Statistics from Altmetric.com
We read with interest the excellent editorial from Goadsby “A triptan too far?”.1 As he points out, both neurologists and general practitioners are now faced daily with the problem of which triptan to choose. However, we think that it is difficult to choose between the different triptans and their dosages from the figures therapeutic gain and therapeutic penalty. The efficacy for 5 mg zolmitriptan, for example, is somewhat higher than for 100 mg sumatriptan, but this is acquired at the cost of more adverse events. To make a choice between the different options, it might be appropriate to relate the increased therapeutic gain with the increase in therapeutic penalty.
It is therefore useful to compare the ratio therapeutic gain/therapeutic penalty (rTG/TP) for the different triptans and their dosages. This direct comparison is allowed as the compounds all belong to the family of the triptans and therefore the characteristics of the adverse events are expected to be comparable.
As Goadsby suggests, the principles he outlines could be applied to the newer triptans when they become available. We propose the addition of a figure in which the therapeutic gain on the x axis is plotted against the therapeutic penalty on the y axis (figure). Here as well, it is possible to indicate a reference for the rTG/TP. Compounds which are situated below the reference line have a relatively favourable rTG/TP, whereas for those situated above the reference line, rTG/TIP is relatively unfavourable.
We have found a difference in the data mentioned in the text and the ones shown in figure 3.1 For sumatriptan 100 mg the therapeutic penalty is 17% (95% Cl 10%-24%) according to the text, but in figure 3 the penalty shown is about 8% (95% Cl 4%-12%). Because sumatriptan 100 mg is used as a reference, this difference is even more important. In our figure we have used the sumatriptan 100 mg data as mentioned in the text.
I read Goadsby’s opinion about “therapeutic gain” as the difference between the response to an active compound and placebo with great interest.1-1 To my knowledge the placebo response in a controlled trial is to assess the characteristics of the group of patients participating in the study and, with a confidence of 95%, to ascertain the therapeutic effect of a treatment.
In the editorial: A triptan too far?questions arise about the way the author tries to compare different drugs in different studies taking the different placebo responses also into consideration. In my opinion it is inappropriate to use the placebo response as a qualitative measure and using it as a variable to compare drugs from different studies with different placebo responses. If there are different percentages of placebo response, the only conclusion can be drawn that the data from different studies cannot be compared.
Moreover, using the range of the SD of one drug to judge the other also seems to me a methodological failure.
Goadsby’s editorial2-1 says many sensible things, and considers the important and difficult question of how to meaningfully compare the outcomes of different treatments for migraine studied in separate clinical trials. However, in promoting the notion of “therapeutic gain” for between study comparisons he has not fully heeded the caveats he suggests for such an approach. I am aware of no compelling data supporting his statements that “the benefit of the active drug is added to that of placebo” or that “the therapeutic gain corrects for the effect of placebo.”
Consider the exceptional drug, megatriptan, which is effective in almost all patients. In one study, in which the investigators are very excited about the medication’s prospects and are also excellent caregivers, the placebo response is 45%, whereas the megatriptan response rate is 95%. In another study, performed at study centres where patients are mechanically processed with little enthusiasm, the placebo response is only 15%, but the megatriptan response remains at 95%. In the second study the therapeutic gain is 80%, whereas in the first it is only 50%. Surely these two figures for therapeutic gain tell us more about the variability of the placebo response than about the response rates of megatriptan.
In irritable bowel syndrome, which has many similarities to migraine, the placebo response rate in clinical trials varies from 0% to over 70%, due to differences in study design, patient and investigator expectations, treatment setting, patient population, and other factors.2-2 Subtracting out the placebo response in such studies does not necessarily help elucidate the true benefit of the active therapy. Indeed, in migraine there are similar examples. In zolmitriptan trials of the 5 mg dose, the 2 hour placebo rate has varied over threefold, from 15%2-3 to around 50% (Diener HC et al, IIIrd European Headache Federation Meeting, Sardinia, 1996). However, in these and other studies the zolmitriptan response rate remained remarkably constant, between 62%–67%.2-4 2-5 Rather than concluding that the therapeutic gain of zolmitriptan is variable, another interpretation would be that there is a ceiling effect for the drug of about 65%, and the placebo response is elastic, depending on various factors. This is not to say that the placebo response should be ignored in interpreting outcomes, but that it must not be mechanically “taken away” from the active treatment response in evaluating either absolute or comparative efficacy. Surely both the active treatment and placebo response rates must be considered as well as the differences between them in fully evaluating migraine studies. And because of differences in the design and execution of such studies, it remains hazardous to draw comparative conclusions from independent studies, as Goadsby states.
I also suggest caution in the use of the concepts of “therapeutic penalty,” and the “number needed to harm,” as discussed by Goadsby. Such approaches treat all adverse events as equally important. However, the frequent occurrence of relatively trivial events such as mild dry mouth is of far less consequence than the infrequent occurrence of, say, myocardial ischaemia. It would not be appropriate to give them the same weight, as is done with both of these approaches.
For truly meaningful comparisons of both efficacy and safety, head to head comparative trials remain the definitive approach. Hopefully, such trials will be forthcoming to allow for truly valid comparisons among the triptans.
I thank van der Kuy and Lohman for their interest in the editorial3-1 and agree that treatment choice is difficult. We have all struggled with a construct that would absorb both the benefit and payment for headache relief and considered a gain/penalty ratio as suggested, being mindful of the issues of measurement that surround the end points currently employed in clinical studies. I have tried to do something similar for number needed to treat and found, as the correspondents show, that it is not very sensitive, and there are issues with the confidence intervals. This is, in part, an indictment of the problems with current end points as they are translated into clinical practice. Perhaps we will only be able to sort these compounds out when more direct head to head comparisons are made. In the meantime the trial data must be mixed with clinical experience. My experience has been that whereas the differences between the triptans are not huge there are differences that matter in practice, and fitting the drug to the clinical problem can be very rewarding in terms of satisfying a patient’s unmet therapeutic needs. Since the editorial a further compound has been approved in Europe, rizatriptan at a dose of 10 mg orally, bringing to four (including sumatriptan, naratriptan, and zolmitriptan), the choices we have for non-ergotamine based specific antimigraine treatments. The data3-2 indicate that rizatriptan will also have an important place in the treatment of acute migraine attacks.
In regard to the numbers that have been used. Unfortunately, the summary data, indeed the field, represent a moving target. The data quoted in the text were from a single large study,3-3whereas those used in the figure were from a meta-analysis that included many more patients and perhaps is likely to be a more accurate representation of the differences (Tfelt-Hansen, personal communication). The problem is that the meta-analysis covers studies that are older and the newer study was more or less contemporaneous with the newer drug developments. One could argue for either data set and we all await with interest the new comparative study between sumatriptan and zolmitriptan that has been completed but not yet formally reported.
In regard to Meijer’s comments, I would agree that the placebo provides an assessment of various things including the therapeutic effect, which I am not sure is different from the therapeutic gain, given that in some settings the treatment may be worse than the natural history of the disorder. In any event the therapeutic gain is merely the reciprocal of the number needed to treat analysis which is a well developed concept.3-4 Certainly in the editorial the point was made that direct comparative studies are lacking; however, by standard criteria the triptan developments are reasonably suited to this methodology.3-5
Klein makes some very interesting points from the perspective of an ex-member of the pharmaceutical industry team that developed zolmitriptan.3-6 He starts by identifying the caveats in the editorial and suggests that I ignored them. I simply presented one way of looking at the trial data having identified potential weaknesses, I am not sure that this equates to ignoring those caveats. Others have found that taking account of placebo responses is useful across a range of clinical disciplines, including neuropathic pain,3-7epilepsy,3-8 and myocardial infarction,3-9although not without criticism.
The example of the zolmitriptan development programme, of which Klein was directly involved, is curious. The very large placebo response quoted, in a study that has never been published (Diener, H-C et al, 3rd European Headache Federation Meeting Sardinia, 1996), may have resulted from a very skewed randomisation in favour of active treatments, so that there were very few patients entered in the placebo arm which was then a noisy data point with wide confidence intervals that contributed little to the meta-analysis. This outcome really makes the point for me that such overall analyses can be very useful when individual studies are problematic to interpret. In regard to making hazardous conclusions, I would suggest that ignoring the placebo effects offers a tool by which efficacy outcomes can be inflated for rather obvious purposes somewhat more hazardous than those employed in the editorial. Concerning side effects, similar comments may be applied. Calculating therapeutic penalty is nothing more than making an overall tolerability index. It does not consider safety and was never suggested by myself to so do. No clinician would weigh myocardial infarction equal with dry mouth and to suggest that this has been done is simply incorrect. Safety is a completely different issue and is considered by practitioners on another level. If there is any confusion about safety and tolerability it is in the mind of industry, not on the clinical side. I would agree that more head to head comparisons are necessary and stated this before commenting on the available triptans.
In addition, I would like to correct the naratriptan logDpH7.4 which is −0.2 not −0.02, a typographical error pointed out to me by an alert neurological colleague (Professor Allan Purdy, Halifax, Canada) to whom I am grateful. Triptans seem to have come of age in that we can have a mature, data driven debate about their use and misuse in clinical practice. Headache as a subspeciality is better for this, as is neurology in general. Patient care is well served by such discussions as we seek to do the best job for this very common clinical problem.