Rapid Responses to:

EDUCATION AND DEBATE: Trevor A Sheldon, Gordon H Guyatt, and Andrew Haines Getting research findings into practice: When to act on the evidence BMJ 1998; 317: 139-142 [Full text]

Rapid Responses: Submit a response to this article

Rapid Responses published:

randomised control trials: gold standard?

Kuoni Jurg   (16 July 1998)

Acting on evidence can can only be done within the context of decision analysis

Benjamin Djulbegovic   (29 July 1998)

randomised control trials: gold standard? 16 July 1998
Kuoni Jurg private practice Send response to journal: Re: randomised control trials: gold standard?	The article "When to act on the evidence" is a strong pleading for evidence based medicine, and I fully agree almost without reservation. Just one restriction: randomised control trials are money-dependent, and most of them are -in one way or another- sponsored. Even if the sponsor does not influence the trial, his influence determines what trial can be performed. An example: there is growing evidence that fluoride-treatment has a positive impact on bone density and can reduce fracture rate --provided that the dose is adequate and the fluoride is a monofluoride. But who funds a trial with a monofluoride? A medicament which will cost about nothing? It is far more interesting to sponsor studies with biphosphonates, which will have a relevant impact on the sales and the balance of an enterprise. Briefly: there is a bias concerning the selection of studies. jürg kuoni, M.D.
Acting on evidence can can only be done within the context of decision analysis 29 July 1998
Benjamin Djulbegovic, Associate Professor of Medicine University of Louisville Send response to journal: Re: Acting on evidence can can only be done within the context of decision analysis	Dear Sir, In their, otherwise, excellent article "When to act on evidence", Sheldon at al. failed to address the critical issue identified by EBM practitioners: how efficacious should a treatment be before it is worth administering. "Is NNT (the number of patients who need to be treated to achieve one favorable outcome )15 a large or a small number of patients that need to be treated to prevent one bad outcome?" (EBM 1996;1:37-38). Intertwined with this question is the question of specific integration of benefits and harms into practical decision, the issue which, so far, has not been successfully tackled within EBM movements. We recently provided the answer to these questions by showing that action about given evidence can be only determined by linking evidence-based medicine (EBM) summary measures to decision analysis (DA) (Cancer Control, in press). By linking EBM to DA we showed that therapeutic benefits and harms are consequences (also referred to as utilities in DA terminology) of our treatment choices. This lack of understanding that EBM treatment indexes are in fact utilities is the main reason why alone they cannot be used in clinical decision making. The magnitude of benefits and harms associated with any single outcome is not necessarily related to the optimal treatment decision. The tenets of DA hold that we should choose a management strategy whose worth or expected value is maximal and not the one associated with highest utility of any particular outcome. By linking EBM to DA, we, for example, showed that in a prophylactic setting the treatment should be given only if NNT<=1/treatment harms. If NNT is bigger than this number, we should never act on that evidence. Likewise, the threshold for action (AT) is equal to AT=harms associated with treatment*NNT, where AT is the treatment threshold probability of a disease, at which physician is indifferent between treating and not treating the patient. If the probability that the patient has the disease is lower than AT treatment is not indicated; if the probability is greater than AT treatment should be given. If, however, we would like to choose between two treatments, or would like to integrate other EBM summary measures of treatment benefits and harms, derivations of AT is somewhat different (Cancer Control, in press). Thus, by linking EBM to DA we also solved the problem of integration of benefits and harms, which so far has not been successfully addressed by EBM practitioners. And, finally by developing relationships between EBM treatment indexes and AT we also show how physicians can tailor their decisions toward individual patients rather than toward "an average" randomized patient (Am J Med 1997;103:529-535). Benjamin Djulbegovic University of Louisville Department of Medicine Division of Medical Oncology/Hematology James Graham Brown Cancer Center 529 S. Jackson St. Louisville, KY, USA 40202 Iztok Hozo Indiana University Northwest Department of Mathematics 3400 Broadway Gary, IN, 46408 USA