How Reliable are Trial-based Prognostic Models in Real-world Patients with Metastatic Castration-resistant Prostate Cancer?

ePCR: an R-package for survival and time-to-event prediction in advanced prostate cancer, applied to real-world patient cohorts

“…As an application use case, we investigated how to best use the ePCR model trained on the clinical trial data from the PCC-DREAM Challenge for prognostic modelling of survival of CRPC patients using independent, real-world hospital registry data (see Supplementary Data ). This investigation was inspired by recent work by Seyednasrollah et al. (2017) , where the authors made use of our original, PCC-DREAM Challenge winning model for prognostic prediction of mCRPC patients treated with docetaxel in the same hospital registry data.…”

“…Model type: H, Halabi et al. (2014) ; S, ePCR model with 81 variables used in Seyednasrollah et al. (2017) ; F, full ePCR model with all the 101 variables available in the real-world cohort; R, reduced ePCR model including those 60 variables that were available for at least 60% of patients in the cohort selected by medical records.…”

Re: Fatemeh Seyednasrollah, Mehrad Mahmoudian, Liisa Rautakorpi, et al. How Reliable are Trial-based Prognostic Models in Real-world Patients with Metastatic Castration-resistant Prostate Cancer? Eur Urol. 2017;71:838–40

“…Seyednasrollah et al [1] posed a relevant question in their recent Research Letter. Once shown to be reliable enough in real-world clinical data, trial-based prognostic models could provide practical guidance in tailoring treatment strategies for patients with metastatic castration-resistant prostate cancer (mCRPC).…”

“…Seyednasrollah et al [1] took the open-source implementations of the three best-performing models from the recent prostate cancer DREAM challenge [2]. It should be obvious that the use of others' models without consulting the model developers will lead to suboptimal results when the models are applied to different types of patient data.…”

“…For instance, in our top-performing prognostic model (Team 1), we implemented and made available an improved model-based imputation method that led to the most accurate predictions in the challenge [2]. It remains unclear why Seyednasrollah et al chose to use simple median imputation instead [ 6 _ T D $ D I F F ] in their[ 7 _ T D $ D I F F ] closed-data results [1]. Further confusion was introduced by simply assuming that all the patients had bone lesions, even though this information was actually missing in this real-world patient data set.…”