Risk Prediction Of Alzheimer’s Disease With Cognitive And Functional Measures:A Joint Modeling Approach

Objective:Our study aims were to explore whether and how the variation of cognitive function and functional impairment affect the onset of Alzheimer’s disease(AD)in patients with Mild cognitive impairment(MCI).Under this context,we analyzed how covariates linked to cognitive function,functional impairment,and the risk of AD,respectively,and then compared the accuracy of different longitudinal measures for risk prediction of AD.Methods:We enrolled 501 participants with MCI at baseline from the Alzheimer’s Disease Neuroimaging Initiative(ADNI)into the survey.The Alzheimer Disease Assessment ScaleCognitive(ADAS-Cog)11,13 and the Mini-Mental State Examination(MMSE)evaluated the cognitive function of patients.And functional impairment was assessed using the Functional Assessment Questionnaire(FAQ).We used joint modeling to examine the association between measures variation and AD progression.The linear mixed model constructed the longitudinal sub-model which described the evolution of a repeated measure over time,and a Cox proportional hazards model constructed the survival sub-model.The goodness of fit measures included log likelihood,Akaike information criterion(AIC)and Bayesian information criterion(BIC).We used Area Under the Curve(AUC)to assess the predictive accuracy of the longitudinal measures.Finally,joint models were used to provide dynamic individual predictions.Results:Among 501 MCI patients,277 patients converted to AD during follow-up.ADASCog11 joint model showed: age(b=0.087,P=0.001),family status(b=-1.570,P=0.002)and ApoEε4 carriers(b=1.426,P<0.001)significantly associated with the cognitive impairment;the carrier of ApoEε4 also significantly associated with the increased risk of AD event(hazard ratio(HR): 1.443,95%CI:1.089~1.913),and ADAS-Cog 11 score increases in one unit would lead to the 1.193 times increase of the AD conversion risk(95% CI:1.159~1.228).In ADAS-Cog13 joint model,age(b=0.126,P=0.001),family status(b=-2.318,P=0.001)and ApoEε4 carriers(b=2.395,P<0.001)were significantly correlated with cognitive decline,ApoEε4 carriers(HR: 1.367,95%CI:1.030~1.815)was a significant predictor for AD,and an increase of one unit in trajectory of ADAS-Cog 13 score increases the risk of AD conversion by 1.153 times(95% CI:1.128~1.178).In MMSE joint model,age(b=-0.042,P<0.001),family status(b=0.478,P<0.05),education(b=0.518,P<0.05),gender(b=-0.389,P<0.05)and ApoEε4 carriers(b=-0.382,P<0.05)did significant effect on cognitive impairment,ApoEε4 carriers(HR:1.554,95%CI:1.180~2.047)was a risk factor for AD,and an unit decrease of MMSE score increases the hazard of AD conversion by 0.797 times(95% CI:0.766~0.829).With respect to FAQ joint model,family status(b=-1.338,P<0.05)statistically significantly associated with functional impairment,ApoEε4 carriers(HR: 1.587,95%CI: 1.197~2.106)and female(HR: 1.366,95%CI=1.040~1.794)were significant predictors for AD,FAQ increases in one unit would lead to the 1.181 times increase of the AD conversion risk(95% CI: 1.153~1.209).Joint models of MMSE and FAQ had better goodness of fit.FAQ had the best predictive accuracy with AUCs ranging from 0.736 to 0.852.ADAS-Cog13 and ADAS-Cog11 did the second best(AUCs: 0.636~0.811 and 0.645~0.806),while MMSE had a lower capability(AUCs: 0.665~0.753).FAQ joint model was used to provide dynamic individual prediction.We considered a married man who entered the cohort at the age of 68,not ApoEε4 carrier,with a high education.After 36 months follow-up,his individual predicted probability of AD conversion at the age of 73 was estimated as 30%(95%CI: 10%~60%).Conclusion:This study showed that age,family status,education,gender and ApoEε4 carriers had significant effect on cognitive impairment;family status would lead to functional impairment;and ApoEε4 carriers,female,cognitive impairment and functional impairment were significant predictors for AD conversion.Moreover,FAQ was the best one in predicting AD.Therefore,stable family status,reading,proper physical and social activity may delay the cognitive and functional impairment among the elders.Using joint modeling to analyze the longitudinal measures and survival data in the AD progression,we could provide theory basis for prevention or intervention of cognitive and functional impairment among the elders,and provide references of longitudinal and survival data analysis for other diseases.