Cognitive Reserve and the Relationship Between Depressive Symptoms and Awareness of Deficits in Dementia

The current study included 66 community-dwelling individuals referred to an outpatient memory clinic for neuropsychological evaluation, who were subsequently diagnosed with questionable or mild dementia. We included patients in the study if they had a global Clinical Dementia Rating (CDR) 13 score of 0.5 (questionable dementia) or 1 (mild dementia) and had a close relative or friend to act as informant. The sample was evenly split between males and females (50% each). Racial composition of the sample was predominantly Caucasian (98%). Participants ranged in age from 56 to 92, with an average age of 74.65 (SD=6.71) years. Educational level ranged from 7 to 20 years, with an average of 13.47 (SD=2.95) years of education. Informants were the participant’s spouse (45%) or a child/other close relative or friend (55%). Diagnosed etiologies of cognitive impairment were: mild cognitive impairment/questionable dementia (N=31), probable Alzheimer disease (N=18), vascular dementia (N=9), mixed Alzheimer disease and vascular dementia (N=4), dementia secondary to movement disorder (N=2), and other (N=2).

We used the 26-item version of the Cognitive Difficulties Scale (CDS; Patient and Informant Versions) 1, 14 to measure awareness. The CDS includes items related to difficulties in attention, concentration, orientation, memory, praxis, language, and daily functioning, and asks the respondent to rate on a Likert scale how often (from 0=“not at all” to 4=“very often”) he or she has had difficulty on various tasks in the past month. It has adequate psychometric properties, including high correlation with performance on neuropsychological measures of memory and attention (r=−0.51), 15 good test-retest reliability (r=0.77), 14 and strong correlation between the 26-item and full length forms (r=0.90). 1 We defined lack of awareness as discrepancy between informant ratings on the family version of the CDS and participant ratings on the self-report version of the CDS (i.e., informant minus patient ratings) 1 because previous research indicates informant ratings of cognitive difficulties correlate more highly with actual performance-based testing than do dementia patient self-ratings. 7 Items missing data from either respondent of the participant-informant pair were omitted from both, and the average item discrepancy of the remaining items was the primary measure of awareness of deficits.

We used the Reading subtest of the Wide Range Achievement Test (WRAT) 16, 17 to measure cognitive reserve. Patients completed the WRAT-R or WRAT-III reading subtest, depending on when they were seen. The two tests are highly interrelated. 17 Psychometric properties are good, with high internal consistency of the reading subtest (WRAT-R r=0.93–0.98; WRAT-III r=0.91–0.97) and excellent test-retest reliability (WRAT-R r=0.94; WRAT-III r=0.91–0.98). Previous research has utilized WRAT reading as a measure of cognitive reserve, 9 and it is the preferred measure of premorbid verbal intelligence, as it better estimates lower ranges of verbal intelligence than other similar measures. 18 A score of 100 is average performance on the WRAT reading subtest.

We used the Beck Depression Inventory, Second Edition (BDI-II) 19 to measure depressive symptomatology. The BDI-II is a 21-item self-report measure that assesses the presence and severity of depression. Items on this scale reflect typical symptoms of depression, including affective, cognitive, and vegetative features. The psychometric properties of the BDI-II are good, with high internal consistency (α=0.92−0.93) and good criterion validity (HRSD-R, r =0.71). 19

Finally, we used the Modified Mini-Mental State Examination (3MS) 20 as a brief measure of current global cognition. The 3MS is an extended Mini-Mental State Examination (MMSE) 21 that includes birthplace and date, semantic fluency, abstract reasoning, and delayed word recall. It has a maximum score of 100 points. The 3MS is a good screening tool for dementia, with excellent test-retest reliability in dementia patients (r=0.93 after 98 days) and 88% sensitivity, 90% specificity, 29% positive predictive value, and 90% negative predictive value. 22

In the overall sample, CDS discrepancy ranged from −1.32 to 3.04, with an average of 0.30 (SD=0.87). 3MS scores ranged from 60 to 98, with an average of 84.53 (SD=8.49). WRAT reading scores ranged from 77 to 122, with an average of 103.03 (SD=12.04). BDI-II scores ranged from 0 to 45, with an average of 6.80 (SD=7.36).

As expected, informant ratings significantly correlated with performance on the 3MS, r=−0.27, two-tailed p<0.05, and MMSE, r=−0.28, two-tailed p<0.05, indicating greater report of cognitive impairment by the informant was associated with poorer cognitive performance. In contrast, patient self-ratings were uncorrelated with performance on 3MS, r=0.01, two-tailed p=ns, and MMSE, r=0.07, two-tailed p=ns.

Hierarchical regression analysis with CDS discrepancy (unawareness) as the dependent variable was used to determine relative contributions of cognitive reserve, depressive symptoms, and their interaction. We entered variables in the following order: 3MS (to control for the severity of dementia), WRAT reading, BDI-II, and the WRAT reading/BDI-II interaction. Variables were centered to reduce multicollinearity.

Analyses demonstrated that WRAT reading accounted for a significant proportion of the variance in CDS discrepancy above and beyond contribution of 3MS, ΔR ² =0.08, F (1, 63)=6.32, two-tailed p<0.01, although BDI-II did not significantly contribute to CDS discrepancy, ΔR ² =0.00, F (1, 62)=0.01, two-tailed p=ns. These findings were qualified by an interaction between WRAT reading and BDI-II on CDS discrepancy, ΔR ² =0.09, F (1, 61)=7.52, two-tailed p<0.05.

To further investigate this interaction, we split the overall sample into high (high cognitive reserve=WRAT reading>100) and low cognitive reserve groups (low cognitive reserve=WRAT reading<100) groups. Fisher’s z[prime] transformation 23 comparison of the depression and CDS discrepancy zero-order correlations between low and high cognitive reserve groups revealed a significant difference (high cognitive reserve: r=−0.34, p<0.05; low cognitive reserve: r=0.35, p=ns), z=2.79, two-tailed p<0.05. These correlations indicate that increased depressive symptoms are associated with decreased discrepancy between the informant and participant (i.e., better awareness) in high cognitive reserve individuals, but this relationship was not significant in low cognitive reserve individuals.

To ensure the robustness of this follow-up finding, we examined demographic differences between the high and low cognitive reserve groups ( Table 1 ). As expected, significant differences between the high and low cognitive reserve groups emerged for level of education, t (64)=−3.17, two-tailed p<0.01, with higher education in the high cognitive reserve group, providing evidence for divergent validity of the groups. The groups did not differ significantly in age, gender, or race. However, CDR scores differed between groups, with individuals meeting criteria for questionable dementia more likely to be in the high cognitive reserve group χ ² (66)=4.29, two-tailed p<0.05. To ensure stability of the findings in light of the group difference in CDR scores, we conducted partial correlations accounting for 3MS performance. The partial correlations indicated the same pattern found in the zero-order correlations holds after controlling for performance on a measure of current global cognition (high cognitive reserve: partial r=−0.33, two-tailed p<0.05; low cognitive reserve: partial r=0.29, two-tailed p=ns). These correlations suggest that even when current global cognition is accounted for, higher levels of depressive symptoms are associated with less discrepancy in ratings of cognitive difficulties between the informant and high cognitive reserve participants. In low cognitive reserve individuals, the relationship between depressive symptoms and lack of awareness of cognitive difficulties remained nonsignificant, even after accounting for performance on a measure of current global cognition.