P1‐557: A METHOD FOR THE DEVELOPMENT OF A DISEASE PROGRESSION COURSE USING TWO SINGLE COHORTS

Characterization of disease progression course is important for prevention and treatment of the disease. However, for the disease with slow progression such as Alzheimer disease (AD), the development of a progression course is not easy because of the difficulty to follow up for a long time. A solution to overcome this limitation is to use multiple single cohorts of successive stages of the disease, for example preclinical AD cohort and MCI cohort for AD. We present a method to integrate the two single cohorts to model a disease progression course over time. We suggested the four steps 1) estimating the model according to follow up time for each cohort (figure 1) 2) generating the predicted outcome and its 95% confidence interval for each subject 3) checking the overlapped region of the predicted values between the two successive cohorts and searching the time to start to overlap between the two cohorts postulating the cohort of the late stage (cohort 2) comes after the cohort of the early stage (cohort 1) at this time (figure 2) 4) finally estimating the linear mixed model of one whole course of the disease using cohort 1 and cohort 2 (figure 3). We examined the validity of our approach using the simulated data. The data of 100 subjects for each cohort was generated as the following setting assuming cohort 2 starts to overlap to cohort 1 at t=4. For cohort 1, ln(Y)=b10+b11*t+ε (t=0∼5), where b10∼N(1.0, 0.12), b11=0.025; ε∼N(0, σ112), σ11=0.1, 0.2. For cohort 2, ln(Y)=b20+b21*t+ε (t=0∼5), where b20∼N(2.2, 0.22), b21=0.03; ε∼N(0, σ212), σ12=0.2, 0.3. For each data of all combinations according to the random variability of intercept and residual in cohort 1 and cohort 2, the time to start to overlap between the two cohorts was estimated to very close to true value of t=4 (range 4.00∼4.03). The disease progression model over t=0∼10 was estimated to ln(Y)=0.920+0.028*t.