GUEST AUTHOR: Nick Kolev

A new study from the Ghebremichael Lab at the Ragon Institute, published in the Biometrical Journal, introduces a statistical method that evaluates biologically linked outcomes jointly rather than relying on separate analyses that ignore their interdependence. The researchers demonstrated their methodology using data from a longitudinal pediatric HIV study examining the efficacy of antiretroviral therapy (ART).

Traditionally, response to ART is evaluated by monitoring two key outcomes independently: suppression of the HIV virus and recovery of the immune system, measured by CD4+ T cell counts. Standard statistical tools, such as the Kaplan–Meier estimator and hazard functions, assess each outcome separately. However, because viral suppression and immune recovery are biologically connected processes, analyzing them in isolation can miss important patterns and introduce bias, resulting in an incomplete understanding of treatment effectiveness.

To address this limitation, the team developed the bivariate median residual life function, which models both outcomes simultaneously. This approach captures complex treatment dynamics that separate analyses cannot, including the detection of patients who achieve viral suppression without adequate immune recovery—or, conversely, those who show immune improvement without fully suppressing the virus.

The findings highlight the limitations of traditional statistical methods and extend classical survival concepts to accommodate multivariate, interconnected outcomes. This work provides researchers and clinicians with a new statistical framework and accompanying software tool for evaluating correlated outcomes. Although HIV data was used in this study, the method and software are broadly applicable to any setting involving interconnected outcomes.