Each patient will receive autografts of two sources of CD34+ cells. Initially two 
aliquots of approximately 10* marrow cells will be processed to select the CD34+ cells. 
One resulting aliquot will be transduced with the GlNa vector and the other with the 
LNL6 vector. One of these marked aliquots will be randomly selected to be incubated 
for 5-7 days with the growth factors ILl, IL6, IL3, and SCF and the other will be 
frozen. At the beginning of the study, sufficiently large batches of the two vectors will 
be obtained to permit the entire investigation to be accomplished with the same batches. 
The randomization is necessary to prospectively adjust for the possible differing 
transduction rates of each vector. For each patient the transduction efficiency of each 
vector will be recorded. After the incubation period the frozen aliquot will be thawed 
and both aliquots of cells reinfused into the patient. Estimates of the numbers of 
reinfused CD34-f cells will be recorded for each aliquot. 
Weekly for 12 weeks following autograft, peripheral blood samples will be drawn to 
count the number of absolute neutrophils (AN) and assayed for the presence of each 
provirus. This can reliably be accomplished if the ANC exceeds 100. The date of the 
sample, ANC and proportions (peo and PbJ of AN’s containing the GlNa and LNL6 
provirus, respectively, will be recorded for each sample even if the absolute neutrophil 
count is less than 100, as will time to engraftment, defined as an ANC greater than 500. 
Bone marrow aspirates will be available on many of the patients at several times 
following autografting. These samples, when available, will contribute accurate estimates 
of the proportion (P^o and PmJ of marrow cells which are marked with the GlNa and 
LNL6 vectors, respectively. 
10. 1 Modeling Time to Engraftment 
The problem is to model the time to engraftment as a function of covariates and in 
particular, to describe the relationship between the proportion of AN’s expressing each 
gene to this time. Censoring is possible. Our first approach will be based on the Cox 
Life Table Regression model (A). Fixed covariates to be investigated include 
transfection rates and numbers of CD34-f- cells reinfused. Time dependent covariates 
to be investigated include time from autograft to ANC first exceeding 100, Pbo> and Pbl- 
The function of the two peripheral blood proportions, which is hypothesized to relate to 
the time to engraftment, is the rate of change of these proportions. For each time point 
following two estimates of a proportion, simple linear regression will be used to provide 
an estimate of these rates. If such an estimate is not possible at a particular point, zero 
will be used in the model as the observed rate. If we are successful in modeling the 
change in proportions over time, we will consider imputing the unknown values instead 
of using zero. Interactions between the two factors as well as with time will be included 
in the model. Other functions will be explored, but the rate of change is expected to 
contain the important information relative to time to engraftment. Other linear models 
to be explored will reflect that the growth of the proportion of cells marking with a gene 
may be exponential and a function of the transfection rates and numbers of cells 
reinfused. After establishing the model using the peripheral blood proportions, a subset 
Recombinant DNA Research, Volume 19 
[497] 
