Abstract
Although the migraine clinical trials literature is enormous, we identified only nine published double-blind studies which compare an oral triptan with a non-triptan acute treatment. Of the nine comparative trials that met inclusion criteria for this review, six compared sumatriptan with other drugs, zolmitriptan was studied in two trials and eletriptan in one trial. In seven of the nine studies reviewed herein, differences between active treatments on the primary endpoints were not dramatic. Experience in clinical practice suggests that, for many patients, oral triptans are superior to non-specific acute treatments, creating a discrepancy between clinical trials results and clinical practice experience. Four possible explanations for the disparities between clinical trials and clinical practice are likely: (i) statistically significant differences may not have emerged because the studies lack adequate statistical power; (ii) patients treated with triptans in clinical practice may be relatively more responsive to triptans and relatively less responsive to other agents than those who participate in clinical trials (patient selection); (iii) headache response at 2 h, as measured in clinical trials, may not fully capture the benefits of triptans relative to other therapies, as assessed in clinical practice; (iv) waiting until pain is moderate or severe, as required in clinical trials, may disadvantage triptans relative to comparators.
Introduction
Clinicians who treat migraine must decide how to select, sequence and combine the available acute treatment options (1, 2). Advances in the treatment and management of migraine have been supported by abundant and diverse clinical trials (3). There is a large body of evidence which establishes the efficacy and safety of medicines used to treat migraine in comparison with placebo (1, 2, 4–6). There is also an ever expanding number of studies comparing triptans with each other (4–6). Surprisingly, few studies compare triptans with acute treatments from other therapeutic classes in head-to-head migraine clinical trials. Standard-comparator studies designed to demonstrate that a novel drug (i.e. a triptan) is better than a standard comparator (a widely used anti-migraine medication with proven efficacy) can support the efficacy and clinical use of a novel drug (7).
Clinical practice experience suggests that, for many patients, triptans provide vastly superior efficacy in comparison with non-specific agents. The American Academy of Neurology survey showed that most neurologists consider that triptans are indicated for moderate to severe migraine (78.9%), and just a few felt that triptans should be reserved for those who have tried and failed at least two other prescription medications (12.2%) (8). Clinical trials, based on per protocol primary endpoints, generally do not reflect the favourable experience with triptans in clinical practice. In this article, we review the published controlled trials which directly compare oral triptans with acute treatments from other classes. We discuss several possible explanations for the disparity between clinical trials results and clinical practice experience. We review evidence which supports each of these hypotheses and close with recommendations for the design of future comparative studies.
Methods
To be eligible for this review, studies had to be prospective, randomized, double-blind trials which compared an oral triptan with a pharmacologic agent from another class for the acute treatment of migraine. Studies had to use the criteria of the International Headache Society (IHS) to define migraine (9). We began with the ‘Evidence Report: Self-Administered Drug Treatments for Acute Migraine Headaches’ prepared by Duke University (1). This group conducted computer-based literature searches, manual reviews of relevant journals and textbooks, and consulted with experts in the field.
We updated their review with a MEDLINE search using the term headache (exploded) and a strategy for identifying clinical trials. Additional search strategies included bibliographical searching of the journals Headache and Cephalalgia. Finally, experts in the field were consulted regarding their knowledge of potentially eligible studies. Eligible studies had to be published in peer-reviewed journals prior to or including 2002. Data published in abstract form only, or after 2002, were not included.
We identified nine eligible studies (Tables 1–3) (10–18), three from the Evidence Report (10, 15, 16) and six published after the deadline for the Evidence Report (11–14, 17, 18). The primary endpoints were based on a 4-point categorical pain scale (none, mild, moderate or severe) in eight of nine studies; one study used a visual analogue scale to measure pain (13). For most studies, 2-h headache response (moderate or severe pain is reduced to mild or no pain) was the primary endpoint. Two-hour pain-free response (moderate or severe pain is reduced to no pain) is a widely reported efficacy endpoint. For studies with multiple attacks, we focus on the first attack because this is usually the per protocol primary endpoint, and because summarizing data across multiple attacks in the same subjects is complex.
Double-blind clinical trials comparing triptans with ergotamine compounds
Ref, reference; N, sample size; DD, double-dummy; PG, parallel-group; MC, multicentre; PC, placebo-controlled; CO, cross-over, NR, not reported.
∗Power to detect statistically significant difference to the 0.05 level (2 tail test), regarding the primary endpoint, using the study result to estimate the effects of active treatment. Power was not calculated when results were not statistically significant under the primary endpoint.
Double-blind clinical trials comparing triptans with combination analgesic
Ref, reference; N, sample size; DD, double-dummy; PG, parallel-group; MC, multicentre; PC, placebo-controlled; CO, cross-over; NR, not reported.
∗Power to detect statistically significant difference to the 0.05 level (2-tail test), regarding the primary endpoint, using the study result to estimate the effects of active treatment. Power was not calculated when results were not statistically significant under the primary endpoint.
Results
Of the nine comparative trials that met inclusion criteria for this review, six compared sumatriptan with other drugs. One study compared sumatriptan 100 mg (in dispersible form) with a combination of ergotamine tartrate 2 mg and caffeine 200 mg (ET + C) (10); this parallel group study evaluated three attacks and did not have a placebo arm. In two studies, sumatriptan (100 mg) was compared with a salicylate plus metoclopramide combination. One used aspirin (900 mg) and metoclopramide (10 mg) (A + M) without a placebo comparator (15). Patients treated up to three attacks in this parallel study. Headache response for the first attack was the per protocol primary endpoint. The other study compared lysine acetyl salicylate 1620 mg and metoclopramide 10 mg (LAS + M) with sumatriptan and placebo (16). This was also a parallel group study; patients treated up to two attacks. Two-hour headache response for the first attack was again the per protocol primary endpoint. In two studies, sumatriptan was compared with non-steroidal anti-inflammatory agents (12, 13). Rapid release tolfenamic acid 200 mg and placebo were the comparators in a two-attack parallel group study (13). For the other study, diclofenac-potassium 50 and 100 mg and placebo were the comparators in a four-period crossover study (13). Finally, sumatriptan was compared with the association of isometheptene mucate, dichloralphenazone and acetaminophen (paracetamol) (17). This study differed from the others because only mild headaches were treated. The primary endpoints were mild or no headache at 2 and 4 h and headache recurrence.
One study compared zolmitriptan 2.5 mg with a combination of A + M (900 mg + 10 mg) (18); this parallel group study evaluated three attacks and did not have a placebo arm. A second study aimed to compare oral ketoprofen and placebo using zolmitriptan 2.5 mg as an active comparator (14); the authors treated moderate or severe pain and used a cross-over design to compare two doses of ketoprofen (75 or 150 mg) with placebo (primary analysis) and zolmitriptan 2.5 mg (secondary analysis) on four consecutive attacks. The last study compared eletriptan 40 mg and 80 mg vs ET + C (2 mg + 200 mg) in a placebo-controlled, parallel group, one-attack study (11).
We first examine the triptan vs ertotamine and caffeine studies (n = 2, Table 1), then the triptan vs NSAIDs studies (n = 4, Table 2), and finally the studies comparing a triptan vs combination analgesics (n = 3, Table 3).
Double-blind clinical trials comparing triptans with non-steroidal medications
Ref, reference; N, sample size; DD, double-dummy; PG, parallel-group; MC, multicentre; PC, placebo-controlled; CO, cross-over.
∗Power to detect statistically significant difference to the 0.05 level (2-tail test), regarding the primary endpoint, using the study result to estimate the effects of active treatment. Power was not calculated when results were not statistically significant under the primary endpoint.
Sumatriptan vs ergotamine tartrate + caffeine (10)
For the first of three attacks, headache response at 2 h was 66% for sumatriptan and 48% for ET + C. Statistically significant differences were demonstrated by the authors (P < 0.001) and in the Evidence Report (OR = 2.1). Pain-free response for sumatriptan at 2 h (35%) was again superior to ET + C (13%; OR = 3.6). The pattern of efficacy results were similar for the subsequent two attacks. Differences in the number of patients experiencing adverse events in the sumatriptan (45%) and the ET + C groups (39%) were not significantly different (Table 1).
Eletriptan 40 and 80 mg vs ergotamine tartrate + caffeine (11)
Both eletriptan 40 mg and 80 mg were statistically superior to ET + C in all endpoints of this single-attack, parallel-group, placebo-controlled study. Headache response at 2 h (primary endpoint) was 54% for eletriptan 40 mg, 68% for eletriptan 80 mg, and 33% for ET + C. The headache response at 1 h was, respectively, 29, 39 and 13%; pain-free rates at 2 h were 28, 38 and 10%. The acceptability of eletriptan was 68% for both doses, higher than ET + C (49%). Interestingly, placebo (53%) had higher acceptability than ET + C. The adverse events were comparable in both active groups (Table 1).
Sumatriptan 100 mg vs tolfenamic acid (200 mg) (12)
Tolfenamic acid and sumatriptan produced similar 2-h headache response rates for attack 1 (sumatriptan 79%vs tolfenamic acid 77%, NS) and attack 2 (sumatriptan 64%vs tolfenamic acid 70%, NS). Two-hour pain-free response for attack 1 (sumatriptan 50%vs tolfenamic acid 37%, NS) and attack 2 (sumatriptan 26%vs tolfenamic acid 16%, NS) were also not significantly different, though trends favouring sumatriptan emerged. The proportion of subjects experiencing adverse events also did not differ (Table 2).
Sumatriptan 100 mg vs diclofenac-potassium (50 and 100 mg) and placebo (13)
This four-period crossover study evaluated head pain using a Visual Analogue Scale. Based on this measure at 2 h, all three active medications were superior to placebo, but not significantly different than each other. It is not possible to evaluate headache response or pain-free response using standard definitions given these measurements. However, diclofenac-K developed statistically significant differences vs placebo at 60 min, while sumatriptan differed from placebo at 90 min. Tolerability profiles did not significantly differ (Table 2).
Zolmitriptan 2.5 mg vs ketoprofen (75 and 150 mg) (14)
Two-hour headache response rates for zolmitriptan, ketoprofen 75 and 150 mg (respectively 66.8 vs 62.6 vs 61.6%) were no different. Two-hour pain-free response rates were, respectively, 36.1, 26.6 and 31.3%. The statistics comparing zolmitriptan and ketoprofen were not presented, but a weak trend favouring zolmitriptan seems present. Escape medication for attacks not relieved at 2 h were used by 76% of the subjects in the zolmitriptan group, 79.7% who used ketoprofen 75 mg and 80.2% in the ketoprofen 150 mg group. Finally, 31.7% of the subjects using zolmitriptan were unsatisfied, compared with 33.1% (ketoprofen 75 mg) and 33.2% (ketoprofen 150 mg) (Table 2).
Sumatriptan vs aspirin plus metoclopramide (15)
For the first attack, 2-h headache response rates for sumatriptan 100 mg were 56 vs 45% for aspirin plus metoclopramide (OR = 1.6, P = 0.078); results were not statistically significant for the primary endpoint. For attacks 2 (sumatriptan 58%, A + M 36%, P < 0.001) and 3 (sumatriptan 65%, A + M 34%, p < 0.001), differences were statistically significant for 2-h headache response. Differences were also significant for 2-h pain-free response for attacks 1 and 3. For attack 1, pain-free rates at 2 h were 26% for sumatriptan and 14% for aspirin plus metoclopramide (OR = 2.2, P = 0.016). Pain-free rates at 2 h for attack 2 did not reach statistical significance (sumatriptan 23%, A + M 15%, NS) but did for attack 3 (sumatriptan 34%, A + M 12%, P < 0.001).
Sumatriptan and A + M were also differentiated based on global ratings. Seventy per cent of the sumatriptan group and 46% of the A + M group said that they would take treatment again (P < 0.001). In addition, 66% of sumatriptan treated patients-rated therapy as reasonable, good or excellent vs 45% of the A + M group (P < 0.001).
There were adverse events in 42% of sumatriptan-treated patients and 29% of patients treated with aspirin plus metclopramide. For this comparison, the difference in the proportion of patients with at least one adverse event is significant: −0.13 (−0.23 to −0.034) (1). Differences in severe adverse events were not statistically significant (Table 3).
Sumatriptan vs lysine acetyl salicylate plus metoclopramide (LAS + M) (16)
Sumatriptan 100 mg and LAS + M, were both superior to placebo but not different than each other on the primary endpoint in this study, for 2-h headache response For the first attack, headache response was 53% for sumatriptan and 57% for LAS + M. For the second attack, differences were again not significantly different (sumatriptan 55%vs LAS + M 43%, P = 0.08). Two-hour pain-free rates for sumatriptan were 30 vs 22% for LAS + M for attack 1 (NS) and 33 vs 24% for attack 2 (NS).
In this study, adverse events were experienced by 30% of patients in the sumatriptan group and 18% of patients in the LAS plus metoclopramide group. This difference in proportions of patients reporting adverse events was significant: −0.12 (−0.22 to −0.019) (Table 3).
Sumatriptan (25 mg + 25 mg) vs isometheptene mucate, dichloralphenazone and acetaminophen (17)
This double-dummy, multicentre, parallel-group study differed from other comparative trials in that patients were not instructed to treat after moderate or severe pain. Instead patients treated as soon as possible, sometimes when pain was mild. Both drugs were encapsulated. The dose of sumatriptan was 25 mg, repeated in 2 h if necessary. This small dose was compared with five capsules of isometheptene combination given in 3 h. Primary efficacy measures were the proportion of patients with mild or no pain at 2 h (68.9 vs 63.1%) and at 4 h (80.3 vs 76.9%), and the number of subjects with recurrence of pain at 24 h (10 vs 11 patients). No statistically significant differences were found in the primary as well as in the secondary endpoints, that included mild or no disability at 2 h (68.9 vs 80%) and presence of associated symptoms (Table 3).
Zolmitriptan 2.5 mg vs ASA 900 mg + metoclopramide 10 mg (18)
Headache response at 2 h occurred in three of three attacks in 33.4% for zolmitriptan and in 32.9% Headache response in at least two attacks was 62.6% for zolmitriptan vs 64.7% for A ± M. These differences were not significant. Headache response at 2 h in the first attack was not also statistically significant (60.4 vs 66.5%). Other secondary endpoints, including 2-h headache response after first dose in at least two attacks, effect on associated symptoms, and recurrence, also did not statistically differ. The pain-free rates in all three attacks were 10.7 vs 5.3% (P = 0.009), the overall pain-free rates were 34.6 vs 27.9% (P = 0.007) and the satisfaction with the treatment was higher among those who used zolmitriptan (83 vs 75%, P = 0.03) (Table 3).
Discussion
In seven of the nine studies reviewed herein, differences in active treatments on the primary endpoints were not dramatic (12–18). Only the studies comparing sumatriptan or eletriptan with ET + C, showed the triptan to be unequivocally superior to a comparator drug (10, 11). Sumatriptan 100 mg was more effective than ET + C for both the 2-h headache response and 2-h pain-free response endpoints (10). Further, the benefits of sumatriptan appeared greater using the more rigorous pain-free endpoint than the less rigorous endpoint, headache response. Eletriptan 40 and 80 mg were also clearly more effective than ET + C for all primary and secondary endpoints (11).
Sumatriptan did not differ from other active comparators on primary and secondary endpoints in four studies (12, 13, 15, 16). Zolmitriptan did not differ from the comparator on the primary or most of the secondary endpoints, though superiority was demonstrated on a few secondary endpoints (14, 18).
In the study comparing aspirin plus metoclopramide and sumatriptan, differences were not statistically significant for 2-h headache response for the first attack (15). Statistically significant differences were demonstrated for 2-h headache response for attacks 2 and 3 and for 2-h pain-free response for attacks 1 and 3. For the comparison of LAS plus metoclopramide, treatment effects were not statistically significant, either for headache response or pain-free response at 2 h for attack 1 or 2 (16). Pooling pain-free data from the two studies, differences were statistically significant in favour of sumatriptan for 2-h pain-free response for the first attack (OR = 1.8, 95% CI: 1.0–3.2). In pooling these data, the underlying assumption is that aspirin + M and LAS + M are sufficiently similar so as to justify combining results. In the sumatriptan-tolfenamic acid studies, there were no statistically significant differences for either 2-h headache response or for 2-h pain-free response for either attack, although trends were suggestive based on pain-free data (12). In the diclofenac-K/sumatriptan (14) and in the isometheptene mucate, dicloralphenazone and acetaminophen/sumatriptan (13) studies, active treatment groups did not differ from each other. Nonetheless, the published evidence suggests that sumatriptan is more effective than ET + C. Evidence for superiority of triptans vs salicylates plus metoclopramide is modest at best and lacking for tolfenamic acid, diclofenac-K and isometheptene compounds.
The results of these clinical trials are inconsistent with clinical practice experience among headache-interested physicians, which suggests that, for many patients, triptans provide vastly superior efficacy in comparison with non-specific agents (8). Because the nine comparative studies were double-blind, randomized clinical experiments, these results cannot be easily dismissed. Furthermore, clinical practice experience is difficult to interpret for several reasons. In practice, patients are not randomly allocated to treatments. Expectations about the efficacy of particular treatments, on the part of both clinicians and patients, may arise from open-label treatment, influencing results. Follow-up is incomplete, outcomes are measured haphazardly, and there is no contemporaneous control group. These very limitations provide the rationale for conducting rigorous experiments (clinical trials). The challenge is to understand the discrepancy between clinical trials results and clinical practice experience.
In the sections below, we consider four possible explanations for this gap between clinical trials results and clinical practice experience. These explanations are neither mutually exclusive (more than one may be true) nor exhaustive. They include:
Statistically significant differences may not have emerged because the studies lack adequate statistical power.
Patients treated with triptans in clinical practice may be relatively more responsive to triptans and relatively less responsive to other agents than those who participate in clinical trials (patient selection).
Headache response at 2 h, as measured in clinical trials, may not fully capture the benefits of triptans relative to other therapies, as assessed in clinical practice.
Waiting until pain is moderate or severe, as required in clinical trials, may disadvantage triptans relative to comparators. Early treatment during mild pain may increase the benefits of triptans vs other classes of treatment.
Issues of statistical power
Insufficient statistical power may explain the lack of significant differences between treatments in a clinical trial (19). The demonstration that one treatment is superior to another depends not only upon the magnitude of the differences in treatment effects, but also upon the variability in the outcome measure and the sample size. Sample size requirements depend upon the primary endpoint selected for study as well as the magnitude of the difference in treatment effect one wants to detect. Sample size requirements rapidly increase in studies designed to detect small differences between treatments (i.e. 5%). A finding of no differences may occur if the treatment effects are similar, if the study is under-powered or if the study is otherwise poorly designed. Studies designed to show that treatments are equivalent generally require very large samples; equivalence cannot be inferred based on the absence of differences in a study designed to detect differences. We performed post-hoc power calculations based on the primary endpoint (Tables 1–3) of the studies presented herewith. In making these calculations we assumed that the observed magnitude of differences represents the actual magnitude of differences between active treatments. Under this assumption, the power (at the P = 0.05 level) ranged from 6 to 63%. In other words, the studies were underpowered to detect difference in the main efficacy measures.
For example, in the A + M vs sumatriptan study (15), if one examines headache response and pain-free response at 2 h for each of the three attacks, there are six endpoints to evaluate. Differences were statistically significant for four of those six endpoints. When non-significant, the pattern of data nearly reached statistical significance. In this study, a larger sample size may have resulted in more consistent separation of active treatments. In two other studies, non-significant trends favour sumatriptan on the 2-h pain-free endpoint. These patterns of results may be explained by lack of statistical power.
Does patient selection contribute to the disparity between clinical trials and clinical practice?
In clinical practice, patients are selected for triptan treatment based on a large number of factors which have not been systematically characterized. Patients who receive triptans, have often first tried and failed many non-specific therapies (20). The relative benefits of triptans may be more apparent in patients selected for non-response to other treatments. In neurologic and headache specialty practice, there is a high level of selection for migraine sufferers with substantial levels of disability and refractoriness to treatment. These migraine sufferers are also likely to experience higher levels of temporary disability.
Results from the Disabilities in Strategies of Care (DISC) study support this hypothesis. In DISC, migraine disability was measured with the Migraine Disability Assessment Scale (MIDAS) questionnaire prior to treatment (20, 21). In one arm of the study, patients with MIDAS grades II (mild or infrequent disability), III (moderate disability) and IV (severe disability) treated their first three moderate or severe migraine attacks with A + M. Treatment success was defined as a headache response at 2 h in two of three or in three of three attacks; these patients continued on aspirin plus metoclopramide in the step-care across attack arm of the study. Those who failed A + M (responded in one of three or zero of three attacks) were escalated to zolmitriptan 2.5 mg. Using this definition, aspirin plus metoclopramide failed in 56.4% of MIDAS grade II patients, 68.9% of MIDAS grade III patients, and in 74% of MIDAS grade IV patients (21). The study showed that the probability of treatment failure increased with MIDAS grade (Chi square test for trend, P = 0.03); the efficacy of A + M declined as disability increased. In MIDAS grade III and IV patients, treatment with a triptan (zolmitriptan 2.5 mg) improved patient outcomes. Based on these findings, we suggest that the results of head-to-head comparator trials may vary with the level of disability in the patient groups studied. We also predict that, within limits, the benefits of a triptan over a non-triptan comparator will be more readily demonstrated in a more disabled, treatment-refractory group of migraine sufferers. A caveat is that in a truly treatment refractory group, it may be difficult to demonstrate treatment benefits at all.
Does 2-h headache response capture the efficacy benefits of triptans vs other agents?
The available evidence suggests that 2-h headache response does not provide an adequate standard for comparing active treatments. In population-based surveys, migraine sufferers in the population report that they want rapid onset of pain relief, complete relief of pain (pain-free), lack of headache recurrence and consistent treatment benefits across multiple attacks (22). Re-analysis of clinical trials data supports these points. Pain-free response at 2 h does a better job of predicting satisfaction with treatment as well as health-related quality of life than does 2 h headache response (23, 24).
If triptans outperform other acute treatments on more stringent endpoints, a focus on 2-h headache response might help explain the apparent gap between clinical trials and clinical practice. In the studies reviewed above, measures of 2-h pain-free response better separate triptans from other therapies than measures of 2-h headache response. In the zolmitriptan vs A + M trial (18), zolmitriptan is statistically superior in the pain-free rates at 2 h overall and for all attacks, but did not differ regarding headache response at 2 h. Comparative data on 24-h pain-free response and intraindividual consistency are not generally available from studies comparing triptans with other classes of drugs, though these endpoints could be partially explored by re-analysing existing data. None-the-less, in the zolmitriptan vs A + M trial there were no differences in intraindividual consistency for headache response if two of three or three of three attacks. More rigorous endpoints, including 2-h pain free may better identify differences between triptans and other agents in comparative studies (6, 25).
Does early treatment during mild pain increase the benefits of triptans vs other classes of treatment?
In most migraine trials, patients are instructed not to treat an acute attack until moderate or severe pain develops (1–6). In practice, patients often treat while pain is mild, if they are reasonably sure they are having a migraine. A recent study showed that treatment of migraine with zolmitriptan 2.5 mg while headache is mild provides higher pain-free response rates in patients with significant disability (26). Similar results were found using sumatriptan (27).
The Spectrum Study provided an analysis of the relative benefits of sumatriptan 50 mg tablets and placebo during mild vs moderate or severe pain. In this study, migraine patients treated up to 10 attacks with sumatriptan 50 mg or placebo in one arm of a crossover study (28, 29). Patients were instructed to treat only moderate or severe pain, but 26 of 249 migraine patients treated 46 migraine attacks while pain was mild. These same 26 patients treated 166 attacks of moderate or severe migraine pain. At 2 h post treatment, pain-free rates for early intervention were 50% for sumatriptan-treated attacks and 0 for placebo-treated attacks. For treatment during moderate or severe pain in these same patients, pain-free rates were 27% for sumatriptan and 6% for placebo-treated attacks (29). Sumatriptan produced much higher pain-free rates when taken during mild pain rather than moderate or severe pain, within the same patients.
To further assess the influence of treating mild pain, a post-hoc analysis of a second study was undertaken. Ninety-two patients treated 118 attacks during mild pain, violating the study protocol which required treatment of moderate or severe pain only (30). Pain-free rates at 2 h were 28% for placebo, 51% for sumatriptan 50 mg and 67% for sumatriptan 100 mg (Fig. 1). Pain-free rates were 21% for placebo, 75% for sumatriptan 50 mg and 90% for sumatriptan 100 mg at 4 h. In this study, rescue medications were not permitted until 4 h. This study confirms the benefit of treating during mild pain on pain-free rates.
Pain-free response to sumatriptan 50 or 100 mg vs placebo during the treatment of mild vs moderate to severe pain at 2 and 4 h [based on Cady et al. (30)].
A post-hoc analysis was also performed regarding a subgroup of patients from a large, open-label, long-term clinical trial wherein 762 migraineurs used almotriptan 12.5 mg for headache attacks of any severity. The study evaluated the efficacy and safety of treatment in those patients who had treated at least three ‘mild’ attacks and three ‘moderate/severe’ attacks. At 1 h following treatment, pain-free status was achieved in 47% of mild attacks vs 14% of moderate/severe attacks (P < 0.001); incidences at 2 h were 84% of mild attacks and 53% of moderate/severe attacks (P < 0.001). The chance of achieving pain-free status at 1 h in at least two of three treated attacks was 45% for mild attacks and 9% for moderate/severe attacks; at 2 h the percentages were 88% for mild attacks and 56% for moderate/severe attacks. Rescue medication was required in 8% of mild attacks and in 13% of moderate/severe attacks (P < 0.01 31).
These post-hoc analyses suggest that triptan treatment during mild pain produces very high pain-free rates relative to placebo and relative to treatment during moderate or severe pain. To account for the clinical trials results, benefits of early treatment with sumatriptan must exceed the benefits of early treatment with ET + C or A + M. To assess these benefits, Cady et al. (30) conducted post-hoc re-analysis in two of the five comparative studies reviewed above (10, 11). In the Multinational Study (10, 30), the 2-h pain-free rates for the treatment of moderate or severe pain were 16% for ET + C and 39% for sumatriptan (Fig. 2) in patients who treated at least one attack while pain was mild. For treatment during mild pain, pain-free rates at 2 h increased from 16 to 34% for ET + C and from 39 to 69% for sumatriptan. These differences were statistically significant both for the treatment of mild pain and for the treatment of moderate/severe pain (Fig. 2). Early treatment produced higher pain-free rates for both ET + C and sumatriptan. However, the increment in pain-free response was 18% for ET + C and 30% for sumatriptan (30). Among patients who treated during mild pain in the OSAM study (11, 30), 2-h pain-free rates for treatment of moderate to severe pain were 10% for A + M and 48% for sumatriptan 100 mg (Fig. 3). Pain-free rates for attacks treated during mild pain increased from 10 to 25% for A + M, a gain of 15%, and from 48 to 73% for sumatriptan, a gain of 25%. Differences were statistically significant and in favour of sumatriptan both for the treatment of mild pain as well as the treatment of moderate/severe pain (Fig. 3).
Pain-free response to sumatriptan 100 mg vs ergotamine plus caffeine in the treatment of mild pain and the treatment of moderate/severe pain at 2 h [based on Cady et al. (30)].
Pain-free response to sumatriptan 100 mg vs aspirin 900 mg plus metoclopramide 10 mg in the treatment of moderate/severe pain at 2 h [based on Cady et al. (30)].
For all of the studies and all of the treatments reviewed above, pain-free rates were higher for the treatment of mild pain. The incremental benefits (on an additive scale) appear much greater for sumatriptan than for the comparator compounds. These findings support the idea that use of early treatment in clinical practice may disproportionately increase the benefits of triptans and contribute to differences between clinical experience and comparative clinical trials results.
One clinical trial reviewed here evaluated sumatriptan in the treatment of mild migraine and failed to demonstrate superiority against the comparator (17). However, this study used a small initial dose of sumatriptan (25 mg), a small final dose of sumatriptan (50 mg reached just after 2 h) and compared sumatriptan with a high initial and final dose of the combination analgesic (five pills, being two while pain was mild and four pills until 2 h). Non-equivalent doses were compared and unnecessary complexity was aggregated. Additionally, both drugs (sumatriptan and the comparator) were encapsulated. As the authors suggest, ‘It … is not possible to determine from this study whether a 50- or 100-mg dose of sumatriptan when taken at the first sign of the migraine would have altered the outcome of this trial’.
Experimental studies provide support by which the efficacy of a triptan is enhanced by treating when pain is mild (32). Burstein et al. contend that migraine involves sensitization of both peripheral and central pathways, which are recruited sequentially (33). Central sensitization can occur as early as 1 h after the onset of a migraine. They hypothesize that the triptans ‘… are expected to be the most effective prior to the development of central sensitization …’ which ‘… justifies the common clinical wisdom of recommending that patient use these medications immediately at the onset of the migraine attack’. According to this hypothesis, data from 128 studies conducted with sumatriptan from 1987 trough 1998 were reviewed and the outcomes were calculated as a function of baseline pain severity (34). This study concluded that treatment of mild pain with sumatriptan was associated with significant therapeutic benefit over treatment of moderate or severe pain.
Designing clinical trials comparing triptans and other classes of drugs: recommendations
1. Studies should be powered to detect a clinically meaningful difference in the primary endpoint
Clinical judgement determines the magnitude of difference in treatment effects that makes a difference. It is vitally important to have adequate power to detect clinically meaningful differences in comparative studies. Many studies comparing sumatriptan with non-triptans are small in comparison with the studies that compare triptans with each other (7).
2. Primary endpoints should fully reflect differences that matter to patients and clinicians.
Two-hour headache response may not be an appropriate primary endpoint for comparing active drugs. We recommend multiple attack studies which examine clinically important endpoints. Endpoints to consider include 2-h pain-free response, 24-h pain-free response, intraindividual consistency across multiple attacks, and disability time per treated attack among others (35). In multiple attack studies, information from multiple attacks can be combined by using statistical methods that take within-subject correlation into account to increase power. Preference and satisfaction studies provide a useful way for combining information about efficacy and tolerability.
3. Select patient groups for study that resemble the clinical groups to whom results will be applied.
Disability is an important predictor of treatment effect, at least for aspirin plus metoclopramide. As a consequence, in patients with low MIDAS scores, it may be difficult to separate aspirin-metoclopramide from a triptan. In patients with high levels of disability (MIDAS Grade IV), we hypothesize that it should be easier to identify differences in treatment effects between triptans and non-triptans. Clinical trials comparing non-specific and specific treatment should enroll patients in whom either therapeutic option would be seriously considered, that is, a patient population to whom the results might be applied in practice. Stratified analysis based on disability (measured with MIDAS, for example) might also help clinicians more meaningfully apply clinical trial results to patients in their practices (21).
4. Use dosing strategies that reflect actual patterns of treatment in practice.
Clinical trials should assess treatment during mild pain, not just treatment of moderate or severe pain. Stratified analyses based on pain status (mild, moderate, severe) at the time of treatment may be helpful. It is also desirable to explore the range of clinically relevant doses for the drugs being compared. This will help distinguish differences in efficacy from differences in dose and facilitate and assessment of efficacy-tolerability trade-offs.