been in the market for a long time and may have a
lot of AE reports in the database) and/or the larger
the number of cases of a particular AE (a common
AE), the larger the expected ‘E’ will be. The larger
that ‘E’ is, the smaller the EBGM will be. A new
drug or a very rare AE would represent lower
proportions of the total database and thus the
expected ‘E’ would be lower.
40.3 Safety surveillance
As discussed in the chapter on Drug Surveillance,
the safety surveillance mission is to implement the
systematic review of spontaneous post-marketing
data for proactive risk identification and assess-
ment. In general, signal generation is done using
clinical trials data, the medical literature, knowl-
edge of class effects and spontaneous reports.
There are numerous challenges with sponta-
neous report databases, including the fact that
they are numerator-based, subject to many report-
ing biases, can be hard to place in population con-
text, clearly dependent on coding practices and
given the granularity of the MedDRA Dictionary,
there can be a dilution of the signal.
Additionally, spontaneous post-marketing safety
surveillance databases were developed for regula-
tory reporting, as such, differences that exist with
the national reporting requirements can alter the
type, frequency and number of post-marketing
safety surveillance reports that get entered into a
database. Also, different companies may interpret
the regulations differently, resulting in differential
reporting of post-marketing safety surveillance
reports. Furthermore, changes take place over time
with dictionary version, with reporting standards
and with product labeling. Data migration may
cause sufficient changes to take place so that data
conversion and legacy data can be lost. Moreover,
causality assessment is rarely consistent.
There are several factors that affect both the
quality and quantity of postmarketing reports.
This is sometimes referred to as the ‘Weber effect’,
where the newness of a drug to market results in a
peak in post-marketing reports during the second
year of being marketed. Additionally, if a drug is
the first in its class to be marketed, as opposed to
being the second or third drug in a class to be
marketed, there can be higher reporting rates of
post-marketing safety surveillance reports. In addi-
tion, items such as publicity, whether it is from a
regulatory action such as a Dear Doctor Letter,
litigation or coverage in the media, can all result
in increased post-marketing reports.
In addition, there are some countries like the
United States, which allow for consumers to
report AEs, whereas other countries only allow
healthcare professionals to make such AE reports.
This can result in higher numbers of reports,
though the information that is received may not
be completely valuable or beneficial when search-
ing for safety signals. Consumer reports can also
be increased as a result of direct consumer adver-
tising, especially when consumer hotlines are
published.
There are two key approaches to safety surveil-
lance. First, the intraproduct signaling, which
seeks to identify changes in the overall AE pattern
for specific products over time. This monitors
selected AEs for a specific product over time to
determine changes in the frequency and severity of
AE reports. The other type of approach is the
interproduct signaling which compares a specific
product with all products in the database.
This interproduct signaling is data mining and
essentially it determines a disproportionality score
to detect drug–event combinations that are distinct
or stand out from the background rate. Both
approaches should be used to systematically screen
large data sets to identity and analyze drug–event
associations. These are, however, hypothesis-
generating approaches and the idea is to search
for new, preventable, serious AEs with potential
public health importance. In addition, the surveil-
lance program should be set up to evaluate new and
emerging safety issues.
Intraproduct signaling essentially looks at a com-
pany’s own database to determine whether the fre-
quency of a particular AE has been increasing, after
appropriately adjusting for sales. Interproduct sig-
naling uses computer-assisted application of statis-
tical algorithms to measure disproportionality. It
tries to identify drug combinations that occur more
frequently than expected. It is important to remem-
ber that such signal scores are measures of statistical40.3 SAFETY SURVEILLANCE 547