Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Developing Performance-Guaranteed Biomarker Combination Rules with Integrated External Information under Practical Constraint

Albert Osom, Camden Lopez, Ashley Alexander, Suresh Chari, Ziding Feng, Ying-Qi Zhao

TL;DR

The paper tackles learning biomarker-based clinical decision rules under a clinically meaningful PPV constraint, aiming to maximize true positives while controlling the screening burden. It introduces a PPV-constrained benefit function and develops two practical methods: DOOLR, which guarantees optimality within the linear rule class using a smooth surrogate, and IT-DOOLR, which can safely incorporate external risk information through a tunable penalty. Theoretical guarantees show asymptotic consistency and PPV control, and extensive simulations demonstrate favorable finite-sample performance, including robustness to external information mis-specification. The methods are applied to PDAC screening in new-onset diabetes (PANDORA data) with external ENDPAC risk information, yielding interpretable linear rules that achieve high sensitivity while maintaining a targeted PPV, illustrating tangible clinical utility for risk-guided screening programs.

Abstract

In clinical practice, there is significant interest in integrating novel biomarkers with existing clinical data to construct interpretable and robust decision rules. Motivated by the need to improve decision-making for early disease detection, we propose a framework for developing an optimal biomarker-based clinical decision rule that is both clinically meaningful and practically feasible. Specifically, our procedure constructs a linear decision rule designed to achieve optimal performance among class of linear rules by maximizing the true positive rate while adhering to a pre-specified positive predictive value constraint. Additionally, our method can adaptively incorporate individual risk information from external source to enhance performance when such information is beneficial. We establish the asymptotic properties of our proposed estimator and compare to the standard approach used in practice through extensive simulation studies. Results indicate that our approach offers strong finite-sample performance. We also apply the proposed methods to develop biomarker-based screening rules for pancreatic ductal adenocarcinoma (PDAC) among new-onset diabetes (NOD) patients.

Developing Performance-Guaranteed Biomarker Combination Rules with Integrated External Information under Practical Constraint

TL;DR

The paper tackles learning biomarker-based clinical decision rules under a clinically meaningful PPV constraint, aiming to maximize true positives while controlling the screening burden. It introduces a PPV-constrained benefit function and develops two practical methods: DOOLR, which guarantees optimality within the linear rule class using a smooth surrogate, and IT-DOOLR, which can safely incorporate external risk information through a tunable penalty. Theoretical guarantees show asymptotic consistency and PPV control, and extensive simulations demonstrate favorable finite-sample performance, including robustness to external information mis-specification. The methods are applied to PDAC screening in new-onset diabetes (PANDORA data) with external ENDPAC risk information, yielding interpretable linear rules that achieve high sensitivity while maintaining a targeted PPV, illustrating tangible clinical utility for risk-guided screening programs.

Abstract

In clinical practice, there is significant interest in integrating novel biomarkers with existing clinical data to construct interpretable and robust decision rules. Motivated by the need to improve decision-making for early disease detection, we propose a framework for developing an optimal biomarker-based clinical decision rule that is both clinically meaningful and practically feasible. Specifically, our procedure constructs a linear decision rule designed to achieve optimal performance among class of linear rules by maximizing the true positive rate while adhering to a pre-specified positive predictive value constraint. Additionally, our method can adaptively incorporate individual risk information from external source to enhance performance when such information is beneficial. We establish the asymptotic properties of our proposed estimator and compare to the standard approach used in practice through extensive simulation studies. Results indicate that our approach offers strong finite-sample performance. We also apply the proposed methods to develop biomarker-based screening rules for pancreatic ductal adenocarcinoma (PDAC) among new-onset diabetes (NOD) patients.
Paper Structure (20 sections, 2 theorems, 32 equations, 1 figure, 11 tables)

This paper contains 20 sections, 2 theorems, 32 equations, 1 figure, 11 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. PPV-constrained benefit function
  4. Estimation of the optimal decision rule
  5. Direct optimization for optimal linear decision rule
  6. Incorporating existing information to improve linear combination rule
  7. Theoretical properties
  8. Numerical studies
  9. Linear decision rule
  10. Nonlinear decision rule
  11. Incorporating existing risk information
  12. Application to PANDORA study data
  13. Discussion
  14. Appendix A (Optimality of decision rule)
  15. Appendix B (Optimization algorithm and justification)
  16. ...and 5 more sections

Key Result

Theorem 1

Under conditions (A1)–(A5), we have that as $h \rightarrow 0$

Figures (1)

  • Figure 1: Comparison of zero-one indicator with $\Phi{(x/h)}$ for different values of $h$. We can observe that, when $h$ approaches zero, the approximation of the zero-one indicator with $\Phi{(x/h)}$ becomes more accurate.

Theorems & Definitions (6)

  • Remark 1
  • Theorem 1
  • Lemma 1: Lemma 2 from meisner2021combining
  • proof
  • proof
  • proof