Добавил:
Sekretar
kiopkiopkiop18@yandex.ru
t.me/Prokururor I Вовсе не секретарь, но почту проверяю
Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз:
Предмет:
Файл:Ординатура / Хирургия / Библиотека им академика М.И. Перельмана / Книга_5910_Библиотеки_им_академика_М_И_Перельмана
.pdf
3 Predisposing Factors forAdverse Drug Reactions
59
practitioners ignore the alerts provided by the
system due to the total volume of the generated
alerts [69, 70]. Furthermore, there is a lack of
consistency and standardization in the information provided by different resources. The availability of drug–drug interaction evidence in a
standardized format with adequate literature
support and inclusion of clinical management
options can improve prescribing decisions [71].
The use of the number needed to treat to have a
clinical interaction, as suggested by MartínPérez et al., would be a helpful tool for clinicians [72].
4.4.2 Polypharmacy
The probability of ADRs increases as the number
of medications increases in a patient. The risk
may be higher due to the basic increase in the
probability of any ADR as more drugs are
included as well as the increased risk of drug–
drug interactions. The effects of multiple drug
use are not simply additive but are rather likely to
be synergistic. At the same time, the concept of
confounding by multiple disease states must be
borne in mind [21].
The likelihood of a serious ADR increases
with multiple drug exposures/the number of
drugs received, and this relationship becomes
statistically signicant when patients are
exposed to four or more drugs [64]. In a systematic review conducted by Saedder etal. to
evaluate the evidence of the relationship
between patient-related risk factors and the risk
of serious ADRs, the total number of drugs was
the most consistent correlated risk factor found
in both univariate and multivariate analyses in
both the general adult population and the
elderly [73].
5 Tools toIdentify
Predisposing Factors
As previously discussed, a predisposing factor is
a component of a constellation of causes, in addition to the drug, which constitutes the sufcient
cause inducing the ADR. Such predisposing factors may be biologically acting independently of
the drug, simply summing its contribution, or,
often, may interact with the drug, resulting in an
overall effect different than the addition of their
independent effects. In the latter case, epidemiologists say that there is an “interaction of causes”
or “an effect modication.” Both terms are essentially synonyms [74], but “effect modication” is
more appealing when, according to the hypothetical causal model, it is clear which variable is the
main exposure of interest and which is the potential effect modier. Using this terminology, many
predisposing factors of drug safety problems can
be conceived as effect modiers.
In drug safety, an effect modier is a third
variable, distinct from both the drug and the
event, which modies the magnitude of the drug–
event association when present (for dichotomous
effect modiers) or when it changes its category
(for polytomous effect modiers) or level (for
ordinal and quantitative effect modiers). For
simplicity’s sake, we will consider the example
of a dichotomous effect modier in the discussion that follows. When the modication moves
the effect measure (or more properly the measure
of association2) away from its null value, the
effect modier strengthens the association (either
positive or negative). This situation is also called
“synergistic interaction”. When the effect modier brings the measure of association closer to its
null value, the effect modier weakens the asso-
4.4.3 Pharmaceutical Factors
The pharmaceutical aspects of a particular dosage
form can lead to predictable ADRs because of
alterations in the quantity of the drug present or of
its release characteristics [21]. For example, a
rate-controlled preparation of indomethacin was
withdrawn after reports of localized intestinal
bleeding and perforation that occurred as a consequence of the delayed-release mechanism [28].
2
When the causality is not yet established, the term “association” or “measure of association” is more appropriate
than the term “effect” or “effect measure,” as the latter
always implies a causal relationship. For this reason, we
consider it preferable to use “association” throughout this
chapter. However, the term “effect modier” or “effect
modication factor” is commonly used in the epidemiological textbooks and literature, and, for the sake of consistency, we follow this terminology, but the reader should
have this caveat present.

60
Relative Risk (RR)
of factor Z
of factor P
J. Jose and F. J. de Abajo
3
3.0 (2.6–3.5)
2
1.9 (1.7–2.1)
1
0.5
RR in the absence
of factors Z or P
Fig. 3.3 Graphical representation of the two types of
effect modiers that are of special interest in pharmacovigilance: a predisposing factor (Z) and a preventive factor
(P). A hypothetical example. The effect of a drug on an
event is estimated through the relative risk (the ratio of the
incidence among the exposed divided by the incidence
among the nonexposed). In people who have neither factor Z nor factor P, the exposure to the study drug (as compared to those nonexposed) increases by 90% the risk of
the event (a relative risk of 1.9, with a 95% condence
interval (95% CI) of 1.7–2.1). Among people who have
RR in the presence
ciation (either positive or negative). This is usually called “antagonistic interaction”.
On the additive scale, a positive association is
operatively dened when the risk difference
between the exposed and the unexposed is >0 (in
this case, the risk difference is usually known as
attributable risk) and is negative when it is <0
(usually called absolute risk reduction).3 On the
multiplicative scale, a positive association is
dened when the measure of association (relative
risk [RR], rate ratio, hazard ratio [HR], or odds
ratio [OR]) is >1 and is negative when it is below
(i.e., between >0 and <1). From a conceptual
point of view, a positive association means that
the probability (or risk4) of presenting an event
increases in the presence of the drug, whereas a
negative association means that the probability
3
The measures of association on the additive scale (attributable risk and absolute risk reduction) are more frequently denominated as “impact measures” or “population
impact measures.”
4
Risk always implies causality.
1.1 (0.7–1.5)
RR in the presence
factor Z (but not factor P), the exposure to the drug further
increases the risk of the event up to a relative risk of 3.0
(with a 95% CI of 2.6–3.5), whereas among people who
have factor P (but not factor Z), the effect of the drug
weakens or practically disappears (showed by a relative
risk of 1.1, with a 95% CI of 0.7–1.5). A predisposing factor would operatively act as factor Z. A preventive factor,
a minimization measure, or an antagonist would operatively act as factor P. In this representation, for simplicity’s sake, both factors Z and P are assumed to be
dichotomous
(or risk) of presenting an event decreases when a
drug is present. An ADR, by denition, implies
that the risk of presenting the event increases,
and, then, in pharmacovigilance, we usually deal
with positive associations, unless we are evaluating the effectiveness of a risk minimization measure, in which case a negative association is
expected. In this context, a predisposing factor
for experiencing an ADR induced by a certain
drug is an effect modier that strengthens a positive association. On the contrary, a preventive
factor or a risk minimization measure weakens a
positive association (see Fig.3.3 for a graphical
representation of these concepts).
According to this conceptual framework, a
predisposing factor can only be formally identied through study designs that allow estimating
valid measures of association (usually called
“analytical” studies). In “nonanalytical” or
“defective” study designs, the existence of a predisposing factor can be suggested but not formally assessed.

Expected
No No
RRRR
AZ AZ AZ+−
=+−
3 Predisposing Factors forAdverse Drug Reactions
61
5.1 Identifying Predisposing
Factors Through Analytical
Study Designs
An analytical design is a study that allows computing measures of association that, under optimal conditions of comparability (validity), can be
interpreted as causal effects. The most reliable of
all analytical methods is the randomized clinical
trial, actuallythe paradigm of clinical research,
but its interventional nature makes it hard to be
used in drug safety assessment (unless the outcome is a surrogate variable, as the area under the
curve in the case of pharmacokinetic drug–drug
interactions), also their limited sample size and
usually short duration make the clinical trials
inefcient to detect ADRs oflow frequency and
long induction periods. For this reason, in the
pharmacovigilance arena (as well as in public
health in general), we rely on observational methods and, among them, cohort and case–control
studies are the most widely used (the so-called
traditional designs). Notwithstanding, in the last
two decades, case-only designs have been
increasingly used in pharmacoepidemiology to
assess drug safety issues, provided that events are
sudden and transient and, importantly, that drug
exposure varies over time [75, 76].
For assessing effect modiers, the measures
obtained on the multiplicative scale (relative risk
[RR], rate ratio [RR], hazard ratio [HR], or odds
ratio [OR]) are by far the most widely used. The
reason is twofold: (1) the relative measures can
be estimated in all analytical epidemiological
studies and (2) the regression models, which are
the most commonly used methods for confounding control are built on the multiplicative scale
and allow directly assessing the effect modication by including a multiplicative term in the
model. However, it may happen that an effect
modier is only detected on the additive scale,
and, from a pharmacovigilance standpoint, it is,
certainly, of interest. In the following, we will
review both.
5.1.1 Assessing Eect Modiers
ontheMultiplicative Scale
The simplest way to assess an effect modier on
the multiplicative scale is through a stratied
analysis in which the population is divided in two
strata: those who present the potential effect
modier and those who do not. In each stratum,
the drug–event association is assessed and, then,
the differences between the two measures of
association obtained are compared using an interaction test. A practical way to do this is by using
the interaction test described by Altman and
Bland [77] in which the ratio of relative risks
(RRRs) is calculated along with their respective
95% CIs. When the 95% CI does not embrace the
null value, the difference between the two RRs is
considered statistically signicant. Based on this
method, Hutchon [78] developed an ofine calculator that easily allows estimating the RRR and
its 95% CI (see the example provided in
Table 3.5). In case–control studies, we would
compute the ratio of ORs instead of RRRs.
Alternatively, researchers may opt to assess the
presence of potential interactions on the multiplicative scale by introducing a product term in the
regression models [74, 79].
5.1.2 Assessing Eect Modiers
ontheAdditive Scale
Assuming A to be the drug and Z to be the potential predisposing factor, the expected absolute risk
in patients who used A and had factor Z, assuming
no effect modication, would be as follows [79]:
where Expected R
=expected absolute risk of
A+Z
(3.1)
the drug A in patients with the predisposing factor Z; RA = observed absolute risk of the drug
itself (in the absence of Z); and RZ = observed
absolute risk of the predisposing factor Z itself
(in the absence of A). As R
is the back-
NoANoZ
ground risk and is counted twice, in both RA and
RZ, it is necessary to be subtracted once.

62
Expected
No No No No No No
RR RR RR
AZ AZ AAZZAZ+
−
()
=−
[]
+−
[]
Expected
AZ AZ+
=+
Table 3.5 Assessment of a predisposing factor on the multiplicative scale. In this hypothetical example, a cohort study
was carried out to evaluate the risk of upper gastrointestinal bleeding (UGIB) in patients using nonsteroidal antiinammatory drugs (NSAIDs). The researchers examined whether the antecedents of previous UGIB were a predisposing factor for having a new episode. To do this, they grouped patients into two strata: those with UGIB antecedents and
those without, estimated the rate ratios in each stratum, and compared them using the interaction test described by
Altman and Bland [77]. The results show that having a history of UGIB signicantly increases by almost three times
the rate ratio of presenting a new episode when exposed to NSAIDs (from 2.47 to 7.14)
Current users of
NSAIDs Nonusers of NSAIDs Rate ratio (95% CI)
All patients
UGIB cases 100 32 –
Person-years (p-y) 10,000 10,000 –
Incidence rate 10.0 per 1000 p-y 3.2 per 1000 p-y 3.13 (2.08–4.81)
Stratum 1: Patients with a history of UGIB
UGIB cases 50 14 –
Person-years 1000 2000 –
Incidence rate 50.0 per 1000 p-y 7.0 per 1000 p-y 7.14 (3.89–13.99)
Stratum 2: Patients with no history of UGIB
UGIB cases 50 18 –
Person-years 9000 8000 –
Incidence rate 5.6 per 1000 p-y 2.3 per 1000 p-y 2.47 (1.42–4.50)
Ratio of rate ratios (95%
CI) (RR of stratum 1/RR of
stratum 2)
a
Having antecedents of UGIB has an independent effect for a new episode to occur and can be estimated in the cohort
of nonusers by dividing the rate of patients with antecedents by the rate of patients without, that is, 7.0 per 1000 p-y/2.3
per 1000 p-y=3.04. Thus, having an antecedent of UGIB increases the risk of a new one by 3 times (from 2.3 to 7.0 per
1000 person-years), but if the patient uses an NSAID, then his/her risk increases by 21 times (from 2.3 to 50 per 1000
person-years)
b
RRR, ratio of rate ratios (or ratio of relative risks) calculated using the Hutchon calculator [78]
a
RRR = 2.89 (1.22-–6.84)
Z-score: 2.4151
p-value=0.016
J. Jose and F. J. de Abajo
b
We may express Eq. (3.1) as attributable risks (ARs), by subtracting the baseline risk (when neither
A nor Z is present) from each absolute risk, as follows:
which may be simplied as follows:
AR AR AR
(3.3)
Then, the expected effect5 of the combination
of A and Z is simply the sum of their independent
effects, under the assumption of no effect modication (or no interaction).
Now, the observed AR
the study) is compared with the expected AR
(the result found in
A+Z
A+Z
A difference>0 would denote the presence of an
effect modication and would allow us to conclude that factor Z is an effect modier.
Table 3.6 shows the results using the data from
Table 3.5. The observed AR of NSAIDs in
patients with antecedents of UGIB is 47.7,
whereas the expected AR of NSAIDs in these
patients is 8 (the sum of the independent ARs of
.
NSAIDs and antecedents of UGIB). As AR is a
measure of the effects of the respective factors
5
Attributable risk can be considered an estimation of the
effect of the exposure concerned, assuming that there is no
bias.
studied on UGIB, the interpretation is that the
(3.2)

RERI
=−+−
()
+AZ AZ
3 Predisposing Factors forAdverse Drug Reactions
Table 3.6 Assessment of a predisposing factor on the additive scale. Data were obtained from Table3.5. The predis-
posing factor assessed is the antecedents of UGIB, and the drugs of interest is NSAIDs. To calculate the attributable
risks, we subtract the corresponding incidence rate from the incidence rate of reference (that of nonusers who had no
UGIB antecedents)
NSAIDs users Nonusers
(a) Incidence rates
Antecedents of UGIB 50.0 7.0
No antecedents of UGIB 5.6 2.3
(b) Attributable risks
Antecedents of UGIB 47.7 4.7
No antecedents of UGIB 3.3 0.0
ARA (independent effect of NSAIDs when there are no antecedents of UGIB)=5.6−2.3 (per 1000 person- years)=3.3
(per 1000 person-years)
ARZ (independent effect of antecedents of UGIB when NSAIDs are not present: nonusers)=7.0−2.3 (per 1000 personyears)=4.7 (per 1000 person-years)
Expected AR
Observed AR
=3.3+4.7=8.0
A+Z
=47.7
A+Z
63
NSAIDs’ effect (on UGIB) is much greater when
the patients had antecedents of UGIB than when
they did not. Then, it is demonstrated that having
an antecedent of UGIB is a predisposing factor
for presenting a new one when the patient is
exposed to NSAIDs.
The assessment on the additive scale can also
be carried out using relative measures, which is
where Observed RR
ObservedRR Observed RR ObservedRR
is the observed risk in
A+Z
patients having the two factors (users of NSAIDs
with UGIB antecedents) relative to the reference
(situation in which none of the two factors are
present) and RRA and RRZ are the observed independent risks of factors A and Z, respectively, relative to the reference (Eq. (3.4) derives from Eq.
(3.1) when we divide all terms by R
NoA–NoZ
to obtain
the respective relative risks).6 To apply this equation, it is necessary to build a variable with four
6
Expected R
(R
NoA–NoZ/RNoA–NoZ
Expected RR
RERI= Observed RR
RR
A+Z
A+Z/RNoA–NoZ
A+Z
− (Observed RRA+Observed RRZ − 1).
=(RA/R
); then,
=Observed RRA+Observed RRZ − 1;
− Expected RR
A+Z
NoA–NoZ
)+(RZ/R
= Observed
A+Z
NoA–NoZ
) −
highly convenient as ARs cannot be derived from
all analytical studies, but all of them do provide
relative measures. The method is described by
Rothman [79] and consists of estimating the “relative excess risk due to interaction” (RERI), with
the equation to assess the interaction between
two factors A (the drug in our case) and Z (the
potential predisposing factor) as follows:
1
(3.4)
categories (for two dichotomous variables): neither A nor Z, A alone, Z alone, and A + Z, and
obtain the RR of each category as compared to the
reference (NoA–NoZ). The sum of the observed
RRA and RRZ would be the expected RR
, assum-
A+Z
ing that there is no interaction or effect modication, but as the null value of the RR is 1 and it is
summed two times, we should subtract 1 to avoid
duplication. An RERI >0 means that the observed
relative effect of the combination of the two factors is greater than the sum of the independent
relative effects, denoting that there is an interaction or an effect modication on the additive scale
between A and Z. In other words, assuming that A
is the drug and Z is the potential predisposing factor, an RERI >0 means that the potential predis-

64
RERI
8
Relative Risk
antecedents
+−
()
()
..
()()
()
J. Jose and F. J. de Abajo
7
6
5
4
3
2
1
0
Fig. 3.4 Assessing the antecedents of UGIB as a potential predisposing factor for presenting a new episode of
UGIB associated with NSAIDs using relative measures
on an additive scale. The data were obtained from
Table3.5. The null value of the relative risk (value=1) is
presented in gray, the excess relative risk associated with
NSAID use (RR-1=1.47) in yellow, the relative excess
Null
posing factor Z modies the relative effect of A
beyond the mere additivity of effects. The equa-
RERI RR RR
=−
drug used by patientswiththe PPF drug used by patientswi
Applying this to the case study shown in
Table3.5:
RERI =− +−
NSAIDs use
=714247 3131 254..
UGIB antecedents
risk associated with the antecedents of UGIB
(RR-1=2.13) in blue, and the relative excess risk due to
the interaction (RERI=2.54) in green. The last bar represents the RR observed in NSAID users with UGIB antecedents, which is 2.54 units greater than the sum of the
independent relative effects of NSAIDs and antecedents
of UGIB
NSAIDs use and UGIB
tion can be adapted to the particular case of a
potential predisposing factor (PPF) as follows:
tthout the PPF PPF itself
RR
1
episode associated with NSAID use, and itcan be
considered a predisposing factor for this drug–
event association. The result can begraphically
represented as shown in Fig.3.4.
With the help of the gure, the effect of the
This result suggests that having an antecedent
combination of A+Z would be as follows:
of UGIB increases the risk of presenting a new
ObservedRR nullvalue gray Observed excessRRyellow
=
1
AZ A+
++
Observed excessRRblueRERIgreen
+
()+()
Z

RERI
=−+−
()
+AZ AZ
RERI
=−+−
()
+AZ AZ
RERI
=−
+AZ
1
3 Predisposing Factors forAdverse Drug Reactions
65
Obtaining the 95% CI for RERI is not straightforward, but Richardson and Kaufman [80] have
developed a program in SAS and STATA that can
provide it.
Observed OR ObservedOR Observed OR
5.1.3 The Case–Control Study Nested
inaCohort ofDrug Users: ASpecial
Design toDirectly Estimate
Predisposing Factors
In pharmacogenetics, a special case–control
study is often used to identify genetic susceptibility factors involved in acute ADRs, like hypersensitivity reactions. The particularity of this
case–control study is that both cases and controls
are exposed to the drug of interest (both are
“nested” in a cohort of drug-exposed subjects):
cases are those who took the drug and experienced the event (and the drug is considered the
culprit of the reaction) and controls are patients
who were exposed to the drug over a sufcient
period of time to have presented the ADR but
they did not (“tolerant controls”). Analytically,
In case–control studies, the computed OR can
be used to estimate the RR and the formula can
be applied to this design by simply replacing RR
by OR
1
(3.5)
researchers proceed as in any case–control study:
they identify the carriers of the genetic factor
among the cases and controls and, then, compute
the odds ratio. When the odds ratio is over 1 and
statistically signicant, it can be interpreted as
the times the odds of presenting the ADR induced
by the drug increases in the presence of the
genetic biomarker as compared to its absence.
Conceptually speaking, with this design, we can
directly estimate the effect of the effect modication factor (assuming causality). In other words,
the computed OR (actually, OR
) estimates the
A+Z
RERI explained before (Eq. 3.4) because the
odds ratio of the drug (ORA) is 1, as all patients
are exposed, and the odds ratio of the genetic factor (ORZ) is also 1, as it has no effect per se on the
absence of the drug:
Observed OR ObservedOR Observed OR
Observed OR
Applications of this method can be found in
Ramírez et al. [81] and Bellón et al. [82]. In
Table3.7, an example is presented, in which this
method was used to assess whether the HLAA*32:01 was a predisposing factor for experiencing a drug reaction with eosinophilia and systemic
symptoms (DRESS) syndrome with vancomycin.
The results showed that HLA-A*32:01 was 13
times more frequent among DRESS cases
1
Table 3.7 Example of the use of the case–control
approach to assess HLA-A*32:01 as a predisposing factor
for DRESS (drug reaction with eosinophilia and systemic
symptoms) syndrome in patients treated with vancomycin. All patients were exposed to vancomycin. Data were
obtained from Bellón etal. [82]
Tolerant
HLAA*32:01
Yes 5 1
No 9 24
DRESS cases
(N=14)
controls
(N=25)
(3.6)
OR (95%
CI)
13.33
(1.36–
130.3)

66
J. Jose and F. J. de Abajo
exposed to vancomycin than among tolerant controls, which strongly suggests that it is a relevant
predisposing factor and could be a potential biomarker (to be assessed in future studies).
This method can be used in other contexts. For
instance, Martín-Pérez etal. [83] used it to identify predictors of over-anticoagulation in a cohort
of warfarin users. They compared patients presenting an International Normalized Ratio
(INR)≥4 (cases) with patients presenting an INR
≤3 (controls) and identied renal failure (30–
44mL/min/1.73m2, OR=1.45; 1.14–1.85; and
<30mL/min/1.73m2, OR=1.85; 1.23–2.78) as a
predisposing factor for over-anticoagulation in
warfarin users (assuming that renal failure has no
effect per se on the INR).
5.2 Identifying Predisposing
Factors Through Defective
Study Designs
When a study design does not allow deriving
effect measures (or measures of association) that
can be interpreted as causal effects under optimal
conditions of comparability, it may be called
“defective” (or nonanalytical), that is, a design
that is faulty because something is missing. In a
series of exposed people (also called a “registry
of exposed” when the series is exhaustive in a
certain geographical area) and in a series of cases
(or a “registry of cases” when the series of cases
is exhaustive in a certain geographical area), they
are both defective because the reference control
group is lacking (the unexposed group or unexposed person-time in the former and the non-case
group in the latter). However, with complementary information from the underlying source population, it is possible to convert a registry of cases
in a case-population study that allows estimating
measures of association and assessing potential
effect modiers [84, 85]. Moreover, in a series of
exposed people, it is also possible to apply a
nested case–control study to identify potential
predisposing factors among the exposed people,
as shown in Sect. 5.1.3.
Prescription event monitoring is an example
of a series (or cohort) of exposed people that mer-
its additional comments. Although in this method,
there is no formal unexposed group, researchers
have managed to identify potential predisposing
factors. First, they use a case/non-case evaluation
within the cohort of exposed people (equivalent
to a case–control nested in a cohort of drug users,
as described in Sect. 5.1.3) to detect associations
between potential predisposing factors and the
event. Once a factor is identied, researchers
compute the incidence rate of the event of interest
in people with such a factor in the cohort (for
instance, males) and compare it with the incidence rate among people not having such a factor
(e.g., females), adjusting for other covariates in
the multivariate analyses. With this methodological approach, Layton etal. identied that a history of peripheral edema was the most important
predictor of having peripheral edema associated
with vildagliptin [86]. Moreover, they found that
among those with such antecedents, the elderly
males had a much higher hazard ratio than elderly
women. The problem of this approach is that the
independent effect of the predisposing factor cannot be estimated (as there is no unexposed
cohort), and, then, it is not possible to disentangle
which part is due to the predisposing factor per se
and which one is due to the interaction between
the drug and the predisposing factor. If we can
reasonably assume that the predisposing factor
has no effect per se, then the estimations are
entirely imputable to the interaction.
Spontaneous reporting schemes continue to be
the cornerstone of pharmacovigilance to identify
new potential risks. However, its simplicity is
also its Achilles’ heel to formally assess drug–
event associations and to identify potential predisposing factors because it is the most defective
of all study designs. Expressing the data from
spontaneous reporting in a two-by-two table, we
can easily observe that the information from
three out of four cells is lacking and, from the one
that we do have information, the data are incomplete and probably biased (due to both underreporting and selective reporting) (Table 3.8).
Bearing this in mind, the limitation of spontaneous reporting to assess the presence of any potential predisposing factor is obvious. For this
reason, this method can only be used for

3 Predisposing Factors forAdverse Drug Reactions
Table 3.8 A two-by-two table showing the information known through spontaneous reporting schemes
Cases Non-cases Total
Exposed a (only those suspected and reported) b (unknown) a+b
Nonexposed c (unknown) d (unknown) c+d
Total a+c b+d a+b+c+d
67
hypothesis- generating purposes, either with a
qualitative interpretation (e.g., there are more
ADR reports in patients having a certain factor
such as sex, age, underlying disease, etc., though
this sole information may be misleading without
using the distribution of such factors in the population who uses the drug) or, better, through a
quantitative approach (using subgroup analysis
with measures of disproportionality or using drug
consumption data to calculate reporting rates by
subgroups) [87, 88].
Finally, the potential for big data and articial
intelligence to assist in identifying new ADRs
and predisposing factors is a growing area that, in
the years to come, may be an important new
source of information for pharmacovigilance
practice [89].
6 Management
ofPredisposing Factors
Once a predisposing factor has been identied
using any of the methods discussed in the previous sections, it is time to use such information for
ADR prevention or minimization. This can be
done at two levels: (1) public health and drug
regulation and (2) clinical practice.
6.1 Management ofPredisposing
Factors at Public Health
andRegulatory Levels
The identied predisposing factors must be mentioned in the appropriate sections of the drug
label (the Summary of Product Characteristics in
the European Union) in order to inform physicians
and in patient information leaets (to
informpatients) on which circumstances, situations, or conditions, or in which special populations, an important ADR is most probable to
occur and which measures should be checked
before the drug is used (e.g., renal function, liver
function, ion concentration in serum,and genetic
susceptibility biomarkers) and during treatment
(e.g., the introduction of a new drug, which may
interact with the ones that the patient already
uses). For this reason, it is of utmost importance
that the marketing authorization holders and regulatory authorities alike are diligent in including
new information in the drug label concerning
potential predisposing factors and informing
about the measures that can be applied to reduce
their impact (dose adjustment, monitoring of certain parameters, etc.). When the ADR is serious
enough, it should appear specically in the section of Warnings and Precautions, especially
when it has been identied as a risk minimization
measure that should be applied in all patients or
in certain groups. The ADR may also be the origin of a formal contraindication to avoid the risk
(e.g., use of the drug in patients with an antecedent of a serious allergic reaction caused by the
drug).
Sometimes, the inclusion of the information in
the drug label is not enough and more proactive
measures are needed to prevent the impact of predisposing factors. This is the aim of the additional
risk minimization measures included in the Risk
Management Plans of the European Union [90]
and the Risk Evaluation and Mitigation Strategies
of the United States [91]. However, when the risk
minimization measures fail to prevent a serious
safety concern, restrictive regulatory measures
should be implemented, including the market
withdrawal of the drug. A striking example of
this is the withdrawal of cerivastatin from the
world market in August 2001, due to a risk of
rhabdomyolysis. The risk was 16 times higher
than the one estimated with other statins, which
was worsened by concomitant use with gembrozil, another cholesterol-lowering drug,
which was contraindicated. The implemented

68
J. Jose and F. J. de Abajo
risk minimization methods failed to prevent coprescribing of the two drugs, leading to the withdrawal of cerivastatin, showing the difculties of
changing clinical practice[92].
6.2 Management ofPredisposing
Factors in Clinical Practice
Taking into account that ensuring safe outcomes
is the primary goal of drug use in patients, predisposing factors for ADRs require careful consideration at various steps of drug use in clinical
practice. Accordingly, the various stakeholders
have a signicant role to play and a multidisciplinary approach is critical. Appropriate knowledge of providers, clinicians, pharmacists, and
nurses about the general and specic predisposing factors for ADRs, is extremely essential to
promote safer use of drugs by avoiding preventable ADRs to the maximum.
6.2.1 Prescription/Selection ofDrug
Therapy
Clinicians must review the details from patient
records, patient interviews, and other sources to
identify the presence of predisposing factors
before making a decision of prescribing a specic
medication. In situations where dysfunctions such
as renal, hepatic, and blood abnormalities could
be a predisposing factor, specic laboratory investigations should be conducted before using the
particular medication in the patient. If the patient
has a specic predisposing factor, then the possibility of selecting alternative drugs that may be
safer to the patient should be considered.
Physicians should avoid inappropriate prescribing
and unnecessary polypharmacy [93].
Careful consideration should be given to special groups such as the elderly and children.
Older patients are often more sensitive to the
effects of medications than are their younger
counterparts because of altered pharmacodynamic responses such as excessive sedation and
confusion with morphine and increased anticoagulant effect with warfarin when used at “standard” treatment doses. Such pharmacodynamic
responses are generally predictable and can be
minimized by starting at the lowest possible
dose and titrating to response. Good clinical
practice for detecting and predicting ADRs in
vulnerable patients includes detailed documentation and regular review of prescribed and OTC
medications through standardized medication
reconciliation. Deprescribing should occur at an
individual level when drugs are no longer benecial or indicated or when safer alternatives
exist [93]. Understanding the different effects of
patient factors on ADRs enables health-care
professionals to give the best advice to their
patients to prevent or minimize these ADRs.
Pharmacogenomic principles and facilities
should be utilized in clinical practice, wherever
they have shown to be useful and practically
possible, to help appropriate drug selection or
dose titration. Information technology such as
computer physician order entry with decision
support helps check for problems such as drug
allergies and drug–drug interactions, providing
dosage adjustment calculations, checking for
appropriate baseline laboratory results, updating the prescriber with the latest drug information, and thereby helping use of medications
safely with careful consideration of predisposing factors [94].
Patient involvement is required for making
decisions on the safe use of medications through
appropriate discussions with the patient or caregiver. It should be borne in mind that, though the
predisposing factors for specic ADRs may not
be present at the time of prescribing, there is a
possibility of its development during drug use,
especially for chronic medications. For instance,
development of renal dysfunction in a patient on
digoxin during the course of drug use can increase
the risk of digoxin toxicity. Prescribers should be
knowledgeable and vigilant for ADRs and their
contributory risk factors, togive the best advice
to their patients.
6.2.2 Dispensing andCounseling
Pharmacists should, before dispensing medications, carefully consider the presence of any predisposing factors for ADRs in the patient, the need
for any specic requirement according to the drug
label, and the presence of safer alternatives(to be
discussed with the prescriber). Appropriate counseling should be provided to the patient, which
Соседние файлы в папке Библиотека им академика М.И. Перельмана
