Pharmacological treatments for acute migraine: quantitative systematic review

Transcription

1 Pain 97 (2002) Pharmacological treatments for acute migraine: quantitative systematic review Anna D. Oldman, Lesley A. Smith, Henry J. McQuay, R. Andrew Moore* Pain Research, Nuffield Department of Anaesthetics, Oxford Radcliffe Hospital, University of Oxford, The Churchill, Headington, Oxford 0X3 7LJ, UK Received 14 November 2001; received in revised form 16 November 2001; accepted 14 January 2002 Abstract We wanted to compare the analgesic efficacy and adverse effects of pharmacological treatments for acute migraine through a systematic review of randomised controlled trials in patients with acute migraine pain of moderate to severe intensity. Trials were identified from systematic searching of bibliographic databases. For eletriptan information from all trials was supplied by Pfizer Inc. Outcomes sought were headache relief at 1 and 2 h, patients pain free at 2 h and sustained relief over 24 h for treatments compared with placebo. Numbers-neededto-treat (NNTs) were calculated, together with relative benefit. Information on adverse effects was also collected. Comparisons of relative efficacy used the same definition of headache, the same degree of pain at the start of treatment and the same definitions of outcomes, and always compared with placebo. Forty-eight publications reporting on 54 trials were included in the meta-analyses, with 79 placebo comparisons for the primary outcome of headache relief at 2 h. Information on any outcome was available for nine oral medications, two intranasal medications and subcutaneous sumatriptan in 21,022 patients. For headache relief at 2 h NNTs ranged from 2.0 for subcutaneous sumatriptan 6 mg to 5.4 for naratriptan 2.5 mg. For patients pain free at 2 h NNTs ranged from 2.1 for subcutaneous sumatriptan 6 mg to 8.6 for aspirin 900 mg plus metoclopramide 10 mg. For sustained relief over 24 h NNTs ranged from 2.8 for eletriptan 80 mg to 8.3 for rizatriptan 5 mg. It was not possible to systematically review adverse effects data. Most interventions are effective. There is considerable information on relative efficacy for a number of outcomes. q 2002 International Association for the Study of Pain. Published by Elsevier Science B.V. All rights reserved. Keywords: Systematic review; Migraine; Acute treatment; Triptans; Number-needed-to-treat 1. Introduction Migraine affects 18% of women and 6% of men (Lipton and Stewart, 1993), with about 34 attacks per year for men and 37 for women (Hu et al., 1999). The significant impact of migraine, on pain, disability, social functioning, quality of relationships, emotional well being and general health (Edmeads et al., 1993; Osterhaus et al., 1994; Solomon and Price, 1997), results in a huge burden for the individual, health services and society (Clarke et al., 1996; Solomon and Price, 1997; Ferrari, 1998; Hu et al., 1999). The annual US economic burden, including missed work days and lost productivity, is US$14 billion (Hu et al., 1999). Successful treatment of acute migraine attacks can reduce use of healthcare resources and disability, increase productivity and improve health-related quality of life (Jhingran et al., 1996; Cady et al., 1998; Lofland et al., 1999). While such * Corresponding author. Tel.: ; fax: address: andrew.moore@pru.ox.ac.uk (R.A. Moore). successful treatment is the most important goal for the patient (Rasmussen et al., 1991; Edmeads et al., 1993; Lipton and Stewart, 1999), only a proportion of migraine sufferers seek professional advice about treating attacks, with 60% of severe migraine sufferers relying solely on over-the-counter medications (Lipton and Stewart, 1993). Direct treatment comparisons in large randomised controlled trials would best determine relative efficacy of available treatments. In their absence we have to rely on the indirect comparisons in systematic reviews of how each drug performed against placebo to determine which treatment is best (McQuay and Moore, 1998). Systematic review methods like these have only recently been used to evaluate the effectiveness of migraine treatments (Tfelt-Hansen, 1993, 1998; Goadsby, 1998; Tfelt-Hansen, 1998; Ferrari et al., 2001; Gawel et al., 2001; Oldman et al., 2001). The more recent reviews have used number-needed-to-treat (NNT) compared with placebo to describe treatment efficacy in terms of the therapeutic effort required to produce a successful outcome. NNT is treatment specific. It describes the difference between active treatment and /02/$20.00 q 2002 International Association for the Study of Pain. Published by Elsevier Science B.V. All rights reserved. PII: S (02)

2 248 A.D. Oldman et al. / Pain 97 (2002) control in achieving a particular clinical outcome (McQuay and Moore, 1997). Low NNTs indicate high treatmentspecific efficacy. An NNT of 1 says that a favourable outcome occurs in every patient given the treatment but in no patient in a comparator group. This would be the perfect result in, say, a therapeutic trial of an antibiotic compared with placebo with a sensitive organism. NNTs of 2 5 are indicative of high efficacy, as, for instance, with analgesics in acute pain (McQuay and Moore, 1998). For migraine there are few direct drug comparisons of adequate size and power. A particular problem is that there is no universal agreement on which outcome measure is most important. Patients rated complete pain relief, no headache recurrence, rapid onset of pain relief and no side effects as the four most important outcomes (Lipton and Stewart, 1999). The standard long-term measure of efficacy is headache recurrence within 24 h, which has limitations (Goldstein et al., 1999; Millson et al., 1999), and a better alternative may be sustained relief over 24 h (Ferrari, 1999). Sustained pain free has been suggested as the final goal in drug development (Tfelt-Hansen et al., 2000). Many randomised controlled trials have been carried out in acute migraine, making systematic reviews possible, as for sumatriptan (Tfelt-Hansen, 1998). A series of systematic reviews using the same methods allows ranking of drug relative efficacy, as in postoperative pain relief (McQuay and Moore, 1998). Combining data from a number of trials can also make sense for safety information, because individual trials are too small for accurate estimation of adverse effect incidence (Edwards et al., 1999). This report contains meta-analyses of the analgesic efficacy of any acute treatment for migraine, evaluating each of several short- and long-term outcomes. Each intervention had been compared with placebo in similar patients experiencing similar pain, treating a single migraine attack with a single dose of medication. 2. Methods This review includes data from two sources: Pfizer Inc. supplied data on all randomised placebo-controlled trials of eletriptan. For the other interventions, full journal publication of single-dose, randomised, double-blind placebocontrolled trials were sought. Different search strategies were used to identify eligible reports from MEDLINE (1966 to July 2000), EMBASE (1980 to June 2000), Cochrane Library (Issue 3, 2000) and the Oxford Pain Relief Database ( ) (Jadad et al., 1996a). A series of free text searches were undertaken, using generic and trade names for each medication indicated for migraine and all widely available analgesics. No restrictions to language were made. Neither individual authors nor other pharmaceutical companies were contacted for unpublished reports. Inclusion criteria for this review were: pharmacological treatment of acute migraine; randomised allocation to treatment groups including a placebo group; double-blind design; International Headache Society diagnostic criteria for migraine with or without aura (Headache Classification Committee of the International Headache Society, 1988); single migraine attack; single-dose treatment at standard doses; adult population; baseline pain of moderate or severe intensity using a four-point standardised rating scale (0 ¼ no pain, 1 ¼ mild pain, 2 ¼ moderate pain and 3 ¼ severe pain); dichotomous or percentage data for at least one efficacy outcome. The choice of main outcome measures for this review was made pre hoc by taking into consideration scientific rigour, availability of data and patient preferences (Lipton and Stewart, 1999). The efficacy outcomes sought using a four-point standardised rating scale (0 ¼ no pain, 1 ¼ mild pain, 2 ¼ moderate pain and 3 ¼ severe pain) were: (i) Headache relief (headache pain reduced from moderate or severe (2 or 3) to mild or none (1 or 0)) at 2 h. (ii) Headache relief (headache pain reduced from moderate or severe to mild or none) at 1 h. (iii) Pain-free response (headache pain reduced from moderate or severe (2 or 3) to none (0)) at 2 h. (iv) Sustained relief over 24 h, (headache relief at 2 h, sustained for 24 h after treatment i.e. pain did not return to moderate or severe, and without use of rescue or second-dose medication). (v) Pain free over 24 h, (pain free at 2 h, sustained for 24 h after treatment i.e. pain did not return, and without use of rescue or second-dose medication). We looked for adverse effects within 24 h of treatment in the form of: (i) Incidence of major harm (defined as withdrawal from study due to adverse effect). (ii) Incidence of any adverse effect (defined as any undesirable experience considered related to the study drug (Stephens et al., 1998). (iii) Incidence of particular adverse effects Data extraction and statistical analyses Each report was read independently by A.D.O. and L.A.S., and scored using a three item, 1 5 score, quality scale (Jadad et al., 1996b). Points were awarded to studies according to whether they were randomised and double blind and mentioned withdrawals or drop outs from the study. An additional point was awarded if the method of randomisation or double blinding was described and was appropriate. From each trial we extracted the number of patients treated per group, dosing regimes, details of study design (parallel or crossover, multiple attack or single attack) and timing/ type of rescue medication. We did not extract data contami-

3 A.D. Oldman et al. / Pain 97 (2002) nated by rescue or second study dose medication. Where sustained relief data over 24 h was not explicit it was calculated by subtracting the number of patients whose headache had recurred by 24 h from the number who had had a headache relief at 2 h. Adverse effects information was extracted from reports together with information on method of data collection, time over which data was collected and whether or not this was contaminated with rescue and/or second study dose medication. For headache relief, pain-free headache relief and sustained relief over 24 h, the proportion of patients achieving the outcome in the active and placebo groups was used to calculate relative benefit (RB) and NNT. RB estimates were calculated with 95% confidence intervals (CIs) (Morris and Gardner, 1995) and were used to test for a significant difference between drug treatment and placebo. A fixed effects model was used because the trials meeting the inclusion criteria were clinically homogeneous, and because heterogeneity tests lack power (Gavaghan et al., 2000). Publication bias was not assessed using funnel plots as these tests have been shown to be unhelpful (Tang and Liu, 2000; Sterne et al., 2000). NNTs were calculated with 95% CI (Cook and Sackett, 1995; McQuay and Moore, 1997) when the RB was statistically significant. A statistically significant difference from placebo was assumed when the 95% CI of the RB or harm did not include one. To assess the effect of dose on efficacy, significance differences between NNTs were tested by calculating the z statistic (Tramèr et al., 1997). CIs (95%) were calculated for proportions using a standard method (Gardner and Altman, 1986). Calculations were performed using Microsoft Excel 98 on a Power Macintosh 8500/ Results Of 129 identified reports of randomised trials, 81 were excluded. Reasons for exclusion were: baseline pain not moderate/severe or not stated (31) use of non-standard outcomes (8) data not presented separately for first attack (8) medication given only at non-standard doses (7) no extractable data (6) medication given in a formulation that is not available (5) duplicate publication (4) migraineurs did not meet IHS criteria (4) randomisation inadequate (3) data collection period less than 1 h (2) not adequately blinded (1) data contaminated by concomitant medication (1) patient sample was biased (1). Forty-eight publications reporting on 54 trials were included, with 79 placebo comparisons for headache relief at 2 h. Information was available for nine different oral drugs, two intranasal drugs and subcutaneous sumatriptan in a total of 21,022 patients for at least one outcome. The median size of treatment groups was 126 patients (range ). Quality scores of included trials ranged from 2 to 5 (median 4). Forty-six trials scored at least three points, and eight scored less than three. Trials were demographically similar in terms of age, sex and type of migraineur. Details of included and excluded trials are available from the authors or at painres/migraine/migrev1.html. Eletriptan protocol has subsequently been published (Goadsby et al., 2000). There were no trials that met the inclusion criteria for the following migraine-indicated medications: isometheptene mucate; Midrin w (isometheptene mucate plus paracetamol plus dichloralphenazone); Midrid w (isometheptene mucate plus paracetamol); aspirin; Migravess w (effervescent aspirin plus metoclopramide); paracetamol; Paramax w (paracetamol plus metoclopramide); Migraleve w (codeine plus buclizine hydrochloride); paracetamol plus codeine combination; Domperamol w (paracetamol plus domperidone); ergotamine compounds other than Cafergot w (ergotamine tartrate plus caffeine). No trials of the following drugs met the inclusion criteria: ibuprofen, naproxen, diclofenac, ketoprofen, oxycodone, dextropropoxyphene, tramadol, ketorolac, piroxicam, indomethacin, pirprofen, lignocaine, pethidine, butorphanol, mefenamic acid. Some guidelines suggest the use of simple analgesics such as aspirin, paracetamol or NSAIDs in the treatment of acute migraine (Foord-Kelcey, 1999). Information on Excedrin w (paracetamol 500 mg plus aspirin 500 mg plus caffeine 130 mg) was available from three trials. Because these trials excluded a subset of patients (migraineurs with severely debilitating attacks usually requiring bed rest, or who vomited more than 20% of the time), the NNT for Excedrin w is not strictly comparable with those for other interventions, and this information has not been included in graphs. No NNT was calculated for tolfenamic acid 200 or 400 mg because data could not be extracted for individual doses Headache relief at 2 h Information was available for 13 oral interventions, two intranasal interventions and subcutaneous sumatriptan (Table 1). Mean response rates for treatment ranged from 34 to 79%, and for placebo from 22 to 35%. Cafergot w was not significantly different from placebo for this outcome (RB 1.5, 95% CI ), but all other treatments had significant RB over placebo (Table 1). NNTs for headache relief at 2 h compared with placebo ranged from 2.0 ( ) to 5.4 ( ). The most effective intervention was subcutaneous sumatriptan 6 mg. Fig. 1 shows the interventions ranked according to efficacy. Subcutaneous sumatriptan 6 mg, eletriptan 80 mg and rizatriptan 10 mg were all significantly more effective than

4 250 Table 1 Summary of relative benefit and number-needed-to-treat for headache relief at 2 h for acute migraine interventions Drug and dose Number of comparisons Number improved Relative benefit NNT Percentage improved With treatment With placebo With treatment mean With placebo mean Subcutaneous Sumatriptan 6 mg 8 379/ / ( ) 2.0 ( ) 79 (69 91) 28 (13 41) Intranasal Dihydroergotamine 2 mg a 1 66/105 23/ ( ) 2.5 ( ) Sumatriptan 20 mg 6 571/ / ( ) 3.4 ( ) 63 (60 78) 34 (25 42) Oral Aspirin 900 mg 1 Metoclopramide 10 mg 3 214/376 95/ ( ) 3.2 ( ) 57 (54 59) 25 (23 29) Cafergot (Ergotamine Tartrate 2 mg 1 Caffeine 200 mg) 1 58/169 19/ ( ) Not calculated Eletriptan 80 mg 6 763/ / ( ) 2.6 ( ) 62 (56 69) 25 (21 40) Eletriptan 40 mg 6 724/ / ( ) 2.9 ( ) 59 (53 63) 25 (21 40) Excedrin (Paracetamol 500 mg 1 Aspirin 3 358/ / ( ) 3.8 ( ) 59 (56 64) 500 mg 1 Caffeine 130 mg) b Naratriptan 2.5 mg 2 154/340 61/ ( ) 5.4 ( ) 45 (40 48) 27 (22 30) Rizatriptan 10 mg / / ( ) 2.7 ( ) 68 (48 77) 31 (18 40) Rizatriptan 5 mg 4 548/ / ( ) 3.9 ( ) 59 (45 62) 33 (18 40) Sumatriptan 100 mg / / ( ) 3.3 ( ) 58 (46 79) 28 (17 43) Sumatriptan 50 mg 6 532/ / ( ) 4.1 ( ) 51 (42 60) 27 (17 40) Tolfenamic acid rapid release 200/400 mg c 1 33/43 12/41 Not calculated Zolmitriptan 5 mg 4 583/943 85/ ( ) 3.1 ( ) 62 (62 67) 30 (15 34) Zolmitriptan 2.5 mg 2 279/438 74/ ( ) 3.5 ( ) 64 (62 65) 35 (34 36) A.D. Oldman et al. / Pain 97 (2002) a b Drug given over 15 min. Patients with severely debilitating migraine excluded. A further 200 mg given at 60 min if headache not improved (28% of patients required only 200 mg). c

5 A.D. Oldman et al. / Pain 97 (2002) There was a significant dose response for rizatriptan 10 vs. 5 mg (Table 2) Pain-free headache relief at 2 h Fig. 1. Relative efficacy for headache relief at 2 h. The bars represent the 95% CI of the NNT compared with placebo for each treatment and dose. Headache relief is headache pain reduced from moderate or severe pain to mild or no pain. SC denotes subcutaneous administration, and IN intranasal; all other treatments were oral. rizatriptan 5 mg, sumatriptan 50 mg and naratriptan 2.5 mg. For higher doses of oral triptans NNTs for headache relief at 2 h ranged from 2.6 ( ) to 3.3 ( ), with little difference between them. For lower doses NNTs ranged from 2.9 ( ) to 4.1 ( ), again with little difference. Naratriptan, which is only available at 2.5 mg, was less effective than all high-dose triptans, and the lower dose of eletriptan. Based on a select patient group of migraineurs (i.e. those not usually experiencing severe migraine-related disability or vomiting), Excedrinw was significantly better than placebo, with a NNT of 3.8 ( ). There was a significant dose response for rizatriptan 10 vs. 5 mg and for sumatriptan 100 vs. 50 mg (Table 2) Headache relief at 1 h Information was available for 11 oral interventions, two intranasal interventions and subcutaneous sumatriptan (Table 3). Mean response rates for treatment ranged from 14 to 70%, and for placebo from 9 to 30%. Cafergotw was not significantly different from placebo for this outcome (RB 1.1, 95% CI ), but all other treatments had significant RB over placebo (Table 3). No NNT was calculated for Cafergotw, as the RB indicated no significant difference from placebo. For the other drugs there was a wide range in the NNT for headache relief at 1 h, with subcutaneous sumatriptan 6 mg clearly the most effective, NNT 2.1 ( ) and Excedrinw the least, NNT 10 (7.3 17). Information was available for 13 oral interventions, and intranasal and subcutaneous sumatriptan (Table 4). Mean response rates for treatment ranged from 11 to 60%, and for placebo from 5 to 12%. Cafergotw was not significantly different from placebo for this outcome (RB 2.4, 95% CI ), but all other treatments had significant RB over placebo (Table 4). No data were available for intranasal dihydroergotamine 2 mg. NNTs for pain-free headache relief at 2 h ranged from 2.1 ( ) to 8.6 (6.2 14). Subcutaneous sumatriptan 6 mg was clearly the most effective, and aspirin 900 mg plus metoclopramide 10 mg the least. There was a significant dose response for all four oral triptans studied (Table 2) Sustained relief over 24 h Because of the wide variation in the design of migraine trials, it was not possible to extract data for this outcome for a number of interventions. Information was available for eight oral interventions and subcutaneous sumatriptan (Table 5). Mean response rates for treatment ranged from 30 to 53%, and for placebo from 14 to 27%. Zolmitriptan 5 mg was not significantly different from placebo for this outcome (RB 1.5, 95% CI ), but all other treatments had significant RB over placebo (Table 5). For sustained relief over 24 h (headache relief at 2 h, and no headache recurrence within 24 h), NNTs ranged from 2.8 ( ) to 8.3 (6.0 14) (Fig. 2). The most effective intervention on this outcome was eletriptan 80 mg, and the least effective was rizatriptan 5 mg. There was a significant dose response for eletriptan 80 vs. 40 mg and for rizatriptan 10 vs. 5 mg (Table 2) Pain free over 24 h No trials reported information to allow the calculation of this outcome measure Adverse effects We were unable to carry out a meaningful analysis of Table 2 Dose response of triptans for efficacy outcomes a Dose comparison Headache relief at 1 h Headache relief at 2 h Pain free at 2 h Sustained relief over 24 h Eletriptan 80 vs. 40 mg Rizatriptan 10 vs. 5 mg 0.014,0.001, Sumatriptan 100 vs. 50 mg , Zolmitriptan 5 vs. 2.5 mg , N/A a Probabilities for the comparison of NNTs for high and low doses of each triptan calculated according to Tramèr et al., The values in italic indicate statistical differences between doses.

6 252 Table 3 Summary of relative benefit and number-needed-to-treat for headache relief at 1 h for acute migraine interventions Drug and dose Number of comparisons Number improved Relative benefit NNT Percentage improved With treatment With placebo With treatment mean With placebo mean Subcutaneous Sumatriptan 6 mg / / ( ) 2.1 ( ) 70 (58 88) 22 (5 33) Intranasal Dihydroergotamine 2 mg a 1 50/105 18/ ( ) 3.4 ( ) Sumatriptan 20 mg 5 374/ / ( ) 5.6 ( ) 47 (45 60) 30 (28 42) Oral Aspirin 900 mg 1 Metoclopramide 10 mg None Cafergot (Ergotamine Tartrate 2 mg 1 Caffeine 1 23/169 11/ ( ) Not calculated mg) Eletriptan 80 mg 6 408/ / ( ) 5.0 ( ) 33 (31 39) 13 (10 21) Eletriptan 40 mg 6 378/ / ( ) 5.8 ( ) 31 (21 35) 13 (10 21) Excedrin (Paracetamol 500 mg 1 Aspirin 3 113/602 55/ ( ) 10 (7.3 17) 19 (16 22) 8.9 (8 9) 500 mg 1 Caffeine 130 mg) b Naratriptan 2.5 mg 1 59/213 17/ ( ) 8.5 (4.8 38) 28 Rizatriptan 10 mg 7 761/ / ( ) 4.9 ( ) 43 (25 55) 22 (14 30) Rizatriptan 5 mg 4 349/ / ( ) 7.2 (5.4 10) 37 (30 41) 23 (20 27) Sumatriptan 100 mg 7 501/ / ( ) 7.6 (5.9 10) 31 (18 39) 18 (12 27) Sumatriptan 50 mg 4 268/934 77/ ( ) 11 (7.1 22) 29 (22 40) 19 (13 27) Tolfenamic acid rapid release 200/400 mg c None Zolmitriptan 5 mg 4 355/943 59/ ( ) 5.9 ( ) 38 (24 44) 21 (15 26) Zolmitriptan 2.5 mg 2 173/438 55/ ( ) 7.3 (4.7 16) 39 (33 44) 26 A.D. Oldman et al. / Pain 97 (2002) a b Drug given over 15 min. Patients with severely debilitating migraine excluded. A further 200 mg given at 60 min if headache not improved (28% of patients required only 200 mg). c

7 Table 4 Summary of relative benefit and number-needed-to-treat for pain-free headache relief at 2 h for acute migraine interventions Drug and dose Number of comparisons Number improved Relative benefit NNT Percentage improved With treatment With placebo With treatment mean With placebo mean Subcutaneous Sumatriptan 6 mg 6 206/345 41/ ( ) 2.1 ( ) 60 (20 76) 12 (3 17) Intranasal Dihydroergotamine 2 mg None Sumatriptan 20 mg 3 182/536 33/ ( ) 4.6 ( ) 34 (31 42) 12 (4 20) Oral Aspirin 900 mg 1 Metoclopramide 10 mg 3 69/378 25/ ( ) 8.6 (6.2 14) 18 (15 21) 6.7 (5 8) Cafergot (Ergotamine Tartrate 1 19/169 4/ ( ) Not calculated mg1Caffeine 200 mg) Eletriptan 80 mg 6 386/ / ( ) 3.7 ( ) 32 (25 37) 4.9 (3 11) Eletriptan 40 mg 6 335/ / ( ) 4.5 ( ) 27 (19 32) 4.9 (3 11) Excedrin (Paracetamol 500 mg 1 Aspirin 3 128/602 43/ ( ) 7.0 ( ) 21 (17 26) 7.0 (5 9) 500 mg 1 Caffeine 130 mg) a Naratriptan 2.5 mg 1 44/213 9/ ( ) 8.2 (5.1 21) 21 8 Rizatriptan 10 mg 7 720/ / ( ) 3.1 ( ) 40 (26 45) 8.4 (3 10) Rizatriptan 5 mg 4 284/933 65/ ( ) 4.7 ( ) 30 (22 35) 9.1 (3 10) Sumatriptan 100 mg 8 415/ / ( ) 4.7 ( ) 28 (17 50) 7.1 (4 13) Sumatriptan 50 mg 4 131/711 21/ ( ) 7.8 (6.1 11) 18 (16 19) 5.9 (3 11) Tolfenamic acid rapid release 200/400 mg b 1 16/43 3/41 Not calculated Zolmitriptan 5 mg 4 298/943 17/ ( ) 3.9 ( ) 32 (14 39) 6.0 (1 13) Zolmitriptan 2.5 mg 2 109/438 17/ ( ) 5.9 ( ) 25 (22 27) 8.0 (7 10) a b Patients with severely debilitating migraine excluded. A further 200 mg given at 60 min if headache not improved (28% of patients required only 200 mg). A.D. Oldman et al. / Pain 97 (2002)

8 254 Table 5 Summary of relative benefit and number-needed-to-treat for sustained relief at 24 h (headache response at 2 h and no recurrence within 24 h) Drug and dose Number of comparisons Number improved Relative benefit NNT Percentage improved With treatment With placebo With treatment mean With placebo mean Subcutaneous Sumatriptan 6 mg 2 70/144 28/ ( ) 3.2 ( ) 49 (32 67) 18 (14 22) Oral Cafergot (Ergotamine Tartrate 1 60/203 15/ ( ) 6.6 (4.1 16) mg1Caffeine 200 mg) Eletriptan 80 mg 5 618/ / ( ) 2.8 ( ) 53 (49 62) 18 (12 31) Eletriptan 40 mg 5 526/ / ( ) 3.6 ( ) 45 (40 55) 18 (12 31) Rizatriptan 10 mg 3 420/ / ( ) 5.6 ( ) 41 (38 43) 23 (18 32) Rizatriptan 5 mg 3 280/ / ( ) 8.3 (6.0 14) 35 (31 39) 23 (18 32) Sumatriptan 50 mg 2 157/362 47/ ( ) 6.0 (4.0 12) 43 (36 50) 27 (21 31) Sumatriptan 100 mg 4 472/ / ( ) 6.7 ( ) 40 (36 42) 25 (18 32) Zolmitriptan 5 mg 1 180/498 14/ ( ) Not calculated A.D. Oldman et al. / Pain 97 (2002)

9 A.D. Oldman et al. / Pain 97 (2002) Fig. 2. Sustained relief over 24 h. The bars represent the 95% CI of the NNT compared with placebo for each treatment and dose. Sustained relief over 24 h is headache relief at 2 h, sustained for 24 h after treatment i.e. pain did not return to moderate or severe, and with no stated use of rescue or seconddose medication. SC denotes subcutaneous administration, and IN intranasal; all other treatments were oral. adverse effects over 24 h because adverse effects were usually collected over 7 10 days. Methods of data collection and reporting were not of sufficient quality, and most trials presented information on the adverse effects of the study drug that was contaminated with rescue and/or second study doses. 4. Discussion This review shows that many currently available treatments effectively relieve the headache of acute migraine, and highlights the absence of high quality clinical trials for many therapies recommended for treatment of acute migraine. Most information was available for the efficacy of triptans, though there was little information on adverse events in the 24 h after treatment. Ranking the drugs by efficacy for a particular outcome showed that some treatments were better than others. Considering all outcomes, at the doses used subcutaneous sumatriptan and the newer oral triptans except naratriptan provided the best efficacy. Our approach was to use explicitly defined outcomes and to exclude reports that did not meet those definitions unequivocally. Only studies likely to be free of bias were included. They had to be randomised and double-blind, avoiding bias from known sources (Schulz et al., 1995). Quality scores were 3, 4 and 5 in 85% of trials, so avoiding the potential for bias from poor reporting quality found in studies with scores of 2 or less (Khan et al., 1996; Moher et al., 1998). Trials also tended to be large, avoiding the potential for bias through small numbers or random chance effects (Moore et al., 1998a,b). The median size of treatment groups was 126 patients, compared with only 40 patients in acute pain studies (Moore et al., 1998a). It is likely that there are unpublished clinical trials, probably done for registration by pharmaceutical companies but not published. We were able to obtain unpublished information from a number of unpublished studies of eletriptan. Meta-analysis of unpublished trials is rare, though it has been done for tramadol in acute pain (Moore and McQuay, 1997); meta-analysis of new treatments before launch is even rarer (Edwards and Moore, 1999). Obtaining unpublished information from pharmaceutical companies is not easy, but is an issue as more new triptans become available. For most of the established drugs large amounts of information was available. Overall there was information on over 21,000 patients for at least one outcome in high quality clinical trials. Comparing different migraine drugs is complicated by the different outcome measures and reporting used in the trials. Debate continues about the best outcome measure, and which provides the best overview of a drug s efficacy. We chose to calculate efficacy based on a number of short-term measures, and the number with a headache relief at 2 h who remain pain free over the succeeding 22 h. The most useful approach may be the number of patients who are pain free at 2 h and remain pain free over 24 h without additional analgesia (Ferrari, 1999), but this outcome was not reported in any trials. Sustained relief and sustained pain free have emerged in the past few years as increasingly important and clinically relevant treatment outcomes (Goldstein et al., 1999). The guidelines for controlled trials of drugs in migraine (Tfelt-Hansen et al., 2000) cites both these sustained outcome measures as being appropriate. The Guidelines state that sustained pain free is the ideal response to a drug for treatment of a migraine attack and should be the final goal in drug development. Recent studies (Gruffyd- Jones et al., 2001), as well as recent meta-analyses (Ferrari et al., 2001; Oldman et al., 2001), have begun to use both sustained headache relief and sustained pain free as clinically important outcome measures. One reason the concept of sustained response is so important is because of its pharmacoeconomic implications (Cady et al., 2001). Patients who both respond and do not relapse require no additional medication. Sustained relief may therefore be considered the most important outcome from the standpoint of the cost of treatment. Patients expect swift, complete and lasting relief (Lipton and Stewart, 1999). They also desire a lack of adverse effects (Lipton and Stewart, 1999), but trials generally collected adverse effects information using poorly defined methodologies and over long periods compared with the measurement of efficacy, and we judged adverse event information to be uninterpretable. Efficacy was derived in trials of high quality, in patients who had the same definition of headache and the same degree of pain at the start of treatment, with similarly defined outcomes, and always compared with placebo. This could be provided for four outcomes. NNTs were lower (better) for headache relief at 2 h than 1 h (Tables 1 and 3). This is expected because the time the treatments have to work is doubled. NNTs were higher (worse) for

10 256 A.D. Oldman et al. / Pain 97 (2002) pain free than headache relief at 2 h (Tables 1 and 4). This is expected because the outcome is harder to attain. For these short-term outcomes relative efficacy was generally similar between treatments, though the non-triptan treatments aspirin plus metoclopramide and paracetamol plus aspirin plus caffeine were much less effective for the harder outcome of pain free at 2 h (Tables 1 and 4). Corroborating support for the results comes from three sources. First the range of responses to placebo in the different studies was relatively narrow for each outcome (Tables 1, 3 5). Differences were those that might be expected by chance (Moore et al., 1998a). This confirms the clinical homogeneity of the included trials. Second, there was a dose response relationship for each triptan on at least one of the outcomes when information was available for two doses of a drug (Table 2). The dose response was shown best on the pain free at 2-h outcome (Table 2), rather than on headache relief. This confirms the validity of the biological assays and the differentiating nature of harder outcomes. Third, the results obtained in this analysis were similar to those in other systematic reviews for sumatriptan (Tfelt- Hansen, 1998), and for several other triptans (Ferrari et al., 2001; Gawel et al., 2001). Relative efficacy from indirect comparisons is supported not only by dose response relationships, but by results of large direct comparisons. Some of the direct comparisons were included in the indirect comparison, so cannot be cited as additional evidence. Others have not been included. For instance, a comparison of rizatriptan 10 mg and zolmitriptan 2.5 mg in 776 patients concluded that rizatriptan was better (Pascual et al., 2000). But zolmitriptan 2.5 and 5 mg was broadly comparable to sumatriptan 50 and 100 mg in 1445 patients (Gallagher et al., 2000) and 1522 patients (Gruffyd- Jones et al., 2001). These results of direct comparisons are in agreement with the indirect comparisons. Patients want rapid and complete pain relief and pain that does not recur (Lipton and Stewart, 1999). The outcome of sustained response defined in this review approximated patient needs. Closer would be the outcome of pain free over 24 h (pain free at 2 h, sustained for 24 h after treatment i.e. pain did not return, and with no stated use of rescue or second-dose medication). There was little or no information in trials reports from which to calculate this outcome. The lack of high quality trials for simple analgesics, NSAIDs, a number of preparations indicated specifically for migraine and for over-the-counter preparations was surprising. We were unable to calculate NNTs for any of these except for Excedrinw, available only in the USA, and those trials omitted patients with the most severe migraine headaches. A popular UK guideline (Foord-Kelcey, 1999) recommends aspirin or paracetamol, followed by oral and rectal non-steroidal anti-inflammatory drugs with or without antiemetics. Few if any of the clinical trials underpinning the guideline met our inclusion criteria because of poor methodological quality. Those trials that did have adequate designs often failed to collect useful adverse effect data. Outcomes of most interest, pain relief soon after headache occurrence and sustained relief over 24 h, were used inconsistently. Determining the kind of evidence needed to differentiate between new and established drugs is a challenge (Goadsby, 1998). Efficacy measurement has to be rigorous but encompass what is important to the sufferers, and better defined long-term outcomes are required, such as the proportion who are both pain free at 2 h and remain pain free for the next 24 h. Newer oral triptans give complete relief of pain at 2 h to about 30 40% of sufferers. The challenge for harm is to measure it. The fact that a reliable assessment of minor or major adverse effects could not be made from these migraine trials hinders progress. Acknowledgements The study was supported by Pain Relief Funds, and by an educational grant from Pfizer UK, Ltd. A.D.O. was supported by the BUPA Foundation. References Cady RC, Ryan R, Jhingran P, O Quinn S, Pait DG. Sumatriptan injection reduces productivity loss during a migraine attack. Arch Intern Med 1998;158: Cady RK, Sheftell F, Lipton RB, Kwong WJ, O Quinn S. Economic implications of early treatment of migraine with sumatriptan tablets. Clin Ther 2001;23: Clarke CE, MacMillan L, Sondhi S, Wells NEJ. Economic and social impact of migraine. QJM 1996;89: Cook RJ, Sackett DL. The number needed to treat: a clinically useful measure of treatment effect. Br Med J 1995;310: Edmeads J, Findlay H, Tugwell P, Pryse-Phillips W, Nelson RF, Murray TJ. Impact of migraine and tension-hype headache on life-style, consulting behaviour, and medication use: a Canadian population survey. Can J Neurol Sci 1993;20: Edwards JE, Moore RA. Implanon: a review of clinical studies. Br J Fam Plann 1999;24:3 16. Edwards JE, McQuay HJ, Moore RA, Collins SL. Reporting of adverse effects in clinical trials should be improved. Lessons from acute postoperative pain. J Pain Symptom Manage 1999;18: Ferrari M. The economic burden of migraine to society. Pharmacoeconomics 1998;13: Ferrari M. How to assess and compare drugs in the management of migraine: success rates in terms of response and recurrence. Cephalalgia 1999;19:2 8. Ferrari MD, Loder E, McCarroll KA, Lines CR. Meta-analysis of rizatriptan efficacy in randomized controlled clinical trials. Cephalalgia 2001;21: Foord-Kelcey G, editor. Guidelines p. 9. Gallagher RM, Dennish G, Spierings EL, Chitra R. A comparative trial of zolmitriptan and sumatriptan for the acute oral treatment of migraine. Headache 2000;40: Gardner MJ, Altman DG. Confidence intervals rather than P values: estimation rather than hypothesis testing. Br Med J 1986;292: Gavaghan DJ, Moore RA, McQuay HJ. An evaluation of homogeneity tests in meta-analyses in pain using simulations of individual patient data. Pain 2000;85: Gawel MJ, Worthington I, Maggisano A. A systematic review of the use of triptans in acute migraine. Can J Neurol Sci 2001;28:30 41.

11 A.D. Oldman et al. / Pain 97 (2002) Goadsby PJ. A triptan too far? J Neurol Neurosurg Psychiatry 1998;64: Goadsby PJ, Ferrari MD, Olesen J, Stovner LJ, Senard JM, Jackson NC, Poole PH, Eletriptan Steering Committee. Eletriptan in acute migraine: a double-blind, placebo-controlled comparison to sumatriptan. Neurology 2000;54: Goldstein DJ, Offen WW, Moster MB. Efficacy definitions for migraine studies. Cephalalgia 1999;19: Gruffyd-Jones K, Kies B, Middleton A, Mulder LJ, Rosjo O, Millson DS. Zolmitriptan versus sumatriptan for the acute oral treatment of migraine: a randomized, double-blind, international study. Eur J Neurol 2001;8: Headache Classification Committee of the International Headache Society. Classification and diagnostic criteria for headache disorders, cranial neuralgias and facial pain. Cephalalgia 1988;8: Hu XH, Markson LE, Lipton RB, Stewart WF, Berger ML. Burden of migraine in the United States. Arch Intern Med 1999;159: Jadad AR, Carroll D, Moore A, McQuay H. Developing a database of published reports of randomised clinical trials in pain research. Pain 1996a;66: Jadad AR, Moore RA, Carroll D, Jenkinson C, Reynolds DJM, Gavaghan DJ, McQuay HJ. Assessing the quality of reports of randomized clinical trials: is blinding necessary? Control Clin Trials 1996b;17:1 12. Jhingran P, Cady RK, Rubino J, Miller D, Grice RB, Gutterman DL. Improvements in health-related quality of life with sumatriptan treatment for migraine. J Fam Pract 1996;42: Khan KS, Daya S, Jadad AR. The importance of quality of primary studies in producing unbiased systematic reviews. Arch Intern Med 1996;156: Lipton RB, Stewart WF. Migraine in the United States: a review of epidemiology and health care use. Neurology 1993;43:S6 S10. Lipton RB, Stewart WF. Acute migraine therapy: do doctors understand what patients with migraine want from therapy? Headache 1999;39(Suppl 2):S20 S26. Lofland JH, Johnson NE, Batenhorst AS, Nash DB. Changes in resource use and outcomes for patients with migraine treated with sumatriptan: a managed care perspective. Arch Intern Med 1999;159: McQuay HJ, Moore RA. Using numerical results from systematic reviews in clinical practice. Ann Intern Med 1997;126: McQuay HJ, Moore RA. An evidence-based resource for pain relief, Oxford: Oxford University Press, Millson D, Tepper S, Dahlof CGH. Issues relating to the assessment of migraine recurrence following triptan therapy. Eur J Neurol 1999;6: Moher D, Pham B, Jones A, Cook DJ, Jadad AR, Moher M, Tugwell P, Klassen TP. Does quality of reports of randomised trials affect estimates of intervention efficacy reported in meta-analyses? Lancet 1998;352: Moore RA, McQuay HJ. Single-patient data meta-analysis of 3,453 postoperative patients: oral tramadol versus placebo, codeine and combination analgesics. Pain 1997;69: Moore RA, Gavaghan D, Tramer MR, Collins SL, McQuay HJ. Size is everything large amounts of information are needed to overcome random effects in estimating direction and magnitude of treatment effects. Pain 1998a;78: Moore RA, Tramèr MR, Carroll D, Wiffen PJ, McQuay HJ. Quantitive systematic review of topically-applied non-steroidal anti-inflammatory drugs. Br Med J 1998b;316: Morris JA, Gardner MJ. Calculating confidence intervals for relative risk, odds ratios and standardised ratios and rates. Statistics with confidence confidence intervals and statistical guidelines. Gardner MJ, Altman DG, editors. Br Med J 1995: Oldman AD, Smith LA, McQuay HJ, Moore RA. Rizatriptan for acute migraine (Cochrane Review). The Cochrane Library, Issue 3, Oxford, Osterhaus JT, Townsend RJ, Gandek B, Ware JE. Measuring the functional status and well-being of patients with migraine headache. Headache 1994;34: Pascual J, Vega P, Diener HC, Allen C, Vrijens F, Patel K. Comparison of rizatriptan 10 mg vs zolmitriptan 2.5 mg in the acute treatment of migraine. Cephalalgia 2000;20: Rasmussen BK, Jensen R, Schroll M, Olesen J. Epidemiology of headache in a general population a prevalence study. J Clin Epidemiol 1991;44: Schulz KF, Chalmers I, Hayes RJ, Altman DG. Empirical evidence of bias: dimensions of methodological quality associated with estimates of treatment effects in controlled trials. J Am Med Assoc 1995;273: Solomon GD, Price KL. Burden of migraine. A review of its socioeconomic impact. Pharmacoeconomics 1997;11:1 10. Stephens MDB, Routledge PA, Talbot J. The detection of new adverse drug reactions, 4th ed. New York, NY: MacMillan, Sterne JA, Gavaghan D, Egger M. Publication and related bias in metaanalysis. Power of statistical tests and prevalence in the literature. J Clin Epidemiol 2000;53: Tang J, Liu JL. Misleading funnel plot for detection of bias in meta-analysis. J Clin Epidemiol 2000;53: Tfelt-Hansen P. Sumatriptan for the treatment of migraine attacks a review of controlled clinical studies. Cephalalgia 1993;13: Tfelt-Hansen P. Efficacy and adverse events of subcutaneous, oral, and intranasal sumatriptan used for migraine treatment. A systematic review based on number needed to treat. Cephalalgia 1998;18: Tfelt-Hansen P, Saxena PR, Dahlof C, Pascual J, Lainez M, Henry P et al. Ergotamine in the acute treatment of migraine. A review and European consensus. Brain 2000;123:9 18. Tramèr MR, Reynolds DJM, Moore RA, McQuay HJ. Impact of covert duplicate publication on meta-analysis: a case study. Br Med J 1997;315: