Interpreting Frailty-Depression Overlap When Cardiovascular Disease Enters the Frailty Construct

To the Editor, I was interested in the population-based analysis by Pérez Covo et al examining the overlap of frailty and depression in relation to cardiovascular disease (CVD) among 18,973 older Colombian adults. 1 The authors should be commended for focusing on an underexplored geriatric phenotype in a nationally important Latin American dataset and for presenting both individual and joint associations.The large sample and clinically intuitive framing make the study relevant to cardiovascular prevention in older adults.Three methodological issues, however, deserve clarification because they bear directly on the interpretation of the reported overlap effect.First, the exposure and outcome may not be fully independent.Frailty was measured with the 5-item FRAIL scale, whose "Illnesses" component is based on the burden of chronic diseases.2 Standard implementations of this item include cardiovascular conditions such as myocardial infarction/heart attack and stroke among the illnesses counted toward the frailty score.In the present study, CVD was itself defined as a self-reported history of myocardial infarction or stroke. 1 Thus, for some participants, the outcome can contribute directly to the exposure classification.This form of criterion overlap can mechanically strengthen the association between frailty and CVD, particularly near the prefrail/frail thresholds, even without any additional biological relationship.This issue is especially important because the paper's central comparison concerns the joint frailty-depression phenotype.I would therefore ask the authors to repeat the analysis using a modified FRAIL score in which myocardial infarction and stroke do not contribute to the Illnesses item, or, preferably, using a frailty definition based only on the remaining non-CVD components.If the frailty-depression association with CVD remains of similar magnitude, the main conclusion would become substantially more convincing.Reporting how many participants changed from frail to prefrail, or from prefrail to robust, after removing CVD-related illnesses would also quantify the extent of this potential incorporation.Second, the observation that the combined frailty-depression category has a larger prevalence ratio than either syndrome alone does not by itself demonstrate that the two conditions interact or produce a synergistic cardiovascular phenotype.Joint categories can yield a larger relative measure simply because two associated exposures coexist.Formal interaction should be evaluated on the multiplicative scale with an exposure-product term and, if the clinical question concerns excess burden attributable to coexistence, on the additive scale using measures such as the relative excess risk due to interaction or attributable proportion.3 A table showing CVD prevalence and adjusted prevalence ratios for the four joint exposure groups -neither condition, depression only, frailty only, and both-would allow readers to distinguish simple co-occurrence from statistical interaction.Third, the study repeatedly describes the analytical sample as nationally representative, yet the statistical methods do not clearly state whether SABE Colombia sampling weights, strata, and clustering were incorporated into prevalence estimation and regression. 1 The SABE survey used a complex population sampling design to obtain national estimates.4 Analyses that ignore such design features can alter both point estimates and, particularly, standard errors and confidence intervals.5 Because the main result for frailty plus depression after adjustment is relatively imprecise (PR 2.05, 95% CI 1.05-3.97),surveydesign-based re-estimation could be clinically consequential.Clarification of whether design weights and clustering were used, and presentation of weighted prevalence ratios if they were not, would strengthen the population-level interpretation.These points are distinct from the limitations already acknowledged by the authors regarding cross-sectional temporality, self-reported CVD, complete-case selection, and the age of the dataset.