View PDF Version - RePub - Erasmus Universiteit Rotterdam
View PDF Version - RePub - Erasmus Universiteit Rotterdam
View PDF Version - RePub - Erasmus Universiteit Rotterdam
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Discriminant analysis using a MLMM with a normal mixture 155<br />
diction uses only the latent characteristics of the subjects with the between subjects<br />
variability.<br />
Application to PBC data<br />
As an illustration, we have applied our methodology to the Dutch PBC study data<br />
described in the introduction. Of the 375 patients, 178 patients are known to be alive<br />
at T = 10 years without needing a liver transplantation. They will be further referred<br />
to as prognostic group 0. A total of 41 patients died from a liver related cause or<br />
needed a liver transplantation during the first 10 years and will be further referred as<br />
prognostic group 1. The remaining 156 patients can be divided in three categories<br />
for whom the prognostic group is unknown because loss of follow-up (14 patients),<br />
or is unknown because the follow-up time was less than 10 years (112 patients).<br />
The third category consists of 30 patients who died in the first T = 10 years<br />
from another than a liver related cause. These 156 patients could be considered<br />
as the third prognostic group in our methodology. However, the results of our<br />
discrimination procedure can be better exemplified with G = 2 groups by means of<br />
sensitivity, specificity and receiver operating curves (ROC). Since the main objective<br />
of this section is to illustrate and explore the performance of our methodology, we<br />
will therefore consider only the two prognostic groups mentioned above. Hence<br />
our training data consist of 178 + 41 patients. The markers used to predict the<br />
prognostic group include values of: bilirubin, albumin, alkaline phosphatase (AP).<br />
Hence, R = 3 in model (1). To better satisfy mixed model (1), the natural logarithm<br />
of AP was used as the marker instead of the original AP value. Figure 1 shows the<br />
observed longitudinal profiles of the markers, separately for Group 0 and Group 1.<br />
We can observe that the bilirubin levels in Group 0 are rather low and stable over<br />
time whereas they start to increase dramatically from a specific time point for many<br />
patients in Group 1. Albumin levels are in general higher in Group 0 than in Group 1<br />
while the reverse is true for AP. For practically all patients in Group 1, log(AP)<br />
is almost never negative whereas in Group 0 negative values of log(AP) are quite<br />
frequent.<br />
Although not required with our approach, the mixed models (1) for the PBC data<br />
will have the same structure in both prognostic groups. Namely, for each of three<br />
markers (r =1, 2, 3) and both groups (g =0, 1), the X g<br />
i,r matrix contains two<br />
columns: age of the patient at start of treatment (median 54.7) and a dose of<br />
ursodeoxycholic acid (UDCA) in mg/day (median 750). Consequently, the vector<br />
αg of fixed effects has a length of 6 for g = 0, 1. Matrices Z correspond to