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

Prostate Biopsy Collaborative Group

Research output: Contribution to journalArticle

4 Citations (Scopus)

Abstract

Abstract Objectives We aimed to compare nonlinear modeling methods for handling continuous predictors for reproducibility and transportability of prediction models. Study Design and Setting 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). Results 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). Conclusion 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.

Original languageEnglish (US)
Article number8766
Pages (from-to)426-434
Number of pages9
JournalJournal of Clinical Epidemiology
Volume68
Issue number4
DOIs
StatePublished - Apr 1 2015

Fingerprint

Prostate
Prostatic Neoplasms
Biopsy
Area Under Curve
Logistic Models
Prostate-Specific Antigen
ROC Curve
Calibration

Keywords

  • Calibration
  • Discrimination
  • External validation
  • Internal validation
  • Nonlinear modeling
  • Prediction models

ASJC Scopus subject areas

  • Epidemiology

Cite this

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

In: Journal of Clinical Epidemiology, Vol. 68, No. 4, 8766, 01.04.2015, p. 426-434.

Research output: Contribution to journalArticle

@article{3d76796285d6443cadf196afc186e0f9,
title = "Nonlinear modeling was applied thoughtfully for risk prediction: The Prostate Biopsy Collaborative Group",
abstract = "Abstract Objectives We aimed to compare nonlinear modeling methods for handling continuous predictors for reproducibility and transportability of prediction models. Study Design and Setting 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). Results 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). Conclusion 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.",
keywords = "Calibration, Discrimination, External validation, Internal validation, Nonlinear modeling, Prediction models",
author = "{Prostate Biopsy Collaborative Group} and Daan Nieboer and Yvonne Vergouwe and Roobol, {Monique J.} and Ankerst, {Donna P} and Kattan, {Michael W.} and Vickers, {Andrew J.} and Steyerberg, {Ewout W.}",
year = "2015",
month = "4",
day = "1",
doi = "10.1016/j.jclinepi.2014.11.022",
language = "English (US)",
volume = "68",
pages = "426--434",
journal = "Journal of Clinical Epidemiology",
issn = "0895-4356",
publisher = "Elsevier USA",
number = "4",

}

TY - JOUR

T1 - Nonlinear modeling was applied thoughtfully for risk prediction

T2 - The Prostate Biopsy Collaborative Group

AU - Prostate Biopsy Collaborative Group

AU - Nieboer, Daan

AU - Vergouwe, Yvonne

AU - Roobol, Monique J.

AU - Ankerst, Donna P

AU - Kattan, Michael W.

AU - Vickers, Andrew J.

AU - Steyerberg, Ewout W.

PY - 2015/4/1

Y1 - 2015/4/1

N2 - Abstract Objectives We aimed to compare nonlinear modeling methods for handling continuous predictors for reproducibility and transportability of prediction models. Study Design and Setting 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). Results 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). Conclusion 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.

AB - Abstract Objectives We aimed to compare nonlinear modeling methods for handling continuous predictors for reproducibility and transportability of prediction models. Study Design and Setting 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). Results 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). Conclusion 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.

KW - Calibration

KW - Discrimination

KW - External validation

KW - Internal validation

KW - Nonlinear modeling

KW - Prediction models

UR - http://www.scopus.com/inward/record.url?scp=84925341651&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84925341651&partnerID=8YFLogxK

U2 - 10.1016/j.jclinepi.2014.11.022

DO - 10.1016/j.jclinepi.2014.11.022

M3 - Article

C2 - 25777297

AN - SCOPUS:84925341651

VL - 68

SP - 426

EP - 434

JO - Journal of Clinical Epidemiology

JF - Journal of Clinical Epidemiology

SN - 0895-4356

IS - 4

M1 - 8766

ER -