Discussion

DISCUSSION

Compared with the pharmacoepidemiological methodologies available today, the measures and methods employed in monitoring the adverse effects of nomifensine were those used in the normal clinical and laboratory assessments of haemolytic anaemia, more specialised immunological investigations into the relationship between the nomifensine and the dyscrasia and epidemiological observations. The last of these was not straightforward in the case of nomifensine because, as discussed, there was a very low rate of reported cases up until the increase in the mid-1980s. This undoubtedly led to the delay in establishing a cause-and-effect relationship between nomifensine and haemolytic anaemia.

It was considered that the rapid escalation of sponta-neous reporting could have been because of increased awareness among doctors resulting from reports of haemolytic anaemia in the literature, changes in the data sheets and encouragement to make use of the CSM’s Yellow Card system. It could also have been partly because of the increased promotion, sale and market share of nomifensine that occurred at the time. Furthermore, an impetus may have come from the withdrawal from the market of benoxaprofen (and increasing recognition of problems associated with other non-steroidal anti-inflammatory agents) during the early to mid-1980s. The neurological prob-lems caused by zimeldine also occurred during the early 1980s and may have contributed to heightened concern over, and increased reporting of, adverse reac-tions in general (Edwards, 1997a).

Altogether, nomifensine was associated with eight deaths before its withdrawal in the United Kingdom. Three of these were associated with haemolytic anaemia (a fourth haemolytic anaemia-associated fatality occurred after the product was withdrawn), and one each with a cardiac arrhythmia, an over-dose of nomifensine in conjunction with lithium, the Stevens–Johnson syndrome, hepatic necrosis and a cerebrovascular accident.

The ‘CSM Update’ in 1986 outlined the basis for the risk–benefit discussion, which took place late in 1985 when for the first time the incidence of haemolytic anaemia in the United Kingdom was greater than 1 in 10,000 prescriptions. However, reports of other adverse events remained modest.

Hoechst UK’s total database was only 55% of the CSM’s, but it contained 76% of all haemolytic anaemia reports and 88% of evaluable reports. For hepatic events, the company had 46 reports of which 44 (96%) were thought to be associated with the drug. This compared with the CSM’s 51 (86%) reports, but for fever the company had only 25 reports compared with the CSM’s 48 (52%). Thus, for perceived serious events, it appeared that prescribers felt more compelled to contact the company directly. For haemolytic anaemia and hepatic events, the manufacturer received over 75% of the reports that formed the CSM’s database compared with approximately 50% for all other adverse reactions, including fever. This supports the view that the company was aware of a greater proportion of seri-ous events than the average reporting rate to the company.

Despite some obvious associations such as increased prescribing, increased awareness of nomifensine after the launch of the 100 mg single daily dose tablet, and literature and media reports, the exact reason for an increase in reports of haemolytic anaemia during 1984 and 1985 was never established nor were the reasons for the timespan of around 9 years from the first introduction of nomifensine to the emergence of a drug safety warning signal that could reasonably be acted upon.

It is possible to compare side effect evaluation between 1976–86 and subsequent years. The current system of evaluation with its heightened awareness amongst healthcare professionals (and indeed society at large) of drug safety risks of marketed products has, at least in part, been the result of the lessons learnt first-hand from problems with former products. These include nomifensine.

The evaluation of nomifensine relied wholly on spontaneous reporting systems with their known inad-equacies of incompletely reported data, lack of popu-lation data to allow for the calculation of incidence rates and estimates of subgroups at risk; poor inter-national co-ordination of drug safety databases; and the need for confidentiality hampering collaboration between the manufacturer and the regulatory authori-ties at least in the early stages.

Nevertheless, the risk–benefit appraisal of nomifen-sine was made through a continuing dialogue between the company and the regulatory authority, taking into account time-honoured but rudimentary indicators of risk and benefit. For the company, these included the general properties of nomifensine in relation to older and newer antidepressants, overdose data, market uptake of the single daily dose, crude adverse drug reaction incidence calculations from prescriptions and sales volume, publications in the medical literature and media reports and comparisons with other drug classes. Specific aspects of nomifensine that were of special concern included the rising incidence of reports of acute immune haemolytic anaemia and the incidence of fatalities.

Of some interest today is what might have been the true effect of a consideration of overdose data on the risk–benefit appraisal of nomifensine, had the successful appeal against the threatened suspen-sion of mianserin using such data been heard earlier (Brahams, 1990). Concern was expressed with mianserin over the number of reports to the CSM of granulocytopenia and agranulocytosis occurring during treatment with this antidepressant, and it was at risk of being suspended. However, it was given a reprieve because of a comparative Prescription Event Monitoring study that was unable to detect any drug-attributable blood dyscrasias and concluded that if mianserin did cause them, then the incidence would probably be in the range of one per 10,000 to one per 100,000 patients. It was also shown that the risks of overdose of mianserin were considerably less than that of amitriptyline (Inman, 1988, 1991).

Between 1977 and 1984, 74 patients taking an overdose of nomifensine, 28 of them nomifensine alone, were reported to the London Centre of the National Poisons Information Service, Guys Hospital (Ali and Crome, 1984). The most common symp-tom, either with nomifensine alone or in combinations with other drugs (benzodiazepines, alcohol and/or tricyclic antidepressants), was drowsiness. There were no reports of convulsions or cardiac arrhythmias in those who took nomifensine alone, and all cases made satisfactory recoveries. It was concluded that nomifen-sine overdose had few clinical sequelae and that there was a notable absence of the complications seen with tricyclic antidepressants.

The nomifensine appraisal might have benefited in a small way, too, from today’s pharmacoepidemiolog-ical databases and case–control studies. These would have added strength at an earlier stage to incidence calculations and allowed the incidence to be compared with the background incidence in the community. However, even today, there is no rare disease registry that provides the background incidence of haemolytic anaemia in the general population.

Since the mid-1980s, the computerisation of data in the international pharmaceutical industry and the regulatory agencies has greatly facilitated the estab-lishment of drug safety databases and the speed and extent of international reporting, accrual and comparison of pharmacovigilance data. Pharmacoepi-demiological databases, such as the Prescription Event Monitoring of the Drug Safety Research Unit (DSRU), the General Practice Research Database (GPRD), the Medicines Evaluation and Monitoring Organisation (MEMO) and Record Linkage, are now available to study, contemporaneously and retrospec-tively, the cause-and-effect relationship of apparently drug-linked events (Mann, 2001).

There has been a concomitant increase in regulation and legislation concerning the formal recording and reporting of suspected adverse events. The application of Good Clinical Practices (GCP), through the Inter-national Conference on Harmonisation (ICH), today formalises all aspects of clinical trials of medicines both before and after licensing. It remains hypothet-ical, however, whether these would have aided the assessment of nomifensine between 1977 and 1986.

A further area of development has been the increased awareness within companies of the need to develop issues management strategies and teams to co-ordinate the response to matters such as specific drug safety alerts. These bring together all the rele-vant company resources from medical, regulatory, manufacturing, quality assurance, legal and commer-cial departments at a local and an international level to address matters raised by, e.g., the increased report-ing of a rare side effect. This enables a much more co-operative and proactive relationship to develop between the company, the regulatory authorities and the media to resolve the issues in a timely and dili-gent manner. Whilst such an approach was taken in the case of nomifensine, it was perhaps more reactive than might be the case today. It remains speculative whether a more formal and rehearsed international issue and relationship management strategy would have helped to shorten the timescale from first alert to the final withdrawal of the drug.

Nomifensine was associated with a rising incidence of a serious life-threatening type B reaction, namely acute immune haemolytic anaemia. The reasons for the rising incidence are not known, although greater doctor recognition and willingness to report, possibly stimulated by literature reports and the media, were undoubtedly factors involved. The immunology was uncertain throughout because of the variety of case presentations, severity and outcomes and conflicting laboratory findings.

Because of difficulties in predicting the haemolytic reaction, distinguishing its initial symptoms from those of other disorders and the variable serological findings, it was impossible to offer firm advice on early diagnosis and treatment.

The drug was withdrawn from sale in the inter-ests of patient safety, even though nomifensine was a well-established antidepressant in many countries, in some of which the problems were thought to be an ‘acceptable’ risk when seen in relation to the drug’s benefits. The decision to withdraw nomifensine was made by physicians employed by the company when, despite the uncertainty, the severity and clinical seque-lae of the haemolytic reaction were fully appreciated. It is arguable whether the science of pharmacoepi-demiology or the procedures of pharmacovigilance as practised today would have impacted on that decision either in January 1986 or indeed in 2001.

Whilst the professionals who make judgements about risk and benefit of a medicine must be aware of both population statistics and individual patient concerns, the decisions on action to ensure the contin-ued safety for some patients without denying the benefits of an established medicine for others will always be demanding. In this context, the decisions of a company to withdraw its product from the market or of a regulatory authority to revoke the marketing authorisation will remain the most difficult of all.