Predicting Risk of Cognitive Decline in Very Old Adults Using Three Models: The Framingham Stroke Risk Profile; the Cardiovascular Risk Factors, Aging, and Dementia Model; and Oxi-Inflammatory Biomarkers

Stephanie L. Harrison, PhD; Anton J. M. de Craen, PhD; Ngaire Kerse, PhD; Ruth Teh, PhD; Antoneta Granic, PhD; Karen Davies, PhD; Keith A. Wesnes, PhD; Wendy P. J. den Elzen, PhD; Jacobijn Gussekloo, PhD; Thomas B. L. Kirkwood, PhD; Louise Robinson, MD; Carol Jagger, PhD; Mario Siervo, PhD; Blossom C. M. Stephan, PhD

Disclosures

J Am Geriatr Soc. 2017;65(2):381-389. 

In This Article

Results

After excluding participants with a history of stroke or dementia at baseline, the analytical sample included 616 participants in the Newcastle 85 + Study, 444 in the Leiden 85-plus Study, and 396 in the LiLACS NZ Study. The baseline characteristics of each sample are shown in Table 1.

Framingham Stroke Risk Profile

Global Cognitive Function. For the individual studies, the HRs for greater risk of impaired global cognitive function were all greater than 1 for the highest FSRP groups ( Table 2 ). Meta-analysis of pooled results showed a greater risk of impaired global cognitive function with a higher FSRP (MMSE: HR = 1.46, 95% CI = 1.08–1.98, P = .01), although in the linear mixed models, none of the associations between the FSRP and global cognitive function (cross-sectional results) or global cognitive decline were significant (P > .05, Table 3 ).

Domain-specific Cognitive Function. Upon examining the specific cognitive domains, which were available only for the Newcastle 85+ and Leiden 85-plus studies, higher FSRP scores were associated with greater risk of impaired speed of reaction times for the Newcastle 85+ Study (HR = 1.42, 95% CI = 1.06–1.91, P = .02) (Table S2). In the linear mixed models, the FSRP was cross-sectionally associated with speed scores for the Leiden 85-plus Study only (Letter Digit Coding Test: (B = −2.19 (standard error (SE) 0.913), P = .02) (Table S3).

The CAIDE Model

Global Cognitive Function. Higher CAIDE scores were associated with an increased risk of impaired global cognitive function in the meta-analysis of pooled MMSE results (HR = 1.53, 95% CI = 1.09 to 2.14, P = .01). For the linear mixed model results, the CAIDE model was cross-sectionally associated with global cognitive function in the Leiden 85-plus Study [B (SE) = 0.493 (0.112), P < .001] and the LiLACS NZ Study [B (SE) = 0.348 (0.109), P = .001].

Domain-specific Cognitive Function. The CAIDE model was not longitudinally associated with any of the domain specific cognitive test scores for the Newcastle 85+ or Leiden 85-plus studies for either the Cox proportional hazard models or the linear mixed models. However, the linear mixed models showed that the CAIDE model was cross-sectionally associated with speed, attention and memory scores for the Leiden 85-plus Study.

Oxi-inflammatory Load

Global Cognitive Function. Higher oxi-inflammatory load was associated with incident global cognitive impairment in the meta-analysis of pooled results (Newcastle 85+ Study and Leiden 85-plus Study only: HR = 1.73, 95% CI = 1.04–2.88, P = .04) (Figure 1). In linear mixed models, oxi-inflammatory load was not longitudinally associated with global cognitive decline, but a cross-sectional association was observed between higher oxi-inflammatory load and poorer global cognitive function scores at baseline (B = 0.320 (SE 0.127), P = .01) in the Leiden 85-plus Study.

Figure 1.

Forest plots to show the pooled risk for impaired global cognitive function, measured using the Mini-Mental State Examination (MMSE) in all three cohorts, associated with higher Cardiovascular risk factors, Aging and Dementia model (CAIDE) and Framingham Stroke Risk Profile (FSRP) risk scores and oxi-inflammatory load.

Domain-specific Cognitive Function. Higher oxi-inflammatory load was associated with greater risk of incident impairment of attention in the Newcastle 85+ Study (HR = 1.58, 95% CI = 1.05–2.36, P = .03) and Leiden 85-plus Study (HR = 2.18, 95% CI = 1.27–3.74, P = .01) (Table S2). Higher oxi-inflammatory load was also associated with greater risk of impairment of speed in the Newcastle 85+ Study (HR = 1.85, 95% CI = 1.24–2.75, P = .003), but this association was not observed in the Leiden 85-plus Study. In the linear mixed models, higher oxi-inflammatory load was not longitudinally associated with any of the domain-specific cognitive test results for the Newcastle 85+ or Leiden 85-plus studies, although the linear mixed models showed that higher oxi-inflammatory load was cross-sectionally associated with better attention scores for the Newcastle 85+ Study (B = 0.058 (SE 0.028), P = .04) (Table S3).

Adding Oxi-inflammatory Load to the FSRP and CAIDE Model

To determine whether oxi-inflammatory load could be used to improve prediction of the FSRP or CAIDE models, oxi-inflammatory load scores were added to FSRP and CAIDE scores, as described in Methods. In the meta-analysis of pooled results (Newcastle 85+ Study and Leiden 85+ Study only), adding oxi-inflammatory load scores improved the prediction of the CAIDE (HR = 1.93, 95% CI = 1.39–2.67, P < .001) and FSRP scores (HR = 1.65, 95% CI = 1.17–2.33, P < .001).

Sensitivity Analysis

The Leiden 85-plus data set did not include years of education and instead used a measure based on the education system in the Netherlands. Therefore, the analysis of the CAIDE score was repeated with education excluded from the calculation to determine whether differences in the operationalization of education were the reason for the discrepant results between the studies. When education was excluded, the result changed, and the CAIDE score was no longer associated with global cognitive impairment (Leiden 85-plus: highest- vs lowest-risk tertile: HR = 1.25, 95% CI = 0.81–1.92, P = .30).

Analyses for the Cox proportional hazards models were repeated using tertiles for oxi-inflammatory load (middle category as reference) (MMSE: HR = 1.36, 95% CI = 0.96–1.92, P = .07).

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....