3.1 Demographic Characteristics and performance of CMMSE and DRS-CV
Demographic information and the scores of CMMSE and DRS-CV among four groups including NC, MCI, mild AD and moderate AD were presented in Table 1. No difference was found in age (F = 2.428, p = 0.065), sex distribution (χ2 = 0.647, p = 0.886) and education (F = 1.500, p = 0.214) among the four groups.
The total scores of CMMSE and DRS-CV were significantly lower in patient groups. The scores of different cognitive domains were ranked in the order of NC >MCI > mild AD > moderate AD group. Subscales of conceptualization and memory was affected at a relatively earlier stage.
3.2 Correlations between demographic variables and DRS total and subscale scores
We analyzed the effects of age, sex and education on DRS-CV total and subscale scores for NC. There was no difference on DRS-CV total and subscale scores between male and female (all p > 0.05). Correlation analysis revealed that age was correlated to DRS-CV total score (r = -0.264, p = 0.002) and initiation/perseveration score (r = -0.223, p = 0.009), and education was correlated to DRS-CV total score (r = 0.254, p = 0.003), attention score (r = 0.363, p < 0.001) and initiation/perseveration score (r = 0.271, p = 0.001) (Table1).
3.3 ROC analysis of DRS-CV total and subscale scores for discriminating MCI from NC, MCI from mild AD and mild AD from moderate AD (Supplementary Figure 1)
The most appropriate cutoffs of DRS-CV were 131 in the discrimination between MCI and NC (sensitivity 65.3%, specificity 67.6%, AUC 0.708), 120 in the discrimination between MCI and mild AD (sensitivity 84.5%, specificity 85.0%, AUC 0.924), 103 in the discrimination between mild and moderate AD (sensitivity 79.7%, specificity 78.4%, AUC 0.846) (Table 2). Using these cutoff scores, the DRS-CV showed excellent sensitivity and specificity in the discrimination between MCI from mild AD and mild AD from moderate AD, but poor sensitivity and specificity in the discrimination between MCI and NC, suggesting that the predictive information captured by the DRS total score was only reasonably good to detect AD but not MCI.
The memory subscale showed good sensitivity and specificity in the discrimination between MCI from mild AD and mild AD from moderate AD, similar to the DRS total score. In comparison, the AUC of the memory subscale for discriminating MCI from mild AD was similar to the initiation/perseveration subscale (0.885 versus 0.845, p = 0.201) and larger than other three subscales (attention: 0.885 versus 0.681, p < 0.001; construction: 0.885 versus 0.605, p < 0.001; conceptualization: 0.885 versus 0.695, p < 0.001). For discriminating mild AD from moderate AD, the AUC of the memory subscale was lager AUC than other four subscales, but not significant (attention: 0.823 versus 0.740, p = 0.100; initiation/perseveration: 0.823 versus 0.755, p = 0.179; construction: 0.823 versus 0.720, p = 0.051; conceptualization: 0.823 versus 0.670, p = 0.005). All DRS-CV subscale scores had poor sensitivity and specificity for discriminating MCI from mild AD.