Risk grouping versus risk continuum in patients with clinically localized prostate cancer: a taxometric test

Two commonly used risk estimation approaches for clinically localized prostate cancer are nomograms and risk grouping. The basic distinction between these 2 approaches is that risk grouping assigns patients to distinct categories while nomograms align patients along a continuum or dimension. We used the taxometric methods developed by Meehl to compare the competing models of risk grouping vs risk continuum in patients with clinically localized prostate cancer.

The study sample consisted of 80,304 patients from the Surveillance, Epidemiology and End Results database from 2004 to 2006. The 3 clinical variables analyzed were serum prostate specific antigen, Gleason score and American Joint Committee on Cancer T stage. Three taxometric procedures were used in analysis, including maximum covariance, mean above minus mean below a cut and latent mode. The comparison curve fit index was calculated for each procedure and the 3 results were averaged. A priori thresholds for the mean comparison curve fit index were that values greater than 0.55 were considered evidence of categorical structure and values less than 0.45 were considered evidence of dimensional structure. Values between 0.45 and 0.55 were deemed ambiguous.

Maximum covariance, mean above minus mean below a cut and latent mode analyses yielded a comparison curve fit index of 0.168, 0.401 and 0.465, respectively (mean 0.345).

Results favor a dimensional rather than a categorical model of risk estimation and provide further support for using nomograms over risk grouping in patients with clinically localized prostate cancer.