Eur J Clin Pharmacol
DOI 10.1007/s00228-017-2223-5
PHARMACOEPIDEMIOLOGY AND PRESCRIPTION
Lack of essential information in spontaneous reports of adverse
drug reactions in Catalonia—a restraint to the potentiality
for signal detection
Lorraine Plessis 1 & Ainhoa Gómez 1,2 & Núria García 1,3 & Gloria Cereza 1,3 &
Albert Figueras 3,4
Received: 9 December 2016 / Accepted: 17 February 2017
# Springer-Verlag Berlin Heidelberg 2017
Abstract
Purpose The aim of this study is to analyze the quality of the
information contained in the adverse drug reactions (ADR)
reports and to describe the magnitude and characteristics of
the lacking information.
Methods All reports of serious ADR received by the Catalan
Center of Pharmacovigilance in 2014 were analyzed using the
VigiGrade and a more clinical and qualitative approach.
Results Up to 824 reports describing serious ADR were in-
cluded in the study; of them, 503 (61.0%) were sent by health
care professionals (HPs) and the remaining 321 (39.0%) came
from pharmaceutical companies (PhC). More than 80% of
missing variables such as ‘onset date’ or ‘time-to-onset’ of
the ADR were from PhCs reports. ‘Onset of treatment date’
was not filled in 28 (22.2%) of the reports including an ‘addi-
tional monitoring’ medicine, and ‘end of treatment’ date was
not completed in 53 of those reports (42.1%). In summary,
39% of the reports involving a black triangle medicine sent
by PhCs lacked some essential information such as the onset
date of treatment.
Conclusions More than one third of the reports coming from
manufacturers did not include information that is considered a
limiting factor to evaluate any causal relationship, and can be
* Albert Figueras
afs@icf.uab.cat
1
Catalan Pharmacovigilance Center, Barcelona, Spain
2
Clinical Pharmacology Service, University Hospital Vall d’Hebron,
Barcelona, Spain
3
Fundació Institut Català de Farmacologia, University Hospital Vall
d’Hebron, P. Vall d’Hebron, 119-129, E-08035 Barcelona, Spain
4
Departament de Farmacologia, de Terapèutica i de Toxicologia,
Universitat Autònoma de Barcelona, Barcelona, Spain
an issue for the detection of safety signals. To take advantage
of this huge amount of potentially important information that
is almost useless at present, data mining tools and new algo-
rithms should be developed and tested with the aim of finding
formulas to deal with a huge amount of low quality data with-
out losing it, nor generating a number of false associations.
Keywords Pharmacovigilance . Pharmaceutical industry .
Database . Data mining
Introduction
The early postmarketing period is critical for any newly
marketed medicine as well as for the patients exposed to it,
mainly because this is the moment to build-up knowledge
about its efficacy and toxicity profile in non-experimental
conditions. Although the number of medicines withdrawn or
restricted after marketing represent about 1% of approved
products [1], the discovery of previously unknown side effects
to new active ingredients force the modification of SPCs, the
inclusion of new precautions or contraindications of use, or
even the limitation of its prescription in many cases.
Additionally, the discovery of safety problems is more likely
for medicines approved in a hurry, immediately before the
deadline than for those approved after a more calm review
of the available evidences [2].
Useful information that helps in building-up this knowledge
on the new medicines is obtained from different sources such
as case-reports, observational studies, randomized trials, sys-
tematic reviews or analyses of big databases containing clin-
ical outcomes for individual patients [3]. Notwithstanding
this, some studies suggest that it could be possible to foresee
the occurrence of many adverse drug reactions (ADRs) at the

time of licensing [4], especially if transparency in the publica-
tions of clinical trials was greater [3].
Spontaneous reporting (SR) is an extended surveillance
system aimed at the detection of potential associations be-
tween a treatment and an ADR in clinical practice, and its
main objective is to contribute to define and strengthen the
safety profile of medicines and eventually to detect new asso-
ciations (safety signals). Despite the acknowledged limitations
and biases, SR has demonstrated its strength in detecting new
safety issues, an outcome that has increased with the introduc-
tion of data mining techniques [5].
Safety signals are the result of identifying a number of
reports including a suspicion of an ADR produced by a spe-
cific medicine. There are different valuable methods for signal
detection and amplification, from manual observation to pro-
portional reporting ratios. Additionally, different systems have
been studied to try to reduce the noise due to nonsense or
spurious associations [6].
One of the key points in the whole process of signal detec-
tion is the quality of the information included in the original
reports, the strength of the causal relationship of each case
reported to the pharmacovigilance system assessed with the
aid of any of the causality algorithms, and the accuracy of its
inclusion in the databases. The Uppsala Monitoring Center
carried out a study of the WHO Pharmacovigilance
Database (VigiBase) that revealed that only 10.6% of the re-
ports received in 2000 included essential information such as
the indication of the treatment, the outcome or the chronology
of the reaction [7, 8], thus a tool was developed (VigiGrade)
which attributes a score according to the quality of the infor-
mation of the reports included in VigiBase [9]. In line with
the score and the analysis of the available data, the authors
found that only 13% of the reports had a good level of com-
pleteness of information.
The aim of the present study was to analyze the quality of
the information contained in the reports received in the Catalan
Pharmacovigilance Center (CPvC) during 2014, as well as to
describe the magnitude and characteristics of the lacking
information.
Methods
To conduct the present study, all suspicions of serious ADRs
gathered by the CPvC and included into the database of
Spanish System of Pharmacovigilance (SSPv) since January
1, 2014 to December 31, 2014 were selected and analyzed.
The CPvC is a large regional center that was the spark of the
future SSPv in 1983. As the remaining 16 regional centers, the
CPvC receives and processes all spontaneous reports origi-
nated in its reference region. All reports are individually
assessed; additional information is requested to the reporter
if necessary, and the causality is evaluated by using a
Eur J Clin Pharmacol
modified Karch and Lasagna algorithm [10] that takes into
consideration five questions regarding the suspected medi-
cine, the characteristics of the ADR, the patient, his/her dis-
eases and comorbidities, as well as concomitant treatments, in
order to establish the ‘temporal sequence’ and the ‘previous
knowledge’ of the association, the ‘evolution after stopping
the suspected medicine’, the ‘effect of the re-exposure to the
suspected medicine’, and the presence of alternative causes
that could also explain the symptoms. Once evaluated, each
report is uploaded to the SSPv database (FEDRA, for
Farmacovigilancia Española Datos de Reacciones
Adversas). The CPvC contributed with approximately 18%
of the serious ADRs reports included in the Spanish database
in 2014.
In Spain, reports come from different sources: health care
professionals, pharmaceutical companies, and citizens. For
pharmaceutical companies, reporting all serious ADRs is
compulsive and these reports should be sent to the SSPv with
a maximum 15 days delay since they had the first sign of the
case [11, 12].
Some medicines require a special monitoring as defined by
the EMA [13] due to the lack or poor safety information, are
described as being under ‘additional monitoring’ (AM) and
are signaled by a ‘black inverted triangle’ on the leaflet of the
product.
Exclusion criteria
The following reports were excluded of the present analysis:
(1) those that were not spontaneous (e.g., ADRs identified
from clinical studies, special safety programs for vaccines
and medical errors, or reports obtained from descriptions of
cases in medical publications); (2) reports sent by citizens, and
(3) non serious ADRs. Non serious ADRs have been exclud-
ed in order to be able to compare all the information reported
by health professionals and by pharmaceutical companies,
which only report serious effects.
Study variables
All variables present in the yellow card form (Targeta groga)
had been taken into account. The data collected in the
reporting form is similar to the data included in most reporting
forms of many national pharmacovigilance programs, and
have been described elsewhere [14]. In summary, these vari-
ables include information of the patient characteristics (age
and gender), information about the suspected and concomitant
medicines (name, dosage, administration route, starting and
ending dates of treatment, and indication of use), information
about the ADRs (starting and ending dates of the reaction,
outcome and treatment of the reaction), origin of the report,
and profession of the reporter. An additional variable has been
created for a better analysis: ‘time-to-onset’ defined as the
Eur J Clin Pharmacol

period between the starting of the treatment and the appear- partially completed (15%); partial ‘ADR onset’ data (15%),
ance of the reaction. information on age group, but not specific age (10%); partially
completed dosage (5%). VigiGrade considered a report as
Relative weight of the study variables ‘well-documented’ if the final score exceeded 0.8. This limit
has been determined, taking into account that a lack of infor-
Despite the interest of all the variables included in the report mation regarding two ‘supportive’ variables is accepted, but
form, not all have the same importance for the evaluation of the penalty is 50% for ‘essential’ and 30% for ‘important’
the causality relationship between the suspected medicine and variables. As the present study only included reports from
the ADR described in the report. This is particularly relevant Catalonia, the country of origin was not taken into account.
for the application of the causality algorithm, and also for the Finally, a combined variable was created to analyze the report-
detection and evaluation of safety signals. Thus, not having er background and clinical setting (hospital or primary care); if
information about the age of the patient or the potential treat- this information was missing, the penalty was 10% (see
ment received to alleviate the ADR seems not as important as Table 1).
knowing the exact onset dates of the suspected medicine and The total score was calculated for each report and varied from
the ADR in order to elucidate if there was a clear temporal 0.07 and 1 [9]. In the present study, the minimum completeness
relationship. was: 1 × 0.5 × 0.7^4 × 0.9^2 = 0.097, depending on the different
For the present study, a double complementary approach penalties. Thus, the score for each report ranged 0.097 to 1.
has been carried out to define the weight of the different study
variables. On the one hand, the VigiGrade score has been
applied; on the other hand, a more clinical and qualitative Results
approach is proposed. For each variable in each report, com-
plete information scored ‘1’ and incomplete information Among the 2002 reports identified in 2014, 824 met the in-
scored ‘0’. For some variables, it has been necessary to attri- clusion criteria and were considered the study sample (Fig. 1);
bute a score of ‘0.5’ (e.g., when the variable ‘age of the pa- up to 503 (61.0%) were sent by health care professionals
tient’ was only completed with the age group [i.e., adult, child, (HPs) and the remaining 321 (39.0%) came from
elderly...]).
With the aim of estimating the lack of efficiency and Table 1 Different penalties attributed to the lack of information in
to improve the pharmacovigilance data recovery system, variables according to the VigiGrade [9] and the present study
the WHO created the VigiGrade score [9]. This score is Variable Penalty according to Penalty for the present
based on the amount of information provided, and eval- VigiGrade study
uates the dimensions of the report rather than the spe-
cific elements included in it. The most important point Time-to-onset 50% 50%
of the VigiGrade is the determination of a penalty for 30% if the uncertainty 30% for uncertainty
exceeds 1 month
each incomplete variable: within each item of a report, 10% otherwise
the absence of information is granted by a penalty de-
Indication of use 30% 30%
pending on the importance of the variable. An adapted
15% for partially
version of this score has been calculated for this study. completed reports
According to VigiGrade, the penalties defined when the ADR outcome 30% 30%
information of a specific variable is not filled in are: 15% for information
partially completed
– Time-to-onset of the reaction: 50% penalty if the variable Sex 30% 30%
was not present in the report; 30% penalty if the uncer- Age 30% 30%
tainty exceeds 1 month, and 10% otherwise. 10% if only the group of 10% if only the group of
– Indication of use of the medicine ADR outcome, sex and age was present age was present
age: 30% penalty. Dose 10% 10%
– Dosage of the suspected medicine, reporter country, pri- 5% if the dose was
mary reporter, and Bcomments^: 10% penalty (see partially completed
Table 1). Reporter country 10% Not applicable
Primary reporter 10% 10% corresponding to
variables: profession of
the reporter + origin of
The following VigiGrade scores have been slightly adapted report
to the characteristics of the local yellow card forms: uncertain- Comments 10% Not evaluated
ty of the ‘time-to-onset’ (30% penalty); ‘indication of use’

ALL REPORTS DURING 2014:
2014
2002
SPONTANEOUS REPORTS:
1107
SPONTANEOUS NON CITIZEN REPORTS:
1022
SPONTANEOUS NON CITIZEN SERIOUS REPORTS:
824
Fig. 1 Selection of the sample and exclusion criteria
pharmaceutical companies (PhC). These 824 reports included
a description of 1438 ADRs and involved 1198 suspected
medicines.
Incomplete information
Among the 824 reports analyzed, the age was incomplete
in 71 reports (7.9%) and sex of the patient was not filled in
13 reports. Regarding the variables that collect information
about the suspected medicines, ‘onset of treatment’ date
was not filled in 118 reports (14.3%) and ‘end of treatment’
date was not filled in 227 reports (27.5%). The treatment
dosage and the administration route were unfilled in 322
(39.1%) and 120 (14.6%) reports of the study sample, re-
spectively. The indication of use was not filled in 76
(9.2%) reports. A number of variables refer to the ADR.
The ‘onset of reaction’ date was not completed in 95 re-
ports (11.5%). ‘Offset of reaction’ date was not present in
413 reports (50.1%) and information about the ‘outcome’
was not available in 16.5% of them. In more than 61.3% of
the reports (505), no information about potential treatments
for the ADR was filled. Thus, ‘time-to-onset’ period could
not be calculated in 143 (17.5%) reports.
Eur J Clin Pharmacol
NON SPONTANEOUS REPORTS: (including
publicaons and reports coming from special
i
programs as vaccines d medical
and ) 895
di l errors):
CITIZEN REPORTS: 85
NON-SERIOUS REPORTS: 198
Origin and completeness of the reports
Table 2 shows the differences in missing information accord-
ing to the characteristics of the senders, mainly health profes-
sionals (HPs) and pharmaceutical companies (PhCs). Results
showed that 97% of the reports with missing information
about the age of the patient and all the reports with missing
information about gender or place of origin came from PhCs.
More than 80% of the missing onset date, administration
route, and the time-to-onset also came from PhCs.
Chi-square tests showed statistically significant differences
between HPs and PhCs in the completeness of all the variables
included in the report, being the reports of the PhCs more
incomplete than those coming from HCs (p < 0.001).
Reports including ‘additional monitoring’ medicines
Out of 824 reports received in 2014, 126 (15.3%) included at
least one medicine which required a special attention. It is to
be highlighted that 61.1% of those reports came from PhCs. In
fact, one quarter of the reports sent by PhCs involved one AM
product. Those most frequently reported in the study sample
were telaprevir (20 reports) and infliximab (19 reports).
Eur J Clin Pharmacol

Table 2 Missing information in

variables of the yellow cards Yellow Card variables All reports PhCs reports HPs reports
depending on the origin of the n = 824 (%) n = 321 (%) n = 503 (%)
report: pharmaceutical companies
(PhCs) or health professionals Patient variables Age 71 (7.9) 69 (21.5) 2 (0.4)
(HPs) Sex 13 (1.6) 13 (4) 0 (0)
Origin 234 (28.4) 234 (72.9) 0 (0)
Suspected medicine variables Onset date of treatment 118 (14.3) 95 (29.6) 23 (4.6)
Offset date of treatment 227 (27.5) 174 (54.2) 53 (10.5)
Indication 76 (9.7) 46 (14.3) 30 (6.0)
Dose 322 (39.1) 164 (51.1) 158 (31.4)
Route 120 (14.6) 102 (31.8) 18 (3.6)
ADR variables Onset date 95 (11.5) 95 (29.6) 0 (0)
Offset date 413 (50.1) 211 (65.7) 202 (40.2)
Treatment 505 (61.3) 264 (82.2) 241 (47.9)
Outcome 136 (16.5) 96 (29.9) 40 (7.9)
Time-to-onset 143 (17.5) 120 (37.4) 23 (4.6)

‘Onset of treatment’ date was not filled in 28 AM reports
(22.2%) and ‘end of treatment’ date was not completed in 53
reports (42.1%). The treatment dosage and the administration
route were not filled in 59 (46.8%) and 23 (18.3%) of the AM
reports sample, respectively. The indication of use was incom-
plete in 12 (9.5%) of those reports.
Regarding the description of ADRs; the ‘onset of reaction’
date was not completed in 30 AM reports (23.8%) and ‘time-
to-onset’ period could not be calculated in 37 reports (29.4%).
‘Offset of reaction’ date was not present in 81 reports (64.3%),
and information about the ‘outcome’ was not available in
26.2% of them. It should be highlighted that no information
about potential treatment for the ADR was filled in more than
75% of the reports (96).
A comparison of the completeness of the most important
variables included in the report, according to its origin is pre-
sented in Table 3. In summary, 39% of the reports involving
an AM medicine sent by PhCs lacked some essential informa-
tion such as the onset date of treatment. Regarding the reports
sent by PhCs, it should also be highlighted that, for each
variable, the lack of information was greater when an AM
drug was involved in the report, except for gender, route of
administration, and indication of use for the treatment.
Chi-square tests were used for those variables with five or
more reports in each category, statistically significant differ-
ences of completeness between PhCs and HPs reports were
found (p < 0.005).
VigiGrade score
In the present study, the same criteria as the VigiGrade were
used, considering a report ‘well-documented’ when the score
exceeded 0.8. The results showed that 350 reports (42.5% of
the study sample) met these criteria. After classifying the re-
ports according to their origin, more than half of those coming
from HPs (54.7%) had a score above 0.8, but only 23.4% of
those reported by PhCs scored above 0.8 (chi-square test
=78.61; p < 0.01).
Regarding the subsample of reports involving AM medi-
cines, 41 (32.5%) presented at least a score of 0.8; 15 were
reported by PhCs and 26 by HPs. It should be highlighted that
only 19.4% of the AM reports sent by PhCs had a score above
0.8, while 53.0% of the AM reports sent by HPs scored above
0.8 (chi-square test =14.26; p = 0.0016).
Discussion
The present study analyzed the quality of the serious sponta-
neous reports from PhCs and HPs received by the Catalan
Center of Pharmacovigilance during 2014, according to the
completeness of the information filled in the report forms. The
main result was that more than one third of the reports coming
from manufacturers did not include information to assess the
appropriateness of the time elapsed between the exposure to
the suspected medicine and the appearance of the adverse drug
reaction. The lack of this information is considered a limiting
factor to evaluate any causal relationship, and can be an issue
for the detection of safety signals. It should be noted that
among the reports coming from health care professionals,
the information regarding time-to-onset was better filled, as
only 4.6% of the reports missed this information.
Regarding the attributed score, the results showed that al-
most one out of two reports had a score under 0.8, which is
considered as not well-documented report. This ratio is better
than the global proportion obtained by the WHO study [9], but
it should be noted that only serious reports have been evalu-
ated in our sample, thus it is possible that reporters were less
cautious to complete reports of non serious ADR. On the other
hand, this ratio is lower than that obtained by Spain in the
 Eur J Clin Pharmacol

Table 3 Results of the missing

information comparing Variables PhCs HPs
‘additional monitoring’ (AM)
medicines reports and non AM Non AM reports AM reports Non AM reports AM reports
reports depending on the sender (n = 244) (n = 77) (n = 454) (n = 49)
(PhCS or HPs)
Gender 11 (4.5) 2 (2.6) 0 (0) 0 (0)
Age 52 (21.3) 17 (22.1) 2 (0.4) 0 (0)
Place of origin of the 177 (72.5) 57 (74.0) 0 (0) 0 (0)
report
Onset date of the 65 (26.6) 30 (39.0) 0 (0) 0 (0)
reaction
Offset date of the 155 (63.5) 56 (72.7) 177 (39.0) 25 (51)
reaction
Treatment of the 196 (80.3) 68 (88.3) 213 (46.9) 28 (57.1)
reaction
Outcome 70 (28.7) 26 (33.8) 33 (7.3) 7 (14.3)
Onset date drug 69 (28.3) 26 (33.8) 21 (4.6) 2 (4.1)
Offset date drug 127 (52.0) 47 (61.0) 47 (10.3) 6 (12.2)
Route 81 (33.2) 21 (27.3) 16 (3.5) 2 (4.1)
Dose 120 (49.2) 44 (57.1) 143 (31.5) 15 (30.6)
Indication 37 (15.2) 9 (11.7) 27 (6.0) 3 (6.1)
Time-to-onset 85 (34.8) 35 (45.5) 21 (4.6) 2 (4.1)

WHO study; this can be explained because our sample only
included data from Catalonia during 1 year, and was more
recent than the WHO sample. A more detailed analysis should
be carried out to determine if there is a progressive and wide-
spread decline in the quality of reports.
On the other hand, less than a quarter of the reports coming
from laboratories presented a Bwell-documented^ level of
completeness. So, the quality of the available information in-
cluded in the report forms is not enough to apply the algorithm
and determine the causality between the suspected medicine
and the serious ADR in most received reports. In order to try
to overcome this great limitation, the PhV centers should con-
tact the reporters (PhCs) that, in turn, should contact the health
professionals which originally informed them about the sus-
picion. This process uses to be long and not always results in
new relevant information. So, in addition to the great volume
of work for the PhV center, not collecting data accurately at
origin increases imprecision as it promotes recall bias either
by patients or health care professionals. This bias is especially
relevant regarding exposure dates.
‘Additional monitoring’ medicines
In relation with the special attention given to medicines la-
beled with an AM, globally speaking, more reports came from
laboratories (60%) than professionals (40%). The relative in-
terest of PhCs to report suspected ADRs in patients taking a
AM medicine can be better expressed by saying that almost a
quarter of the PhC reports analyzed (24%) included one of
these interest drugs, while AM reports only represented
9.7% of the HPs reports included in the study.
Sending more AM reports is a good point, as more safety
information is required for those active ingredients due to their
novelty or lack of safety information. But a careful analysis
showed that those reports were generally less completed than
other reports. This result is shocking, as the purpose of the AM
status is to try to ensure a closer monitoring and priorization.
Moreover, regarding the quality of AM reports, similar results
to non AM reports were found: there was a significant differ-
ence of completeness of the yellow card variables between
PhCs and HPs. For example, the onset date of the drug was
not completed in one third of the AM reports coming from
laboratories and only in 4.1% of those coming from profes-
sionals. This result accentuates the problem, as these are drugs
which need a special attention. In summary, the quality of the
variables of the AM reports was worse than that of all other
reports, and to our opinion, this point should be addressed by
health authorities and PhCs.
Since the beginning of the pharmacovigilance systems at
the end of the 1960s, they were based upon reporter’s suspi-
cions that were summarized in the reporting forms, the ‘yel-
low cards’. Traditionally, these forms were designed to in-
clude the minimum information to assess the causal relation-
ship by using a causality algorithm and, with the aim of being
practical, that information was minimal in order to increase the
easiness of reporting, thus increasing the sensibility of the
spontaneous reporting system despite losing some specificity.
The main idea was to quickly identify signals from suspicions
that could be confirmed later as a way to avoid risky exposures
Eur J Clin Pharmacol
before establishing a clear causality by means of long and
expensive observational studies.
Comparison with previous studies
Different studies have been conducted to determine the quality
of the reports of adverse drug reactions. With a similar objec-
tive and using their national pharmacovigilance database, a
group of Italian researchers carried out a study evaluating
the quality of ADRs received in 2009 [15]. They created an
algorithm taking into account the seriousness of ADRs but
also other clinical or social criteria, even if the report was
not considered serious. They considered four important ele-
ments to determine the quality of the reports: ‘causality’, ‘no-
toriety’, ‘clinical relevance’, and ‘completeness’ of the re-
ports, and found that only 21.6% of reports had a Bhigh^
quality according to the sensitivity and specificity values.
This study is in accordance with the method we used, and
the results were quite similar to ours.
Another study conducted in France compared the quality
of reports coming from laboratories or regional
pharmacovigilance centers as done in the present study, al-
though it focused on two drugs subject to risk management
plans (sitagliptin and exenatide) [16]. The authors created a
tool called ATHE score for the quality evaluation and took
into account: associated medications, time-to-onset, medical
history, and evolution; so, the higher is the score, better is the
completeness of the report. A comparison of PhCs and centers
scores also showed that reports from centers were better filled
than those from PhCs, and the maximal ATHE score was
obtained in 66% of the reports coming from centers and only
in 44% of the reports coming from PhCs, with a significant
difference between both groups (p < 0.001).
A study carried out by the Quarter Watch project team
(Institute for Safe Medication Practices) [17] analyzed reports
from the US Food and Drug Administration’s Adverse Event
Reporting System (FAERS). In 2009, 96.6% of all the reports
collected by the FAERS came from manufacturers and, within
those reports, only 49.6% had the minimum information (age,
sex, and the event date) which permitted to consider those
reports as valid. In comparison, the proportion for reports from
HPs and consumers with the minimum information rose to
85%.
To our knowledge, the analyses conducted with the objective
of describing the quality of reports are in line with the present
study and concluded that the quality of ADR reports is in
general not as good as desired, and all studies which specifi-
cally focused on manufacturer reports noticed that the quality
in terms of completeness was even worst. A deeper analysis
could be performed in order to distinguish the proportion of
missing information in variables only partially completed.
The present study included a sample of serious adverse
drug reactions. In theory, reports involving those adverse
effects should be more carefully filled, due to the severity of
the clinical condition and the potential consequences for the
patient. A bias in the sample selection could be a limitation of
this study that, in addition, was carried out in a selected time
period. Notwithstanding this, it should be highlighted that the
CPvC began its activities in the first 1980s, so its evaluating
experience had already been consolidated. Additionally, many
of the reporting PhCs were transnational companies with great
international experience in reporting. On the other hand, PhCs
should be interested in an optimal reporting practice in
order to ensure the best knowledge for their medicines, as this
is the best way to help patients to benefit from them.
At this point, the question is perhaps how to take advantage
of such a huge amount of potentially important information in
order to help to transform ADRs into reports in new knowl-
edge about the safety profiles of newly marketed medicines.
Reporting has evolved from the first PhV era; the reporters are
also different and more conscious of the potential risks asso-
ciated with any product, access to information has dramatical-
ly changed, as well as the capacity of data management. So, to
our opinion, the urgent question is whether data mining tools
and new association algorithms should be developed and test-
ed with the aim of finding formulas to deal with a huge
amount of low quality data without losing it, nor generating
a number of false associations. There is a good research area
there. Meanwhile, health authorities should recommend a real
implication of PhCs as actors in the postmarketing surveil-
lance farther than the bureaucratic risk management plans,
especially in the present moment, where medicines are pushed
to be marketed earlier.
Acknowledgements This work was supported in part by the
Departament de Salut de la Generalitat de Catalunya, a government insti-
tution of Catalonia.
Contributions of authors Lorraine Plessis designed the initial study
protocol, retrieved and selected the reports, analyzed and discussed the
results and prepared the first draft of the manuscript.
Ainhoa Gómez contributed to the data analyses and prepared tables
for the discussion of results. She also participated in the discussion of the
manuscript draft, and improved the last version of the manuscript.
Núria Garcia trained LP in the use of the FEDRA Database, contrib-
uted to the data analyses and prepared tables for the discussion of results.
She also participated in the discussion of the manuscript draft, and im-
proved the last version of the manuscript.
Gloria Cereza participated in the design of the study protocol, coordi-
nated the meetings to discuss the assessment of the included reports,
indicated the results to be included in the manuscript and actively partic-
ipated in the first draft review.
Albert Figueras proposed the idea of the study, contributed to the
study protocol, participated in the general results discussions and
reviewed the second draft of the manuscript and prepared the final draft.
Compliance with ethical standards
Conflict of interest The authors declare that they have no conflict of
interest.

References
1. Wysowski DK, Swartz L (2005) Adverse drug event surveillance
and drug withdrawals in the United States, 1969–2002: the impor-
tance of reporting suspected reactions. Arch Intern Med 165:1363–
1369
2. Carpenter D, Zucker EJ, Avorn J (2008) Drug-review deadlines and
safety problems. N Engl J Med 358:1354–1361
3. Onakpoya IJ, Heneghan CJ, Aronson JK (2015) Delays in the post-
marketing withdrawal of drugs to which deaths have been attribut-
ed: a systematic investigation and analysis. BMC Med 13:26. doi:
10.1186/s12916-015-0270-2
4. Aagaard L, Soendergaard B, Stenver DI, Hansen EH (2008)
Knowledge creation about ADRs—turning the perspective from
the rear mirror to the projector? Br J Clin Pharmacol 65:364–376
5. Härmark L, van Grootheest AC (2008) Pharmacovigilance:
methods, recent developments and future perspectives. Eur J Clin
Pharmacol 64:743–752
6. Grundmark B, Holmberg L, Garmo H, Zethelius B (2014)
Reducing the noise in signal detection of adverse drug reactions
by standardizing the background: a pilot study on analyses of pro-
portional reporting ratios-by-therapeutic area. Eur J Clin Pharmacol
70:627–635
7. Lindquist M (2008) VigiBase, the WHO global ICSR database
system: basic facts. Drug Inf J 42:409–419
8. Lindquist M (2003) Seeing and observing in international
pharmacovigilance: achievements and prospects in worldwide drug
safety. Uppsala Monitoring Centre, Uppsala
9. Bergvall T, Norén GN, Lindquist M (2014) vigiGrade: a tool to
identify well-documented individual case reports and highlight sys-
tematic data quality issues. Drug Saf 37:65–77
Eur J Clin Pharmacol
10. Karch FE, Lasagna L (1977) Toward the operational identification
of adverse drug reactions. Clin Pharmacol Ther 21:247–254
11. European Medicines Agency (2014) Guideline on good
pharmacovigilance practices (GVP) Module VI. EMA http://
www.ema.europa.eu/docs/en_GB/document_library/Scientific_
guideline/2014/09/WC500172402.pdf. Accessed 7 Dec 2016
12. BOE (2007) Real Decreto 1344/2007, de 11 de Octubre, por el que
se regula la farmacovigilancia de los medicamentos de huso
humano. Boletin oficial del Estado http://www.boe.es/boe/dias/
2007/11/01/pdfs/A44631-44640.pdf. Accessed 7 Dec 2016
13. European Medicines Agency-Medicines under additional monitor-
ing (2016) List of medicines under additional monitoring. EMA
http://www.ema.europa.eu/ema/index.jsp?curl=pages/regulation/
document_listing/document_listing_000366.jsp. Accessed 7
Dec 2016
14. Sempere E, Palop V, Bayón A, Sorando R, Martínez-Mir I (2006)
Calidad de la publicación de reacciones adversas a medicamentos
en la sección de Cartas al Director de cuatro revistas españolas de
medicina interna y medicina general. Aten Primaria 37:187–193
15. Tuccori M, Giustarini G, Blandizzi C et al (2013) Quality of
Adverse Drug Reaction (QADRA) reports: an algorithm to appraise
t h e effi c i e n cy o f s p o n t a n e o u s r e p o r t i n g s y s t e m s i n
pharmacovigilance. J Public Health 21:365–372
16. Faillie JL, Robin P, Bres V, Pinzani V, Bos-thompson MA, Hillaire-
buys D (2013) The Athe score: a new quality score for spontaneous
adverse drug reaction reports. Fundam Clin Pharmacol 27:106–107
17. McCarthy M (2015) Drug makers’ adverse event reports are often
incomplete, US report finds. Br Med J 350:651. doi:10.1136/bmj.
h651