Signal generation in a non-EU country

ORIGINAL REPORT

Signal Generation in a Non-EU Country

K. HARTMANN, C. CIORCIARO AND M. KUHNSwiss Drug Monitoring Center SANZ, Neubruchstrasse 37, CH-7000 Chur, Switzerland

SUMMARY

One of the most important objects of post-marketing surveillance is the early detection of serious,unknown and unexpected adverse events. As spontaneous reported adverse events vary considerably intheir signal-generating value the Swiss Drug Monitoring Center SANZ implemented an early signaldetection system in 1991. With a set of screening criteria the individual case reports are automaticallychecked by the system and read into a signal ®le. Subsequently these potential signals are processed toquantify the signal generating value of the particular case. According to the screening criteria 2072(53%) of the reported cases were considered as potential signals. After validating and quantifying theirsignal value 27% (565) were drug-related and unlabelled, 8% (173) were also rated as serious. On thebasis of these cases 114 labelling changes were made. Some of the most interesting cases are discussed.This computer-assisted early signal generation model in a small country with 7 million inhabitants hasa high detectability of new, rare, serious and quality of life a�ecting adverse drug reactions (ADRs).These newly detected ADRs must be interpreted in cognizance of the limitations and restrictions ofspontaneous reporting and do not allow pharmacoepidemiologic conclusions. # 1997 John Wiley &Sons, Ltd.

Pharmacoepidemiology and Drug Safety, 6 Suppl. 3: S13±S19, 1997

No. of Figures: 4. No. of Tables: 9. No. of Referencs: 14.

KEY WORDS Ð post-marketing surveillance; spontaneous reporting; signal generation; Switzerland

INTRODUCTION

Switzerland is a small country with 7 millioninhabitants and with approximately 15,000 practis-ing physicians, how can signals be generated onsuch a scale?

Spontaneous reporting schemes (SRS) aremainly intended to produce signals about potentialnew adverse drug reactions (ADRs) and to detectrarely occurring adverse events (AE). Detectingsignals from SRS is certainly a question ofnumbers, i.e. the treated population size. To detectan ADR with the incidence of 1 : 10,000 we musttreat 30,000 patients and to see three similar cases65,000 patients have to be treated; this is withoutconsidering the background incidence.1 So thequestion arises, whether it is possible to generatesignals from the SRS in such a small country.

BACKGROUND

In Switzerland the reporting of AE by healthprofessionals, mainly physicians but also bypharmacists and dentists is voluntary. Thesereports are of suspected adverse drug events linkedto one or more drugs. It is not a strict eventreporting system. There is at least some concern onthe part of the doctor that the event is drug related.

In the seventies there were some discussionsin Switzerland of introducing SRS. Finally in1981 a private institution called by its acronymSANZ was founded by the society of the Swissmedical association and the Swiss pharmaceuticalindustry. In 1991 the regulatory authority IKS,the Intercantonal O�ce for the Control ofMedicine, started the coordination of the surveil-lance of adverse drug events in Switzerland, the

CCC 1053±8569/97/S30S13±07$17.50# 1997 John Wiley & Sons, Ltd.

PHARMACOEPIDEMIOLOGY AND DRUG SAFETY, VOL. 6 Suppl. 3: S13±S19 (1997)

Pharmacovigilance Center. There is a mutual trans-mission between SANZ and IKS of all seriousand unlabelled reports as well as all reports ofnewly marketed drugs.2 In January 1996 SANZextended its activities to the surveillance of bloodproducts. The Federal Institute of Health, theBAG, authorized SANZ to monitor the unstableblood products.3

The SANZ database contains more than 12,000reports and increases annually by approximately1200 reports. All reports in the data base areanalysed, validated and assessed for their causality.The data base is very homogeneous and with a highdata quality. A signal detection system was imple-mented in 1991.4

SIGNAL DETECTION

Spontaneous reports are certainly of an anecdotalnature but still they can have a possible signalfunction. We assume that a possible signal casemust have some speci®c characteristics. We there-fore set up a set of criteria to automatically scan allincoming reports (Table 1).The diagram in Fig. 1 shows how all incoming

reports are handled for signal generation: afteranalysing, evaluating and coding each case the key

data are fed into the data base and all cases are readinto the case report ®le. In addition, all cases areautomatically scanned for possible signal casesaccording to the screening criteria. All cases whichmeet at least one of these criteria are read into thesignal case ®le. The potential signals in this ®ledi�er greatly in their capacity to generate a signal.They have di�erent signal values. To estimate thevarious signal values of these potential signals weexamine the case-related or intrinsic value and thedocumentation status of a possible adverse drugevent or the extrinsic value (Tables 2 and 3): thedi�erential diagnosis is checked Ð could the eventbe explained by a cause other than the drug; howlikely is a drug-related cause; what is the timerelationship, was there a dechallenge or a rechal-lenge? Important are the quality and consistency ofthe reported data: what is the clinical relevance, the

Table 1 Ð Screening criteria

. Serious adverse events (CIOMS criteria)

. Medically serious events

. New chemical entities (NCE)

. Non-labelled adverse events

. First report to SANZ

. New application form

. Well-documented single case report

Fig. 1 Ð Signal generation

S14 K. HARTMANN, C. CIORCIARO AND M. KUHN

# 1997 John Wiley & Sons, Ltd. PHARMACOEPIDEMIOLOGY AND DRUG SAFETY, VOL. 6 Suppl. 3: S13±S19 (1997)

severity and the seriousness? For each potentialsignal case these variables are assessed and scored.The next step is to assess and score the documenta-tion status. Standard reference books are checkedsuch as AHFS Drug Information, Martindale,Meyler's Side E�ect of Drugs and the weeklyjournal Reactions by Adis Press. The labellingstatus and the SANZ data base is checked. Afterhaving assessed and scored the intrinsic as well asthe extrinsic value of a potential signal case thesignal value is estimated (Fig. 2). The signal value iscalculated by subtracting the result of the extrinsicscoring from the result of the intrinsic scoring.According to the calculated signal value the reportcan go back to the case report ®le with no actioninitiated, it can be set on a watch list or an in-housealert list, or a labelling change or publication in aSwiss or international medical journal can be theresult.

As SANZ is a private foundation it does nothave any regulatory competence, but alerts arediscussed with the companies as well as with the

authorities and depending on the situation variousactions are suggested.

From 1992±1995 a total of 4393 cases werereported. Of these cases 3911 were automaticallyscanned to extract potential signals. During these4 years SANZ received approximately 400 reportsfrom IKS within the mutual transmission ofreports. Therefore not all incoming reports werescreened for signals. More than half of the reportsproved to be potential signals and were read intothe signal case ®le. More than 25% of thesereports were unlabelled and drug-related, 1/3 ofthem also being serious. Of all screened cases14.5% were unlabelled and drug-related and 4.5%of all cases also unlabelled, were drug-related andserious (Table 4).

From 1992 until August 1996 we knowof 114 labelling changes in Swiss product inform-ation, largely due to generated signals fromSANZ.

Table 2 Ð Intrinsic evaluation

. Di�erential diagnostic analysis

. Plausibility of an association

. Chronology/dechallenge/rechallenge

. Quality and consistency of reported data

. Clinical relevance

. Degree of seriousness

Table 3 Ð Extrinsic evaluation

. Reference in the SANZ data base

. Reference in the product informationÐ labelled/unlabelled

. Reference in the medical literatureÐ detailed and comprehensive referencesÐ standard reference booksÐ isolated anecdotal reports, biologically expectedÐ not previously reported

Fig. 2 Ð Estimation of the signal value

SIGNAL GENERATION IN A NON-EU COUNTRY S15


Table 5 shows some examples of alerts fromSANZ. These alerts were published in national andinternational medical journals.5 ±13

EXAMPLES

Timeliness of an alert

In the beginning of 1992 terbina®ne, an oralantimycotic agent was launched in Switzerland.As early as 3 months after the launch the ®rst caseof a taste disorder was reported to SANZ, someweeks later a second case. In June 1992 the ®rst

case report of a taste disorder was published.14 Upto November 1992 SANZ had received six casesreporting dysgeusia with a similar time course andwith similar reversible symptoms. Thus an alertwas raised.7 In the 1993 Swiss terbina®ne productinformation this ADR was labelled. The manufac-turer states an incidence of 0.125% (Table 6). In theSANZ spontaneous reporting scheme taste dis-orders count for approximately 1/3 of all reportedadverse events with terbina®ne (Fig. 3).

Selective underreporting

Underreporting is a well-known fact in the SRS,although it is realized that underreporting di�ersgreatly with regard to the nature of an adverseevent. Experiences from SANZ show that medi-cally important events are far less underreportedthan non-serious events: in May 1994 SANZreceived a report of a massive aggravation of amyasthenia gravis. One hour after taking 500 mgazithromycin for an in¯uenza syndrome a 25-year-old female patient with myasthenia gravis experi-enced an acute deterioration and progressivemuscle weakness and 1 h later the patient had tobe hospitalized because of dyspnoea and amyasthenic crisis. There was a strong time relation-ship and in the medical literature there are a fewreports of an association between erythromycinand myasthenia gravis. By examining the patient'smedical records we found that in 1986 during ahospitalization due to thymectomy, the patient hadalready experienced worsening of her conditionseveral times after parenteral applications oferythromycin. In 1986 this was not regarded asan adverse erythromycin event, the exacerbationswere considered within the context of her disease(Table 7). Also in this case an alert was raised andthe case published.12 In December 1995 a case of ademasking of a previously unknown myastheniagravis by clarithromycin was reported (Table 8).The prevalence of myasthenia gravis is variously

estimated to be between 2 and 10 per 100,000 witha peak incidence for females in the third decade,while in males the incidence is higher in the sixthand seventh decades. Considering this low pre-valence rate the reporting rate of rare and medicallyserious events must be high.

Quality of life a�ecting AE

The third example leads to the assumptionthat quality of life a�ecting adverse events are

Table 4 Ð Analysis of the signal cases

Total reports on ®le 1992±1995:N � 4393Automatically screened by 3911 100%

the systemPotential signals 2072 53%Unlabelled and 565 14.5%

drug-related AESerious unlabelled and 173 4.5%

drug-related AE

Table 5 Ð Signals from SANZ

1992 Fluconazole Anaphylactic reactionRetinoids Teratogenicity study

1993 Lysozyme Allergic reactionsTerbina®ne Taste disorders

1994 Omeprazole Joint disordersHalofantrine Rhythm disordersMethotrexatelow-dose Lethal PC pneumonia

1995 Tiaprofenic acid CystitisAzithromycin Aggravation of myastheniaGlucocorticoids Hypersensitivity reactions

Table 6 Ð Timeliness of the alert to terbina®ne andtaste disorders

! February 1992 Launch of terbina®ne inSwitzerland

! May 1992 First report of tastedisturbance to SANZ

! June 1992 First report published (Lancet)! November 1992 Manuscript `Taste disorders

and terbina®ne' sent to BritishMedical Journal

! July 1993 Publication of the alert inBritish Medical Journal

! 1993 Labelled in the Swiss productinformation (incidence: 0.125%)



reported with a higher rate. Meanwhile gingivahyperplasia is known to be a class e�ect ofcalcium channel blocker agents. With nifedipineSANZ received the majority of gingiva hyperplasiareports. It was assumed that this is becausenifedipine is most widely used in Switzerland. In1990 amlodipine was introduced onto the Swissmarket and 1.5 years later the ®rst case wasreported to SANZ. In 1992 the manufacturerstated that they knew of only three additionalreports worldwide, so no labelling change wasconsidered necessary though gingiva hyperplasia isa class e�ect. As new cases were reported a labellingchange in Switzerland will now be considered andin 1997 it should be labelled (Fig. 4). Table 9 showsthe labelling status of gingiva hyperplasia forcalcium channel blocker agents and their year ofmarketing in Switzerland.

CONCLUSION

We conclude that the SANZ signal detectionsystem is sensitive and a high rate of non-labelled,drug-related and serious AE have been detected.This signal detection system seems to have goodspeci®city: in particular medically serious and

quality of life a�ecting AE can be detected andrare AE were detected in a timely manner.

The carefully selected screening criteria helpto extract possible signals from the data pooland the signal value scoring system draws atten-tion to possibly relevant signals. The carefulreevaluation of these signal cases can give rise toan alert.

Most important for the signal generation in theSRS is the good clinical observation and criticaljudgement of reporting physicians. The detectionof new, rare and unexpected adverse drug reactionsis to a large extent dependent on alert physiciansand their awareness of and cooperation with ADRreporting.

Table 7 Ð Aggravation of myasthenia gravis by a macrolide1;2

May 1986 Diagnosis of myasthenia gravis Ð treatment: thymectomy, pyridostigmine, prednisone,azathioprine

December 1993 Mild deterioration (diplopia) Ð dosage adjustment23 March 1994 Deterioration because of an infection Ð dosage adjustment5 April 1994 Slight muscular weakness, fever

21 h: 500 mg azithromycin22 h: acute deterioration, progressive muscle weakness23 h: severe dyspnoea, myasthenic crisis, hospitalization

Strong time relationship Ð no similar case in the SANZ data base Ð few reports in the medical literature witherythromycin! examination of the patient's medical records

Table 8 Ð Demasking of a previously unknown myasthenia gravis by a macrolide

Until August 1995 No clue to a myasthenia gravisHealth status in accordance with age (66 years)

In September 1995 During 7 days treatment of an upper respiratory tract infection with clarithromycin250 mg bid

30 September 1995 Clinical symptoms: weakness of ocular muscles, subclinical: generalized symptoms ofweakness, diagnosis of myasthenia gravis, no thymoma

! prevalence is estimated to be between 2 and 10 per 100,000

Table 9 Ð Labelling of gingiva hyperplasia in Swissproduct information

Verapamil Labelled 1964Nifedipine Labelled 1975Diltiazem Labelled 1981Nitrendipine Labelled 1985Felodipine Labelled 1988Isradipine Not labelled 1989Amlodipine Not labelled 1990Nisoldipine Not labelled 1991Lacidipine Labelled 1993



REFERENCES

1. BeÂ gaud, B., Tubert-Bitter, P., Chaslerie, A. andHaramburu F. Signal generation in pharmaco-epidemiology: a new rule of three. Pharmaco-epidemiology and Drug Safety, 1995; 4: S67.

2. Kuhn, M. and Sahli, H. R. Zusammenarbeitzwischen der SANZ un der der IKS. SchweizerischeAerztezeitung 1992; 73: 10.

3. Brand, B., Hartmann, K. and Kuhn, M. HaÂ mo-vigilanz, eine neue Dienstleistung der SANZ im

Auftrag des Blutspendedienstes (BSD) SRK.Schweizerische Aerztezeitung 1996; 77: 18.

4. Hartmann, K., Kuhn, M. and Gartmann, J.Erfassung von neuen, unbekannten und seltenenunerwuÈ nschten Arzneimittelwirkungen im Spontan-meldsystem. Schweizerische medizinische Wochen-schrift 1992; 12: 1409±1413.

5. Hartmann, K., Kuhn, M. and Gartmann, J.Erfahrungen mit neuen Arzneimitteln: Omeprazol,Fluconazol, neue ACE-Hemmer. SchweizerischeAerztezeitung 1992; 73: 423±424.

Fig. 3 Ð Terbina®ne and taste disorders

Fig. 4 Ð Gingiva hyperplasia due to calcium channel blockers



6. Hartmann, K., Kuhn, M. and Gartmann, J.Allergische Reaktionen auf lysozymhaltige Arznei-mittel. Schweizerische Aerztezeitung 1993: 74: 1880.

7. Beutler, M., Hartmann, K., Kuhn, M. and Gart-mann, J. Taste disorders and terbina®ne. BritishMedical Journal, 1993; 307: 26.

8. Beutler, M., Hartmann, K., Kuhn, M. and Gart-mann, J. Arthralgias and omeprazole. British Med-ical Journal 1994; 309: 1620.

9. Hatz, Ch., Loutan, L., Raeber, A.-A., Hartmann,K. and Kuhn, M. Halofantrin, ein neues Malaria-medikament. Schweizerische Aerztezeitung 1994; 75:346±347.

10. Wyss, E., Kuhn,M., Luzi, H.-P. and Reinhart, W. H.Fatal verlaufende Pneumocystis-carinii-Pneumonie

unter low-dose-Methotrexat und Prednison-Therapiewegen chronischer Polyarthritis. SchweizerischeRundschau fuÈr Medizin (Praxis) 1994; 83: 449±452.

11. Stoller, R., Buclin, T., Hartmann, K. and Kuhn, M.Schwere Zystitiden unter TiaprofensaÈ ure (Surgam1).Schweizerische Aerztezeitung 1995; 76: 459±460.

12. Cadisch, R., Streit, E. and Hartmann, K. Exazerba-tion einer Myasthenia gravis pseudo-paralytica nachAzithromycin. Schweizerische medizinische Wochen-schrift 1996; 126: 308±310.

13. Hartmann, K., Kuhn, M. and Ciorciaro, C. Hyper-sensitivity reactions to parenteral glucocorticoids.Pharmacoepidemiology and Drug Safety 1995; 4: S17.

14. Juhlin, L. Loss of taste and terbina®ne Lancet 1992;339: 1483.



Documents

Signal generation in a non-EU country