Validated ECOG Proxies Could Make Real-World Oncology Evidence More Representative

In an interview with Pharmacy Times, Carole R. Berini, PhD, research scientist at Ontada, and Saamir Pasha, MPH, senior biostatistician at Ontada, discussed how validated proxies for missing Eastern Cooperative Oncology Group (ECOG) performance status could strengthen real-world oncology evidence. Berini explained that ECOG data may not be missing at random. Patients with documented scores can differ systematically from those without scores because patient or clinical characteristics may influence whether performance status is recorded. Restricting an analysis to patients with documented ECOG status can therefore reduce statistical power and exclude types of patients, limiting how well the findings represent community oncology practice.

A validated proxy could allow researchers to include patients who would otherwise be excluded. However, determining whether this approach improves representation of historically underrepresented groups requires researchers to identify who is missing ECOG data and who is reintroduced through the proxy. Pasha clarified that patients may have an ECOG score elsewhere in their electronic health record, but not at the specific treatment time point required for a study. Previous ECOG measurements could help predict later scores and address these time-specific gaps.

For oncology pharmacists, more complete functional-status data could make real-world findings easier to interpret across a patient population. Berini said this could help pharmacists assess whether evidence reflects the patients they encounter when evaluating treatment intensity, supportive care, toxicity monitoring, and medication management.

The investigators addressed whether the approach could extend beyond advanced prostate cancer. Prior studies have examined performance-status proxies in advanced non–small cell lung cancer, bladder cancer, and melanoma. Broader use would require transparent model development, external validation across tumor types, health systems, and patient populations, and sensitivity analyses tailored to each intended use. The investigators emphasized that stronger ECOG documentation and better electronic health record workflows would improve the data available for model development.