BACKGROUND: Research into the pathophysiology of age-associated cognitive function and decline requires a valid estimate of cognitive function. However, this estimation can be grossly influenced by a selective loss to follow-up. METHODS: We investigated the influence of health selection on the estimated age-associated cognitive function and decline by studying the effect on this estimation of study design and of the handling of multiple and missing data. We used linear regression analyses and linear mixed models to assess cognitive function from cross-sectional and longitudinal data. Repeated measures of cognitive function (assessed with dedicated neuropsychological tests) were carried out in 2 independent population-based cohort studies: the Rotterdam Study (3719 participants; mean age 71 years) and the Leiden 85-plus Study (369 participants; age 85 years). RESULTS: The effect of age on cognitive function was greater in cross-sectional analyses when all participants were included than when analyses were restricted to participants with repeated measurements. The decline in cognitive function over 4.6 years of follow-up was intermediate between the cross-sectional estimates from the total sample and from the restricted sample. Moreover, the estimated decline in cognitive function was larger when using a short follow-up than when using the complete follow-up over 5 years. The estimated decline using linear mixed models was similar to analyses including those with a complete follow-up over 5 years. CONCLUSION: Selection for health and survival results in better age-specific cognitive test scores and less cognitive decline. Statistical methods handling multiple and missing data do not fully correct for this bias.
BACKGROUND: Research into the pathophysiology of age-associated cognitive function and decline requires a valid estimate of cognitive function. However, this estimation can be grossly influenced by a selective loss to follow-up. METHODS: We investigated the influence of health selection on the estimated age-associated cognitive function and decline by studying the effect on this estimation of study design and of the handling of multiple and missing data. We used linear regression analyses and linear mixed models to assess cognitive function from cross-sectional and longitudinal data. Repeated measures of cognitive function (assessed with dedicated neuropsychological tests) were carried out in 2 independent population-based cohort studies: the Rotterdam Study (3719 participants; mean age 71 years) and the Leiden 85-plus Study (369 participants; age 85 years). RESULTS: The effect of age on cognitive function was greater in cross-sectional analyses when all participants were included than when analyses were restricted to participants with repeated measurements. The decline in cognitive function over 4.6 years of follow-up was intermediate between the cross-sectional estimates from the total sample and from the restricted sample. Moreover, the estimated decline in cognitive function was larger when using a short follow-up than when using the complete follow-up over 5 years. The estimated decline using linear mixed models was similar to analyses including those with a complete follow-up over 5 years. CONCLUSION: Selection for health and survival results in better age-specific cognitive test scores and less cognitive decline. Statistical methods handling multiple and missing data do not fully correct for this bias.
Authors: Frederick W Unverzagt; Lin T Guey; Richard N Jones; Michael Marsiske; Jonathan W King; Virginia G Wadley; Michael Crowe; George W Rebok; Sharon L Tennstedt Journal: J Int Neuropsychol Soc Date: 2012-03-09 Impact factor: 2.892
Authors: Cyrus A Raji; Oscar L Lopez; Lewis H Kuller; Owen T Carmichael; William T Longstreth; H Michael Gach; John Boardman; Charles B Bernick; Paul M Thompson; James T Becker Journal: Neurobiol Aging Date: 2011-09-23 Impact factor: 4.673
Authors: F W Unverzagt; L A McClure; V G Wadley; N S Jenny; R C Go; M Cushman; B M Kissela; B J Kelley; R Kennedy; C S Moy; V Howard; G Howard Journal: Neurology Date: 2011-11-08 Impact factor: 9.910
Authors: Sarah R Gillett; Evan L Thacker; Abraham J Letter; Leslie A McClure; Virginia G Wadley; Frederick W Unverzagt; Brett M Kissela; Richard E Kennedy; Stephen P Glasser; Deborah A Levine; Mary Cushman Journal: Clin Neuropsychol Date: 2015-05-15 Impact factor: 3.535
Authors: Jennifer Weuve; Eric J Tchetgen Tchetgen; M Maria Glymour; Todd L Beck; Neelum T Aggarwal; Robert S Wilson; Denis A Evans; Carlos F Mendes de Leon Journal: Epidemiology Date: 2012-01 Impact factor: 4.822
Authors: Virginia G Wadley; Frederick W Unverzagt; Lisa C McGuire; Claudia S Moy; Rodney Go; Brett Kissela; Leslie A McClure; Michael Crowe; Virginia J Howard; George Howard Journal: Ann Neurol Date: 2011-05-26 Impact factor: 10.422