Between October 1, 2004, and September 1, 2007, a total of 1969 non-demented participants were randomly selected and gave their consent for the study. There were 329 subjects with MCI and 1940 cognitively normal subjects. Neuropsychiatric data were available on 319 of the 329 participants with MCI (97.0%) and 1590 of the 1640 cognitively normal persons (97.0%). summarizes the demographic data. There were almost equal number of men and women among cognitively normal subjects; however, there were more men in the MCI group. As expected, subjects with MCI were older than cognitively normal subjects. Hence, we controlled for age (as a continuous variable) by entering it as a covariate in the multivariate analysis. Within the MCI group, there were more subjects with amnestic MCI (232/319 = 72.7%) than with non-amnestic MCI (87/319 = 27.3%). Within the amnestic MCI group, 61.2% were men whereas within the non-amnestic MCI group, 47.1% were men.
Demographics Characteristics of Study Participants
The median education of the cognitively normal group was 13 years and that of the MCI group was 12 years (P< 001). The difference between the 2 groups remained significant when education was dichotomized at 12 years of education (P< 001). All the analyses that compared the OR of neuropsychiatric symptoms between the MCI and cognitively normal group were adjusted by age, sex, and education.
displays the frequency of neuropsychiatric symptoms in MCI and cognitively normal elderly participants, along with odds ratios, associated 95% CI, and P values. About 51% of MCI subjects and 27% of cognitively normal subjects had 1 or more neuropsychiatric symptoms (). The prevalence of neuropsychiatric symptoms in MCI subjects was significantly higher than in cognitively normal persons; however, there was no difference between the 2 groups regarding hallucination and aberrant motor behavior. Symptoms were ordered by descending magnitude of the population attributable risk which takes into consideration both the frequency of a symptom and the magnitude of the OR. The most distinguishing neuropsychiatric features between MCI and cognitively normal controls were apathy (OR, 4.53; 95% CI, 3.11–6.60; P<.001), followed by agitation (OR, 3.60; 95% CI, 2.18–5.92; P<.001), anxiety OR, 3.00; 95% CI, 2.01–4.48; P<.001), irritability (OR, 2.98; 95% CI, 2.11–4.22; P<.001), and depression (OR, 2.78; 95% CI, 2.06–3.76; P<.001). Delusions, hallucinations, and euphoria were rare events in MCI and virtually absent in the cognitively normal group. For instance, delusion was present in 11 out of 319 MCI subjects (3.4%), and in 6 out of 1590 cognitively normal subjects (0.4%). Thus, the OR of delusion was large and the corresponding CI was wide (OR, 8.12; 95% CI, 2.92–22.6; P<.001). However, the population attributable risk for delusion was only 2.62% as compared to 14.60% for apathy. There were relatively more events for disinhibition, with 15/319 (4.7%) in MCI and 26/1590 (1.6%) in the cognitively normal group. There was no difference between the 2 groups regarding hallucinations (P=.693).
Prevalence of Neuropsychiatric Symptoms in Cognitively Normal Subjects and Subjects with MCI
Distribution by number of neuropsychiatric symptoms in cognitively normal subjects and subjects with MCI.
displays stratified analyses by MCI subtypes. These analyses were conducted to explore whether the prevalence of neuropsychiatric symptoms varied by MCI subtype. There were 232 subjects with amnestic and 87 subjects with non-amnestic MCI. The comparison between amnestic and non-amnestic MCI was made by computing ORs and the corresponding 95% CIs for each neuropsychiatric domain. The OR was computed by comparing subjects with a specific MCI subtype with all the cognitively normal subjects. Hallucinations were not significantly associated in either group. The OR and 95% CI for euphoria approached significance in the amnestic MCI group (OR, 2.44; 95% CI, 0.49–12.2; P=.276) whereas it was significant in the non-amnestic MCI group (OR, 6.64; 95% CI, 1.33–33.1; P=.021). However, one should interpret this finding with caution because euphoria was a rare event (2 persons with euphoria out of 232 individuals with amnestic MCI and 2 persons with euphoria out of 87 individuals with non-amnestic MCI).
Prevalence of Neuropsychiatric Symptoms in Subjects with Amnestic MCI or with Non-Amnestic MCI
The OR and 95% CI for apathy were higher in amnestic MCI (OR, 5.17; 95% CI, 3.44–7.77; P<.001) than in non-amnestic MCI (OR, 2.82; 95% CI, 1.42–5.58; P=.003). Similarly, the OR for agitation and irritability were slightly higher for amnestic MCI than for non-amnestic MCI. On the other hand, the OR for depression, anxiety, disinhibition, and delusion were higher in non-amnestic than in amnestic MCI. The OR of delusion for the non-amnestic group (OR, 12.7; 95% CI, 3.70–43.6; P<.001) was almost twice that for the amnestic MCI type (OR, 6.65; 95% CI, 2.07–21.4; P=.001); however, the symptom was rare in both groups. The OR for appetite was comparable between the 2 groups.
We used the demographic data reported in this manuscript (age, sex, and education) to compute propensity scores for each subject. We then used these scores in analyses that weighted the data more heavily towards subjects with higher propensity for missing data or for refusal to participate in the study. We performed 2 sets of sensitivity analyses. In the first set of analyses, we adjusted the observed results back to the complete dataset of 1969 subjects who participated in the study. In this analysis, the propensity score reflected the propensity of missing data on a per-variable basis. In the second set of analyses, we adjusted the observed results back to all subjects who were eligible for the study (1969 participants plus 1657 subjects who refused and 669 with partial participation).15
In neither of these assessments did we observe markedly different results before and after propensity adjustment.
We illustrate our findings using the missing data on night-time behavior. The primary analyses showed an odds ratio of 1.80 (95% CI, 1.25–2.60). The propensity-weighted analysis adjusted for missing data yielded an almost identical odds ratio of 1.79 (95% CI, 1.24–2.58). The other adjustments for missing data were even smaller because far fewer observations were missing for the other neuropsychiatric symptoms. In the analyses for refusal to participate, the adjustments back to all the individuals who were eligible for the study produced relatively minor differences in odds ratios. For only 1 of the symptoms (aberrant motor behavior), the adjustment resulted in a qualitatively different conclusion because the odds ratio increased from 2.30 (95% CI, 0.70–7.61; P=.171) to 3.17 (95% CI, 1.04–9.66; P=.042). However, even this difference was relatively small and statistically not significant.