Figure 5 from Late-Stage Outcomes as Surrogates for Mortality in Cancer Screening Trials: A Systematic Review and Meta-analysis

Exploratory evaluation of the utility of the 95% CI RR for late-stage incidence as a “forecast” of the final trial result for cancer-specific mortality. A, The observed RR for mortality (point estimates) versus the “predicted” 95% CI for the RR based on the observed incidence of late-stage cancer in the intervention versus the control arm. B, The data from A are transformed so that observed – predicted mortality is shown by trial, ordered by where on the prediction the observed mortality fell (mortality at top end of range at top and bottom end of range at bottom of the plot). Estimates marked red are from trials in which the observed mortality exceeds predicted range based on the late-stage outcome; those marked green are trials in which the observed mortality is lower than predicted; and observed mortality RR was included within the 95% CI range of the late-stage outcome for all the other trials shown in black. Wide confidence intervals are due to inclusion of smaller trials.