Spontaneous Reporting Schemes

SPONTANEOUS REPORTING SCHEMES

There is widespread agreement that spontaneous reporting schemes are here to stay. They are econom-ical and embrace the entire population of patients and reporters. However, it is important to treat all such reports as hypotheses. Some events will almost certainly be causally linked with the suspect drug, whereas others will turn out not to be so. With secu-lar trends in widening the reporter base to include nurses, pharmacists, other health professionals and now patients, the balance may vary somewhat from region to region and from one group of reporters to another. It is important that standardised procedures are adopted to review and analyse all such sponta-neous reports. In so doing, there is a danger that the output from any review could be made available with-out the benefit of careful clinical and pharmacologi-cal expertise and input with serious consequences to all concerned. The rule here is to appreciate that the raw information from spontaneous reporting schemes are anecdotes – no more and no less. They have to be treated as such. Sophisticated analyses of anecdo-tal data are justified if great care is taken with the subsequent interpretation, otherwise more harm could arise than good. Careful review of all reported suspect reactions to a particular medicine may point to a sub-population at especial risk. Such reviews can rarely be automated, but require careful, time-consuming analy-sis by trained, experienced observers. Such people are in short supply; nevertheless, they are extremely valu-able in the context of logical interpretation of spon-taneous reporting schemes. Signal detection can now be enhanced by advances in information technology, including techniques for interrogation and analysis of databases of spontaneous reports.

The development of Augmented Spontaneous Reporting Schemes whereby potential reporters are contacted about details of outcomes after specific medicines have been prescribed in the hope that they will respond in greater numbers and with better qual-ity information is to be encouraged. These schemes are best developed in New Zealand and by the Drug Safety Research Unit in Southampton (United King-dom). The present authors believe that these schemes should be encouraged and developed further in the coming decade. They are not without their problems, however. This is especially so in the United King-dom at present where there is a severe epidemic of concern within the public psyche about confidentiality of medical information. Whilst no one would disagree with the need to maintain confidentiality when dealing with information on illness, and all would support the need for great care in this area, nonetheless there are circumstances in which the need to link information from several different sources is necessary to ensure appropriate interpretation of the data. This is most marked in the case of cancer registry data, but is also a clear feature of many pharmacovigilance issues.

The problem becomes particularly acute when we observe the controversy about patient confidentiality in the United Kingdom at the present time. The regu-latory authority for prescribers (the General Medical Council) has been rigid in its emphasis of the need for total patient confidentiality. Whilst at first sight this seems entirely reasonable and laudable, observa-tional research could be seriously damaged by such an approach: in particular, studies such as cancer registries and drug safety monitoring studies are uniquely vulnerable since both require coordination of disparate data sources (e.g. demographic data, drug prescription data, hospital records and general prac-tice records) to form a relevant patient record. In the absence of adequate anonymous patient registration numbers to bring these records together, identifiable information may be required solely to coordinate such information. If this can only be undertaken by receiv-ing individual patient consent, an unknown propor-tion of patients (possibly up to 30%) will for one reason or another be unable or unwilling to give such permission. Thus the value of the resulting data set is dramatically reduced as it no longer constitutes a random sample from the population. Warlow and colleagues have recently demonstrated this consent bias in observational research (Al-Shahi, Vousden and Warlow, BMJ 2005; 331: 942–5). In a follow-up study of patients with intracranial arterio-venous malforma-tions, outcomes were clearly different between those who did and did not give consent. Moreover, in the case of prospective databases involving literally millions of patient-years of observations, the practi-calities of obtaining patient approval to use identifi-able information solely to permit record linkage with the objective of furthering public health objectives of potential benefit to all people in the land seem at first sight almost insurmountable as well as being prohibitively expensive. Are we then to cease this type of research? Surely the answer to this must be a resounding ‘No’! We must find other more prac-tical ways of achieving the desired end of maintain-ing quality research into drug safety and into cancer surveillance whilst fulfilling the need for confidential-ity for all patients. We would suggest that a reasonable position to adopt would be one in which it was a recognised duty on patients receiving treatment in the National Health Service to accept that their informa-tion would be used for routine monitoring purposes, including disease incidence and prevalence studies and studies into the safety of medicines. Such studies will require records to be linked across several areas, and identifiers may be needed for this purpose. At all times such confidential information would be kept to the minimum necessary and would be used solely for this purpose. Data can be protected by coding and restriction at various levels in the capture-and-research process. Any breach of this confidentiality would be dealt with severely by fines or suspension of a licence to practise. As far as we know, there is no record of any confidential information being placed in the public domain from such data sets. Thus the obsessive concentration on confidentiality to the exclusion of all other facets of this issue is likely to do substantial harm to world-class research if the issue of post hoc anonymisation cannot be adequately and economically addressed.

In summary, anonymised records should be the usual type of information used by pharmacovigi-lators and pharmacoepidemiologists; however, there are times when, for the public good and because anonymised information is not readily available, iden-tifiable data will be required for linkage purposes. With suitable safeguards in place and enforced, the public can be reassured that such records can and should become a part of participation in NHS treatment, including public health monitoring and research.