*3.4. Effects of Aging*

Linear regression models were performed to evaluate the effects of aging independent of disease status. Scatterplots and the fitted regression line for each cognitive test are presented as Figure S1. There was a significant main effect for aging on nine of the 17 measures, controlling for disease status. The three memory and two language measures that were affected by aging were DSMSE Language, DSMSE Memory, Purdue Pegboard, RADD-2 Expressive Language, RADD-2 Receptive Language, Rivermead Recognition, and SRT (Fs(3, 130) = 3.33 to 25.48, ts = −3.62 to −2.01, *p*s = 0.02 to 0.0001). Since there was only one significant interaction between age and disease status, RADD-2 Sensorimotor (F(3, 130) = 5.04, t = −5.08, *p* = 0.002), we decided not to include age as a covariate in all subsequent analyses in order to minimize the likelihood of a type II error.

### *3.5. Path Model 3.5. Path Model*

Path analysis was used to examine the magnitude and significance of the three domain composites, adjusting for sex and premorbid ID level. Every domain composite was significantly related to MCI-DS status (see Table 4, but the path model only explained about half of the variance between CS and MCI-DS (R<sup>2</sup> = 0.49) and had poor model fit, as the RMSEA was 0.49 (much greater than the 0.08 limit), CFI was 0.51 (not close to the 0.90 threshold), and TLI was −0.94 (far from the 0.90 cutoff). Path analysis was used to examine the magnitude and significance of the three domain composites, adjusting for sex and premorbid ID level. Every domain composite was significantly related to MCI-DS status (see Table 4, but the path model only explained about half of the variance between CS and MCI-DS (R2 = 0.49) and had poor model fit, as the RMSEA was 0.49 (much greater than the 0.08 limit), CFI was 0.51 (not close to the 0.90 threshold), and TLI was −0.94 (far from the 0.90 cutoff).

RADD-2 Expressive Language, RADD-2 Receptive Language, Rivermead Recognition, and SRT (Fs(3, 130) = 3.33 to 25.48, ts = −3.62 to −2.01, *p*s = 0.02 to 0.0001). Since there was only one significant interaction between age and disease status, RADD-2 Sensorimotor (F(3, 130) = 5.04, t = −5.08, *p* = 0.002), we decided not to include age as a covariate in all

subsequent analyses in order to minimize the likelihood of a type II error.

*Brain Sci.* **2021**, *11*, x FOR PEER REVIEW 9 of 15

**Table 4.** Path model comparing CS to MCI-DS adjusted for sex and premorbid ID level (*n* = 144). **Table 4.** Path model comparing CS to MCI-DS adjusted for sex and premorbid ID level (*n* = 144).


MCI-DS status was associated with lower scores on every domain composite, with the language/EF composite being the most affected, followed by the memory composite, then the visuomotor composite. the language/EF composite being the most affected, followed by the memory composite, then the visuomotor composite.

MCI-DS status was associated with lower scores on every domain composite, with

## *3.6. Structural Equation Model 3.6. Structural Equation Model*  Figure 1 displays the relationship between the latent, measurement, and exogenous

Figure 1 displays the relationship between the latent, measurement, and exogenous variables (sex, premorbid ID, and MCI-DS status, respectively). The 17 cognitive measures and their latent variables predicted MCI-DS status substantially better than the composite scores used in the path model. The SEM model had good model fitness since RMSEA = 0.08, CFI = 0.91, and TLI = 0.90. variables (sex, premorbid ID, and MCI-DS status, respectively). The 17 cognitive measures and their latent variables predicted MCI-DS status substantially better than the composite scores used in the path model. The SEM model had good model fitness since RMSEA = 0.08, CFI = 0.91, and TLI = 0.90.

**Figure 1.** Structural equation model of how MCI-DS affected performance on the 17 cognitive measures and 3 latent variables. Solid lines indicate paths that were statistically significant, *p*s = 0.001. Ovals represent the endogenous variables with their indicator variables represented by white rectangles. CFI = Comparative Fit Index, CI = confidence interval, **Figure 1.** Structural equation model of how MCI-DS affected performance on the 17 cognitive measures and 3 latent variables. Solid lines indicate paths that were statistically significant, *p*s = 0.001. Ovals represent the endogenous variables with their indicator variables represented by white rectangles. CFI = Comparative Fit Index, CI = confidence interval, DSMSE = Down Syndrome Mental Status Examination, EF = executive function, Exp = expressive, ID = intellectual disability, mMMSE = Modified Mini-Mental Status Exam, MCI-DS = Mild Cognitive Impairment-Down syndrome, RADD-2 = Rapid Assessment of Developmental Disabilities-2nd Edition, Rec = receptive, RMSEA = root mean square error of approximation, Model Fit Indices: Chi-square = 299.20, CFI = 0.91, RMSEA = 0.08, 90% CI = 0.07–0.09.

Sex was not related to outcomes on any of the cognitive measures. MCI-DS status was associated with greater memory impairment than premorbid ID level was, whereas premorbid ID level was associated with greater language/EF impairment than MCI-DS status was. MCI-DS status and premorbid ID had similar effects on visuomotor performance.

As shown in Table 5, the latent construct of memory was a stronger indicator of MCI-DS status than the latent construct of visuomotor, which was a stronger indictor than language/EF. Although the path and SEM models indicated the same directionality of domain impairments for those with MCI-DS, the magnitude and ranking were different. (The path model identified the language/EF composite to be a stronger indicator of MCI-DS than the memory and visuomotor composites.) In short, giving all tests within the same domain differential weights as in the SEM can lead to considerably different conclusions than when all tests within the same domain are given equal weight as in the path analysis.

**Table 5.** The relationship between sex, premorbid ID, MCI-DS, and cognitive functioning.


CR = critical ratio, EF = executive function, MCI = mild cognitive impairment, PID = premorbid level of ID, SE = standard error.

The largest language/EF differences between the CS and MCI-DS groups were observed on the following tests: Boston Naming, DSMSE Language, RADD-2 Expressive Language, RADD-2 Digit Span Forward, and RADD-2 Similarities (SSs = 0.76 to 0.88). The largest differences between the two groups on memory tests were on DSMSE Memory and SRT (SSs = 0.85 to 0.86). The largest differences between the two groups on visuomotor tests were on Block Design and DSMSE Visual Spatial (SSs = 0.80 to 0.85).
