Clinical and lifetime natural history of oral cancer: implications for screening and surveillance
Menée à partir de données d'un essai portant sur 41 225 chiqueurs, fumeurs ou consommateurs d'alcool ayant participé à au moins un examen de dépistage ou à 2 examens consécutifs, cette étude évalue la sensibilité, la valeur prédictive positive et la valeur prédictive négative d'un examen visuel buccal pour détecter une lésion précancéreuse ou cancéreuse, estime le délai avant progression d'une lésion précancéreuse vers une lésion cancéreuse et conclut que l'intervalle entre deux examens de dépistage peut être porté à 10 ans lorsque deux examens consécutifs n'ont pas révélé d'anomalie
Background : We estimated the clinical and lifetime natural history of oral cancer to inform empirical screening and surveillance algorithms.
Methods : We used data from ever-chewers/smokers/alcohol drinkers in the screening arm of the Kerala Oral Cancer Screening Trial (1996-2010), who attended at least one screen (n = 41,225) or two consecutive screens (n = 28,821). The clinical natural history (sensitivity, positive predictive value [PPV] and complement of negative predictive value [cNPV]) was estimated using Kaplan-Meier methods. The multistate natural history across initiation of chewing/smoking/drinking, precancer, early-stage cancer, late-stage cancer, and competing mortality was estimated using semi-Markov models.
Results : Visual-inspection over 2 consecutive screens (median=3.5 years apart [IQR=2.9-3.9]) had sensitivity=89.2%, PPV=1.55%, and NPV=99.96% for prevalent oral cancers (n = 83). On the first screen, 10.0% were positive with a referable oral lesion/precancer, and 12.9% were positive on either screen. Screen-positivity on both screens provided 32-fold risk-stratification for incident cancers over 11 years of follow-up (PPV=4.13% vs. cNPV in persistent screen-negatives=0.13%). Screen-positives on either screen accounted for 66.1% of incident oral cancers through 11 years of follow-up. In the lifetime/multistate model, an estimated mean=18.2% of ever-smokers/chewers/drinkers would develop oral precancer and mean=6.04% would progress from precancer to early-stage cancer (mean sojourn=19.3 years). The estimated mean sojourn time from early-stage to late-stage oral cancer was 2.7 years.
Conclusions : Oral cancer screening intervals could be extended to 10 years for repeat negatives on 2 consecutive screens. The high PPV and short sojourn time from early-stage to late-stage cancer indicate the need for intensive surveillance among screen-positives with precancer.
Journal of the National Cancer Institute , résumé, 2026