Re: Prognosis research strategy (PROGRESS) 1: A framework for researching clinical outcomes
The PROGRESS group has provided an insightful overview of prognosis research. (1) An additional insight is that statistical models do not predict individual risk of a future outcome; that number does not exist. (2,3) Different statistical models using different information will assign an individual to different risk strata, resulting in a discordance of individual risk estimates. So statistical models allow medicine to be stratified, but not personalized. When the spectrum of risk is categorized, this is described as reclassification. The PROGRESS group provided an example of this much ado about nothing in a table in a subsequent publication. (4) This discordance was recently described as “Prognosis: A Variable Parameter” when the different information was available at different times. (5) Importantly, none of the potential clinical utility of prognostic models is reduced by this insight, only the illusion of precision.
1. Hemingway H, Croft P, Perel P, et al. Prognosis research strategy (PROGRESS) 1: A framework for researching clinical outcomes. BMJ 2012;345:e5595.
2. Kent DM, Shah ND. Risk models and patient-centered evidence. should physicians expect one right answer? JAMA 2012;307:1585-6.
3. Stern RH. Individual Risk. J. Clin Hypertens 2012;14:261-4.
4. Steyerberg EW, Moons KGM, van der Windt A, et al. Prognosis Research Strategy (PROGRESS) 3: Prognostic Model Research. PLOS Medicine 2013;10:e1001381
5. van der Schroeff MP, Steyerberg EW, Wieringa MH, et al. Prognosis: A Variable Parameter. Dynamic Prognostic Modeling in Head and Neck Squamous Cell Carcinoma. Head&Neck 2012;DOI 10.1002/hed
Competing interests: No competing interests