Abstract
This issue of Cephalalgia contains what is likely to be one of the last reports of a clinical trial of telcagepant (1). Telcagepant was one of the first small-molecule calcitonin gene-related peptide (CGRP) receptor antagonists to make it to Phase III trials. Several drugs in this class showed promise in treating individual attacks of migraine, leading to high hopes for a new era in migraine pharmacotherapy. In the case of telcagepant, these hopes were dashed when, in 2011, development of this drug for clinical use was halted after a small number of participants developed elevated liver function tests in trials testing the drug for migraine prophylaxis. In those studies, individuals were exposed to the drug on a daily rather than an intermittent basis, suggesting that frequent use of the drug could produce acute hepatocellular toxicity.
There is still much to be learned from the many clinical trials that were conducted with this drug. The current study by Ho et al. made sensible use of the data from the previous prophylaxis trials, tailoring treatment to the perimenstrual period and limiting the dose to the lowest possible still expected to have a clinical effect, and only for seven consecutive days. This is, in effect, two studies: a large safety study with a nested efficacy study in a subset of the trial population. The target population was an admixture of women with menstrually related migraine (MRM) and pure menstrual migraine (PMM), as defined by International Classification of Headache Disorders, second edition (ICHD-2) appendix criteria. The primary treatment outcome of mean monthly headache days was not statistically significantly different between the telcagepant and placebo groups, so the authors appropriately caution readers that results for the secondary outcomes should be viewed as exploratory at best.
This study provides information on a number of important matters. Of foremost interest is the additional data about the risk of hepatocellular injury with telcagepant. Drug-induced liver injury (DILI) may be viewed as an acceptable risk in relation to the benefits of a drug, particularly for those who treat potentially life-threatening illnesses. In the case of migraine, however, which is a highly disabling but not life-threatening illness, there is very little tolerance for even transient laboratory abnormalities due to medical therapy. Ho et al. note that we cannot yet say whether DILI is a mechanism-related effect or one specific to the chemistry of telcagepant, because the pattern of liver injury has not been entirely consistent. In this study, three patients receiving telcagepant but none receiving placebo had ≥eight-fold elevations in alanine aminotransferase (ALT) levels. We are reminded, however, that the development of another molecule—MK 3207—was stopped even sooner than that of telcagepant when serious but delayed hepatic abnormalities were observed, some of which required immunosuppressive treatment (2).
Some preclinical evidence supports the view that liver injury may be a predictable class effect of the CGRP inhibitors. Kroeger et al., for example, reported that αCGRP inhibited inflammatory hepatic injury in a mouse in vitro system, suggesting that telcagepant’s antagonist effects potentiate hepatic inflammation in humans (3). A summary of convergent lines of research in the same paper concludes that CGRP-containing, capsaicin-sensitive primary afferents (in hepatic tissue) “appear to be required for induction of immune-mediated liver injury,” suggesting that altering CGRP transmission from these hepatic sensory neurons could modulate drug toxicity. In addition, CGRP is a negative immunomodulator, particularly of the innate immune system (4), further suggesting that inhibiting CGRP signaling could potentiate immune-mediated responses. CGRP-containing neurons are not limited to hepatic tissue—CGRP is found in C-fibers and Aδ fibers throughout the enteric nervous system (5). Considered as a whole, such studies suggest that DILI is likely to occur with most or all CGRP antagonists. More recently, other studies appear to confirm a role for CGRP in modulating many organs’ responses to injury, such as playing a key role in peripheral sensitization during experimental osteoarthritis (6). It remains to be seen whether these problems occur in trials testing antibodies to CGRP or its receptor, but in the two small Phase II trials already published, there is no indication of hepatotoxicity (7,8).
The efficacy subgroup analysis is an interesting feature of this trial. The investigators enrolled and treated a large number of individuals, then selected those having a minimum number of attacks in the two months prior to the study period to test efficacy. This effectively enriches the study population in the service of detecting a signal of treatment benefit while at the same time enlarging the pool of treated patients for detection of adverse events. While this might appear to be an ideal model for Phase III testing, in the current study this approach meant that only about a third of randomized patients contributed data relevant to the primary outcome, potentially subverting the benefits of randomization.
The small proportion of participants in the efficacy analysis in the Ho et al. study also highlights important matters regarding recruitment for and reporting of clinical trials of migraine in general: Most patients recruited to this trial inaccurately reported the degree of MRM they experience, thus reducing the power and internal validity of the study by decreasing the testable population; the primary efficacy outcome was not met, apparently because it was too optimistic. In addition, and contrary to the accepted dogma that menstrual migraine occurs without aura, there is a suggestion in the current study that aura is not rare in migraine attacks that occur with menstruation.
The authors of this study are to be congratulated for recognizing and reporting duplicate enrollments: Several patients enrolled in this trial at more than one site. This problem was initially identified when one participant, enrolled in the placebo arm, turned in active drug remainders from another trial site. A review of the study databases across all trial sites discovered 28 individuals enrolled at more than one site (four of those at more than two sites). Clearly, having a given participant’s data in a study more than once violates the concepts of independence of variables, not to mention the potential for toxicity if that person is randomized to the active drug more than once, let alone at the same time. Almost nothing has been written on this subject in relation to Phase III trials, although there is a small amount of literature regarding “professional patients” in Phase I trials. In treatment trials, which by their nature relate to illness and so invoke more stringent patient privacy rules, detection of such duplication is more difficult. Payment for participation creates some incentive for re-enrollment, and the ready availability of clinical trial information on the Internet makes identification of multi-site studies relatively easy for anyone so motivated. Contemporary trialists should take heed of this potential loophole in our privacy protection efforts and guard against the loss in trial power it creates.
The relationship of menstrual cycles to migraine has been a topic of interest for decades, yet it remains unclear whether there is value in stratifying patients or treatments according to the phase of the menstrual cycle (9). An early diary study of MRM noted that, while migraine attacks were more likely to occur from −2 to +2 days of menstruation, the intensity was only slightly greater (10). When analysis of attack characteristics is focused on an enriched subgroup—those with MRM referred to a tertiary headache center—then attacks linked to menses are more clearly more severe, last longer, and are less responsive to treatment (11,12). However, the current study highlights some of the problems inherent in focusing on MRM: Despite the fact these individuals were recruited on the basis of having self-reported MRM or PMM, fewer than half met criteria for inclusion in the efficacy study when their attack profiles were monitored in relation to menses. This discordance exemplifies a fundamental issue that complicates the study of episodic diseases: Recall is biased according to the most recent, or most severe, event. William Osler noted long ago that patients, like physicians, are “constantly misled by the ease with which our minds fall into the ruts of one or two experiences” (Aequanimitas, 35–36). Recall bias is a basic principle of headache epidemiology, dictating the need for diaries of headache activity to objectively describe disease activity, context, and response (13,14). Fortunately, a recent study indicates a careful diagnostic interview has a high positive predictive value (PPV) compared to diary-based diagnosis of MRM (15).
Most studies of migraine in relation to menses note that only migraine without aura occurs during menstruation, yet the current study gives some evidence that this is not an invariable association. However, “aura” was not objectively defined or validated here—it was recorded by patients as an option on a checklist, suggesting that the rate of misinterpretation of prodromal symptoms as “aura” was high.
On the whole, this study is well reported, and these authors are careful to acknowledge that their own “prespecified strategy to adjust for multiplicity stipulated that the primary hypothesis had to be significant … in order for the secondary hypotheses to be formally tested.” However, the primary outcome was not met, and yet we are told the results of testing secondary hypotheses: “ … based on nominal p values, there was a benefit for telcagepant over placebo when looking at monthly on-drug headache days.” Later in the Discussion section, the authors go further: “By extrapolation, and assuming that the menstrual-relatedness of a headache has no impact on efficacy, this would suggest a reduction of approximately one to two headache days per month with daily dosing, which is consistent with efficacy findings from previous migraine prophylaxis studies.” Not everyone will agree with this emphasis on post-hoc analysis of “on-treatment headache days”: Extrapolation of a nominally significant secondary outcome when the primary target was not met is a departure from standard reporting principles; in addition, there is, in fact, evidence in several studies that attacks occurring during menses are less responsive to treatments used at other times in the cycle (op. cit. 11,12). Recognizing that the primary outcome for this trial was negative, we should be cautious about interpretation of analyses of the remaining outcomes.
The thought-provoking study of Ho et al. offers important reminders regarding the unpredictable nature of drugs and of diseases. This study shows no clinical value in “mini-prophylaxis” of menstrual migraine attacks using scheduled seven-day dosing of telcagepant. It also confirms the liver toxicity signals of previous studies, indicating those signals were not artifactual. The current focus of drug development in anti-CGRP therapy is with anti-ligand antibodies directed against CGRP itself; small-molecule receptor antagonists are fading from the list of promising migraine treatments, but important lessons learned in the development program will, we hope, not soon be forgotten.
Footnotes
Conflict of interest
None declared.