Cancer Stage Compared With Mortality as End Points in Randomized Clinical Trials of Cancer Screening: A Systematic Review and Meta-Analysis

Abstract

Importance: Randomized clinical trials of cancer screening typically use cancer-specific mortality as the primary end point. The incidence of stage III-IV cancer is a potential alternative end point that may accelerate completion of randomized clinical trials of cancer screening.

Objective: To compare cancer-specific mortality with stage III-IV cancer as end points in randomized clinical trials of cancer screening.

Design, setting, and participants: This meta-analysis included 41 randomized clinical trials of cancer screening conducted in Europe, North America, and Asia published through February 19, 2024. Data extracted included numbers of participants, cancer diagnoses, and cancer deaths in the intervention and comparison groups. For each clinical trial, the effect of screening was calculated as the percentage reduction between the intervention and comparison groups in the incidence of participants with cancer-specific mortality and stage III-IV cancer.

Exposures: Randomization to a cancer screening test or to a comparison group in a clinical trial of cancer screening.

Main outcomes and measures: End points of cancer-specific mortality and incidence of stage III-IV cancer were compared using Pearson correlation coefficients with 95% CIs, linear regression, and fixed-effects meta-analysis.

Results: The included randomized clinical trials tested benefits of screening for breast (n = 6), colorectal (n = 11), lung (n = 12), ovarian (n = 4), prostate (n = 4), and other cancers (n = 4). Correlation between reductions in cancer-specific mortality and stage III-IV cancer varied by cancer type (I2 = 65%; P = .02). Correlation was highest for trials that screened for ovarian (Pearson ρ = 0.99 [95% CI, 0.51-1.00]) and lung (Pearson ρ = 0.92 [95% CI, 0.72-0.98]) cancers, moderate for breast cancer (Pearson ρ = 0.70 [95% CI, -0.26 to 0.96]), and weak for colorectal (Pearson ρ = 0.39 [95% CI, -0.27 to 0.80]) and prostate (Pearson ρ = -0.69 [95% CI, -0.99 to 0.81]) cancers. Slopes from linear regression were estimated as 1.15 for ovarian cancer, 0.75 for lung cancer, 0.40 for colorectal cancer, 0.28 for breast cancer, and -3.58 for prostate cancer, suggesting that a given magnitude of reduction in incidence of stage III-IV cancer produced different magnitudes of change in incidence of cancer-specific mortality (P for heterogeneity = .004).

Conclusions and relevance: In randomized clinical trials of cancer screening, incidence of late-stage cancer may be a suitable alternative end point to cancer-specific mortality for some cancer types, but is not suitable for others. These results have implications for clinical trials of multicancer screening tests.

Figure 1.. Reduction in Cancer Mortality and Incidence of Stage III-IV Cancer Among 41 Cancer Screening Trials

Positive numbers indicate reductions in mortality or late-stage cancer. Analyses are unweighted. Grey shading indicates 95% CIs for linear regression. For panels A-F, the diameters of the circles are scaled by the number of trial participants.

Figure 2.. Reduction in Cancer Mortality and Incidence of Stage IV Cancer Among 29 Cancer Screening Trials

Positive numbers indicate reductions in mortality or late-stage cancer. Analyses are unweighted. Grey shading indicates 95% CIs for linear regression. For panels A-F, the diameters of the circles are scaled by the number of trial participants. The number of trials is reduced compared with Figure 1 because some publications did not present numbers of stage III and IV cancers separately. The correlation is not presented for cancer types with 2 observations.