretation. The evidence does not establish clinical utility or identify a model suitable for selecting ICIs over alternative treatments. Clinical Trial: PROSPERO CRD420261526885
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.