Treatment Bias Can Skew Results of Observational Studies
MONDAY April 21, 2008 -- Certain biases may compromise the findings of observational studies that compare the outcomes of different cancer treatments, a new study shows.
Researchers at the University of Texas M.D. Anderson Cancer Center in Houston said observational studies should include more thorough information and should be better designed to minimize inaccuracies.
Randomized clinical trials -- in which patients are randomly selected to receive different treatments or a placebo -- are considered the gold standard for determining the effectiveness of new cancer treatments, according to background information in the study. Observational studies involve analysis of population-based data to determine how well patients respond to particular drugs.
In this study, the researchers looked at the effectiveness of different cancer therapies for prostate and colon cancer by examining data from the Surveillance, Epidemiology and End Results (SEER) Tumor Registry, a U.S. national population-based cancer registry.
In all cases, the researchers came up with improbable results. They said this shows how easy it is to generate questionable findings when doing an observational study.
They said the skewed results in their three analyses were caused by selection biases when cancer patients are treated. Selection bias can occur when patients with poorer prognoses are more likely to receive a more effective drug, or when patients with better underlying health are more likely to receive a more toxic treatment, because they're more likely to be able to tolerate it.
The researchers said their findings "suggest that the results of observational studies of treatment outcomes should be viewed with caution."
At a minimum, researchers analyzing observational data should attempt to separate patient outcomes into those that could possibly be due to treatment versus those that could not, the study authors concluded. They noted that many observational studies on cancer treatments only report death rates from all causes and don't specify cancer-related deaths.
The study is published in the June 1 issue of Cancer.
The U.S. National Cancer Institute has more about cancer clinical trials.
Posted: April 2008