Dementia after ischemic stroke

The incidence of dementia after cerebral infarct has not been well studied in specific North American populations, ^[1,2] even though in stroke surveys dementia is common after cerebral infarctions. ^[3-6] In an unselected study of autopsied patients, those without dementia had more infarcts than those with dementia. ^[7] In our previous studies on the incidence of dementia, we estimated that 5% of patients with dementia suffered a cerebral infarct. ^[8] It has been difficult to estimate prevalence of stroke-related dementia, mainly because of lack of common diagnostic criteria. ^[9] Newly published criteria may help in future studies. ^[10,11] We took advantage of the detailed medical information for Rochester, Minnesota, residents to estimate the average annual age- and sex-specific incidence rates of dementia after first cerebral infarcts in this community.

Methods.

Medical diagnoses and surgical procedures for all Rochester, Minnesota, residents seen at any of the health care facilities (including the Mayo Clinic and its affiliated hospitals, Olmsted Medical Group, Olmsted Community Hospital, the former Rochester State Hospital, Veterans' Administration Hospital in Minneapolis, and University of Minnesota Hospitals in Minneapolis), by solo medical practitioners, and in nursing homes are indexed and entered into a computerized database and are available for retrieval by diagnosis. This medical records link-age system was used to retrieve the records of all residents of Rochester who were diagnosed as having stroke or disorders that might have been stroke during the years 1960 to 1984. All potential cases were screened by a trained nurse abstractor, and cases of first cerebral infarction were identified by the following criteria: focal neurologic deficit of acute onset, persisting for more than 24 hours, without clinical characteristics of intracerebral hemorrhage or another disease process, and CT of the head or autopsy, when performed, did not show evidence of intracerebral hemorrhage. The date of onset of infarction was ascertained.

The entire medical records of each person with a diagnosis of stroke were reviewed by a nurse who searched for specific diagnostic criteria for dementia, either before or after the diagnosis of stroke. To be accepted as dementia, there had to be documented evidence of previously normal intellectual and social function; decline of intellectual and/or cognitive and social function that was irreversible with medical or psychiatric treatment; evidence of memory impairment; and dementia sufficiently important to impair age-, education-, and occupation-appropriate lifestyle. Also, there had to be documented evidence for at least two of the following: disorientation; personality or behavioral problems; dyscalculia; aphasia, apraxia, or agnosia; and impairment of judgment or abstract thinking. The year of onset of the symptoms of dementia was also determined after review of all physician notes and all correspondence available in the entire medical record. The medical record was reviewed for >or=to 5 years after the last follow-up date if the patient survived. Therefore, when a diagnosis of dementia was made, the record usually documented the beginning of symptoms several years earlier. If a patient died when mild dementia was present, that patient could have been missed as a case of dementia, but such a patient might not have fulfilled all the criteria for dementia that were noted. Each case history was abstracted, including autopsy results (when available) and death certificates (when applicable). Thus, for this study we identified all patients with a first cerebral infarct occurring between January 1, 1960 and December 31, 1984 who did not have dementia before the onset of cerebral infarction. Each patient was a resident of Rochester, Minnesota, on the date of the cerebral infarction and had lived in Rochester for >or=to 1 year. In this way, we avoided including persons who moved to Rochester specifically for medical care.

We also ascertained the Rankin disability score ^[12] at 3 months after the onset of the cerebral infarct. Patients who were so impaired by their infarct that dementia could not be ascertained, usually because of severe neurologic impairment or severe aphasia, were excluded. Also, those patients who did not survive for 3 months after their cerebral infarct were excluded from the study. One of the investigators (EK) reviewed each medical history to assign a cause of dementia. All available information, including autopsy report, death certificate, and results of ancillary studies, was taken into consideration. The dementia that developed in patients after their first cerebral infarct may have been related to that infarct or a subsequent one or to another condition, for example, Alzheimer's disease (AD). If, on review of the medical care record, it appeared that after an initial cerebral infarct, the patient's condition stabilized or the patient partially recovered only to develop a progressively worsening dementia that could not be attributed to subsequent stroke or to other causes by clinical or autopsy observations, a diagnosis of AD was made.

Results.

During the period studied, 1,138 persons had a first cerebral infarct and survived it for 3 months. Among these persons, 96 were excluded because they had dementia before the infarct, and 71 others were excluded because the infarct caused such severe neurologic impairment or aphasia (or both) that dementia could not be ascertained. Thus, a total of 971 patients were available for study, of whom 485 (50%) were women. CT of the head or autopsy (or both) were performed in 75% of the patients after 1975, but autopsy was performed in 36% before 1975 and thus before the frequent use of CT. During 6,782 person-years of observation starting at 3 months after the stroke, dementia was judged to have developed in 196 patients among whom 47 had autopsies and 16 were still alive. Another 18 patients had some features of cognitive impairment but did not meet our criteria for dementia and so were not considered to be demented.

Cumulative incidence of dementia after a first cerebral infarct is given in Table 1. When the observations were censored at the time of the second stroke, this provided an estimate of the cumulative incidence of dementia associated only with the first cerebral infarct (Table 1). The standardized morbidity ratios in Table 1 are estimates of the excess risk in the cohort compared with the risk in the population with the same age and sex distribution as in the cohort.

(Table 1) indicates that in the first year after the stroke, the risk of dementia in the cohort was nearly nine times greater than in the population. This was modified only slightly when observations were censored at the second stroke. The incidence of dementia after the first cerebral infarct increased strikingly with age (Table 2), but the standardized morbidity ratio was highest in the younger age groups and decreased with increasing age.

If dementia had not developed within 1 year after baseline (3 months after the stroke), it occurred at about twice the annual rate that would have been expected in the community over a 25-year period (Table 3).

Among the variables noted in the Cox model, only those listed in Table 4 were significant in the univariable analyses adjusted for age. All others among those noted previously in the Statistical analysis section were not significant (p > 0.05). For the multivariable analysis, the final model included only age, male sex, second stroke, and mitral valve prolapse (Table 4). None of the cardiovascular variables other than mitral valve prolapse were significant. There also was no significant effect of location or laterality and no difference in the effect of mild stroke (Rankin disability scores 0, 1, and 2; n = 708) from that of moderate-to-severe stroke (Rankin disability scores 3, 4, and 5; n = 263).

Among the persons in the community in whom dementia developed, [approximately]80% had AD as at least a part of their dementia. ^[8,16] Among the patients with dementia in our stroke cohort, 41% had a diagnosis of AD, about half that of the proportion for AD in the community (p < 0.001).

The cumulative incidence of AD after first cerebral infarct in our cohort is shown in Table 5. Standardized morbidity ratios were estimated with expected rates derived from incidence rates of AD in Rochester, Minnesota. ^[8,16] These rates are based on the dementia judged to be partially or wholly due to AD. Generally, there was about a 50% increase in AD in the cohort compared with that in the community after the first year of observation and for 25 years after baseline. Analysis of person-years for the whole period of observation showed a similar result.

Discussion.

Dementia due to cerebrovascular disease has been difficult to define for various reasons: cerebral infarcts and dementia both occur with ever increasing frequency in elderly persons. They frequently coexist, both clinically and neuropathologically. Individual specific criteria for vascular dementia are still elusive despite recent efforts. ^[10,11] Sometimes, cerebrovascular disease and its risk factors may exist without causing acute well-defined episodes of neurologic deficit, leading to a diagnosis of acute cerebrovascular disease. There have not been any previous population-based studies of dementia related to stroke in the United States. Recent studies showed that stroke is an important risk factor for the subsequent development of dementia. Dementia attributed to cerebrovascular disease was surprisingly frequent in a group of very old Swedish retired persons. ^[17]

Our study takes advantage of the availability of epidemiologic data for both cerebral infarction and dementia over a long period in our community, where it is possible to obtain detailed medical evaluations for every person who received medical attention. The incidence of dementia after a cerebral infarct may underestimate vascular dementia in the community because we did not use neuroimaging data for diagnosis and we did not try to identify those causes of vascular dementia that are not associated with an acute temporal profile. The standardized morbidity ratio that we estimated is similar to that obtained by Tatemichi et al ^[4] in their hospitalized clinical series, which was not limited to persons with first stroke. They found an odds ratio of 9.4 for dementia 3 months after onset of stroke. In a more recent report on the risk of dementia, Tatemichi et al ^[5] concluded that after 52 months of follow-up, the relative risk of developing dementia was 5.5. Our study also showed that the very high standardized morbidity ratios became smaller as the interval between stroke and dementia increased. Our estimates of incidence of dementia after first cerebral infarction in Rochester are similar to the estimates from several Scandinavian countries ^[18-20] and to one from Beijing, China. ^[21]

In our study, we evaluated medical care records. Therefore, systematic standardized mental testing was not available, as it would be in a prospective study. However, we had access to data for repeated visits, often encompassing the whole adult life of the patients. This allowed us to assess the development of symptoms and signs over time for evolving and progressing dementia.

In the Rochester, Minnesota, historical cohort of persons who survived their first cerebral infarct, incidence of dementia within the first year after baseline (3 months after stroke) was nearly nine times greater than would have been expected in the population with the same age and sex distribution as the cohort. As shown in Table 1, most instances of dementia that occurred in the first year were noted soon after the stroke; thus, dementia was present at the beginning of observation (3 months after the stroke). Although there was a decreasing standardized morbidity ratio over time after stroke (Table 1), conditional on survival free of dementia for 1 year, the ratio remained at about 2 for observations up to 25 years after a cerebral infarct.

(Table 2) shows an increasing incidence of dementia after first cerebral infarct with increasing age but a decreasing standardized morbidity ratio with increasing age. This indicates the association of dementia with cerebral infarction regardless of age at infarction, but the expected incidence of dementia is low in younger ages, thus the higher standardized morbidity ratio. However, we did not observe any cases of dementia in persons younger than 50 years with cerebral infarction.

When observations after the first cerebral infarct were censored at second stroke, the cumulative incidence of dementia did not change much, indicating that the risk of dementia was very strongly associated with the first cerebral infarct. However, the multivariable Cox model showed that second stroke was also a highly significant variable in predicting dementia.

The significant variables that predicted dementia in the multivariable Cox regression analysis were age, sex, second stroke, and mitral valve prolapse. No cardiovascular variables other than mitral valve prolapse were significant. We did not show either an excess of dementia associated with any location of the infarct in the brain or a difference in the effect of mild stroke (Rankin disability scores 0, 1, and 2) versus severe stroke (Rankin disability scores 3, 4, and 5) judged at 3 months after the first cerebral infarct, which is contrary to the observations of Tatemichi et al ^[22] in a hospitalized cohort of patients with dementia. Only the second of multiple strokes was significant in the model, but the power to detect the independent effect of more strokes was limited.

The finding that mitral valve prolapse was associated with the occurrence of dementia in our cohort in the multivariable analysis was unexpected (Table 4). Of the 14 patients with mitral valve prolapse, dementia developed in 7, 2 of which were soon after onset of cerebral infarction and the others over nearly 11 years of observation. This allowed mitral valve prolapse to be selected in the final model of the Cox multivariable regression analysis. A bootstrap resampling procedure ^[23] using 1,000 random samples from the data for validation of the model confirmed the final model, including mitral valve prolapse as a significant variable (Table 4). We do not have a satisfactory explanation for the relevance of this association.

Among the persons in our cohort in whom dementia developed, the proportion with AD was about half that in the community. However, the Kaplan-Meier method, censoring at dementia from cause other than AD, showed about a 50% annual excess of AD in the cohort as compared with the community for 25 years of observation. Perhaps after a stroke, AD is manifested sooner or there is a detection bias due to closer and more frequent observation of patients. An alternative explanation is that the recorded and indexed health care information, such as we used, may underestimate the incidence of AD in the community.