For amyloid positive cognitively unimpaired individuals, the best predicting compositive measure included gender, and changes in ADAS delayed recall, MMSE, SDMT, and Trailmaking Test B.
Using a cohort of cognitively unimpaired (CU) individuals with and without subjective or mild cognitive impairment (MCI), investigators established minimally clinically important differences (MCIDs) for group-based worsening in test scores for 8 commonly used cognitive tests. The triangulated MCIDs might help guide clinicians and researchers on clinically relevant cognitive decline and formulate future clinical trials with appropriate measures.1
For CU, results showed potentially triangulated MCIDs of –1.5 on Mini-Mental State Examination (MMSE) scores, 1.4 for Alzheimer’s Disease Assessment Scale 10-word delayed recall scores, 5.5 for Stroop Color and Word Test scores, –2.8 for Animal Fluency scores, –2.9 for Letter S Fluency scores, –3.5 for Symbol Digit Modalities Test (SDMT) scores, 11.7 for Trailmaking Test (TMT) A scores, and 24.4 for TMT B scores. Triangulated MCID for those with MCI showed changes of –1.7, 1.1, 9.3, –2.9, –1.8, –3.8, 13.0, and 20.1 for the respective cognitive test scores.
"The present findings are important because there is no prior consensus on MCIDs for cognitive test outcomes in AD trials, yet FDA specifically highlights that a clinically meaningful improvement on cognitive test scores should be shown before approval of the drug,” senior investigator Sebastian Palmqvist, MD, PhD, associate professor, Lund University, and colleagues, wrote.
The study aimed to establish cut-offs for cognitive test changes that show meaningful magnitude of treatment effect, as well as investigate which single and combinations of cognitive test differences best corresponds to a clinically meaningful decline. To do so, investigators first calculated MCIDs associated with a change of at least 0.5 or 1.0 on Clinical Dementia Rating-Sum of Boxes (CDR-SB) for cognitive tests that covered the domains of executive function, attention, episodic and semantic memory, as well as visuospatial function.
The investigators then triangulated MCIDs for clinical use for CU, MCI, and amyloid positive CU patients, using receiver under the curve (ROC) analyses. Designed to capture efficacy and clinical relevance for early AD, a change of 1 in CDR-SB score could be either a change of 0.5 points in two boxes, or a change of 1 point in one box. Increases in CDR-SB score have been previously identified as having both face and predictive validity for identifying people who are later diagnosed with probable AD or another dementia.
The study included 451 CU individuals and 292 participants with MCI. Of those in the CU group, 16% (n = 116) were amyloid positive. Psychometric criterion reliable change index (RCI), used to evaluate whether a change of an individual score is considered statistically significant, were similar for CU and MCI groups for MMSE, ADAS, and Animal Fluency. On the other hand, investigators found larger RCIs for MCI compared with CU in Stroop, Letter S, SDMT, and TMT A and B.
Using logistic regression models, a clinically relevant change was defined as a discrimination between progression of CDR-SB of at least 0.5 vs CDR-SB change of 0 for CU and 0 vs at least 1 for those with MCI. To find the most optimal combination of test differences to estimate a cognitive change, Palmqvist et al examined all cognitive test changes in the model for CU to identify a model with the lowest Akaike Information Criterion (AIC). AIC accounted for the trade-off between model fit and sparsity, with lower numbers indicating a better model. When combined, the best model for CU individuals was a combination of age and changes in ADAS delayed recall, MMSE, and TMT B, which produced an AUC of 0.79 (95% CI, 0.72-0.86) for identifying a clinical change. For those with MCI, the best predicting model included changes in Stroop, MMSE, and age, which had an AUC of 0.82 (95% CI, 0.76-0.88).
In a separate analysis only including amyloid positive CU, changes in ADAS delayed recall, Stroop, SDMT, and TMT B, as well as gender, represented the best predicting cognitive measure. After first removing Stroop because of insignificance (P = 0.12), the best predicting composite cognitive measure included ADAS delayed recall, MMSE, SDMT, TMT B, and gender, which produced an AIC of 130.5 and AUC of 0.87 (95% CI, 0.79-0.94; sensitivity, 75%; specificity, 88%).