still imply that the graft is capable of contributing 
to relapse. If no patients relapse with marked tumor 
cells, unfortunately it will not be possible to 
distinguish poor transduction efficiency of tumor cells 
contaminating grafts from the absence of tumor cells in 
the graft. 
Estimating the probability of detecting a marked 
relapse is very difficult, especially because we are 
transducing a CD34-enriched population. Breast cancer 
cells and myeloraatous plasma cell do not express CD34 
22 but conceivably earlier tumor "stem cells" that we * 
are not yet able to identify do. Thus it is not known 
if the CD34-enrichment procedure will act as a "purge" 
of tumor cells or not. 
Both normal and bcr/abl+ CFLJ-C and long-term bone 
marrow culture initiating cells from CML patients 
express CD3 4 23 ' 24 , thus the CD34-enrichment procedure 
would be expected to • enrich both normal and leukemic 
cells equally. The likelihood of detecting a marked 
relapse depends on a number of variables, including 
transduction efficiency, yield after CD34 enrichment, 
number of cells transduced, and number of cells 
contributing to a relapse. If we transduce 30 % of the 
marrow cells, assume a 50% yield of leukemic stem cells 
after CD34 enrichment, have a transduction efficiency 
of 1% into leukemic cells, and 100 cells contribute to 
a relapse, then we would detect a polyclonal marked 
relapse in 12 patients with a 95% probability. More 
cells contributing to a relapse would increase the 
likelihood of detecting a marked relapse. But there is 
little data to support any of these assumptions, thus 
this type of likelihood analysis may not be predictive. 
[ 24 ] 
Recombinant DNA Research, Volume 16 
