MCI is a heterogeneous clinical entity with incidence rates that vary substantially by age, sex, and subtype. The MCSA showed a higher incidence of MCI in men compared to women that is contrary to a higher risk of dementia in women reported by some studies20
and contrary to a similar incidence rate of dementia in men and women observed in this same Olmsted County population.21
The higher risk of MCI in men was consistent for both aMCI and naMCI. In addition, our study showed higher incidence rates for aMCI than for naMCI. Finally, we observed higher rates of MCI in persons with lower education and who were not married. The observed higher incidence of MCI in men is consistent with the higher prevalence of MCI in men observed in this same Olmsted County population.3
Similarly, the increased incidence rates in subjects with lower education and in subjects who were not married were consistent with the findings for prevalence in our study.3
We were unable to directly compare our incidence rates with some of the population-based studies previously published for several reasons.1
First, unlike MCSA, age- or age- and sex-specific incidence rates were not reported for some studies.2,22,23
Second, the study participants were much younger in one study,22
or had a much different age distribution in another.23
Third, one study was conducted in a healthy cohort where MCI risk factors were an exclusion criterion.24
Finally, some studies only investigated aMCI.22–27
shows a comparison of age- and sex-specific incidence rates from our study and 2 other studies that provided rates for aMCI and naMCI separately: the Leipzig Longitudinal Study of the Aged from Germany,28
the Kungsholmen Project from Sweden,29
and the MCSA from the United States (present study). For aMCI, our incidence rates were higher than in the Swedish study but similar to the rates in the German study (B). The sex pattern in our study was opposite to the German study but was consistent with the Swedish study. For naMCI, our incidence rates were lower than the rates in both the German and the Swedish studies (C). The sex pattern was similar across the 3 studies; however, the sex differences in risk were small in the German study.
Comparison of age- and sex-specific incidence rates for mild cognitive impairment (MCI) across studies
The variability in incidence rates across studies may, in part, be due to differences in study design and implementation of criteria for MCI. Studies that relied solely on the Mini-Mental State Examination or that used an algorithmic categorization based on neuropsychological tests, studies that required a memory complaint from the participant (as in the earlier MCI criteria),30
studies that did not collect information from an informant, studies that did not require the judgment of a clinician, or studies that applied MCI criteria retrospectively to previously collected data may have yielded different patterns. For example, the lower rates of naMCI in the MCSA may be due to the use of a consensus panel diagnosis rather than an algorithmic diagnosis. MCI diagnoses that rely solely on a neuropsychological algorithm will necessarily overestimate the frequency of deficits in several nonamnestic domains compared to a single memory domain. The variability of MCI incidence rates across studies underscores the need to standardize approaches to MCI diagnosis.1
The demographic predictors of incident MCI other than sex that were observed in the MCSA were consistent with those in some other studies. Older age was associated with an increased risk of MCI in several,2,22,28,29,31
but not in all studies.25,26,32
Low education was associated with increased risk in some studies.28,31
Amnestic MCI was associated with older age27,29
and low education.22,27
Nonamnestic MCI was associated with low education.2
Our findings for marital status are consistent with findings in the German study and do not appear to be due to a diagnostic bias.28
Subjects who had lost their spouse at baseline were expected to experience an underdiagnosis or a delayed diagnosis of MCI because of lack of an optimal informant. Indeed, previously married subjects had significantly higher scores on the Clinical Dementia Rating scale or the Functional Activity Questionnaire at the time of first diagnosis of MCI than married subjects (data not shown). The biological basis of this association remains uncertain. Subjects who had lost their spouse had a significantly higher rate of depression than married subjects at baseline15
; however, adjustment for depression at baseline did not modify the association of marital status with MCI incidence noticeably (, footnote d). Therefore, other mechanisms such as changes in dietary habits or lack of daily support in managing existing diseases (e.g., compliance with medications) may also be involved.
In our study, we observed 2 interesting patterns of interaction. The antagonistic interaction of sex and age was significant for aMCI but not for naMCI, whereas the synergistic interaction of sex and education was significant for naMCI but not for aMCI. One possible interpretation is that men with low education have behaviors and life events that may expose them to several risk factors leading to naMCI. Alternatively, men with low education may have been at cognitive disadvantage early in life and their low education was already a marker of limited cognitive reserve.33
These 2 patterns of interaction between sex and age and sex and education seem to differ for aMCI and naMCI, suggesting that risk factors for MCI should be investigated considering both type of MCI and sex separately. Analyses that pool men with women and aMCI with naMCI may fail to identify specific risk factors because the specific associations would be diluted away.
We observed a reversion rate from MCI to normal of approximately 12.3% per year. This rate is lower than or similar to previous estimates,2,34,35
and does not account for subjects who will revert back a second time from normal to MCI in an extended follow-up. Because our consensus diagnoses were made without knowledge of the performance at prior visits, our estimates of conversion to MCI and reversion from MCI were not biased by clinical expectations.
Our study has several strengths. First, it is a large population-based study that was specifically designed to investigate the incidence of MCI and its subtypes using published criteria.5
Second, the comprehensive clinical evaluation and the consensus approach provided reliable diagnoses of MCI and MCI subtypes; in particular, we did not use a multistage evaluation where subjects who screened negative were not fully evaluated. Third, repeated assessments of cognitive status were performed without knowledge of the cognitive status at the preceding evaluation; thus we were able to investigate the stability of MCI diagnosis over time. Finally, the passive ascertainment of incident events through the medical records linkage system for subjects who were lost to follow-up provided more accurate estimates of MCI incidence.6,7,16
Our study had some limitations. First, nonparticipants at the baseline evaluation were more likely to be men, older, or to have type 2 diabetes mellitus and a high Charlson Comorbidity Index compared to participants, as reported elsewhere.4
However, when we adjusted for potential nonparticipation bias using propensity scores, the unadjusted and adjusted estimates were similar. Second, the population of Olmsted County is primarily of European ancestry, thus the findings may not apply to other ethnic groups.