death or toxicity, no adjustment in the primary analysis will be made for the 
sequential design. 
Death during ABMT: 
If five deaths during transplantation are observed in the treatment of the first 
20 patients, accrual to the study will be terminated and a complete evaluation 
of all information made as to whether continuation is warranted. This 
decision rule provides a significance level of 0.17 for the null hypothesis that 
the true death rate during ABMT is at most 0.15. The power for rejecting the 
null hypothesis (terminating accrual) if the true death rate is 0.20,0. 30or 0.40 
is 0.37, 0.76 and 0.95, respectively. 
Toxicity 
If seven episodes of unacceptable (Grade IV) toxicity are observed in the first 
20 patients treated, then accrual to the study will be terminated and all data 
evaluated to determine whether continuation is warranted. This decision rule 
has a significance level of 0.21 for the null hypothesis that the true toxicity 
rate is at most .25. The power of rejecting the null hypothesis (stopping 
accrual) is .75 if the true toxicity rate is 0.40. 
If accrual to the study is not halted after 20 patients have been treated with 
ABMT, the study will continue until a total of 35 patients have been treated 
with ABMT. 
8.3 Evaluation of Purging 
If one assume that the study is not terminated early due to safety concerns, 
then we expect to evaluate the purging techniques in a total of 35 patients. 
Since the transfection rates may vary with the vectors being used, patients will 
be randomized to one of the two possible combinations of purging techniques 
and vectors. A blocking factor of four will be used. 
If the transduction efficiency of a vector is assumed conservatively to be 1 % 
and the reinfused marrow is exposed to either the vector G1 or LN6, then the 
following table provides estimates of the probability of detecting a marked 
relapse as a function of the number of progenitor cells (NC) giving rise to the 
relapse. A cell in the exposed marrow has a 0.01 probability of being 
transfected with one of the vectors. A binomial model has been used to 
calculate the estimates provided in the table. 
[244] 
Recombinant DNA Research, Volume 17 
