Volume 30, Issue 9 pp. 1233-1241
ORIGINAL ARTICLE

Development and evaluation of a proxy for baseline ECOG PS in advanced non-small cell lung cancer, bladder cancer, and melanoma: An electronic health record study

Natalia Sadetsky

Corresponding Author

Natalia Sadetsky

Genentech, Inc., South San Francisco, California, USA

Correspondence

Natalia Sadetsky, Atara Biotherapeutics, 611 Gateway Blvd. Unit 900, South San Francisco, CA 94080, USA.

Email: [email protected]

Search for more papers by this author
Ching-Yi Chuo

Ching-Yi Chuo

Genentech, Inc., South San Francisco, California, USA

Search for more papers by this author
Amy J. Davidoff

Amy J. Davidoff

Yale University School of Public Health, New Haven, Connecticut, USA

Search for more papers by this author
First published: 18 June 2021
Citations: 3

Funding information: Genentech

Abstract

Purpose

Eastern Cooperative Oncology Group performance status (ECOG PS) is an important predictor for receipt of treatment and overall survival (OS) but is often unreported in routine care. We developed a proxy for baseline ECOG PS using electronic health records (EHRs).

Methods

We analyzed patients who were diagnosed with advanced non-small cell lung cancer (aNSCLC), advanced bladder cancer (aBCa), and advanced melanoma (aMEL) between 2011 and 2018 and had a baseline (reported between diagnosis and treatment) ECOG PS in a real-world database. We used stepwise multivariable logistic regression to model associations between baseline ECOG PS good (<2) versus poor (≥2) and sociodemographic, clinical, and laboratory measures in each cancer type. Predictive accuracy of classifying ECOG PS was assessed. We tested the association between OS and observed and predicted ECOG PS.

Results

In total, 20 697 aNSCLC patients, 2627 aBCa patients, and 2558 aMEL patients constituted the study population. Percentage of patients with poor ECOG PS ranged from 15.3% (aMEL) to 28.5% (aNSCLC). Poor ECOG PS was associated with more comorbid conditions, older age, lower body mass index, metastases, and abnormal laboratory indicators. Overall prediction accuracy using a 0.50 cutpoint was 73.3% for NSCLC, 73.8% for aBCa, and 85.4% for aMEL. The association of OS with ECOG PS was consistent between the observed and proxy measures.

Conclusions

In the EHR-derived data, clinical, sociodemographic, and laboratory information were used to assign ECOG PS and enhance the use of real-world data in outcome studies.

CONFLICT OF INTEREST

Natalia Sadetsky and James Ching-Yi Chuo: employees of Genentech Inc. and have salary and stocks from Roche/Genentech. Amy J. Davidoff: grants from Celgene during the conduct of the study and consulting fees with Amgen; a family member received personal fees from AbbVie, Jazz Pharmaceuticals, Kyowa Hakko Kirin, and Tolero Pharmaceuticals and owns stocks from Daiichi Sankyo outside the submitted work.

The full text of this article hosted at iucr.org is unavailable due to technical difficulties.