![[PDF]](/pb-assets/Icons/adobe-pdf-icon-1498649896243.png){kind=icon}
Abstract
Background
Migraine is the most common neurological condition in developed countries. The abortive treatment of migraine attacks is important both for quality of life and costs associated with illness. Triptans, serotonin 5-HT1B/1D receptor agonists, effectively relieve the pain, disability, and associated symptoms of migraine while improving health-related quality of life. Although a number of direct head-to-head triptan comparisons have been made, data for all possible permutations are not available, and unlikely to ever be so, although in clinical practice such information would be useful.
Objective
We aimed to determine the relative efficacy of all available triptans to abort migraine headache among patients with previous adequate response to migraine treatments.
Methods
We included only double-blinded randomized clinical trials comparing triptans to either placebo or another triptan. Our primary outcomes were pain-free response at two hours and 24-hour sustained pain-free response, and our secondary outcomes were headache response at two hours and 24-hour sustained headache response. We used Bayesian multiple treatment comparison meta-analyses of seven triptans used in adult patients to abort migraine attacks. We applied a random-effects analysis with meta-regression adjusting for dose. Results are reported as odds ratios with 95% credible intervals.
Results
We included data from 74 randomized clinical trials. All triptans were significantly superior to placebo for all outcomes, with the exception of naratriptan for 24-hour sustained pain-free response. Eletriptan consistently yielded the highest treatment effect estimates. For the two-hour endpoints, eletriptan was statistically significantly superior to sumatriptan, almotriptan, naratriptan, and frovatriptan for at least one of the two outcomes. Rizatriptan yielded the second highest treatment effects followed by zolmitriptan. For the 24-hour endpoints, eletriptan was statistically significantly superior to sumatriptan, rizatriptan, almotriptan, and naratriptan for at least one of the two outcomes. Frovatriptan data were not available at that endpoint.
Further, the probability that eletriptan is the most likely of all triptans to produce a favorable outcome was 68% for pain-free response at two hours, and 54% for 24-hour sustained pain-free response.
Introduction
Migraine is a chronic neurological disorder affecting about 18% of women and 5% of men in developed countries (1). It is probably the most common neurological condition reported in developed countries, and is very costly to society (2–4). Typical migraine headaches last from four to 72 hours and occur on average about 34 times per year for men and 37 times for women (5,6). Because migraine headaches can often be severe and because migraine is often associated with nausea, vomiting, photophobia, and phonophobia, its impact on quality of life and health-related resource utilization can be large (7,8).
Treatments for migraine aim to prevent attacks or abort migraine once started. Treatments range from complementary therapies to pharmacological interventions, including analgesics, anticonvulsants, calcitonin gene-related peptide receptor blockers, ergotamines, corticosteroids, antiemetics, and triptans (9–11). In the United States (US), triptans, serotonin 5-HT1B/1D receptor agonists, are the most commonly prescribed migraine-specific treatments. The potential mechanism(s) of action of triptans include cranial vasoconstriction and inhibition of peripheral and central trigeminal nerve transmission and altered modulation of nociceptive processing (12,13).
Although seven triptans are available and have been evaluated in randomized clinical trials, the relative efficacy of each triptan is uncertain since relatively few have been evaluated in direct (head-to-head) comparative randomized clinical trials (14). We aimed to determine the relative efficacy of all triptans using a relatively new statistical approach known as a multiple treatment comparison (MTC) meta-analysis, a meta-analysis technique that allows for the integration of direct and indirect comparisons (15,16). MTC meta-analysis is now widely accepted by health technology appraisal agencies worldwide, and are frequently used to inform comparative effectiveness and safety between multiple interventions in the context of decision-making (15,16). MTCs are particularly valuable as they permit inferences into the relative efficacy of treatments that may not have been compared to each other in direct randomized clinical trials, and they allow strengthening of the evidence-base by combining head-to-head with indirect evidence (15,16).
Methods
Trial eligibility
We considered all marketed triptans for our analysis (eletriptan, sumatriptan, rizatriptan, almotriptan, zolmitriptan, naratriptan, and frovatriptan). We included double-blind (patient and provider) randomized clinical trials (both parallel and cross-over designs) investigating the treatment effects of one or more triptans for adults (18–65 years) with episodic migraine with or without aura according to the International Headache Society (IHS) criteria. We included trials in which patients had at least one migraine attack every six weeks. We included both single and multiple migraine attack trials. We included randomized clinical trials in which 1) treatment was given orally in standard encapsulated format; 2) treatment was taken within four to six hours of migraine onset; 3) the patients had a history of moderate to severe migraine attacks; 4) no analgesic or antiemetic medication was given within four to six hours prior to dosing, and no triptan or ergot medication was given within 48 hours.
We excluded trials in which 1) treatment was not given orally or in a nonconventional oral formulation (e.g. wafer); and in which patients 2) suffered from other frequent nonmigrainous headaches; 3) had a history of treatment-resistant migraines (i.e. previous poor response to a triptan); 4) had other clinical significant illnesses; 5) or had standard contraindications to triptans.
Search strategy
We systematically searched the medical literature for relevant randomized clinical trials (from inception to March 2012) using MEDLINE, EMBASE, LILACS and the Cochrane Controlled Trials Register. The online appendix displays our search strategy. We did not exercise any language restrictions. We also searched the websites of the pharmaceutical companies manufacturing the considered triptans to identify any unpublished clinical trials. All identified articles were assessed independently by two reviewers (KT and PW) using predesigned eligibility forms, according to our prospectively defined eligibility criteria.
Outcomes and data extraction
We considered the following outcomes 1) the proportion of patients who had a pain-free response at two hours; 2) the proportion of patients with 24-hour sustained pain-free response; 3) the proportion of patients who had a headache response at two hours; and 4) the proportion of patients with 24-hour sustained headache response. All endpoints were considered with respect to the first migraine attack. Our primary endpoints were comparative effectiveness estimates (odds ratios) between all triptans for pain-free response at two hours, and 24-hour sustained pain-free response (17). Secondary endpoints were comparative effectiveness estimates (odds ratios) between all triptans for headache response at two hours, and 24-hour sustained headache response. The time point of two hours reflects short-term response and 24 hours reflects prevention of recurrence (17). We did not consider endpoints of adverse effects. Although such are important for informing clinical practice, the complexity of MTCs increases exponentially with the addition of endpoints. To retain simplicity of this manuscript, we therefore focused on efficacy endpoints only.
Two reviewers (KT and PW) independently extracted and recorded data in a Microsoft Excel spreadsheet. All data extraction was then checked by a third reviewer (EM). Additionally, for each trial, the following trial characteristics were extracted: countries of origin; number of centers; study design (parallel or cross-over); number of participants; type of migraine; migraine with aura; mean age (and standard deviation); proportion of females; criteria used to diagnose migraine; number with moderate to severe attacks; medication schedule; maximum dosage/day; study duration; and proportion of patients excluded from the analysis (e.g. in modified intention-to-treat analysis). Study quality items included reporting of randomization sequence generation, allocation concealment, reporting of who was blinded, use of intention to treat versus per protocol with justification for approach used, and proportion with >20% loss to follow-up.
Statistical analysis
In order to assess inter-rater reliability on inclusion of articles, we calculated the Phi (Φ) statistic, first developed to provide a measure of inter-observer agreement independent of chance (18).
We first conducted DerSimonian-Laird random-effects pairwise meta-analysis of the direct trials (19,20). We visually inspected the forest plots for possible heterogeneity and calculated I2 values, which provide an estimate of the percentage of variability of effects that is unlikely to be chance (12). The direct meta-analysis outcomes were used to check potential inconsistency with indirect evidence.
MTC meta-analysis
We first plotted the geometric distribution of the network of randomized clinical trial treatment comparisons (21). We used the conventional Bayesian MTC models for binary outcomes to estimate comparative odds ratios and associated 95% credible intervals between all triptans for the considered outcomes (22). All Bayesian MTC meta-analyses were carried out in WinBUGS version 1.4.3 (Cambridge, UK). A more detailed description of the employed statistical model, including key distributional and deterministic relationships as well as the statistical WinBUGS code, is presented in the appendix.
Primary analysis
In keeping with the guidelines for controlled trials of drugs in migraine, third edition (17), for our primary analyses we examined data from patients who had at least one migraine attack, were not lost to follow-up and did not violate the trial protocol. When studies that did not report this number, we used the modified intention-to-treat (ITT) data (i.e. all patients who suffered at least one migraine attack, rather than all patients who were randomized). We analyzed data from the first migraine attack only for crossover or multiple attack studies.
We incorporated the effect of dose in our analysis. As our primary analysis we employed meta-regression to control for the effect of doubling or halving the “common” dosage (23). Our model assumed the same relative change in odds ratios associated with doubling and halving the dose. Accordingly, the dose covariate was constructed to contain the categories “half,” “common,” and “double,” and the “common” was used as the reference category in the regression model. The considered doses for each triptan are presented in Table 1 of the online appendix. For all triptans except for sumatriptan, each “common” dose was the single dose indicated by the Food and Drug Administration (FDA). We chose the “common” dose as the reference dose to facilitate meta-regression on the effect of doubling and halving the dose. As a secondary model we employed MTC meta-analysis without meta-regression on the doses but considered each dose as a different treatment.
|
Table 1. Odds ratios estimates and 95% credible intervals for all treatment comparisons based on the multiple treatment comparison model regressing on dose.
We obtained odds ratios and 95% credible intervals for all comparisons. We also produced treatment rankings in the form of probabilities that each treatment is likely to work best, second best, third best, etc. Technically these probabilities, for each triptan, are the probability that the triptan in question is at highest odds of producing a favorable response, e.g. pain-free response at two hours. As such, treatment probabilities for each treatment were calculated as the proportion of Markov Chain Monte Carlo samples (20,000 in total) where odds ratio was larger than the odds ratio for all other treatments.
Analysis of common indicated doses in the US
We additionally estimated the comparative effectiveness between triptans with respect to the common indicated single dose in the US, which varies slightly from other countries: 100 mg for sumatriptan, and “common dose” for all other doses.
Data consistency check
We examined inconsistency between direct and indirect estimates both visually and using z-tests.
Statistical software
Variable recoding was handled in Microsoft Excel. All statistical analyses were conducted in WinBUGS (MRC Biostatistics Unit, Cambridge, UK) and StatsDirect (StatsDirect Ltd, UK).
Results
We included 67 publications with 74 randomized clinical trials examining triptans for the treatment and prevention of migraine attacks (Figure 1, online appendix Table 9 for included studies and study characteristics). Five publications each included two trials and one publication included three trials. Placebo was compared to eletriptan, sumatriptan, rizatriptan, zolmitriptan, almotriptan, naratriptan, and frovatriptan in 15, 30, 16, 5, 9, 5, and 4 trials, respectively. Fifteen trials included head-to-head comparisons between the seven triptans. Four trials were cross-over designs, of which two trials reported results after the first stage, and two reported results after the second stage. Sixty-three trials reported the outcome of pain-free response at two hours; 25 reported 24-hour sustained pain-free response; 61 reported headache response at two hours; and 16 reported 24-hour sustained headache response. Figure 2 shows the network of available comparisons for the endpoint of pain-free response at two hours, and Figure 3 displays the network for the endpoint of 24-hour sustained pain-free response. Online appendices Figures 1 and 2 display the networks for headache response at two hours, and 24-hour sustained headache response, respectively.
Strength and diversity of treatment networks
Inspection and analysis of direct and indirect estimates did not reveal inconsistency beyond the play of chance for any of the outcomes. Generally, direct and indirect estimates were of similar magnitude and direction. In the few cases in which comparative estimates of different directions occurred, 95% confidence intervals were very wide. Online appendix Table 7 shows the inconsistency checks for triptan standard doses for two-hour pain-free response. The remaining consistency tables are available from the authors upon request. Eletriptan, sumatriptan, and rizatriptan were informed by many placebo-controlled trials. Almotriptan and zolmitriptan were generally informed by a moderate number of placebo-controlled trials, whereas naratriptan and frovatriptan were typically informed only by a few placebo-controlled trials. For the two-hour outcomes, head-to-head comparisons existed for many of the triptans; however, only sumatriptan was compared with other triptans (eletriptan, rizatriptan, and almotriptan) in more than one trial. For the 24-hour outcomes, five head-to-head comparisons were available between triptans for both outcomes, and except for eletriptan versus sumatriptan, each head-to-head comparison comprised only one trial.
Primary analysis
Table 1 displays the comparative odds ratios and 95% credible intervals for all outcomes from the primary MTC model adjusting for dose. Table 2 presents the treatment rankings for all unadjusted and adjusted analysis for each outcome. Figure 4 (a) and (b) display forest plots of the primary endpoints for individual triptans versus placebo based on the MTC findings.
|
Table 2. Treatment rankings.
At the two-hour time point, all triptans were significantly superior to placebo (defined as the 95% credible interval precluding 1.00). At the 24-hour time point all triptans except for naratriptan were significantly superior to placebo. Eletriptan consistently yielded higher odds ratios than the other triptans. For both two-hour outcomes, eletriptan was significantly superior to sumatriptan, almotriptan, and naratriptan. Eletriptan was also significantly superior to zolmitriptan for achieving pain-free response at two hours, and significantly superior to frovatriptan for achieving headache response at two hours. For both 24-hour outcomes, eletriptan was significantly superior to sumatriptan and naratriptan. Eletriptan was also significantly superior to rizatriptan for achieving 24-hour sustained headache response. The probability that eletriptan is the most likely of all triptans to produce a pain-free response at two hours and a headache response at two hours was 67.7% and 71.6%, respectively. The probability that eletriptan is the most likely of all triptans to produce a 24-hour sustained pain-free response and a 24-hour sustained headache response was 54.1% and 87.8%, respectively. The highest probability of being the second best was found with rizatriptan for the two-hour outcomes, and with zolmitriptan for the 24-hour outcomes (online appendix Table 8). The highest probability of being the third best was found with zolmitriptan for the two-hour outcomes, and shared between almotriptan and zolmitriptan for the 24-hour outcomes (online appendix Table 8).The above results apply to the “common” doses of triptans (see online appendix Table 1). Meta-regression incorporating the dose covariate demonstrated a significant effect of doubling (or halving) the dose for both the two-hour outcomes and the 24-hour outcomes. The effect modification by dose was larger for the pain-free response outcomes (log odds ratio coefficient estimates of 0.24 and 0.41) than for the headache response outcomes (log odds ratio coefficient estimates of 0.19 and 0.22).
Inspection of the model fit statistics, deviance information criterion (DIC), revealed that the primary model, the regression model on dose, yielded a notably better fit than the secondary model for the pain-free response at two hours. For the other outcomes, the models yielded similar fits (online appendix Table 2). This suggests that the dose-regression model is well-justified as the primary model.
Secondary analysis
Online appendix Table 3 displays the comparative odds ratios and 95% credible intervals for all outcomes from the primary multiple treatment comparison model adjusting for dose. Generally, the results were highly similar to the primary analysis. However, for 24-hour sustained pain-free response, eletriptan also became significantly superior to rizatriptan and almotriptan (in addition to sumatriptan and naratriptan). For sustained 24-hour headache response, the statistical significance was lost for the comparison with rizatriptan.
Sensitivity analysis
Online appendices Tables 4 and 5 display the results from the primary and secondary MTC models, respectively, where 100 mg was considered the “common” dose for sumatriptan. Online appendix Table 6 displays the treatment ranking from the primary MTC model considering 100 mg sumatriptan as the 'common' dose. Overall, sumatriptan had more favorable comparative effectiveness estimates than in the primary analysis. Eletriptan still yielded larger treatment effect estimates than sumatriptan, albeit not statistically significant. The probabilities of eletriptan being the most likely to produce a favorable outcome did not change with 100 mg being the “common” sumatriptan dose.
Sensitivity analysis including six additional eligible but unpublished trials comparing 50 mg and 100 mg sumatriptan with placebo did not cause noticeable changes in the effect estimates or width of the credible intervals (results not shown).
Discussion
Our study examined whether triptans offer differing effects for the abortive treatment of migraine and found statistically significant differences between available triptans. We found that all triptans offered favorable response compared with placebo both for short-term and sustained pain-free response and headache response. Eletriptan appeared to offer the most favorable outcomes in terms of pain-free response and headache response in relation to other triptans both for short-term and sustained outcomes. Rizatriptan, zolmitriptan and high-dose (100 mg) sumatriptan also appear effective at two hours, whereas only Zolmitriptan and high-dose sumatriptan appear to maintain their efficacy at 24 hours. These findings should be of interest both to clinicians and patients as they provide guidance in the choice of the most favorable drugs available.
MTC meta-analysis permits inferences into the relative effects of treatments even if they have not been directly compared in head-to-head trials. The methods for the conduct and interpretation of MTCs have been extensively reviewed elsewhere (16,21–25) Several important considerations are necessary to allow for valid MTCs. These include that the trials of each triptan are sufficiently similar to combine, that the trials of differing triptans are sufficiently similar in terms of patient populations and outcomes to compare, and that the direct evidence is sufficiently similar in direction and magnitude to the indirect evidence. We believe that these considerations have been met in this analysis.
Our study has several strengths. Our searches were sensitive and we believe that all completed and published trials have been included (26). When we applied a variety of sensitivity analyses, we found that our primary analysis findings were robust. Limitations of this analysis include the quality of reporting at the level of individual trials. Although most trials reported their findings according to established guidelines (17), there is some debate about the use of intention to treat versus per protocol in the analysis of migraine trials as it is necessary that a randomized patient experience at least one migraine attack for them to have an endpoint of pain-free response or headache response at any time point (27). We examined this in a sensitivity analysis and did not find important differences. Our treatment networks display that some treatment comparisons are informed by considerably more randomized clinical trials than others (21). Most treatments had a large number of placebo comparison trials at the two-hour endpoint, yet naratriptan and frovatriptan had much fewer trials informing their comparisons. It is possible that the findings of our analysis would differ if a large number of new head-to-head randomized clinical trials became available. It is also possible that publication bias explains the lower number of trials of specific drugs (26,28). Finally, our study did not examine adverse events and it is possible that different triptans exhibit differing adverse events (29).
Different dosages of drug treatments are frequently employed in migraine treatment. We accounted for the effect of dosage both in our primary and secondary model. Our study displays that treatment dosage does have an important role in the choice of drug and treatment outcomes. Our sensitivity analysis in which the common dose of sumatriptan was considered to be 100 mg further inform comparative effectiveness particularly pertinent to the US.
It should be noted that the results should not readily be generalized to patient populations that were not included in our MTC. Particularly, the comparative efficacy between triptans may be different in patients with a history of a previous poor response to migraine treatments (e.g. to triptans) and patients suffering from frequent nonmigranious headaches or other clinically significant illnesses. Of course, if a patient has previously responded poorly to a triptan, following the results of our analysis, a natural choice in clinical practice could well be to (or if a patient did not respond well to eletriptan, try one of the triptants that appear second or third best).
In conclusion, our study displays a hierarchy of treatment effects offered by the currently available triptans. Eletriptan appears to offer consistently the largest treatment efficacy at two and 24 hours. Rizatriptan appears to offer the second most favorable treatment outcome at two hours, but does not maintain the same degree of efficacy at 24 hours’ efficacy. Zolmitriptan and high-dose sumatriptan offer the third highest chance of pain-free response and headache response, and appear to maintain their efficacy at 24 hours. Clinicians will want to target a triptan that balances efficacy, tolerability, and costs for their patients.
Clinical implications
The relative efficacy of all triptans for the abortive treatment of migraine has remained uncertain as most randomized trials have compared the available triptans to placebo.
A multiple treatment comparison meta-analysis combining placebo and head-to-head trials was used to establish which triptan has the highest odds of producing favorable relief outcomes.
Eletriptan consistently has the highest odds of producing two-hour pain relief, two-hour headache response, 24-hour sustained pain relief, and 24-hour sustained headache response.
Rizatriptan, zolmitriptan, and high-dose (100 mg) sumatriptan also appear effective at two hours, whereas only zolmitriptan and high-dose sumatriptan appear to maintain their efficacy at 24 hours.
Funding
This study was sponsored by Pfizer Inc. With the exception of physician expert Anjan Chatterjee, as well as pre-analysis-stage discussion about overall trial eligibility criteria, the company had no role in the conduct, analysis or interpretation of the results and had no say in where this would be published.
Conflicts of interest
Kristian Thorlund and Edward Mills have consulted with Merck, Pfizer, Novartis, Nycomed, Johnson & Johnson, and GlaxoSmithKline on multiple treatment comparison and systematic review issues. Edward Mills is the owner of MacReviews Inc. Kristian Thorlund is an employee of MacReviews Inc. MacReviews Inc was financially supported by Pfizer Inc to develop and conduct this study and manuscript. In addition, they have received grants from the Canadian Institutes of Health Research (CIHR) and consulted with the Canadian Agency for Drugs and Technology in Health and US Agency for Healthcare research and Quality (AHRQ). Kristian Thorlund’s salary is supported by CIHR. Edward Mills’ salary is supported by a CIHR Canada Research Chair. Edward Mills is the president of MacReviews Inc, and was a paid consultant to Pfizer Inc in connection with the development of this manuscript. Kristian Thorlund is a director at MacReviews Inc, and was a paid consultant to Pfizer Inc in connection with the development of this manuscript. Peter Goadsby has received research grants from Boston Scientific, Medtronic, GSK, MSD, MAP, Johnson & Johnson, and Neuralieve and has received consulting fees or honoraria from Allergan, Almirall, ATI, BMS, Boehringer, Boston Scientific, Coherex, Colucid, Lilly, Medtronic, Minster, MSD, MAP, Neuralieve, NeurAxon, NTP, and Pfizer. Elodie Ramos is currently a full-time employee of Pfizer Inc. Anjan Chatterjee was a full-time employee of Pfizer Inc at the time this study was conducted. Ping Wu and Eric Druyts have no conflicts of interest to report.
Acknowledgments
We would like to thank Mary Almas, Younos Abdulsattar, and Rahul Bhambri from Pfizer for their valuable suggestions during the pre-analysis stage about the design and analysis of this study.
References
| 1. | Goadsby, PJ, Lipton, RB, Ferrari, MD. Migraine—current understanding and treatment. N Engl J Med 2002; 346: 257–270. Google Scholar | Crossref | Medline | ISI |
| 2. | Stewart, WF, Lipton, RB, Celentano, DD. Prevalence of migraine headache in the United States. Relation to age, income, race, and other sociodemographic factors. JAMA 1992; 267: 64–69. Google Scholar | Crossref | Medline | ISI |
| 3. | Andlin-Sobocki, P, Jonsson, B, Wittchen, HU. Cost of disorders of the brain in Europe. Eur J Neurol 2005; 12(Suppl 1): 1–27. Google Scholar | Crossref | Medline | ISI |
| 4. | Stewart, WF, Ricci, JA, Chee, E. Lost productive time and cost due to common pain conditions in the US workforce. JAMA 2003; 290: 2443–2454. Google Scholar | Crossref | Medline | ISI |
| 5. | Headache Classification Subcommittee of the International Headache Society . The International Classification of Headache Disorders: Second edition. Cephalalgia 2004; 24(Suppl 1): 9–160. Google Scholar | Medline |
| 6. | Stewart, WF, Lipton, RB, Liberman, J. Variation in migraine prevalence by race. Neurology 1996; 47: 52–59. Google Scholar | Crossref | Medline | ISI |
| 7. | Lipton, RB, Hemelsky, SW, Kolodner, KN. Migraine, quality of life and depression: A population-based case-control study. Neurology 2000; 55: 629–635. Google Scholar | Crossref | Medline | ISI |
| 8. | Stokes, M, Becker, WJ, Lipton, RB. Cost of health care among patients with chronic and episodic migraine in Canada and the USA: Results from the International Burden of Migraine Study (IBMS). Headache 2011; 51: 1058–1077. Google Scholar | Crossref | Medline | ISI |
| 9. | Goadsby, PJ, Sprenger, T. Current practice and future directions in management of migraine: Acute and preventive. Lancet Neurol 2010; 9: 285–298. Google Scholar | Crossref | Medline |
| 10. | Evers, S, Afra, J, Frese, A. EFNS guideline on the drug treatment of migraine—revised report of an EFNS task force. Eur J Neurol 2009; 16: 968–981. Google Scholar | Crossref | Medline | ISI |
| 11. | Silberstein, SD, Holland, S, Freitag, F. Evidence-based guideline update: Pharmacologic treatment for episodic migraine prevention in adults: Report of the Quality Standards Subcommittee of the American Academy of Neurology and the American Headache Society. Neurology 2012; 78: 1337–1345. Google Scholar | Crossref | Medline | ISI |
| 12. | Goadsby, PJ . The pharmacology of headache. Prog Neurobiol 2000; 62: 509–525. Google Scholar | Crossref | Medline | ISI |
| 13. | Akerman, S, Holland, P, Goadsby, PJ. Diencephalic and brainstem mechanisms in migraine. Nat Rev Neurosci 2011; 12: 570–584. Google Scholar | Crossref | Medline | ISI |
| 14. | Ferrari, MD, Roon, KI, Lipton, RB. Oral triptans (serotonin 5-HT(1B/1D) agonists) in acute migraine treatment: A meta-analysis of 53 trials. Lancet 2001; 358: 1668–1675. Google Scholar | Crossref | Medline | ISI |
| 15. | Mills, EJ, Ioannidis, JP, Thorlund, K. How to use an article reporting a multiple treatment comparison meta-analysis. JAMA 2012; 308: 1246–1253. Google Scholar | Crossref | Medline | ISI |
| 16. | Mills, EJ, Bansback, N, Ghement, I. Multiple treatment comparison meta-analyses: A step forward into complexity. Clin Epidemiol 2011; 3: 193–202. Google Scholar | Crossref | Medline |
| 17. | Tfelt-Hansen, P, Pascual, J, Ramadan, N. Guidelines for controlled trials of drugs in migraine: Third edition. A guide for investigators. Cephalalgia 2012; 32: 6–38. Google Scholar | SAGE Journals | ISI |
| 18. | Cramer, H . Mathematical methods of statistics, Princeton: Princeton University Press, 1946, pp. 282–282. Google Scholar |
| 19. | DerSimonian, R, Laird, N. Meta-analysis in clinical trials. Control Clin Trials 1986; 7: 177–188. Google Scholar | Crossref | Medline |
| 20. | Higgins, JP, Thompson, SG, Deeks, JJ. Measuring inconsistency in meta-analyses. BMJ 2003; 327: 557–560. Google Scholar | Crossref | Medline |
| 21. | Salanti, G, Kavvoura, FK, Ioannidis, JP. Exploring the geometry of treatment networks. Ann Intern Med 2008; 148: 544–553. Google Scholar | Crossref | Medline | ISI |
| 22. | Lu, G, Ades, AE. Combination of direct and indirect evidence in mixed treatment comparisons. Stat Med 2004; 23: 3105–3124. Google Scholar | Crossref | Medline | ISI |
| 23. | Salanti, G, Dias, S, Welton, NJ. Evaluating novel agent effects in multiple-treatments meta-regression. Stat Med 2010; 29: 2369–2383. Google Scholar | Medline |
| 24. | Lu, G, Ades, A. Modeling between-trial variance structure in mixed treatment comparisons. Biostatistics 2009; 10: 792–805. Google Scholar | Crossref | Medline | ISI |
| 25. | Lu, G, Ades, AE, Sutton, AJ. Meta-analysis of mixed treatment comparisons at multiple follow-up times. Stat Med 2007; 26: 3681–3699. Google Scholar | Crossref | Medline | ISI |
| 26. | Tfelt-Hansen, PC . Published and not fully published double-blind, randomised, controlled trials with oral naratriptan in the treatment of migraine: A review based on the GSK Trial Register. J Headache Pain 2011; 12: 399–403. Google Scholar | Crossref | Medline | ISI |
| 27. | Tfelt-Hansen, P . A review of evidence-based medicine and meta-analytic reviews in migraine. Cephalalgia 2006; 26: 1265–1274. Google Scholar | SAGE Journals | ISI |
| 28. | Tfelt-Hansen, PC . Unpublished clinical trials with sumatriptan. Lancet 2009; 374: 1501–1502. Google Scholar | Crossref | Medline | ISI |
| 29. | Pascual, J, Mateos, V, Roig, C. Marketed oral triptans in the acute treatment of migraine: A systematic review on efficacy and tolerability. Headache 2007; 47: 1152–1168. Google Scholar | Crossref | Medline | ISI |
