Single-round performance of colorectal cancer screening programs: a network meta-analysis of randomized clinical trials.

Abstract

Background: Demonstrating mortality reduction in new colorectal cancer (CRC) screening programs through randomized clinical trials (RCTs) is challenging. We systematically reviewed single-round program performance outcomes using a stepwise approach proposed by the World Endoscopy Organization CRC Screening Committee framework.

Methods: The MEDLINE, EMBASE, Central, and Ichushi Web databases were searched until October 28, 2024, to find RCTs comparing guaiac-based and immunochemical fecal occult blood testing (gFOBT and FIT), flexible sigmoidoscopy (FS), computed tomographic colonography (CTC), and total colonoscopy (TCS). Paired reviewers screened studies, extracted data, and assessed bias risk. A Bayesian random-effects network meta-analysis was conducted, and the certainty of evidence was evaluated using the Grading of Recommendations Assessment, Development and Evaluation approach. The primary outcome was advanced neoplasia (AN) detection, and the secondary outcomes were participation and colorectal cancer (CRC) detection, all during the first screening round.

Results: Eighteen RCTs (437,072 invitees) were included. The risk of bias was low or raised some concerns for screening participation, but it was high for detection outcomes. In the network meta-analysis of 15 RCTs not allowing crossover, the FIT-based program had a higher AN detection rate than the gFOBT-based program (relative risk [RR] 2.48; 95% credible interval [CrI] 1.52-4.21; moderate certainty). AN detection rates were not different in the CTC- (RR 1.01; CrI 0.43-2.23; very low certainty) and TCS-based (RR 1.03; CrI 0.54-1.78; low certainty) programs compared with the FS-based program. All the visualization modality programs had higher AN detection rates than the FIT-based program (FS: RR 2.13 [CrI 1.38-3.77]; CTC 2.16 [1.11-4.51]; and TCS 2.19 [1.43-3.48]; all with low certainty). Low event rates precluded definitive conclusions regarding CRC detection (very low to low certainty). The TCS-based program had the worst participation rate (very low to low certainty). Comparative data allowing crossover were limited.

Conclusions: This is the first network meta-analysis that evaluates program-level initial performance indicators. FIT-based programs likely detect more AN cases than gFOBT-based programs, while FS-, CTC-, and TCS-based programs may outperform FIT. Due to limitations in first-round results, long-term outcomes should be assessed after 10-15 years.