Dusetzina SB. Med Care. 2011;00: 000–000.
When the FDA disseminates risk communications about drugs or drug
utilization, the results may vary greatly, according to recent findings.
Researchers from several sites in the United States reviewed literature
pertaining to the FDAs drug-risk communication system. Forty-nine of
1,432 records screened were eligible for analysis. Eligible studies were those
published from January 1990 to November 2010 that empirically evaluated the
effect of an FDA risk communication and contained terms associated with drug
utilization, the FDA and advisories or warnings. Other data extracted included
the drug or drugs analyzed, relevant communications from the FDA, the source of
the data, the method of analysis and the outcome measures assessed.
G. Caleb Alexander, MD, associate professor of medicine at the
University of Chicago, told Infectious Disease News that there were four
types of risk communications evaluated: warnings about serious adverse
events, recommendations against use in specific patient populations, cautions
regarding drug-drug interactions and calls for increased laboratory or clinical
Alexander said the most effective communications were the ones that were
simple, specific, provided alternative therapies or drugs, and were
reinforced over time. He cited an FDA alert linked to the
diabetes drug rosiglitazone that demonstrated many of
Other risk communications were not so effective, according to Alexander.
Drivers of drug use exist within a sea of information, he said.
Predicting whether a risk communication will be effective is difficult,
as our results indicate.
The findings covered 16 medicines or therapeutic classes. Communications
for antidepressants comprised one-third of the studies. There was variety in
the data sources and statistical analysis methods used, but most of the studies
used medical or pharmacy claims. Patientprovider communication,
decision-making or risk perceptions were evaluated in fewer studies. The
data sources we were using were so varied that it was difficult to perform
quantitative analyses, Alexander said.
FDA advisories recommending increased clinical or laboratory monitoring
were linked to decreased drug use, according to the results. However, these
advisories were linked to moderate, transient increases in monitoring.
Communications aimed at specific subpopulations often demonstrated
effect on other subpopulations. The largest decreases in drug utilization were
observed with drug-specific warnings, but it was also observed that the
magnitude of substitution within
therapeutic classes demonstrated variation across
Alexander said an effective risk communication need not necessarily lead
to decreased use of the drug in question.
The ultimate goal is to cause more informed decision-making,
he said. An advisory or communication may be effective, even though no
changes in use of the drug occur.
The researchers concluded that the effect of risk communication is
varied and unpredictable. They said the use of risk communication
to ensure safe use of medications is complex and requires consistent
Identifying factors that are associated with rapid and sustained
responses to risk communications will be important for informing future risk
communication efforts, they wrote.
In this article, the authors seek to assert a causal relationship
between FDA communications regarding drug adverse effects and risks with
subsequent patterns of prescribing behavior and drug usage. They find a
variable, but disappointing small relationship. Unfortunately, the underlying
assumptions of such an analysis are faulty and therefore the findings are
unreliable. Far more variables determine patterns of drug prescribing behaviors
and their changes than merely FDA bulletins and notices. Chief amongst these
confounders are the impact of publicity and media attention afforded each
issue. These gain patient awareness, which in turn may change physician
prescribing behavior. Collateral commentary in professional journals confirming
or minimizing FDA concerns also modulate responses to FDA alerts. Finally, the
subject matter often determines the degree of concern attached to the FDA
communication. For example, comments regarding cancer risks with drugs are
taken more seriously than concerns regarding relatively rare but non-fatal
adverse effects. There are no controls for any of these confounders in this
analysis and therefore the conclusions must be interpreted in light of these
Alan J. Garber MD PhD
Professor, departments of medicine, molecular and cellular biology, biochemistry and molecular biology
Baylor College of Medicine