Negative-Control Calibration and E-Value Bounding of Cancer-Survivor Mortality Contrasts: A Design-Aware Competing-Risks Analysis

Objective: Cardiovascular death among cancer survivors is widely studied in population surveys, yet the survivor-versus-non-survivor contrast those studies rest on is rarely tested for confounding, and a survey cohort of prevalent survivors is left-truncated by design. The objective of this article is to test that contrast itself. Methods: The methods we used built a nationally representative survivor cohort from the National Health and Nutrition Examination Survey (NHANES) 1999–2018 (3826 survivors among 52,398 examined adults) linked to the National Death Index, with 288 deaths from diseases of the heart and 497 cancer deaths, and paired survey-weighted Fine–Gray competing-risks models with designed probes of what these data cannot support. Results: Our results are that accidental death, a negative-control outcome, was raised among survivors (hazard ratio 1.78, 95% CI: 0.89–3.54) and equally among participants whose only malignancy was a skin cancer (2.01, 0.87–4.66). For cardiovascular death, that negative-control exposure reproduced the survivor contrast (0.87, 0.65–1.18 against 0.84, 0.71–1.00), leaving a calibrated contrast of 0.96 (0.70–1.32). Refitting on the time-since-diagnosis scale with delayed entry left the comorbidity index’s association with cardiovascular death unchanged (1.41, 1.22–1.64 against 1.40, 1.24–1.59). Net survival against internally modeled expected mortality, 0.897 (0.869–0.924) at 10 years, agreed with the recorded cancer-death incidence of 0.115 (0.102–0.127). Conclusions: The conclusions drawn are that what such a survey establishes about survivor mortality is narrower than it appears.