A model of the natural history of screen-detected prostate cancer, and the effect of radical treatment on overall survival

Abstract

The lead time and over-detection associated with prostate-specific antigen (PSA) screening, and generational improvements in all-cause mortality, make prostate cancer outcome studies from the pre-PSA era difficult to interpret in a contemporary setting. We developed a competing-risks hazard model to estimate the natural history of screen-detected prostate cancer, and the impact of radical treatment on overall survival. The model of hazard of mortality was fitted to clinical outcome data from the pre-PSA era, and the effects of screening, generational mortality improvements and radical treatment were incorporated. Sensitivities to the choice of baseline data and values of key parameters were assessed. Lead-time estimates in men diagnosed aged 55-59 years were 14.1, 9.3 and 5.0 years for men with Gleason scores <7, 7 and >7, respectively, assuming biennial screening with 100% attendance. Central estimates of 15-year prostate cancer mortality for conservative management of screen-detected prostate cancer ranged from 0 to 2% for Gleason scores <7, 9 to 31% for Gleason score 7 and 28-72% for Gleason scores >7. For men aged 55-59 years at diagnosis, the predicted absolute 15-year survival benefit from curative treatment was 0, 12 and 26% for men with Gleason scores <7, 7 and >7, respectively. Estimates of the survival benefit of radical treatment were relatively insensitive to values of key parameters. The case for curative treatment, rather than conservative management, of screen-detected localised prostate cancer is strongest in men with high-grade disease. This conclusion contrasts with current patterns of care.

Figure 1

Overview of methods.

Figure 2

Projections of survival over 15 years under conservative treatment, using the central model assumptions. Black areas represent deaths due to prostate cancer, grey areas represent deaths due to other causes.