BACKGROUND: The Alzheimer’s disease (AD) composite score (ADCOMS) was developed to measure outcomes among patients with mild cognitive impairment (MCI) due to AD and mild dementia due to AD and was shown to be more sensitive to cognitive changes than the component measures from which it was derived. The component measures include the mini mental state examination (MMSE), clinical dementia rating scale—sum of boxes (CDR-SB), and the AD cognitive subscale (ADAS-Cog). However, the validity of using ADCOMS to stage AD severity has yet to be established in patients with earlier disease.
OBJECTIVE: This study was to derive ADCOMS scores to stage severity of dementia and to evaluate the utility of the derived scores in distinguishing patients across AD stages, especially patients with MCI due to AD from the mild AD dementia stage.
METHODS: Patients enrolled (N = 2,073) in the Alzheimer’s Disease Neuroimaging Initiative (ADNI) (data downloaded on December 13, 2018) were assessed at baseline and split into a derivation sample (n = 1,034; cognitively normal [CN] = 393, MCI = 464, and AD = 177) and a validation sample (n = 1,035; CN = 384, MCI = 474, AD = 177). Data from assessments at visit 24 (N=1,262) were used to determine ADCOMS cutoff scores to stage moderate and severe AD (derivation sample: CN=231, MCI=246, and AD= 54; validation sample: CN=209, MCI=237, AD=185). Stage classification based on the existing criterion measures of CDR global, CDR-SB, ADAS-Cog, and MMSE were used to generate receiver operating characteristic (ROC) curves to indicate the optimal ADCOMS cutoff scores for each disease stage. This analysis was conducted in the derivation sample at baseline and visit 24 for every two adjacent scores on the criterion measures that are indicative of each AD stage (e.g., CDR global scores of 0 and 0.5). The optimal ADCOMS cutoff was established using an empirical estimation method based on maximizing the product of sensitivity and specificity. Equality of the ROC curves between the derivation and validation samples was tested using a χ2 test. The diagnostic accuracy in distinguishing between patients with MCI and mild AD was assessed within the validation sample. This was achieved by restricting the analysis of baseline data to patients within the validation sample whose global CDR score was 0.5 (MCI=471 and AD=84) and calculating an ROC curve using the ADCOMS scores. To evaluate the diagnostic accuracy of the ADCOMS scores in distinguishing between patients with mild AD and moderate/severe AD, this process was repeated using data from visit 24 within the validation sample among patients with a global CDR score of 1 (n = 70 mild AD and n = 24 moderate or severe AD).
RESULTS: The following cutoff ADCOMS scores were identified to stage AD severity: < 0.29 is indicative of normal cognition, 0.29 to < 0.50 is indicative of MCI, 0.50 to 0.80 is indicative of mild AD, and > 0.80 is indicative of at least moderate AD. The reliability of these cutoff ADCOMS scores was supported by the tests of equality between the derivation and validation samples: all χ2 test results indicated no significant difference between the samples. When applied to the validation sample, the diagnostic accuracy test showed that 82% of patients with MCI and mild AD, and 73% of patients with mild and moderate/severe AD were correctly classified against the classification based on CDR-global score. When applied to the data at visit 24, the proportion of correctly classified patients ranged from 72% to 92%. The study is limited by the small sample size at the moderate (n=94) and severe AD stages (n=8).
CONCLUSIONS: The ADCOMS scores can be consistently mapped to the existing criterion measures including CDR-SB and other instruments in AD severity classification. Further validation efforts may help establish ADCOMS as a new diagnostic tool and criterion measure to quantify AD progression and staging.
