Response to Mehmet I. Naharci

To the Editor: In Dr. Naharci's thoughtful comments1 on our recent article in this journal,2 he suggested that frailty, depression, and medications for hypertension and diabetes mellitus could be important considerations in the relationship between inflammation and cognition. Dr. Naharci correctly noted that a significant proportion of the GENOA cohort were taking medications for hypertension and diabetes; 495 (18%) were taking medications for diabetes mellitus and 1,902 (70%) for hypertension. Depressed mood was rated on a 6-point Likert scale by asking “How often have you felt sad or depressed over the past month?” Of 1,965 responders, 1,350 (69%) rated symptoms 0 (almost never) or 1 (seldom); only 79 (4%) rated symptoms 4 (very often) or 5 (constantly). Frailty was not measured, but usual walking speed (mean 1.04 m/s, interquartile range 0.85–1.27), a component of a validated frailty phenotype3 and predictor of mortality in older adults,4 was assessed. We provide updated results adjusting for these variables (Table 1).

Results were similar to the original findings, with some attenuation of relationships, largely due to walking speed. When all additional variables were included, soluble tumor necrosis factor (sTNFR)-2 remained statistically associated with sustained attention and processing speed, language, memory, and marginally global mental status (P = .06) and executive function (P = .10) in African Americans. sTNFR1 retained statistically significant associations with sustained attention and processing speed in European Americans and African Americans and marginally with executive function in African Americans (P = .08). Relationships between C-reactive protein and sustained attention and processing speed in African Americans and between interleukin-6 and executive function in African Americans were no longer statistically significant. Although examining influences of a variety of potential confounders is important, possible causal and mechanistic pathways should also always be considered. If inflammation influences depression, and depression and frailty are linked to cognitive outcomes (previous studies support both pathways), then these variables may lie in a mediation path between inflammation and cognition. Including mediators as adjustors statistically may remove some of the very effects of inflammation we sought to quantify.

We thank Dr. Naharci for his interest in our work. We believe that our primary findings linking inflammation, particularly tumor necrosis factor activity, to cognition in African Americans with prevalent hypertension and vascular risk factors remain robust. It is our hope that ongoing exchange of ideas such as this will continue to spur research into mechanisms of cognitive decline and strategies to protect against it.