Estimating (stage-)sojourn time for multiple cancer-sites: literature review and structured elicitation of expert beliefs

Preclinical cancer sojourn time, defined here as the duration a cancer is undetected but detectable, is important for understanding disease progression and evaluating screening policies. This study aims to robustly characterise empirical evidence and existing knowledge over mean sojourn times across 21 stageable tumour sites, including stage-specific preclinical cancer sojourn times and the sojourn time of circulating tumour DNA (ctDNA)-positive cancers. We updated an existing systematic review through to February 2025 to extract population-level empirical sojourn time estimates derived from mathematical models of primary screening data. To synthesise this heterogeneous literature, quantify uncertainty, and obtain estimates for cancer-sites lacking empirical evidence, we conducted a formal Structured Expert Elicitation involving 15 clinical experts. The elicitation was grounded on the systematic review results, supplemented by an evidence dossier that included survival data and outcomes from relevant ctDNA cancer studies. The systematic review revealed heterogeneity in existing literature, which focused on a small subset of screened cancers (e.g., breast, cervical, colorectal). The elicitation successfully generated comprehensive probability distributions of overall mean sojourn times for all 21 cancer-sites (representing the site of tumour origin), as well as stage-specific sojourn times and overall sojourn times for ctDNA-positive cancers across 14 cancer-sites. This study used robust methodology to quantitatively describe existing evidence and experts' beliefs on the sojourn time of multiple cancer-sites, also describing uncertainty. Such estimates are important for future evaluations of the clinical impact, potential for overdiagnosis and subsequent cost-effectiveness of emerging screening technologies, including multi-cancer detection tests.