-
Views
-
Cite
Cite
Alex Dregan, Robert Stewart, Martin C. Gulliford, Cardiovascular risk factors and cognitive decline in adults aged 50 and over: a population-based cohort study, Age and Ageing, Volume 42, Issue 3, May 2013, Pages 338–345, https://doi.org/10.1093/ageing/afs166
Close - Share Icon Share
Abstract
Objectives: the objective of the present study was to explore the association between cardiovascular risk and cognitive decline in adults aged 50 and over.
Methods: participants were older adults who participated in the English Longitudinal Study of Ageing. Outcome measures included standardised z-scores for global cognition, memory and executive functioning. Associations between cardiovascular risk factors and 10-year Framingham risk scores with cognitive outcomes at 4-year and 8-year follow-ups were estimated.
Results: the mean age of participants (n = 8,780) at 2004–05 survey was 66.93 and 55% were females. Participants in the highest quartile of Framingham stroke risk score (FSR) had lower global cognition (b = −0.73,CI: −1.37, −0.10), memory (b = −0.56, CI: −0.99, −0.12) and executive (b = −0.37, CI: −0.74, −0.01) scores at 4-year follow-up compared with those in the lower quartile. Systolic blood pressure ≥160 mmHg at 1998–2001 survey was associated with lower global cognitive (b = −1.26, CI: −2.52, −0.01) and specific memory (b = −1.16, CI: −1.94, −0.37) scores at 8-year follow-up. Smoking was consistently associated with lower performance on all three cognitive outcomes.
Conclusion: elevated cardiovascular risk may be associated with accelerated decline in cognitive functioning in the elderly. Future intervention studies may be better focused on overall risk rather than individual risk factor levels.
Comments
We are grateful for Prof Sharrett's (1) interest in our paper (2). He suggests that adjusting for baseline could lead to bias and that use of a difference score should be preferred. The use of differences scores has been shown to be associated with greater risk of bias (Vickers, 2004; Senn, 2006). Differences scores suffer from several limitations. One concerns regression to the mean. Participants with high scores will worsen, on average, and those with low scores may improve. Difference scores may sometimes be highly skewed or asymmetrically distributed, violating the assumptions for parametric tests (Bonate, 2000). Difference scores are based on baseline and follow-up measures that include error. If error is sufficiently large relative to the variance between patients, the net result might be to introduce more, rather than less, error into the estimate of the exposure effect. As Dimitrov and Rumrill (2003) observed, where comparison groups cannot be assumed to be equivalent on the baseline, baseline adjustment should be the preferred method for analysis of data. Baseline adjustment has also been suggested to provide a more powerful test of group differences comparing to differences in scores (Fitzmaurice, 2001). We have explored several sensitivity analyses, comparing the 4-year follow-up model with and without baseline adjustment, which led to the same conclusions, differing only slightly in effect size, increasing therefore our confidence in the study results. Thus we are confident that the reported method of analysis should be preferred on theoretical grounds, and that the analyses were robust in sensitivity analyses.
References Sharett AR. Cardiovascular risk factors and cognitive decline. Age Ageing 2013. e-letter Dregan A, Stewart R, Gulliford M. Cardiovascular risk factors and cognitive decline in adults aged 50 and over: a population-based cohort study. Age Ageing 2012m [Epub ahead of print] Vickers AJ and Altman DG. Analysing controlled trials with baseline and follow up measurements. BMJ 2001; 323:1123-1124. Senn S. Change from baseline and analysis of covariance revisited. Stats Med. 2006; 25:4334=4344. Bonate PL. Analysis of Pretest-Posttest Designs. Chapman & Hall, Boca Raton 2000. Dimitrov DM and Rumrill PD. Pretest-posttest designs and measurement of change. Work 2003; 20:159-165. Fitzmaurice G. A conundrum in the analysis of change. Nutrition 2001; 17:360-361.
Conflict of Interest:
None declared
A. Richey Sharrett Johns Hopkins Bloomberg School of Public Health
Re: Dregan A, Stewart R, Gulliford MC. Cardiovascular risk factors and cognitive decline in adults aged 50 and over: a population-based cohort study. Age Ageing 2012. Sir: Dregan et al. report in the English Longitudinal Study of Aging that smoking and other risk factors were associated with 4-year decline in cognitive test scores1. However, the statistical model they used adjusts for baseline cognitive score, thereby permitting a bias which may be substantial and non-conservative. This bias, described by Glymour2, results from the correlation induced by baseline adjustment between errors in baseline score measurement and the exposure, in this case, smoking. If the bias is substantial in the English Longitudinal Study of Aging data, a more appropriate analysis may not in fact support the associations reported. An alternative, namely examining smoking in relation to the simple difference between baseline and follow-up cognitive scores would avoid baseline adjustment bias. The authors' concern for potential confounding by cognitive ability at baseline is appropriate, but use of the difference in cognitive scores as the outcome avoids most confounding by stable subject characteristics, such as completed education or native cognitive ability.
1. Dregan A, Stewart R, Gulliford MC. Cardiovascular risk factors and cognitive decline in adults aged 50 and over: a population-based cohort study. Age Ageing 2012.
2. Glymour MM, Weuve J, Berkman LF, Kawachi I, Robins JM. When is baseline adjustment useful in analyses of change? An example with education and cognitive change. Am J Epidemiol 2005;162(3):267-278.
Conflict of Interest:
None declared