Dataset Information

Nonlinear modeling was applied thoughtfully for risk prediction: the Prostate Biopsy Collaborative Group.

ABSTRACT: We aimed to compare nonlinear modeling methods for handling continuous predictors for reproducibility and transportability of prediction models.We analyzed four cohorts of previously unscreened men who underwent prostate biopsy for diagnosing prostate cancer. Continuous predictors of prostate cancer included prostate-specific antigen and prostate volume. The logistic regression models included linear terms, logarithmic terms, fractional polynomials of degree one or two (FP1 and FP2), or restricted cubic splines (RCS) with three or five knots (RCS3 and RCS5). The resulting models were internally validated by bootstrap resampling and externally validated in the cohorts not used at model development. Performance was assessed with the area under the receiver operating characteristic curve (AUC) and the calibration component of the Brier score (CAL).At internal validation models with FP2 or RCS5 showed slightly better performance than the other models (typically 0.004 difference in AUC and 0.001 in CAL). At external validation models containing logarithms, FP1, or RCS3 showed better performance (differences 0.01 and 0.002).Flexible nonlinear modeling methods led to better model performance at internal validation. However, when application of the model is intended across a wide range of settings, less flexible functions may be more appropriate to maximize external validity.

SUBMITTER: Nieboer D

PROVIDER: S-EPMC4474141 | biostudies-literature | 2015 Apr

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Nonlinear modeling was applied thoughtfully for risk prediction: the Prostate Biopsy Collaborative Group.

Nieboer Daan D Vergouwe Yvonne Y Roobol Monique J MJ Ankerst Donna P DP Kattan Michael W MW Vickers Andrew J AJ Steyerberg Ewout W EW

Journal of clinical epidemiology 20141129 4

<h4>Objectives</h4>We aimed to compare nonlinear modeling methods for handling continuous predictors for reproducibility and transportability of prediction models.<h4>Study design and setting</h4>We analyzed four cohorts of previously unscreened men who underwent prostate biopsy for diagnosing prostate cancer. Continuous predictors of prostate cancer included prostate-specific antigen and prostate volume. The logistic regression models included linear terms, logarithmic terms, fractional polynom ...[more]

PMID: 25777297

Similar Datasets

Project description:ObjectivesTo develop and validate a prostate cancer (PCa) risk calculator (RC) incorporating multiparametric magnetic resonance imaging (mpMRI) and to compare its performance with that of the Prostate Biopsy Collaborative Group (PBCG) RC.Patients and methodsMen without a PCa diagnosis receiving mpMRI before biopsy in the Prospective Loyola University mpMRI (PLUM) Prostate Biopsy Cohort (2015-2020) were included. Data from a separate institution were used for external validation. The primary outcome was diagnosis of no cancer, grade group (GG)1 PCa, and clinically significant (cs)PCa (≥GG2). Binary logistic regression was used to explore standard clinical and mpMRI variables (prostate volume, Prostate Imaging-Reporting Data System [PI-RADS] version 2.0 lesions) with the final PLUM RC, based on a multinomial logistic regression model. Receiver-operating characteristic curve, calibration curves, and decision-curve analysis were evaluated in the training and validation cohorts.ResultsA total of 1010 patients were included for development (N = 674 training [47.8% PCa, 30.9% csPCa], N = 336 internal validation) and 371 for external validation. The PLUM RC outperformed the PBCG RC in the training (area under the curve [AUC] 85.9% vs 66.0%; P < 0.001), internal validation (AUC 88.2% vs 67.8%; P < 0.001) and external validation (AUC 83.9% vs 69.4%; P < 0.001) cohorts for csPCa detection. The PBCG RC was prone to overprediction while the PLUM RC was well calibrated. At a threshold probability of 15%, the PLUM RC vs the PBCG RC could avoid 13.8 vs 2.7 biopsies per 100 patients without missing any csPCa. At a cost level of missing 7.5% of csPCa, the PLUM RC could have avoided 41.0% (566/1381) of biopsies compared to 19.1% (264/1381) for the PBCG RC. The PLUM RC compared favourably with the Stanford Prostate Cancer Calculator (SPCC; AUC 84.1% vs 81.1%; P = 0.002) and the MRI-European Randomized Study of Screening for Prostate Cancer (ERSPC) RC (AUC 84.5% vs 82.6%; P = 0.05).ConclusionsThe mpMRI-based PLUM RC significantly outperformed the PBCG RC and compared favourably with other mpMRI-based RCs. A large proportion of biopsies could be avoided using the PLUM RC in shared decision making while maintaining optimal detection of csPCa.

Project description:BackgroundExternal validation of risk calculators (RCs) is necessary to determine their clinical applicability beyond the setting in which these were developed.ObjectiveTo assess the performance of the Rotterdam Prostate Cancer RC (RPCRC) and the Prostate Biopsy Collaborative Group RC (PBCG-RC).Design setting and participantsWe used data from the prospective, population-based STHLM3 screening study, performed in 2012-2015. Participants with prostate-specific antigen ≥3 ng/ml who underwent systematic prostate biopsies were included.Outcome measurements and statistical analysisProbabilities for clinically significant prostate cancer (csPCa), defined as International Society of Urological Pathology grade ≥2, were calculated for each participant. External validity was assessed by calibration, discrimination, and clinical usefulness for both original and recalibrated models.Results and limitationsOut of 5841 men, 1054 (18%) had csPCa. Distribution of risk predictions differed between RCs; median risks for csPCa using the RPCRC and PBCG-RC were 3.3% (interquartile range [IQR] 2.1-7.1%) and 20% (IQR 15-28%), respectively. The correlation between RC risk estimates on individual level was moderate (Spearman's r = 0.55). Using the RPCRC's recommended risk threshold of ≥4% for finding csPCa, 36% of participants would get concordant biopsy recommendations. At 10% risk cut-off, RCs agreed in 23% of cases. Both RCs showed good discrimination, with areas under the curves for the RPCRC of 0.74 (95% confidence interval [CI] 0.72-0.76) and the PBCG-RC of 0.70 (95% CI 0.68-0.72). Calibration was adequate using the PBCG-RC (calibration slope: 1.13 [95% CI 1.03-1.23]), but the RPCRC underestimated the risk of csPCa (calibration slope: 0.73 [0.68-0.79]). The PBCG-RC showed a net benefit in a decision curve analysis, whereas the RPCRC showed no net benefit at clinically relevant risk threshold levels. Recalibration improved clinical benefit, and differences between RCs decreased.ConclusionsAssessment of calibration is essential to ensure the clinical value of risk prediction tools. The PBCG-RC provided clinical benefit in its current version online. On the contrary, the RPCRC cannot be recommended in this setting.Patient summaryPredicting the probability of finding prostate cancer on biopsy differed between two assessed risk calculators. After recalibration, the agreement of the models improved, and both were shown to be clinically useful.

Project description:Importance:Multiparametric magnetic resonance imaging (MRI) in conjunction with MRI-transrectal ultrasound (TRUS) fusion-guided biopsies have improved the detection of prostate cancer. It is unclear whether MRI itself adds additional value to multivariable prediction models based on clinical parameters. Objective:To determine whether an MRI-based prediction model can reduce unnecessary biopsies in patients with suspected prostate cancer. Design, Setting, and Participants:Patients underwent MRI, MRI-TRUS fusion-guided biopsy, and 12-core systematic biopsy in 1 session. The development cohort used to derive the prediction model consisted of 400 patients from 1 institution enrolled between May 14, 2015, and August 31, 2016, and the validation cohort included 251 patients from 2 independent institutions who underwent biopsies between April 1, 2013, and June 30, 2016, at 1 institution and between July 1, 2015, and October 31, 2016, at the other institution. The MRI model included MRI-derived parameters in addition to clinical variables. Area under the curve of receiver operating characteristic curves and decision curve analysis were performed. Main Outcomes and Measures:Risk of clinically significant prostate cancer on biopsy, defined as a Gleason score of 3?+?4 or higher in at least 1 biopsy core. Results:Overall, 193 (48.3%) of the 400 patients in the development cohort (mean [SD] age at biopsy, 64.3 [7.1] years) and 96 (38.2%) of the 251 patients in the validation cohort (mean [SD] age at biopsy, 64.9 [7.2] years) had clinically significant prostate cancer, defined as a Gleason score greater than or equal to 3?+?4. By applying the model to the external validation cohort, the area under the curve increased from 64% to 84% compared with the baseline model (P?<?.001). At a risk threshold of 20%, the MRI model had a lower false-positive rate than the baseline model (46% [95% CI, 32%-66%] vs 92% [95% CI, 70%-100%]), with only a small reduction in the true-positive rate (89% [95% CI, 85%-96%] vs 99% [95% CI, 89%-100%]). Eighteen of 100 fewer biopsies could have been performed, with no increase in the number of patients with missed clinically significant prostate cancers. Conclusions and Relevance:The inclusion of MRI-derived parameters in a risk model could reduce the number of unnecessary biopsies while maintaining a high rate of diagnosis of clinically significant prostate cancers.

Project description:BackgroundDecipher Biopsy is a commercially available gene expression classifier used in risk stratification of newly diagnosed prostate cancer (PCa). Currently, there are no prospective data evaluating its clinical utility. We seek to assess the clinical utility of Decipher Biopsy in localized PCa patients.MethodsA multi-institutional study of 855 men who underwent Decipher Biopsy testing between February 2015 and October 2019. All patients were tracked through the prospective Michigan Urological Surgery Improvement Collaborative and linked to the Decipher Genomics Resource Information Database (GRID®; NCT02609269). Patient matching was performed by an independent third-party (ArborMetrix Inc.) using two or more unique identifiers. Cumulative incidence curves for time to treatment (TTT) and time to failure (TTF) were constructed using Kaplan-Meier estimates. Multivariable Cox proportional hazard models were used to evaluate the independent association of high-risk Decipher scores with the conversion from AS to radical therapy and treatment failure (biochemical failure or receipt of salvage therapy).Results and limitationsEight hundred fifty-five patients underwent Decipher Biopsy testing during the study period. Of the 855 men, 264 proceeded to AS (31%), and 454 (53%) received radical therapy. In men electing AS, after adjusting for NCCN risk group, age, PSA, prostate volume, body mass index, and percent positive cores, a high-risk Decipher score was independently associated with shorter TTT (HR 2.51, 95% CI 1.52-4.13 p < 0.001). Similarly, in patients that underwent radical therapy, a high-risk Decipher score was independently associated with TTF (HR 2.98, 95% CI 1.22-7.29, p = 0.01) on multivariable analysis. Follow-up time was a limitation.ConclusionIn a prospective statewide registry, high-risk Decipher Biopsy score was strongly and independently associated with conversion from AS to definitive treatment and treatment failure. These real-world data support the clinical utility of Decipher Biopsy. An ongoing phase 3 randomized trial (NCT04396808) will provide level 1 evidence of the clinical impact of Decipher biopsy testing.

Dataset Information

Nonlinear modeling was applied thoughtfully for risk prediction: the Prostate Biopsy Collaborative Group.

Publications

Nonlinear modeling was applied thoughtfully for risk prediction: the Prostate Biopsy Collaborative Group.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets