Development and Validation
Demographic characteristics of the training and test samples from Project FRONTIER can be found in . In the training sample, there were 204 participants consensus judged as cognitively normal and 52 as MCI. In the training sample, the following items were significantly endorsed more often among the MCI group than the normal cognition group: feeling of worse memory problems (χ2=12.39, p<0.001), feeling downhearted and blue (χ2=6.97, p=0.008), feeling worthless (χ2=5.58, p=0.02), frequently feel like crying (χ2=6.50, p=0.01), and trouble concentrating (χ2=7.82, p=0.005). Of note, a positive endorsement on each of these items is in the direction of positive depression, therefore reverse scoring was not needed. The DepE was generated by summing the responses of each person on these 5 items resulting in a score ranging from 0 to 5. The mean DepE score in the training sample was 1.0 (sd = 1.3, range = 0-5).
The test sample consisted of 203 participants designated as cognitively normal on consensus review and 60 MCI. In the test sample, the mean DepE score was 1.0 (sd = 1.4, range= 0.5) with MCI cases scoring significantly higher (1.7, sd = 1.8) than cognitively normal participants (0.8, sd = 1.1) (t=3.59, p=0.001). The DepE was significantly negatively related to RBANS index scores of Immediate Memory (B=-2.22, SE=.37, p<0.001), Visuospatial skills (B=-1.11, SE=0.26, p<0.001), Language (B=-1.03, SE=0.21, p<0.001), Attention (B=-2.56, SE=0.49, p<0.001), and Delayed Memory (B=-1.54, SE = 037, p<0.001). Higher DepE scores were also related to significantly poorer executive functioning (EXIT25; B=0.65, SE=0.19, p=0.001). To illustrate the independent impact of DepE scores on neuropsychological testing, analyses were re-run including GDS total scores (minus DepE items). DepE scores were significantly related to RBANS indices of Immediate Memory, Visuospatial skills, Language, Attention, and Delayed Memory independent of GDS total scores. GDS total scores were only significantly related to Delayed Memory and Attention scores. GDS total scores were significantly related to EXIT scores, but not DepE scores. Therefore, DepE scores appear to be most related to memory abilities with global depression scores related to measures of attention and executive functioning. shows the correlations between DepE scores and demographic and neuropsychological data in the FRONTIER test sample.
Correlation of DepE with demographic variables and cognitive test scores.
Next, logistic regression was utilized to determine the risk of being diagnosed with MCI as a function of DepE scores within the test sample. DepE scores significantly increased risk for MCI diagnosis (odds ratio [OR] = 2.04; 95% CI=1.54-2.69), which was the only significant predictor aside from age (OR=1.09; 95% CI=1.05-1.13) and education (OR=0.82; 95% CI=0.71-0.95). In a conditional stepwise forward logistic regression, age entered into the model first, followed by the DepE and then education; no other variables entered into the model.
Mean DepE scores were significantly higher among AD cases (1.39, SD=1.27, range = 0-5) as compared to controls (0.37, SD=0.71, range 0-3) (F=27.88, p<0.001). MCI patients scored between AD cases and normal controls (0.90, SD=1.20, range = 0-3) though the sample size was too small for statistical comparison. Based on the findings from Project FRONTIER, neuropsychological tests of memory and language were evaluated in TARCC as were scores of global cognition and disease severity. Correlations between DepE scores and demographic and neuropsychological data are presented in .
The results of the linear analyses indicated that DepE scores were significantly related to poorer scores on all measures (see . To illustrate the specific DepE – cognition link independent of global depression scores, these baseline analyses were re-run including GDS total score (minus DepE questions) in the models. With the exception of FAS, results remained unchanged with DepE scores being a significant predictor of cognitive test scores with total GDS scores not being significantly related to any cognitive outcome variables. For FAS, total GDS score was significantly related to test scores (p=0.03) with DepE scores not being significant (p=0.07).
Relation between DepE & DepE2 scores and neuropsychological test scores in TARCC.
Next a logistic regression model was created with AD versus normal control as the outcome variable; age, gender, ethnicity, education, ApoEε4 presence (yes/no), GDS total score and DepE scores were entered as the predictor variables. Age (OR=1.18, 95% CI= 1.12-1.24, p<0.001), ApoEε4 status (OR=2.42, 95% CI=1.13-5.19, p=0.02) and DepE scores (OR=2.49, 95% CI=1.40-4.43, p=0.002) were the only significant predictors of AD status. In the forward conditional stepwise logistic regression, the order of entry into the model was age, DepE, and ApoEε4 status. The DepE alone was a significant predictor of AD status using receiver operating characteristic (ROC) curve analysis (Area Under the Curve [AUC] = 0.74 (95% CI=0.68-0.81), p<0.001). The results remained consistent when predicting AD/MCI status. However, even though DepE index scores were significantly related to AD status in both ApoEε4 carriers (OR=2.81, 95% CI=1.2-6.5, p=0.02) and ApoEε4 non-carriers (OR=8.03, 95% CI=3.06-21.08, p<0.001), the effect was clearly stronger for non-carriers.
The TARCC is a longitudinal cohort and, therefore, preliminary analyses were conducted with data from participants who had available DepE scores at 12- and 24-month follow-up evaluations (DepE2, DepE3). A total of 142 participants (AD n=67, 1 MCI, NC n=70, 4 other) had available information for 12-month analyses. The mean (standard deviation) age, education and DepE2 scores for the participants (44 men and 98 women) was as follows: 72.9 (9.7), 14.5 (3.2), and 0.6 (0.9), respectively. DepE2 scores were significantly associated with poorer scores on all measures (see . DepE2 was also a significant predictor of AD status (compared to controls) at 12-month follow-up visits: OR=4.02 (95% CI=1.73-9.32, p=0.001). A total of 90 participants (AD n=41, NC n=49) had 24-month follow-up data available for calculation of a DepE, which continued to be a significant predictor of AD case status (OR = 3.84, 95% CI=1.13-13.10, p=0.03); age (OR=1.19, 95% CI=1.09-1.30, p<0.001) was the only other variable that remained a significant predictor of case status. Comparisons with neuropsychological test scores were not carried out due to sample size.
In order to conduct preliminary analyses on the impact of baseline DepE scores on 24-month cognitive change, longitudinal change scores were created for the MMSE and CDR scores (due to missing data and sample size, calculations were not possible with all neuropsychological test data). Of the 90 total TARCC participants with 24-month follow-up data, 65 participants (AD n=41, NC n=49) had all available baseline and 24-month follow-up MMSE and CDR sum of boxes (CDR-SB) scores along with DepE scores and appropriate covariates (age, gender, ethnicity, education; ApoEε4 was not available on sufficient numbers to be included as a covariate). Change scores were calculated as follows: ΔMMSE = baseline MMSE -24mo MMSE and ΔCDR-SB = baseline CDR-SB -24mo CDR-SB. Therefore, a positive ΔMMSE was indicative of decline whereas a negative ΔCDR-SB was indicative of decline. The mean (standard deviation, range) of ΔMMSE and ΔCDR-SB were 2.66 (3.37, 0-16) and -2.23 (3.05, -11-0.5), respectively. Baseline DepE scores were significantly related to both change scores (see .