The Predictive Value of a “Positive” ASH Abstract in AML Therapeutics.

ASH abstracts describing outcomes of trials of new therapies (Rxs) are presumably read in anticipation that positive results, although necessarily preliminary, will herald therapeutic advances. Here we examine the predictive value of a “positive” (pos) ASH abstract. We reviewed abstracts, presented in oral or poster sessions from 1993–2001, dealing with early phase trials of new drugs, used alone or together with more standard Rx, in adult AML (APL excepted). A research nurse (RN), a medical editor (ME), and E. Estey (EE) reviewed the abstracts independently, with the latter’s decision final. There was 80–85% concordance among the 3 reviewers, with EE tending to view fewer abstracts as pos. 63/91 abstracts (69%) were judged pos from a therapeutic standpoint (based typically on conclusions that the rx was “active” or “effective”), 14 were considered “negative”, and 14 inconclusive. The 63 pos abstracts involved 37 separate drugs, i.e. several drugs were the subject of > 1 pos abstract, including gemtuzumab, and topotecan combinations (5 abstracts each) and PSC and IL-2 (without histamine) (4 abstracts each). A “pub med” search indicated that 45/63 (71%) pos abstracts, covering 27 of the 37 separate drugs, subsequently appeared in peer-reviewed journals, with half the unpublished abstracts appearing at the 2000 or 2001 ASH meetings. The pos conclusion was unaltered in 38 of the 45 and was changed to negative in 5 and to inconclusive in 2. Three of the 37 pos drugs have either migrated into clinical practice (gemtuzumab) or been pos in a randomized trial (IL-2 + histamine, cyclosporine + infusional daunorubicin + ara-C). Four of the pos drugs (PSC833 + chemotherapy. HuM195, fludarabine + ara-C, troxacitabine + ara-C) have appeared, in some pt subsets, negative in randomized trials. However, 81% (30/37) of the drugs reported pos in ASH abstracts remain, at least 5 years from date of ASH publication, either outside the scope of clinical practice or unevaluated in randomized trials. We suggest that the reasons for this phenomenon be explored.