Neurology -- Correspondence for DeCarli et al., 63 (2) 220-227

Correspondence published:

Memory impairment, but not cerebrovascular disease, predicts progression of MCI to dementia: Jeremy Koppel (27 September 2004)

Reply to Koppel: Charles DeCarli, Dan Mungas, Danielle Harvey, Bruce Reed, Michael Wiener, Helena Chui, William Jagust (27 September 2004)

Memory impairment, but not cerebrovascular disease, predicts progression of MCI to dementia 27 September 2004

Jeremy Koppel,
North Shore-Long Island Jewish Health System
75-59 263rd Street Glen Oaks NY 11004
Send Correspondence to journal:
Re: Memory impairment, but not cerebrovascular disease, predicts progression of MCI to dementia

JKoppel{at}lij.edu Jeremy Koppel

DeCarli et al [1] report a cohort of subjects with mild cognitive impairment (MCI), investigating the impact of cerebrovascular disease on progression of MCI to dementia. The authors conclude that cerebrovascular risk factors did not predict progression to dementia.
Given the methodologic limitations of the data analytic techniques employed in the evaluation of the relationship between cerebrovascular risk factors and progression to dementia, the assertion that cerebrovascular risk factors did not predict progression to dementia is incautious.
First, the sample size in the investigation was limited to 52 subjects with 17 (33%) of those progressing to dementia and 35 (67%) remaining non-demented at the end of the observation period. As the sample was limited to a very small number of MCI converters to dementia, sophisticated statistical analytic tools such as logistic regression analysis can not be used to examine the relationship between increasing cerebrovascular risk and incipient dementia. Instead, the authors employed t-test as a method of comparison of baseline cerebrovascular risk in the groups of dementia converters versus non-converters, a methodology which lacks the sensitivity of logistic regression to evaluate relationships between multiple independent variables and discrete dependent variables (dementia and non-dementia).
Second, the authors created an ordinal scale of composite vascular risk drawn from six cerebrovascular risk factors, and caculated these as a sum from 0 to 6 of the presence of any of these factors. The authors have assumed, given the equal interval nature of their scale, that the difference in risk between having no cerebrovascular risk factors and having two is the same as the difference between having four cerebrovascular risk factors and having six. This may not be true. The authors comment that only 36% of subjects were free of cerebrovascular risk factors and 22% had three or more vascular risk factors , yet the mean composite risk for the dementia converters was 0.94 with SD 1.03 and for the non-converters was 1.67 with SD 1.60. Given the high degree of variance evident from the standard deviations, using the median of the entire sample for evaluation of central tendency and comparing "hi" and "low" interquartile ranges of cerebrovascuar risk non-parametrically for differences in the number of subjects progressing to dementia may have yielded more information.
Larger studies are needed to accurately estimate in the population the extent to which cerebrovascular risk impacts conversion from MCI to dementia.
References
1. DeCarli C, Mungas D, Harvey D, et al. Memory Impairment, but not cerebrovascular disease, predicts progression of MCI to dementia. Neurology 2004; 63:220-227.

Reply to Koppel 27 September 2004

Charles DeCarli,
University of California at Davis
4680 Y Street, Suite 3700, Sacramento, CA 95817,
Dan Mungas, Danielle Harvey, Bruce Reed, Michael Wiener, Helena Chui, William Jagust
Send Correspondence to journal:
Re: Reply to Koppel

cdecarli{at}ucdavis.edu Charles DeCarli, et al.

We thank Dr. Koppel for his comments regarding our preliminary observations on the impact of cerebrovascular disease (CVD) and conversion from mild cognitive impairment (MCI) to dementia. We concur with the need for further study as stated in the manuscript. However, we respectfully disagree with the notion that we were �incautious�, especially with our statistical approach.
We agree with Dr. Koppel�s point that the study was small. However, we believe that Dr. Koppel misunderstands our statistical approach. Our initial analyses that utilized t-tests, Wilcoxon rank sum and chi-square tests were simply descriptive. Our primary analysis utilized Cox proportional hazards models. Small groups do not violate use of proportional hazards they only limit the power to detect associations between test measures and clinical outcome. Importantly, Cox proportional hazards models include information about both whether and when clinical conversion occurred. This additional capacity enabled us to study the pattern of conversion, unlike logistic regression that only allows for associations between independent variables and whether or not conversion took place.
We thank Dr. Koppel for pointing out the limitations of our clinical CVD scale and agree with Dr. Koppel�s underlying concern that the scale is not truly interval. We note, however, that subsequent analysis of median split data did not change the observed relationships. Moreover, we note that clinical CVD data was analyzed solely as a secondary measure and was considered mostly as a confounding variable. Importantly, both the clinical CVD variable and the MRI CVD variables tracked conversion in a similar manner (i.e. evidence of CVD either by imaging or by medical history was associated with a lower frequency of conversion from MCI to dementia) supporting our presumption that CVD-related brain injury likely has a different time course to dementia than does Alzheimer�s disease (AD). Conversely, selection of individuals with predominant memory loss may emphasize AD pathology. Either way, we are hopeful that this preliminary data will stimulate further interest in understanding the combined effects of AD and CVD within individuals destine to develop dementia.