Results

RESULTS

Data analysis in PEM utilises several approaches which combine the application of epidemiological methods with medical evaluation.

INCIDENCE DENSITIES

Since most adverse drug reactions are the so-called ‘type A’ reactions, which are caused by the pharmacological effects of the product, and commonly occur within a short period after exposure, comparing the rates of events occurring soon after exposure with subsequent periods provides a useful means to generate possible drug safety signals.

PEM provides a numerator (the number of reports) and a denominator (the number of patient-months or  patient-weeks of exposure), both collected within a known time frame (the difference, for each patient, between the start and stop dates of the drug being monitored).

The incidence density (ID) for a given time period, t, for each of the event terms in the DSRU dictionary is calculated as follows:

ID_t = [ number of events during treatment for period t / number of patient-months (weeks) of treatment for period t ] ×1000

The IDs per 1000 patient-months (or patient-weeks) of treatment are then ranked to give estimates of the ‘real-world’ frequency of reported events.

While events with higher incidence densities in the period after exposure compared with subsequent periods are considered safety signals for the product under study, such events may be due to the effects of a product taken before the drug under  study was started. For example, cough occurring soon after starting an angiotensin-II (A-II) receptor antagonist (e.g. losartan) may have been caused by an angiotensin-converting enzyme inhibitor taken before starting the A-II antagonist.

REASONS FOR STOPPING

The green form asks the doctor to specify the ‘Reason for stopping’ the drug being monitored if treatment was stopped. Thus, the ranked ‘Reasons for stopping’ (in terms of the number of reports of each event) is another source for generating signals and can be compared with the ranked IDs for the first month of therapy in each individual patient. As examples, data for the most frequently reported events with the two anti-epileptic drugs, lamotrigine and vigabatrin, are given in Table 24.3.

In general, there appears to be a high degree of correlation between these two sets of values. These values can be used to compare drugs within one ther-apeutic class: for example with anti-epileptic drugs it shows that rash is the most frequently reported event likely to be a drug side effect with lamotrigine, whereas rash is far less common with vigabatrin; simi-larly, respiratory tract infection (which occurs month in and month out in all cohorts and which is, with many drugs, unlikely to be related to either the drug or disease being treated) is fairly common among the ID values but virtually never appears among the common reasons for drug withdrawal.

GENERATION AND EXPLORATION OF SIGNALS

Signals are generated by an event having an unusually high ID or ranking in the list of ‘Reasons for stop-ping’ the drug being monitored or being considered medically important by the Research Fellow. While comparisons of incidence densities nearly always utilise the differences between the incidence density in the first month and subsequent months, it has been possible to use the difference between incidence densities in month 6 with months 1–5 in a 6-month study to generate signals for delayed adverse reac-tions such as gynaecomastia with finasteride (Wilton et al., 1996), a product used for benign prostatic hypertrophy.

Many signals have been generated in PEM, exam-ples include visual field defects in patients taking vigabatrin (Wilton et al., 1999), gastrointestinal intol-erance due to acarbose (Mackay et al., 1997a), oesophageal reactions with alendronate (Mackay et al., 1997b), aggression, agitation and abnormal dreams with donepezil (Dunn et al., 2000), diarrhoea in the elderly with lansoprazole (Martin et al., 2000), and serotonin syndrome with antidepressants (Mackay et al., 1999).

FOLLOW-UP OF IMPORTANT EVENTS

Analysis and evaluation of pharmacoepidemiologi-cal data should include medical assessment, both to improve the understanding of signals raised by epidemiological techniques and to raise (and evaluate) new signals or hypotheses by using medical judge-ment with appropriate systems for causal inference.

Medical evaluation of individual case reports and clusters of reports is an important part of PEM. Impor-tant safety signals have been generated in this way. In the PEM study of the antiepileptic drug vigaba-trin, following published case reports of visual field defects associated with the use of the product, four cases of visual field defects were identified initially in the PEM cohort. In view of the importance of the signal, 7228 patients who were reported to be taking the product by the end of the study were followed up by sending a simple questionnaire to the GP to ask whether any serious adverse events or changes in vision had been reported since the initial green form had been returned. In addition, if the patient has been seen by an ophthalmologist for visual problems, the ophthalmologist was asked to complete a ques-tionnaire giving details of visual field testing before and during treatment with vigabatrin. The follow-up information revealed an additional 29 cases of visual field defects which were considered by the ophthal-mologist to be probably or possibly related to viga-batrin, giving an incidence of risk of 7.00 per 1000 patients (Wilton et al., 1999). The follow-up exercise in the PEM study of vigabatrin contributed to the understanding of this important adverse reaction and provided a method to compute the reported rate of the adverse reaction in real clinical use which was not possible with spontaneous reporting or in clinical trials.

THE OUTCOME OF EXPOSED PREGNANCIES

All pregnancies reported during PEM studies are followed up by the medical and scientific staff of the DSRU in order to determine the outcome in those babies exposed during pregnancy to the drugs being monitored.

A review (Wilton et al., 1997) showed that 2508 pregnancies have been followed up in 34 PEM studies. The study drug was known to have been dispensed during 904 of these pregnancies (839 during the first trimester and 65 during the second/third trimesters). The first trimester pregnancies produced 553 live births among which 20 (3.6%) abnormal-ities were reported. The findings are little different from the proportion of abnormalities reported in the general population in the United Kingdom. Thus, these observational data may be of value to those who need to advise pregnant women exposed to newly marketed medicines. The pregnancy database of PEM is expanding. Moreover, the DSRU is currently analysing the pregnancy exposure data with the appli-cation of comparative statistical methods between products in the PEM database or with external data, e.g. national statistics of congenital abnormalities, and the results will be published in due course.

LONG LATENCY ADVERSE REACTIONS

Delayed reactions can be investigated by sending out further green forms relating to those patients shown in the initial PEM survey to be receiving long-term medication. One such study has provided reassuring data on the safety of long-term use of lamotrigine in epilepsy (MacKay et al., 1997c).

COMPARING DRUGS IN THE SAME THERAPEUTIC CLASS

The size of the PEM database (78 completed stud-ies with a total of one million patients) and advances in information technology are providing increasing opportunities to compare the safety profiles of prod-ucts in the same therapeutic class. In the last few years many comparative studies (Table 24.4) have been conducted using PEM data which contributed to the understanding of the safety of many products.

Comparisons in PEM have included the appli-cation of nested case–control methodology (Dunn et al., 1999). Nested case–control design appears to have useful applications to PEM and will be applied increasingly in the future. Another method that is currently being developed for signal generation in PEM is the routine application of comparative report-ing rates for reported events in PEM.

INVESTIGATION OF SAFETY SIGNALS FROM OTHER SOURCES

The DSRU monitors the literature and the World Wide Web for important drug safety signals gener-ated elsewhere, particularly those that cause public health or regulatory concerns. The Unit also receives requests from regulatory authorities and manufactur-ers to investigate drug safety signals in the PEM database. Whenever possible the DSRU conducts retrospective analyses (which usually include follow-up of reports for the drug in question and compara-tor drugs). Such analyses contribute to the debates on these signals and to regulatory and public health decisions.

One example is the study on sertindole (Wilton et al., 2001). Sertindole is an atypical antipsychotic known to be associated with prolongation of the QTc interval. The product was withdrawn from markets in the European Union following reports of sudden death and serious cardiac arrhythmias. The compara-tive analyses of the PEM studies of sertindole and two other atypical antipsychotics, risperidone and olanzap-ine, studied cardiovascular events, deaths from cardio-vascular events as well as deaths from other causes such as suicide. The report of the comparative anal-ysis was considered to be a very important source of information for the regulatory decision on the matter.

Another example of a retrospective analysis of a PEM study is the analysis conducted on the associa-tion between selective serotonin re-uptake inhibitors (SSRIs) and bleeding, which showed a possible weak association (Layton et al., 2001).

While such comparisons produce valuable additions to the understanding of the safety of medicines, it is important to emphasise that comparisons of inde-pendent cohorts are subject to bias and confounding, which must be taken into consideration in the anal-ysis and evaluation process. However, the paucity of post-marketing safety studies in large populations makes the information provided by these compara-tive studies very useful. Real benefit can only be achieved when not only the limitations of any post-marketing safety study are taken into consideration but when its results are considered in relation to other studies that had been conducted on the same product.

COMPARISON WITH NATIONAL DATA

Where appropriate, comparisons are made between event rates in PEM studies and other data resources, e.g. national statistics. An example is the analysis of cardiovascular events of the PEM study on silde-nafil (a product used for erectile dysfunction) (Shakir et al., 2001). Reported deaths from myocardial infarc-tion and ischaemic heart disease in users of sildenafil in the PEM study were found to be no higher than expected according to national mortality statistics. The precautions with regard to possible sources of bias and confounding also apply to external comparisons.