NINDS AIREN neuroimaging criteria do not distinguish stroke patients with and without dementia

Methods.

Results.

One hundred thirty-seven patients were recruited. Of these, 103 were free of cognitive decline post stroke and 34 had developed dementia. Twelve participants were excluded (three: movement, four: no evidence of cerebral infarction on MR, one: unable to tolerate the MR examination, one: tumor, three: dementia pre stroke). The final study population consisted of 125 patients. Ninety-eight were stroke patients without dementia and 27 with poststroke dementia.

The patients with dementia were more likely to be male and more likely to have a history of myocardial infarction but were not older. As would be expected, the dementia group had significantly greater impairment of cognitive function. A full description of the characteristics of the sample is shown in table 1.

None of the patterns of cerebrovascular disease (including large infarcts, multiple infarcts, specific strategic infarcts, and extensive white matter lesions) highlighted as essential for a diagnosis of probable VaD within the NINDS AIREN criteria was significantly different between stroke patients with or without dementia (table 2). Of particular interest, 13 (13%) of the stroke patients without dementia had single infarcts >50 mm, although they had significantly greater impairment of memory and executive function than patients with smaller infarct volumes (table 3).

In a more extensive comparison of the number of infarctions in specific regions by dementia status, none of the other comparisons was significant (frontal cortical infarcts: Z = 0.04, p = 0.96; parietal cortical infarcts: Z = 1.4, p = 0.17; occipital cortical infarcts: Z = 0.4, p = 0.72; temporal cortical infarcts: Z = 0.8, p = 0.41; total cortical infarcts: Z = 0.7, p = 0.48; frontal white matter infarcts: Z = 0.04, p = 0.97; parietal white matter infarcts: Z = 0.3, p = 0.78; occipital white matter infarcts: Z = 0.5, p = 0.60; total white matter infarcts: Z = 0.7, p = 0.48; frontal cortical <9 mm: Z = 0.7, p = 0.47; parietal cortical <9 mm: Z = 0.7, p = 0.46; occipital cortical <9 mm: Z = 0.7, p = 0.46; temporal cortical <9 mm: Z = 0.9, p = 0.36; total cortical <9 mm: Z = 0.2, p = 0.81; frontal white matter lacunes: Z = 0.7, p = 0.46; parietal white matter lacunes: Z = 0.2, p = 0.85; occipital white matter lacunes: Z = 1.0, p = 0.33; temporal white matter lacunes: Z = 0.7, p = 0.48; total white matter lacunes: Z = 0.6, p = 0.52).

In contrast, the stroke patients with dementia had more MTA than those without (right: mean 2.13 ± 1.23 vs 1.46 ± 0.95, Z = 2.5, p = 0.01; left: mean 1.96 ± 1.12 vs 1.29 ± 1.0, Z = 2.5, p = 0.01; see table 2).

Regression analysis was used to look at the relationship between extensive periventricular white matter lesions, age, and hippocampal atrophy. Right-sided MTA was associated with extensive white matter lesions (p < 0.028), whereas there was a trend for greater white matter lesions with left-sided hippocampal atrophy (p < 0.053). The inclusion of age in the regression analysis showed similar results (right MTA: p < 0.016; left MTA: p < 0.059).

Discussion.

The current study focuses on 125 older stroke patients (mean age 80) with (n = 27) or without (n = 98) dementia from representative hospital-based stroke registers in the North East of England. The substrates of cerebrovascular disease highlighted as critical for a diagnosis of VaD according to the NINDS AIREN criteria were not significantly more frequent in stroke patients with dementia than those without. Furthermore, in this cohort of older stroke patients, there was no association between large, multiple, strategic, or bilateral cortical infarcts, white matter infarcts, or extensive white matter disease (as defined within the NINDS AIREN criteria) and the presence of dementia. In contrast, MTA was significantly associated with dementia and is perhaps a more substantial contributor to poststroke dementia in older stroke patients than has been previously recognized. Dementia was evaluated by an expert clinician according to operationalized criteria, based on a detailed clinical and neuropsychological evaluation at 3 months post stroke; hence, the absence of dementia is a robust diagnosis. In the people with dementia, medical records and a clinical history from an informant were utilized to ensure that dementia did not precede the stroke. The quality of information was generally good, but the possibility that a small proportion of these individuals had dementia before the stroke cannot be dismissed. In addition, the sample size of patients with dementia is modest, which widens the confidence intervals for the percentages in table 2. Of particular interest, there were 13 patients with infarct volumes >50 mm who did not have dementia, although these individuals did have subtle cognitive impairments with mild deficits of memory and attention. This refutes the suggestion that 50 mm of infarct volume is invariably or even usually a sufficient substrate for dementia.

Although a number of studies indicate that the NINDS AIREN criteria have excellent specificity at separating patients with VaD from those with AD, the current data support a previous small neuropathologic study in older patients with cerebrovascular disease in suggesting that there are no clear-cut differences in the magnitude or distribution of cortical infarcts between patients with and without dementia in the context of cerebrovascular disease.24 The findings are also consistent with important recent work suggesting that subcortical ischemic VaD is associated with hippocampal and cortical atrophy but not the number or distribution of lacunes.33 Our results indicate that review of the criteria is appropriate. However, given the limited evidence base, available specific recommendations would be inappropriate. Although further studies are needed to inform the revision of criteria, accumulating evidence suggests there should be increased emphasis on the severity and distribution of white matter lesions.

When interpreting these findings, it is, however, important to consider the possibility that some of these individuals with hippocampal atrophy in the absence of dementia may be in the “preclinical” stages of a developing dementia. For example, many individuals meeting diagnostic criteria for mild cognitive impairment have some hippocampal atrophy (but not dementia),34 and several studies have indicated that this may be a predictor of future dementia in these people.35 In addition, many older people with cerebrovascular disease have some concurrent neurodegenerative changes, although this was not verified at autopsy in the current study. It is likely that many of these individuals had some “prestroke” hippocampal atrophy in the absence of dementia. In the context of a number of studies highlighting the interactive effects of neurodegenerative and vascular pathology,36,37⇓ it is likely that pre-existing “subclinical” atrophy may have made them more vulnerable to experiencing cognitive deficits after a subsequent stroke. As this is a common clinical presentation in older patients with cerebrovascular disease, diagnostic criteria need to be applicable to these individuals.

Although there was no association between the WMH criteria within the NINDS AIREN and the presence of dementia, the majority of evidence from previous studies indicates the importance of WMH as a substrate of dementia and as an association of key neuropsychological deficits. Perhaps of particular importance, a key neuropathologic study15 highlighted the importance of WMH but reported that many of the patients had insufficient WMH to meet the 25% criteria within the NINDS AIREN criteria. As this threshold was not empirically derived and does not seem useful in discriminating between stroke patients with and without dementia, there is a particular need to redefine this aspect of the criteria.

At first consideration, there appears to be a discrepancy between these data and some previous reports, but this is probably explained by the age of the cohorts. The size and number of infarcts may be an important association of dementia in younger stroke patients but may be less important in older individuals. However, this is only a hypothesis, and it is a limitation of the study that there was no younger comparison group. It is a concern that the distribution of vascular lesions identified in the NINDS AIREN criteria as indicative of VaD does not appear to distinguish people with and without dementia, as it therefore suggests that the relatively precise descriptions of qualifying lesions may be unnecessarily restrictive and may hence lead to “false negatives,” not only in studies of patients with cerebrovascular disease but in other contexts where the differential diagnosis of dementia is the key. We suggest that revised criteria or a caveat within the existing criteria may be needed for stroke patients with dementia in general and for older patients with dementia in the context of cerebrovascular disease. These criteria will need to relax the requirements for specific types of cortical infarction and improve the operationalization of what constitutes sufficient WMH. In addition, the potential importance of concurrent MTA in these patients should be discussed. Improving the sensitivity of operationalized criteria for VaD is a particular priority now that specific pharmacologic treatments are becoming available for these patients. Hence, a review of the neuroimaging specifications is especially timely.