How to avoid bias when comparing bone marrow transplantation with chemotherapy.

It is important to know whether the survival of patients receiving an allogeneic bone marrow transplant (BMT) is better than that of patients receiving "conventional" treatment and, if BMT is better, to know how much better. Unfortunately, this information is surprisingly difficult to obtain accurately. Most studies that have attempted to define the value of BMT have been subject to varying degrees of bias because of the problems of identifying a "conventional chemotherapy" control group with which the outcome of patients who have received BMTs can be compared. A common bias arises when disease-free survival of BMT patients is compared with that of all other remitters. Early failures are then automatically assigned to the chemotherapy group even if they have donors and so would have gone on to a BMT had they not relapsed or died. Since some patients receive BMTs many months (or even years) after achieving remission-when their prognosis is already much improved-the definition of "early failure" is problematic. Nevertheless, although it is very rarely used, an adequate statistical method does exist to overcome this problem. Careful analysis cannot, however, overcome the problem of selection bias: patients selected for BMT are likely to have better (or worse) prognosis than patients who are treated conventionally. The only really satisfactory way of assessing the value of BMT is to conduct randomised trials comparing BMT with no BMT - or with extra chemotherapy. Several such studies are currently being undertaken assessing the role of autologous BMT in AML.(ABSTRACT TRUNCATED AT 250 WORDS)