Portal de Periódicos da CAPES

Drug regulation & therapeutic efficacy

2008; Wiley; Volume: 65; Issue: 6 Linguagem: Inglês

10.1111/j.1365-2125.2008.03208.x

1365-2125

James M. Ritter,

Pharmaceutical studies and practices

Modern medicines regulation has a venerable ancestry [1]. Surprisingly, evidence of efficacy is a more recent requirement of drug regulators (e.g. FDA, EMEA) than evidence related to toxicity and safety. (Even today, labels on some 'traditional medicines' can list uses to which the medicine has traditionally been put without evidence of efficacy.) The explanation of this seemingly 'cart before the horse' situation is historical. Drug legislation in the UK and USA began in the mid-nineteenth century with restrictions on the sale of poisons and later of addictive substances. The British Pharmacopoeia of 1864 contained few pharmacologically active preparations but these included some drugs of considerable potential toxicity. Chloroform was introduced as an anesthetic in England in 1847 and anesthetic fatalities followed in short order: well over 100 by 1864 [2]. Oliver Wendell-Holmes wrote in 1860 that '. . . if the whole materia medica, as now used, could be sunk to the bottom of the sea, it would be all the better for mankind – and the worse for the fishes.' By the end of the nineteenth century the pharmaceutical industry took off, as a result of a convergence of scientific medicine with organic chemistry and with the chemical dyestuffs and medical supplies industries, but drug regulation of medicinal products proceeded at a leisurely pace [3]. Its initial focus was on direct toxic effects, and regulatory legislation remained rudimentary both in the USA and in Europe during the first part of the twentieth century. This complacency was shattered by two public health disasters: Elixir Sulfanilamide, and thalidomide. These led to the introduction of major regulatory controls throughout much of the world. Elixir Sulfanilamide was formulated with diethylene glycol, causing 107 deaths in the USA (mainly of children). This led to Congress passing the 1938 Federal Food, Drugs and Cosmetics Act. Thalidomide was introduced in Germany in 1957 and in the UK in 1958. There were early reports of peripheral neuropathy but (as mentioned in a recent Editor's View [4]) it was vigorously promoted as a safe alternative hypnotic to barbiturates, including as a sedative during pregnancy and subsequently also to control nausea during the first trimester. The USA avoided the resulting pandemic of phocomelia because of the FDA's reservations regarding neurotoxicity [2]. Legislation stemmed from the tragedy in many countries including the USA (the Kefauver-Harris Amendment, 1962), the UK (The Medicines Act 1968), European Union, Australia and New Zealand. For the first time, regulatory authorities were empowered to monitor all phases of drug development including efficacy as well as safety studies. The main way of establishing clinical efficacy is by clinical trial against a comparator. Where no effective treatment is available the comparator is usually placebo, but where effective treatment is available then the comparator is usually an active drug (the 'current standard of care' drug). The hope must be that a new treatment (A) is better than existing treatment (B), so common sense dictates that the hypothesis under test should be that one of the treatments differs significantly from the other as to clinical outcome (i.e. that the statisticians' null hypothesis is rejected). But common sense is not always the same as commercial sense, and many large trials are now powered to detect not 'superiority' of A over B but 'equivalence' of A to B, a question that has been addressed with smaller (and hence shorter and much less expensive) clinical trials than would be needed to detect realistic and clinically meaningful differences. If A is truly equivalent to B, and B is of proven value, then regulatory bodies may accept this as evidence of effectiveness of A leading to the grant of a license, aggressive marketing and profit for the shareholders even if there is no benefit to patients who are exposed to increased risks of idiosyncratic adverse drug reactions inherent in treatment with a novel molecular entity. In the present issue a distinguished group from the Mario Negri Institute (Milan) has assessed the reliability of such equivalence trials [5]. They set up a simulated 'trial' to test the absurd hypothesis that placebo is equivalent to active treatment with thrombolysis in myocardial infarction – an impeccably secure instance of clinical effectiveness. They used the GISSI-1 control group and found that with random sampling to construct three subsets, death rates in patients treated with placebo were similar enough to those treated with active thrombolysis to satisfy a conclusion of equivalence in one of the subsets. Two thirds of 100 replications of such sampling gave the same result. They conclude that equivalence trials as currently designed and powered are scientifically unreliable and should not be accepted by drug regulatory authorities. One might further ask how research ethics committees (hopefully supported by professional statistician members) can approve such studies when these are commercially rather than scientifically or ethically driven. The term 'efficacy' is used by pharmacologists attempting to understand the consequences of agonist versus antagonist binding to receptors of endogenous mediators. Beyond this molecular conundrum pharmacologists describe biochemical and physiological effects (e.g. on serum lipids, airways conductance etc.) in terms of dose response relationships – for an example in this issue see the study of rac-formoterol in patients with COPD by Whale et al.[6]. Modern quantitative imaging techniques permit in vivo receptor occupancy data to be gathered in parallel with dose response findings as in the elegant PET study of Tashiro et al.[7]. Surrogate markers such as blood pressure or bone mineral density are of great importance in understanding mechanisms as well as in early (Phase I, II) drug development. Clinical efficacy is defined by epidemiologists as the effect as determined in a randomized controlled trial (phase III) with clinical endpoints (such as strokes, fractures, deaths etc). How is effectiveness best expressed and quantified? This depends on the use to which the information is to be put. Relative risk reduction (RRR) informs readers reliably about comparative drug efficacy (i.e. efficacy compared to another drug or placebo). For example the effectiveness of aspirin in reducing arterial thrombosis, which is approximately 25%–30% risk reduction across a wide range of age and underlying risk. RRR is however much less helpful to the clinician faced with advising a patient, where the risk of harm versus likelihood of benefit varies very markedly with the underlying population risk. Number needed to treat (NNT) has considerable attractions as a point estimate in such circumstances, both for clinicians and for many patients, and has proved extremely popular with clinicians. NNT does however have its own limitations. It can be infinite (where the drug has no beneficial effect) or negative (when it causes harm rather than benefit: the number needed to harm rather than to benefit 1 patient), but it cannot be zero (1 patient is the minimum number need to treat to prevent one event). This causes problems when a confidence interval rather that a point estimate is required, especially when a treatment has not had a significant effect. In this situation, two part confidence intervals are needed: one from a positive value to +∞, the other from a negative value to −∞[8], a clumsy and unsatisfactory state of affairs. In the present issue Salmi, Suissa, Chêne and Salamon review various different measures of clinical effect [9]. They champion the 'clinical result ratio', a measure described by Babbs [10] and which they re-christen the 'attained effect' (AE). They argue that this is especially useful when the background risk in the control group is small resulting in a ceiling to demonstrable effect. Attained effect gives insight into the proportion of effectiveness that remains as a target for future interventions, and, they argue, can usefully complement the NNT as a clinically informative measure of effect. We hope that it will facilitate future clinical trials of superior (not 'equivalent') new drugs for areas of currently unmet clinical need.