Beta-blockers for the prevention of headache in adults, a systematic review and meta-analysis

Background Headaches are a common source of pain and suffering. The study’s purpose was to assess beta-blockers efficacy in preventing migraine and tension-type headache. Methods Cochrane Register of Controlled Trials; MEDLINE; EMBASE; ISI Web of Science, clinical trial registries, CNKI, Wanfang and CQVIP were searched through 21 August 2018, for randomized trials in which at least one comparison was a beta-blocker for the prevention of migraine or tension-type headache in adults. The primary outcome, headache frequency per month, was extracted in duplicate and pooled using random effects models. Data synthesis This study included 108 randomized controlled trials, 50 placebo-controlled and 58 comparative effectiveness trials. Compared to placebo, propranolol reduced episodic migraine headaches by 1.5 headaches/month at 8 weeks (95% CI: -2.3 to -0.65) and was more likely to reduce headaches by 50% (RR: 1.4, 95% CI: 1.1–1.7). Trial Sequential Analysis (TSA) found that these outcomes were unlikely to be due to a Type I error. A network analysis suggested that beta-blocker’s benefit for episodic migraines may be a class effect. Trials comparing beta-blockers to other interventions were largely single, underpowered trials. Propranolol was comparable to other medications known to be effective including flunarizine, topiramate and valproate. For chronic migraine, propranolol was more likely to reduce headaches by at least 50% (RR: 2.0, 95% CI: 1.0–4.3). There was only one trial of beta-blockers for tension-type headache. Conclusions There is high quality evidence that propranolol is better than placebo for episodic migraine headache. Other comparisons were underpowered, rated as low-quality based on only including single trials, making definitive conclusions about comparative effectiveness impossible. There were few trials examining beta-blocker effectiveness for chronic migraine or tension-type headache though there was limited evidence of benefit. Registration Prospero (ID: CRD42017050335).

Tension-type headache is more common than migraine; up to 90% of adults experience one at some time in their life [18][19][20][21][22]. In any given month, a tension-type headache occurs in 46% of adults [22]. Most tension-type headaches are managed with over the counter medications, consequently most do not seek medical attention. However, tension-type headache reduces the quality of life [23], results in up to a fifth of all missed work days [24], and costs EUR 21 billion annually in Europe [25].
The 2012 American Academy of Neurology guideline recommends beta-blockers, specifically propranolol and metoprolol, as first line therapy for preventing migraines [34]. Specific medications commonly used in prophylaxis has not been well described. In Europe, commonly prescribed prophylactic agents include antiepileptics, beta-blockers, flunarizine, pizotifen and TCAs [35]. Other studies found that specialists are twice as likely to prescribe antiepileptics than primary care providers [36], that treatment persistence is low [37] and that use of prophylactic medications has increased [38], though none of these three characterized the specific medications used.
The purpose of this study is to assess the efficacy of beta-blockers in the prophylaxis of migraine and tension-type headache. Two previous systematic reviews focused on the use of beta-blockers in migraine headaches, both are more than 15 years old [39;40], and included limited outcomes, though both suggest benefit of beta-blockers compared to placebo. There are two more recent comparative effectiveness analyses of headache management that included beta-blockers. Shamliyan reviewed pharmacologic treatment for episodic migraine (V15.1 College Station TX). A priori, the analytic plan was to pool data at 4, 8, 12, 16, 20 and 24 weeks. Studies that reported outcomes at different time points were combined at the closest time point available (for example, 9-week outcomes were pooled with the 8-week group). For comparative effectiveness trials with 2 or fewer studies, outcomes were reported at the last point reported. Small study effects (publication bias) was assessed using the methods of Peters [52] for dichotomous outcomes and Egger [53] for continuous outcomes if there were a sufficient number of studies. Trial sequential analysis was performed for the comparisons of propranolol to placebo for headache frequency (at 8 weeks), using the O'Brien-Fleming method of alpha-spending function to robustness of the pooled estimates against type 1 and type 2 error [54], using TSA software (Copenhagen, Denmark).
A network meta-analysis was performed for beta-blockers that were compared with placebo at 8 and 12 weeks using the residual maximum likelihood with a modification to the coefficients' estimated variance using the Kapp and Hartung approach [55] that had a minimum of 2 studies. Both 8-and 12-week results were pooled including all beta-blockers using multivariate random-effects meta-analysis using the network package in STATA [56].
Finally, the quality of evidence was assessed using the GRADE (Grades of Recommendation, Assessment, Development and Evaluation) system to rate the quality of the evidence (GRADEPro GDT 2015) following Cochrane guidelines [45]. Grade assesses quality in four levels: High (further research is unlikely to change estimate of effect); Moderate (further research may impact effect estimate); Low (further research is likely to have important impact on estimate); Very low (any estimate of effect is very uncertain).
The 108 included studies ranged from 4 to 64 weeks in duration (average: 12.9). Fifty-one were parallel in design and 57 had a crossover design. Among crossover trials, 43 were randomized, with washout periods ranging from zero to four weeks. Twenty-five different countries (Table 1) and four languages (Chinese (n = 15), English (n = 86), German (n = 6), Polish (n = 1) were represented. The average age was 38.6 years and 77% of participants were women. Ten different beta-blockers were studied. Propranolol (n = 74) and metoprolol (n = 21) were the most commonly evaluated beta-blockers. Atenolol, nadolol, pindolol and timolol had two studies each. Several beta-blockers were evaluated in only a single trial (acebutolol, alprenolol, bisoprolol and oxprenolol). Study characteristics for included trials are provided in Table 1 and quality ratings are given in S2 Table. Studies had a number of common quality problems (S3 Table) including high drop-out rates (16.1%, range 0-51%), lack of intention to treat analysis (76%), inadequate sequence generation (83%), lack of evidence of concealed allocation (90%) and inadequate blinding (60%). Twenty-three studies assessed compliance (21%). Fifty-one (47%) studies reported all collected outcomes. Sixteen trials (15%) were sponsored by industry. All comparisons that had only a single study were graded as low-quality evidence.

Episodic migraines
The primary outcome was headaches per month. Outcomes from placebo-controlled trials for all beta-blockers and time-points are provided in Tables 2-3. Outcomes at 8 weeks was the most commonly reported time point. Among patients with episodic migraines (Table 2), the average number of headaches at baseline was 4.9 headaches/month (95% CI: 4.4-5.4). The best studied beta-blocker was propranolol, which was more effective than placebo at 8 and 12 weeks (8 weeks: -1.5 ha/month, 95% CI: -2.3 to -0.65); 12 weeks: -1.2 ha/month, 95% CI: -1.8 to -0.60, Fig 2). Propranolol outcomes at 8 and 12 weeks were both graded as high-quality evidence. TSA analysis of propranolol vs. placebo for headache frequency demonstrated that it is unlikely that these results are due to a Type 1 error (Fig 3). Other beta-blockers that were more effective than placebo at 8 weeks (Fig 4) included bisoprolol (-0.70 ha/month, 95% CI: -1.4 to -0.05, low quality), metoprolol (-0.86 ha/month, 95% CI: -1.4 to -0.34, moderate quality) and Baseline Frequency 0.00 (-0.14 to 0.14) -Low timolol (-0.77 ha/month, 95% CI: -1.4 to -0.12, moderate quality). The remaining beta-blockers, in single trials did not significantly reduce headache frequency (Fig 3). There was a similar pattern at twelve weeks. Among secondary outcomes, the majority of trials studied subjects with episodic migraine headaches (Table 3). Propranolol was the most commonly studied beta-blocker. Propranolol was more likely to reduce headaches by 50% than placebo at 12 weeks (RR: 1.4, 95% CI: 1. Network meta-analysis. For the primary outcome, headache frequency, the network meta-analysis found no difference at 8 weeks (p = 0.27) in effectiveness in comparisons between propranolol (n = 9) compared to bisoprolol (n = 2), metoprolol (n = 3) and nadolol (n = 3). Similarly, at 12 weeks, there was no difference (p = 0.84) n effectiveness in comparisons between propranolol (n = 9) compared to bisoprolol (n = 2), nadolol (n = 3) and timolol (n = 2). The 8-and 12-week analysis confirmed this lack of difference between all beta-blockers (Fig 5), including those with single trials (Fig 6) Comparative effectiveness trials. There were 83 randomized trials that included at least one comparison to a non-placebo treatment. Propranolol was the most commonly compared beta-blocker (n = 72, 87%). Propranolol was compared to pharmacologic interventions in 50 trials. Eleven comparisons that were from single trials (5-hydroxytryptophan, aspirin, atenolol, candesartan, clonidine, cyclandelate, mefenamic acid, naproxen, nifedipine, nimodipine, nortriptyline, pindolol, riboflavin). Comparisons with more than one study included: amitriptyline (n = 2), femoxetine (n = 2), flunarizine (n = 11), metoprolol (n = 4), nadolol (n = 3), nimodipine (n = 2), timolol (n = 2) tofenamic acid (n = 2), topiramate (n = 3) and valproate (n = 3). Several combinations were tested including propranolol + amitriptyline vs.  amitriptyline (n = 1), and propranolol + flunarizine was compared to flunarizine. Propranolol combined with topiramate was compared with topiramate alone. Nonpharmacologic interventions were compared to propranolol in three trials (acupuncture, behavioral management, biofeedback). Metoprolol was assessed in 14 trials, 8 were comparisons with pharmacologic interventions (aspirin, bisoprolol, clomipramine, clonidine, flunarizine, nebivolol, pizotifen) and one with nonpharmacologic (acupuncture). Metoprolol was compared to a combination of metoprolol and fluoxetine in 2 trials and the combination of metoprolol and flunarizine was compared to flunarizine alone in 2 trials. The primary outcome comparison (headache frequency) is provided in Table 4, and secondary outcomes are in Table 5. The majority of comparisons were single trials, making definitive conclusions difficult. Metoprolol, but not propranolol, was more effective than aspirin. Metoprolol was more effective than clomipramine, though comparable to acupuncture, bisoprolol, clonidine, flunarizine, nebivolol and pizotifen. Adding fluoxetine to metoprolol or flunarizine to either propranolol or metoprolol did not improve headache frequency. All comparisons were graded as low-quality. Propranolol was more effective than femoxetine, mefenamic acid, naproxen, nifedipine and very low-dose (40mg) nortriptyline, but comparable to 5-hydroxytryptophan, acupuncture, atenolol, behavioral management, biofeedback, candesartan, clonidine, cyclandelate, flunarizine, metoprolol, nadolol, naproxen, nimodipine, riboflavin, timolol, tolfenamic acid, topiramate and valproic acid. All comparisons were single trials and were rated as low-quality evidence with the exception of the comparisons to flunarizine and metoprolol at 8 weeks that were graded as moderate or high-quality. The network meta-analysis confirmed these findings, but suggested that metoprolol was also superior to naproxen (SMD: -1.2, 95% CI: -1.6 to -0.78).

Chronic migraine
There were four trials that evaluated beta-blockers for chronic migraine headaches, none were placebo controlled. (Table 6). Propranolol was compared to flunarizine [151], nortriptyline [76], valproic acid [105] and to the combination of propranolol and flunarizine [151]. In addition, a combination of propranolol and topiramate was compared to topiramate alone [129]. Propranolol was no better than valproic acid or flunarizine and the combinations (propranolol Beta-blockers for preventing headaches + topiramate and propranolol + flunarizine) was no better than topiramate and flunarizine alone ( Table 6).

Tension-type headache
There was only one trial evaluating tension-type headache, comparing the combination of pindolol and amitriptyline to placebo and to amitriptyline alone [57]. The combination of pindolol and amitriptyline was more effective than placebo at reducing headache frequency at 4 and 8 weeks (Table 6) and in reducing headaches by at least 50% (RR: 3.8, 95% CI: 1.5-9.3), but equally effective with amitriptyline.
Propranolol was the only beta-blocker with sufficient numbers of studies to perform sensitivity analysis. There was no evidence of publication bias for propranolol's reduction of headache frequency at 8 weeks (Egger's p = 0.77) or at 12 weeks (p = 0.62). There was no evidence of an effect of quality (p = 0.97), age (p = 0.71), percent women (p = 0.28), percentage of dropouts (p = 0.55), dose (p = 0.61), intention to treat analysis (p = 0.35), concealed allocation (p = 0.38) or appropriateness of blinding (p = 0.98).

Discussion
This review included one hundred and eight randomized controlled trials. Nearly all evaluated the efficacy of beta-blockers for episodic migraine headaches. Compared to placebo, Beta-blockers for preventing headaches propranolol was effective in reducing episodic migraine frequency. The effect began as early as four weeks. Migraine headache sufferers experienced an average reduction of 1.3 headaches/ month; this translates to a reduction from 4.8 to 3.5 headaches a month. Subjects given propranolol were more likely to report at least 50% reduction in headaches and to reduce their consumption of analgesic medications. In addition to reducing the number of headaches, the residual headaches were less severe and shorter in duration compared to those receiving placebo. Outcomes from the propranolol comparisons to placebo were rated as high-quality evidence. In three trials, metoprolol also reduced headache frequency, though the reduction was less than 1 headache a month. Conclusions regarding the efficacy of other beta-blockers is less certain, as most were studied in just one trial each. Atenolol, bisoprolol and timolol had weak evidence of benefit. Acebutolol, alprenolol and nadolol appeared to be ineffective in migraine prophylaxis. This is unlikely to be due to properties of the beta-blockers. Propranolol is nonselective as is alprenolol and nadolol. Metoprolol, also effective is a ß-1 selective drug as is atenolol, bisoprolol and acebutolol. Given that acebutolol, alprenolol and nadolol were only studied in one trial each, it is possible that this may be either random variation in outcomes or a Beta-blockers for preventing headaches    problem with the trials (such as dose or duration). The network analysis suggests that the benefit of beta-blocker may be a class effect.

Baseline
The literature comparing beta-blockers to other modalities consisted mostly of single-trials with the exception of the comparison of propranolol to metoprolol (moderate quality, no difference) and to flunarizine (high quality, no difference). Flunarizine, not available in the Beta-blockers for preventing headaches          Beta-blockers for preventing headaches United States, is well-established as effective in treating migraine headache. Universally, betablockers were associated with bradycardia and with lower average pulse rates than placebo trials. This is not surprising given their impact on chronotropy. Other side effects more common among participants taking beta-blockers included dizziness and fatigue, though subjects on beta-blockers were not more likely to withdraw from the studies. While these conclusions are similar to previous reviews, this analysis is a unique contribution in many ways, first it included many more trials than previously reported, including Chinese trials that had previously not been included. Secondly, all beta-blockers were carefully parsed by type of headache (tension v. migraine, episodic v. chronic). Third, this study examined outcomes at the specific times reported, it is common for meta-analyses to pool trials at the last time point, regardless of whether there were significant differences in that time-point. Fourth, trial sequential analysis that demonstrated the adequacy of the current database for propranolol, suggesting that there is low likelihood of type 1 error in the conclusions. Fifth, the  Beta-blockers for preventing headaches network meta-analysis didn't show clear benefit of one beta-blocker over another, suggesting a class effect, though other beta-blockers have weaker evidence for benefit. An important question, unanswered in this review, is how effective beta-blockers are compared to other commonly used prophylactic regimens. Propranolol, metoprolol and flunarizine appear to have similar efficacy. The other comparisons were all single-trial comparisons, making definitive conclusions impossible. In a previous review, tricyclic antidepressants resulted in a reduction in headache frequency for patients experiencing chronic migraines of 1.3 SMD, compared to placebo, a large effect [17]. In this study, there was only had one trial on chronic migraines, and the calculated SMD was 0.58, about half of the effect previously reported for TCAs. In a network meta-analysis of chronic migraines, tricyclic antidepressants were more effective than propranolol [17] but propranolol was similar in efficacy to antiepileptics and flunarizine, similar to findings in this study, though it is important to note that the majority of trials for beta-blockers are for episodic rather than chronic migraines. Definitive answer to this comparative effectiveness question requires additional studies that directly compare the different prophylactic modalities.
This review has several important limitations. First, while the quality of evidence for the comparison between propranolol and placebo was high, in general, the remaining comparisons were of low quality, consisting mostly of underpowered single randomized trials. While all the included trials were randomized, there were significant methodologic problems; combining poorly designed studies can lead to questionable results. It is important to note that most of the comparisons were graded as being of low-quality evidence because of the paucity of studies and concern about study problems. Secondly, studies were inconsistent in reporting outcomes, so even when there were more than one trial available, specific outcomes may only be provided by a single study. Moreover, there were significant problems with selective reporting of outcomes and many studies did not collect information on headache frequency, the measure preferred by the International Headache Society. Third, the number of studies available precluded sensitivity analyses, such as assessing for publication bias or exploring sources of heterogeneity. Fourth, beta-blockers have been studied almost exclusively in the management of episodic migraine headaches. Their benefit for chronic episodic or tension-type headaches is uncertain. Fifth, because of the paucity of trials for most beta-blockers the network analysis was underpowered to show differences between beta-blockers, Sixth, the average age of participants was 38, and mostly female. While this reflects the demographics of headache, it limits applicability to older adults.

Conclusions
Propranolol is effective in reducing the burden of patients with episodic migraine headaches, reducing headaches from 5 to 3 headaches a month. This means that migraine sufferers given propranolol will have substantial residual headache burden. Propranolol reduces headaches by more than 50% as well as reducing the number of analgesic medication doses required. It also reduces the severity or duration of the headaches experienced. Propranolol and metoprolol exert similar effects and propranolol is as effective as flunarizine. The data for other betablockers and other comparisons are less clear.