coding and the lack of ex-US, serious, labeled
reports, making this an invalid comparison.
Since the signaling data sources are post-mar-
keting reports, factors that influence reporting will
affect the signaling output. With rarely reported
AEs, a disproportionality will become magnified,
because even one reported case against an expected
background of zero may be statistically significant
even though its clinical significance may be
unknown.
Special caution should be exercised with any
comparison of reporting ratios across different
products, for example comparison with competitor
drugs. Differences in company interpretation and
applications of the international health authority
regulations, coding and case processing standards
and practices, and factors such as time on the
market, labeling, product publicity and total
patient exposure may be important factors in
explaining apparent differences. The latency
between agency receipt of data and public avail-
ability of the data may obscure differences between
older and newly marketed products.
Data obtained by proportional reporting evalua-
tions or EB05 scores should be reviewed in light of
clinical experience with the product and, in gen-
eral, shouldbevalidated usingan external sourceof
data such as exposure estimates, clinical trial
results, review of the scientific literature and clin-
ical and epidemiological studies.
For all of these reasons, these types of reports
should be used solelyfor the purpose ofinitial signal
identification. Causality cannotbedetermined using
this instrument, and its primary utility is for gener-
ating hypotheses for further evaluation. These data
cannot be used alone to make safety decisions or
recommendations about safety issues, as signal ana-
lysis cannot prove or disprove a causal association
between a specific drug and an AE, in the absence of
other compelling evidence. Appropriate regulatory
and legal guidance should be sought concerning
other uses of these types of reports.
40.6 Regulatory guidance
According to two FDA guidance documents issued
in 2005, and presented in synoptic form below,
applying data mining techniques to large AE data-
bases, such as FDA’s AERS or VAERS, can
enhance risk identification and assessment. Data
mining, by systematically examining reported
AEs, may be able to provide additional information
about the existence of an excess of specific
AEs reported for a product, warranting further
investigation. FDA Guidance for Industry
(2005).
This technique can be used to supplement exist-
ing signal detection strategies, but does not estab-
lish a causal association between the drug–event
pairs being studied. However, it can be used for
assessing patterns, time trends and events asso-
ciated with drug interactions. Data mining can be
improved by adjusting for aspects of reporting (e.g.
cumulative reporting by year) or characteristics of
the patient (e.g. age or gender), or limiting the
analysis to drugs of a specific class or for those
used to treat a particular disease. The score gener-
ated by data mining quantifies the disproportion-
ality between the observed and expected values for
a given product–event combination.
Although it is recognized that all of these
approaches are inherently exploratory or hypoth-
esis generating, they may provide insights into the
patterns of AEs reported for a given product rela-
tive to other products in the same class or to all
other products. The FDA urges caution when
making such comparisons among products,
because voluntary AE reporting systems such as
AERS or VAERS are subject to a variety of report-
ing biases.
As of now, the use of data mining techniques is
not a required part of signal identification by reg-
ulatory authorities; however, if data mining results
are submitted to the FDA, it is expected that they
will be presented in the larger appropriate clinical
epidemiological context, to includea description of the database used;
a description of the data mining tool used (e.g.
statistical algorithm, and the drugs, events and
stratifications selected for the analyses) or an
appropriate reference;
a careful assessment of individual case reports
and any other relevant safety information related
to the particular drug–event combination of40.6 REGULATORY GUIDANCE 553