Adaption of the Memorial Sloan Kettering Cancer Center Nomograms for the Prediction of Prostate Cancer–specific Death in Sweden: A Population-based Cohort Study

Figures

• 

  Fig. 1. Discrimination of the pre- and postoperative models predicting prostate cancer–specific and other-cause mortality at 10 yr. Vertical lines present 95% confidence intervals for C-index estimates. Horizontal lines present the optimism-corrected C-index for the pre- and postoperative models. The cT2 and cT3 substages, which are included in the preoperative MSKCC nomogram but are not recorded in the National Prostate Cancer Register of Sweden, were calculated using models developed in the study of Zelic et al [1]. Apparent (MSKCC) refers to the preoperative model discrimination in the population with nonmissing variables needed for the calculation of cT2 and cT3 substages (N = 29 259). MSKCC = Memorial Sloan Kettering Cancer Center.

• 

  Fig. 2. Calibration plot for the pre- and postoperative models predicting prostate cancer–specific and other-cause mortality at 10 yr. The apparent predicted risk is a risk predicted for the study population using estimates from the model developed using the study population. The mean bootstrap risk and 2.5th and 97.5th percentiles of bootstrap risks are calculated from 500 risks predicted for the study population using estimates from the models developed using the 500 bootstrap samples. Smooth calibration curves were created by modelling the outcome of interest as a function of the predicted risk transformed as log(-log[1-risk]) in a Fine and Gray model. The predicted risks were modelled using restricted cubic splines with four degrees of freedom. The predictions from this model were used as the observed outcome proportions. For each of 500 bootstraps, we averaged the estimated 10-yr risks in the groups defined by the deciles of the apparent predicted risk among individuals who had the event of interest, and plotted their mean and 2.5th and 97.5th percentiles against the Aalen-Johansson estimate. The cT2 and cT3 substages, which are included in the preoperative MSKCC nomogram but are…

• 

  Fig. 3. The distribution of the preoperative model–predicted 10-yr risk of prostate cancer–specific mortality in the levels of modified NCCN risk groups. Risk groups were defined as follows: low risk: cT1–2 and Gleason score ≤6 and PSA <10; intermediate risk: not cT3–4 and Gleason score 7 and PSA <20, or not cT3–4 and Gleason score ≤6 and 10 ≤ PSA < 20; high risk: cT3 and any Gleason score and PSA <50, or not cT4 and Gleason score 8–10 and PSA <50, or not cT4 and any Gleason score and 20 ≤ PSA < 50; regional metastases: PSA ≤100 and N1, or cT4 and PSA ≤100, or 50 ≤ PSA ≤ 100; and distant metastases: PSA >100. NCCN = National Comprehensive Cancer Network; PSA = prostate-specific antigen.