Figure 2 from Late-Stage Outcomes as Surrogates for Mortality in Cancer Screening Trials: A Systematic Review and Meta-analysis

Bubble plots to evaluate the strength of association between the trial-level effect of screening on cancer mortality and the effect on either incidence of late-stage cancer or the proportion of cancers that are late stage across all trials. These are the primary results from the meta-analysis. A, Incidence of late-stage cancer. B, Proportion of cancers diagnosed at late stage. Each data point in A and B is a trial arm comparison. C, Incidence of late-stage cancer. D, Proportion of cancers diagnosed at late stage. Each data point in C and D is from a class of screening test. The same type of plot is used for A–D. E, Incidence of late-stage cancer. The scales (1-RR) from C are inverted and SEs are added to better show uncertainty. F, Associated z-statistics from E. Each data point in E and F is from a class of screening test. In all plots, the bubble sizes are proportional to the inverse variance of the screening effect on the mortality outcome. a, the intercept from the regression analysis; b, the regression coefficient from the regression analysis; ρ, the estimated correlation coefficients. Dotted lines: axes of no screening effect on late-stage and mortality outcomes. Dashed line: diagonal of equal screening effect on late-stage and mortality outcomes. FOBT, fecal occult blood test; FS/C, flexible sigmoidoscopy/colonoscopy; LDCT, low-dose CT.