Why do Clinical Trials Anyway?

[Guest guest]

[From Karl's nhl-info list]

Acting on Imperfect Evidence: How Much Regret

Are We Ready to Accept?

Djulbegovic H. Lee Moffitt Cancer Center and

Research Institute at the University of South Florida, Tampa, FL

Frohlich and L. , Brown VA

Medical Center and the VA Midwest Center for Health

Services Research and Policy Studies, the Division of

Hematology/Oncology of the Department of Medicine, the Institute

for Health Services Research and Policy Studies, and the

H. Lurie Comprehensive Cancer Center of Northwestern

University, Chicago, IL

....First, one should consider the overall purpose of

(clinical) research. The purpose of clinical research is to

address uncertainties about the effects of competing treatments:

if there were no uncertainties, there would be no need to

conduct research.8,9 Scientifically, nothing new would be

learned, and ethically we would be exposing patients to risks

associated with potentially inferior treatment.

Uncertainty comes in various forms, ranging from simply not

knowing, to a high degree of uncertainty. The choice of

scientific methodology and particular research design should

reflect the level of uncertainty.10

Therefore, the choice of randomized versus nonrandomized

design should be viewed from the perspective of tailoring a

research design to varying levels of uncertainty. For

example, few would argue that the bleeding patient should be

randomly assigned to blood transfusion versus no blood

transfusion to prove that transfusion is superior, though clinical

trials have evaluated low versus high hemoglobin thresholds

for transfusion

in initiations.

When the effects of treatments are dramatic and readily

recognizable, we do not need randomized trials to demonstrate

that a particular intervention is effective.11 However,

when estimated uncertainties about benefits and harms of

competing treatment alternatives are high, we can never be sure

if the observed findings were due to treatment(s) or

because of other factors related to patient selection or biases

associated with the chosen methods of study. Under these

circumstances, there is simply no better method than a

randomized controlled trial to resolve these uncertainties.

Second, we should be cognizant of the fact that there are

always possibilities that ineffective treatments are

recommended or even that effective treatments are not recommended.

In other words, we may ultimately regret our decisions and

recommendations.

Formally, regret for not recommending effective treatments

becomes acceptable when chances of successes and expected

treatment benefits

are low.12,13 Committing to treatment will make regret

tolerable only if the chances of treatment success are higher

than the harms associated with these therapies (which may

include costs).12,13

Third, our " tolerance toward error " 8,12,13 depends in large

part on the goals of specific treatments.12,13 When the

goal is prevention of disease, such as with tamoxifen as

prophylaxis against breast cancer or small improvements in

expected survival durations, we should insist on the highest

level of evidence. This would typically include systematic

reviews based on large-scale randomized evidence. However,

if the goals of treatment are cure or improvements in years

of expected survival, with high perceived benefits and low

or unknown extent of perceived risks, a lower

level of evidence may be acceptable.

If the goal of the treatment is less meaningful

prolongation of survival (eg, days to months as is often seen in

lung

cancer chemotherapy trials), with uncertain benefits and

harms, then our practice should be based on the highest

evidentiary standards possible. In evaluating symptomatic

treatments associated with high costs and/or quality-of-life

benefits, a high quality of evidence is needed to facilitate

societal decision making.

....

full text with tables:

http://www.jco.org/cgi/reprint/JCO.2005.06.007v1