Applying a linear offset procedure to correct for bias in cardiovascular age models within the evaluation cohort itself introduces methodological bias.
This commentary refers to ‘Sex-specific body fat distribution predicts cardiovascular ageing’, by V. Losev et al., https://doi.org/10.1093/eurheartj/ehaf553 and the discussion piece ‘Approaches to age bias correction’, by D. O’Regan and V. Losev, https://doi.org/10.1093/eurheartj/ehaf801. We read with great interest the article by Losev et al.,1 describing sex-specific effects of adiposity phenotypes on cardiovascular ageing using cardiac magnetic resonance imaging. This work highlights the translational promise of imaging-derived ‘cardiovascular age’ biomarkers. However, we wish to raise a methodological concern regarding the procedure used to correct for bias in cardiovascular age-delta (CA-Δ). The authors report correcting the correlation between predicted cardiovascular age and chronological age via a linear offset procedure. They cite prior work from their group (Shah et al., Nat Commun 2023), where the methodological details and code are provided.2,3 From inspection of that repository,4 it appears they implement bias correction by regressing predicted age on chronological age within the evaluation cohort itself and then using that same regression to ‘correct’ predictions in the same set. It is therefore reasonable to infer that the current article employs the same procedure.
Lebowitz et al. (Thu,) conducted a letter in Cardiovascular ageing. Linear offset procedure for bias correction was evaluated. Applying a linear offset procedure to correct for bias in cardiovascular age models within the evaluation cohort itself introduces methodological bias.