oa Editorial [Hot Topic: When Can a Drug be Declared “Safe”? (Guest Editors: Frank M.C. Besag & Seetal Dodd)]

Thalidomide is the most notorious example of a drug with serious adverse effects [1] that was initially declared to be safe [2]. This disastrous episode had a major influence on the stringency of pharmaceutical regulation and had a similarly great effect on what could be advertised legally. However, the same stringency does not seem to apply to scientific journals reporting on the safety of drugs. Investigators are allowed to state, in reputable peer-reviewed journals, that a drug is “safe” when they do not have the justification for doing so. Some examples follow. These are from the field of antiepileptic medication but similar examples could be drawn from a number of other therapeutic areas. It must be emphasised that the examples given are from publications in very reputable journals by highly-respected colleagues; the quality of the work is not being criticised but we are advocating a change in terminology and attitude. Felbamate is a broad-spectrum antiepileptic drug that was granted marketing approval in the USA in 1993. The initial publications stated, in good faith, that the drug was safe [3]. The problem is that the statement on safety was based on a trial of only 56 patients treated over a few weeks. Subsequently the potentially life-threatening adverse effects of liver toxicity and aplastic anaemia emerged [4, 5], with the result that the drug is now infrequently prescribed. Vigabatrin was said to be “a safe drug for the treatment of partial seizures” [6]. In 1997, the same year that the first report of serious visual field defects with vigabatrin was published [7], another report stated: “vigabatrin has been shown through several studies to be safe and effective” [8]. Prescribing fell sharply after this adverse effect was reported [9]. Buccal midazolam is now the standard emergency treatment for seizures in paediatric practice in the UK. A very valuable and informative study published in the Lancet, one of the most reputable international medical journals, showed that buccal midazolam was statistically significantly superior to rectal diazepam in the emergency room treatment of status epilepticus in children [10]. This was a landmark paper but why did the authors also conclude, on the basis of only 109 children, that this treatment was safe? It must be emphasised that there is no evidence to contradict this statement but there is also inadequate evidence to support it. If a drug is safe in about 100 patients the mortality rate might still prove to be unacceptably high when information from larger numbers of patients becomes available. It is reassuring to know that no such data of concern have emerged with regard to buccal midazolam but it could be argued that the original statement claiming safety for this treatment was, nevertheless, unjustified. Physicians and patients need to know how safe a drug is but the evidence seems to indicate that we might never be able to declare that a drug is truly “safe”. Infrequent serious adverse effects, including mortality, might not become evident until the drug has been used for several years. We strongly advocate a change in both terminology and attitude. Arguably, to refer to a drug as “safe” is the reassuring terminology of marketing, not of science. Scientific method requires observation and “a safe drug” is not a variable that can be observed. We have many important pre-clinical experiments that look for specific toxicities. In clinical studies we can compare reported adverse events in drug-treated and placebo-treated participants. However, we have no scientific experiment that can prove that a drug is safe. With post-marketing pharmacovigilance, generalisations are made about the tolerability and adverse-event profile of a drug based on a large number of exposures. Conclusions can be proposed about the likelihood of adverse events. However, these conclusions remain probabilities, which may be very unlikely for some drugs, but are never certainties. Journals should refuse to publish papers that claim to have established the safety of a drug on the basis of treatment of a few hundred patients or less. Instead they should insist that authors make more accurate statements. The term “safe” may not be appropriate to describe a drug in scientific writing.“Well tolerated” is more appropriate terminology as it refers to an observation rather than suggesting an inherent characteristic of the drug. For example, instead of “This drug appeared to be safe in the treatment of epilepsy in the 550 patients included in clinical trials” authors should write “This drug was well tolerated in the treatment of epilepsy in the 550 patients included in clinical trials”. If the number of patient years of treatment is large, typically over 100,000, and if this experience is spread over several years, it might be justifiable to state that extensive experience has revealed no serious adverse effects, implying that these are very unlikely to occur. Pharmaceutical companies should also state clearly what the percentage risks of serious adverse effects are, so that the relative safety can be assessed. For example, if an anti-cancer drug increases short-term mortality from 5% to 10% but decreases longterm mortality from 80% to 30%, the patient and physician might still want to give this drug serious consideration. Clinicians prescribing medication have a duty to their patients to provide accurate information so that informed decisions can be made with regard to be risk and benefit. However, they cannot provide this information if they do not have it. The most reliable source of information is generally considered to be reputable peer-reviewed journals. We recommend that journals should not accept unqualified statements about the safety of a drug but should insist that the limitations of the data, particularly with regard to the number of patients treated and the duration of the treatment, should always be specified.