Participant characteristics
Participants’ sociodemographic characteristics are summarized in Table 1. The 26 incident cases developed dementia after a mean of 33.96 months (SD=25.71), and the 88 controls remained free of dementia for a mean follow-up duration of 45.39 months (SD=27.46). Follow-up length was significantly longer in controls (χ2=4.955, p=.026). Age at Time0 (χ2=2.970, p=.085) and education (χ2=3.325, p=.068) were comparable between groups. The incident dementia group included a significantly greater proportion of males (χ2=14.954, p<.001). A significantly greater proportion of incident dementia cases reported concern about their memory compared to controls (χ2=4.769, p=.029).
Table 1. Mean (SD) characteristics of incident dementia cases and controls.
|
Incident dementia cases
(n=26)
|
Controls
(n=88)
|
Statistic
|
p
|
Age at Time 0
|
75.50 (7.67)
|
71.61 (11.48)
|
2.798a
|
.094
|
Years of education
|
15.38 (3.03)
|
16.61 (2.76)
|
3.325a
|
.068
|
Sex (% males)
|
76.9%
|
34.1%
|
14.954a
|
<.001
|
Mean follow-up length (years)
|
2.85 (2.22)
|
3.75 (2.29)
|
4.466a
|
.035
|
Subjective sense of decline at any time (% yes)
|
69.2%
|
44.8%
|
4.769a
|
.029
|
Baseline MMSE
|
27.42 (2.28)
|
28.57 (1.74)
|
15.758b
|
<.001
|
Baseline NPI symptom severity
|
76.33 (13.78)
|
83.2 (13.93)
|
2.592b
|
.110
|
Baseline NPI symptom distress
|
3.00 (2.15)
|
1.96 (2.16)
|
2.617b
|
.108
|
Baseline Hachinski score
|
1.08 (1.53)
|
0.77 (1.03)
|
0.626b
|
.430
|
Baseline smoking history (years)
|
11.54 (17.45)
|
13.64 (16.59)
|
0.228b
|
.634
|
Baseline alcohol abuse (% never)
|
80.8%
|
76.1%
|
1.525a,b
|
.219
|
Notes. SD=Standard deviation. NPI=Neuropsychiatric Inventory.
aχ² was used because variables were non-normally distributed (Mann-Whitney U) or binary.
bF statistic. Refers to longitudinal linear mixed models. Means refer to baseline only.
Risk factors for cognitive decline
The presence of neuropsychiatric symptom distress (F= 6.952, p=.009) and severity (F= 6.686, p=.010) was statistically different between groups, but the Group × Time interactions terms in the NPI distress model (F=0.542, p=.462) and the NPI severity model (F=0.314, p=.576) were not. Vascular risk factors were similar between incident cases and controls (F=0.009, p=.925), and there was no Group x Time interaction (F=1.995, p=.159). The groups did not differ on smoking (main effect: F=0.380, p=.539, interaction: F=0.042, p=.838) or alcohol abuse (main effect: F=1.619, p=.206, interaction: F=1.062, p=.304) nor on two-week prior use of any psychotropic medication after correcting for multiple comparisons (all main and interactions effects p>.002).
Cognitive outcomes
Of the 26 incident dementia cases, 17 (65.4%) reported memory as the predominant symptom first recognized as a decline in cognition, five (19.2%) reported executive difficulties (judgment, planning, problem-solving), and three (11.5%) reported language symptoms. One participant observed no change. BD was determined to be the primary etiology contributing to cognitive impairment in two participants (7.7%), one of whom also had depression and Parkinson’s disease as primary contributing factors. Fifteen participants (57.7%) had AD as a primary etiologic diagnosis, three (11.5%) had fronto-temporal lobar degeneration, two (7.7%) had Lewy body dementia. One participant had alcohol-related dementia, one had vascular dementia and one had normal-pressure hydrocephalus. In one participant, etiology was undetermined. These etiologic diagnoses were presumptive and determined by clinicians within individual ADCs based on clinical presentation.
Sex-adjusted linear mixed model results are illustrated in Figure 1. A main effect of Group was apparent in models involving immediate (F=41.762, p=.000) and delayed short story recall (F=44.440, p<.001), digit span forward (F=12.193, p=.001) and backward (F=12.741, p<.001), animal fluency (F=23.582, p<.001), vegetable fluency (F=25.283, p<.000), trails A (F=22.214, p<.000) and B (F=13.245, p<.000), DSS (F=22.079, p<.001) and confrontation naming (F=26.180, p<.001). A main effect of Time was apparent in models involving immediate (F=4.662, p=.032) and delated story recall (F=5.698, p=.018), animal fluency (F=18.701, p<.001), vegetable fluency (F=16.707, p<.001) and trails A (F=7.770, p=.006) and B (F=10.745, p=.001), but not digit span forward (F=1.440, p=.231) or backward (F=0.020, p=.887), DSS (F=1.531, p=.217) or naming (F=3.071, p=.081). A significant Group × Time interaction was present in models involving immediate (F=10.054, p=.002) and delayed story recall (F=11.096, p=.001), animal fluency (F=11.643, p=.001), vegetable fluency (F=5.625, p=.018), trails A (F=6.942, p=.009) and DSS (F=4.184, p=.042). No interactions were significant in digit span forward (F=3.092, p=.080) or backward (F=0.148, p=.700), trails B (F=1.888, p=.170), or naming models (F=3.071, p=.081).
To determine the earliest point at which between-group differences appeared in models yielding a significant interaction term, yearly differences between incident cases and controls were examined using sex-adjusted ANCOVA. Only data from Time-6 to Time0 were included, due to insufficient data at prior time points. Effect sizes are reported as partial eta squared (ηp), where values .01, .06, and .14 are considered small, medium and large, respectively (30). Group differences in DSS were first apparent five years before the incident group’s dementia diagnosis (F=5.514, p=.027, ηp=.187). Group differences in immediate (F=7.519, p=.009, ηp=.165) and delayed story recall (F=4.297, p=.045, ηp=.102) emerged four years before diagnosis. Animal fluency differences emerged three years prior to dementia diagnosis (F=5.555, p=.023, ηp=.114). Group differences in Trails A emerged two years before diagnosis (F=8.978, p=.004, ηp=.112). Vegetable naming group differences only became statistically significant at Time0 (F=27.787, p<.001, ηp=.353).
Because DSS and Logical Memory were the first to be affected in the incident dementia group, their usefulness as potential screening tools for dementia in BD was tested in exploratory post-hoc analyses. The presence of impairment (yes/no) on the DSS or Logical Memory immediate story recall at any time during the study (prior to their last visit) was first determined for each participant. Impairment was defined as performance below -1.5 SD based on published normative data(26). Sensitivity, specificity, accuracy, positive (PPV) and negative predictive value (NPV) were then calculated. Results indicated that DSS impairment at any time prior to dementia onset was associated with 0.54 sensitivity (95% CI [0.33—0.74]), 0.87 specificity (95% CI [0.77—0.93]), 0.80 accuracy (95% CI [0.71—0.87]) 0.54 PPV (95% CI [0.33—0.74]) and 0.87 NPV (95% CI [0.77—0.93]). Short story immediate recall impairment was associated with 0.46 sensitivity (95% CI [0.26—0.67]), 0.81 specificity (95% CI [0.71—0.88]), 0.73 accuracy (95% CI [0.64—0.81]), 0.41 PPV (95% CI [0.23—0.61]) and 0.84 NPV (95% CI [0.74—0.91]).
Neuroimaging
Intracranial volumes were compared across groups to ensure consistency of the automatic tracings. Intracranial volumes did not differ significantly between the incident dementia cases (M=1,540 cm3, SD=261 cm3) and controls (M=1,476 cm3, SD=211 cm3; t (10)=.477, p=.644). With the limited amount of imaging data provided for participants, hippocampal atrophy comparisons were largely underpowered. Therefore, hippocampal atrophy was only compared at the level of marginal means, providing a purely descriptive approach. A 7.4% reduction in hippocampal volume was found in the incident dementia cases (M=3.255 cm3, SE=.274 cm3) compared to controls (M=3.516 cm3, SE=.230 cm3) while controlling for age, intracranial volume, and number of days between Time0 and the date of MRI acquisition. This is on par with previous literature suggesting that mild to medium levels of dementia should result in a 5% to 10% reduction in hippocampal volume, respectively (31).