Are non-cancer diagnoses missed opportunities in diagnosing cancer? Observational case-control studies

Background Initial non-cancer diagnoses that plausibly explain cancer symptoms may be underexplored missed opportunities in cancer diagnosis. Aim To characterise "interim diagnoses" representing potential missed opportunities to diagnose cancer; namely, non-cancer conditions diagnosed before cancer that are associated both with cancer and better and/or worse cancer outcomes. Design and Setting Observational case-control and case-only analyses of registry-linked Clinical Practice Research Datalink data from English primary care. Methods Cancer cases (cancer registry diagnoses, 1/1/2012-31/12/2018) were matched to cancer-free controls on birth year, gender and general practice. Incident non-cancer diagnoses sharing cancer symptoms were sought in the year before the case's diagnosis. Conditional logistic regression characterised non-cancer diagnoses associated with cancer (Bonferroni-corrected p<0.05). Associations between non-cancer diagnoses and cancer outcomes were tested using logistic regression (stage, route to diagnosis) and parametric regression accelerated failure-time models (1-year survival). Results Analyses of 223,422 patients yielded evidence of 37 "interim diagnoses" across 7 cancer sites that were consistently associated with worse (n=18) or better (n=8) cancer outcomes, or a mixture of worse and better outcomes within or between cancer sites (n=11). For example, gastro-oesophageal reflux disease, gastritis, oesophagitis, or peptic ulcer were associated with worse outcomes in pancreatic, colorectal and ovarian cancers, with better outcomes in stomach cancer, and a mixture in oesophageal cancer. Conclusion We characterise 37 "interim" diagnoses as potential missed opportunities in cancer diagnosis. Safety-netting around these non-cancer diagnoses, focussing on management of clinical uncertainty and robust patient follow-up, may help avoid missed opportunities to diagnose cancer and improve outcomes.