Prognostic models for breast cancer: a systematic review

Area of Science:

• Oncology
• Biostatistics
• Epidemiology

Background:

• Breast cancer is a leading cause of cancer in women globally, with significant variability in patient outcomes.
• Accurate prediction of breast cancer outcomes is crucial for patients, clinicians, researchers, and policymakers.
• Numerous prognostic models have been developed and evaluated across diverse settings.

Purpose of the Study:

• To systematically review prognostic models developed and/or validated for female breast cancer patients.
• To identify commonly used predictors and methodologies in breast cancer prognostication.
• To assess the performance of these models in development and independent validation cohorts.

Main Methods:

• Systematic search of four electronic databases, oncology websites, and bibliographies of relevant studies.
• Inclusion of original studies published before January 1, 2017, focusing on clinico-pathological factors for mortality and/or recurrence prediction.
• Identification and analysis of 58 prognostic models from 96 selected articles.

Main Results:

• The 58 models predicted mortality (28), recurrence (23), or both (7), utilizing predictors like nodal status, tumor size, tumor grade, age, and estrogen receptor status.
• Models were developed globally, with Cox proportional hazards regression being the most common development method.
• While models performed well in development cohorts, accuracy diminished in some independent populations, particularly for high-risk, young, and elderly patients.

Conclusions:

• A multitude of breast cancer prognostic models exist, yet few achieve widespread validation across different settings.
• The predictive performance of these models is often suboptimal in independent populations, especially for specific patient subgroups.
• The Nottingham Prognostic Index demonstrated consistent predictive ability across most independent populations, serving as a notable exception.