Pizotifen for the treatment of migraine. A systematic review and meta-analysis

Introduction Pizotifen is an oral drug developed many years ago for the prophylaxis of migraine. Trials on pizotifen are decades old, and there has never been a systematic review and meta-analyses of data from these clinical studies. Methods This is a systematic review and meta-analyses on pizotifen's efficacy and safety for prophylactic migraine treatment. We considered for inclusion only randomized clinical trials (RCTs). A comprehensive electronic search was performed without language, date or publication status restrictions in the formal electronic databases, clinical trial registration platforms and grey literature. Results There were eight RCTs of pizotifen compared either to placebo or to other drugs. Very low certainty of evidence showed that pizotifen seems to be superior to placebo regarding clinical symptoms improvement (Relative risk [RR] 6.00; 95% Confidence interval [CI] 1.63 to 22.03; p = 0.007), but not inferior to naproxen, flunarizine, valproate or clonidine. Weight gain was the most frequent adverse event of pizotifen but there was no difference with placebo (RR 1.92; 95% CI 0.30 to 12.38; 2 RCTs; 142 participants; I2 = 67%; p = 0.49). Conclusion The RCTs of pizotifen were decades old. It is a safe and potentially efficacious inexpensive drug that deserves a better designed, modern clinical trial before being dismissed as an option for migraine therapy. PROSPERO Register: CRD42020194347. Yara Dadalti Fragoso Universidade Metropolitana de


Introduction
Pizotifen is an oral drug developed many years ago for the prophylaxis of migraine. Trials on pizotifen are decades old, and there has never been a systematic review and meta-analyses of data from these clinical studies. Methods This is a systematic review and meta-analyses on pizotifen's efficacy and safety for prophylactic migraine treatment. We considered for inclusion only randomized clinical trials (RCTs). A comprehensive electronic search was performed without language, date or publication status restrictions in the formal electronic databases, clinical trial registration platforms and grey literature.

Results
There were eight RCTs of pizotifen compared either to placebo or to other drugs. Very low certainty of evidence showed that pizotifen seems to be superior to placebo regarding clinical symptoms improvement (Relative risk [RR] 6.00; 95% Confidence interval [CI] 1.63 to 22.03; p = 0.007), but not inferior to naproxen, flunarizine, valproate or clonidine. Weight gain was the most frequent adverse event of pizotifen but there was no difference with placebo (RR 1.92; 95% CI 0.30 to 12.38; 2 RCTs; 142 participants; I2 = 67%; p = 0.49).

Conclusion
The RCTs of pizotifen were decades old. It is a safe and potentially efficacious inexpensive drug that deserves a better designed, modern clinical trial before being dismissed as an option for migraine therapy. PROSPERO Register: CRD42020194347.

Introduction
M igraine affects over 10% of adults and can limit their activities both at home and at work. 1 The socioeconomic costs of migraine are remarkably high, and patients tend to overuse medications for the pain if the headache is not properly treated. 2 This disease is an important cause of absenteeism and presentism at work, missed days at school and excessive numbers of medical consultations and examinations. 3 Although several medications have been used for prophylaxis of migraine attacks, only two classes of drugs presently commercialized have been specifically developed for this condition. One of the drugs is the relatively old pizotifen and the other is the monoclonal antibody (MAb) anti-CGRP. Small oral molecules with anti-CGRP effect are undergoing clinical trials.
Anti-CGRP monoclonal antibodies have a strong placebo and weak nocebo effect. 4 This, in addition to their good safety profile [Hou], makes these drugs successful in controlling migraine. The complicating aspect of this treatment is its cost of circa USD 200/month per patient. In places with high per capita income and/or in countries with healthcare systems providing reimbursement of the drug, use of anti-CGRP MAbs can thrive.
Pizotifen, on the other hand, costs little. The only bothersome adverse event caused by pizotifen is weight gain, which can be tolerated by some patients if they know what to expect. It is hard to find pharmacies selling branded pizotifen: it is mostly available online and at compounding pharmacies. Pizotifen may decrease migraine attacks at a fraction of the price of some other drugs. There has never been a systematic review and meta-analysis on pizotifen for the treatment of migraine.
Apart from pizotifen and anti-CGRP MAbs, other drugs are used in migraine prophylaxis, like tricyclic antidepressants, calcium channel blockers, betablockers and anticonvulsants. For all these drugs, the profile of adverse events is worse than that of pizotifen or anti-CGRP MAbs.
Should pizotifen prove to be an efficient and safe prophylactic drug for migraine, many individuals who cannot afford the expensive new therapy could benefit from the older one. Thus, the objective of this systematic review was to assess the effects (benefits and harms) of pizotifen for treating migraine in adults.

Methods
This systematic review followed the methodological recommendations of the Cochrane Handbook for Systematic Reviews of Interventions 5 and the PRISMA statement (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) to ensure the quality of the report. 6 This systematic review protocol was prospectively registered in the PROSPERO (International Prospective Register of Systematic Reviews) platform, under the number CRD42020194347.

Criteria for including studies for this review
We considered randomized clinical trials (RCTs) with parallel design, assessing the effects (benefits and harms) of pizotifen for treating migraine in adults (over 18 years), who reported episodic or chronic migraine with or without aura. Studies that included any other headache condition were excluded if data were not presented separately for patients with migraine. The RCTs included compared any dose or scheme of pizotifen with placebo, no intervention, or another active drug treatment.

Types of outcome measurements
Primary outcomes -Reduction of frequency and/or severity and/or duration of migraine attacks.
-Reduction of medications taken to treat a migraine attack.
-Adverse events: proportion of participants with at least one adverse event resulting from the use of pizotifen (for example, any gastrointestinal events or allergy). Secondary outcomes: -Patients' satisfaction and preferences.
-Tolerability of weight gain. We considered all time points reported by the RCTs, but we only pooled similar time points: short term (up to 8 weeks of treatment), intermediate-term (9 to 16 weeks of treatment) and long term (over 16 weeks of treatment).

Data collection and analysis
Selection of studies and data extraction All titles and abstracts obtained through the search strategies were included in a reference management program (Endnote ) where duplicates were removed. Two authors independently selected titles and abstracts of the references retrieved using the software Rayyan. 7 All references classified as 'potentially eligible' were read in full text to confirm their eligibility.
The data extraction process was carried out by three independent authors using a pre-established data extraction form. Two other authors resolved all discordance in the selection and extraction process. When necessary, the authors of the trials included in the review were contacted for additional information.
Risk of bias (quality) assessment Two independent authors assessed the methodological quality of the studies included using the Cochrane Risk of Bias (RoB) table 5 , which classify each of the following domains as presenting a high, low or unclear risk of bias: (1) random sequence generation; (2) allocation concealment; (3) blinding of participants and personnel; (4) blinding of outcome assessors; (5) incomplete outcome data; (6) selective reporting of outcomes; and (7) other potential sources of bias (for example, baseline imbalances). A third author was consulted in cases of disagreement.

Data synthesis
We planned to calculate risk ratios (RR) for dichotomous outcomes and mean differences (MD) for continuous outcomes, or the standard mean difference (SMD), if any outcome of interest had been measured using different scales or questionnaires (95% confidence interval). When possible (if data were available and homogeneous), treatment effects were combined using a randomeffects model meta-analysis in the Review Manager 5.4.1 software. The heterogeneity between the studies included was evaluated according to the clinical and methodological characteristics. Statistical heterogeneity was assessed through visual inspection of forest plots. We calculated a chi² test, considering p > 0.1 as indicative of statistical heterogeneity, and an I² test for measuring inconsistency across studies (we defined I² > 50% as indicative of significant inconsistency). 8 Subgroup and sensitivity analysis We planned to assess subgroups for all primary outcomes, comparing separated results between adults and children. We also planned to conduct a sensitivity analysis in which RCTs with a high risk of bias (selection, detection and attrition bias) would be excluded from the meta-analysis.
Publication bias assessment Publication bias would be investigated through analysis of funnel plots if there had been meta-analyses with at least ten studies.
Assessment of the certainty of the evidence Two independent authors assessed the certainty of the body of evidence for primary outcomes from the main comparison: pizotifen versus placebo, using the GRADE approach (Grading of Recommendations, Assessment, Development and Evaluations). 9 We summarized the evidence in the 'Summary of findings table' (SoF table) through the GRADEpro GDT platform.

Search results
The results from the search strategies, retrieved 749 records. After removing 104 duplicates, 645 were analyzed using the title and abstract. Of these, 623 did not fulfil the inclusion criteria and were excluded. Twentytwo studies were analyzed using the full text, and nine [10][11][12][13][14][15][16][17][18] were excluded (crossover trials). Five studies [19][20][21][22][23] were classified as 'awaiting classification studies' because their randomization was not clear, and we did not find the authors' contact details to request additional information. Contacts with the editors of the journals in which these papers were published were fruitless. Therefore, in the end, eight RCTs were included in this review. [24][25][26][27][28][29][30][31] Figure 1 summarizes the study selection process.  Characteristics of the studies included Table 1 summarizes the main characteristics of the eight RCTs included.  Regarding selection bias, only one study 31 reported using an adequate method for generating the randomization sequence, and this study was judged as presenting a low risk of bias. In the other studies, insufficient information was provided, and these were classified as having an unclear risk of bias.
Regarding performance bias, three studies 24,25,28 were classified as having high risk of bias because they were single-blinded. Only one study 27 was considered as having a low risk of bias regarding blinding of participants, personnel and outcome assessors.
Four studies [24][25][26]29 had substantial losses of participants during the study (16 to 44%) and were classified as presenting high risk of attrition bias. Trial register protocols were not available for any of the studies included, leading to an overall unclear risk of reporting bias. Lastly, two studies 27,29 did not describe the baseline characteristics between groups and were judged as having unclear risk of other sources of bias.

Effects of interventions
Comparison 1. Pizotifen versus placebo -Frequency, intensity and duration of episodes Two studies 26,29 evaluated these outcomes, and it was not feasible to pool their results in a meta-analysis due to their clinical diversity and the unavailability of data. When possible, estimated effects were calculated using individual study data.
Bellavance et al. 26 (Figure 3). Moreover, there was a slight statistical heterogeneity (67%), which can be explained as possible clinical differences between participants, including in the treatment scheme with pizotifen.
-Patients' overall evaluation In one study 26 , good or excellent ratings in the patients' overall evaluation were reported by 68% of pizotifen-treated participants and 36% of placebo-treated participants (p = 0.005).

Comparison 2: Pizotifen versus flunarizine -Frequency, intensity and duration of episodes
Regarding the severity of episodes measured on a pain intensity scale, in one study 27 an improvement of 81% in the flunarizine group was observed, compared with 40% in the pizotifen group (p < 0.01). The episodes' duration was not significantly changed by either pizotifen or flunarizine (mean 8.5 versus 31, respectively). Louis et al. 30 used a 4-point scale to assess the severity of the episodes, and no difference was noted between pizotifen and flunarizine, considering the number of participants with migraine grade 1 (mild) after four months of treatment (RR 0.14; 95% CI 0.02 to 1.06; p = 0.06). This effect was uncertain, given the breadth of the confidence interval. These authors also stated that the episodes' duration was not changed by either drug (no numerical data provided).
Three studies 27,30,31 assessed these outcomes, but it was not possible to group their data in a meta-analysis or calculate the estimated effects for most individual studies due to missing numerical data.
One study 31 (75 participants) presented a mean reduction in the number of migraine episodes of 54% for the flunarizine group and 45% for the pizotifen group after four months. The authors stated that there was a significant difference between the groups (p < 0.001). Cerbo et al. 27 (30 participants) reported that there was no difference between drug treatments regarding the mean reduction in the number of episodes per month (mean 2.67 pizotifen versus 3.56 flunarizine). Rascol et al. 31 (35 participants) reported that the reduction in the flunarizine group (65%) was slightly greater than in the pizotifen group (45%), but the intergroup difference was not significant after two months of treatment (p = 0.10).
In two studies 27,30 , daytime sedation among the participants in both groups was reported. The results from a metaanalysis showed no difference between groups (RR 0.49; 95% CI 0.24 to 1.01; 2 RCTs; 105 participants; I 2 = 0%; p = 0.48). Rascol et al. 31 found no difference between the groups regarding occurrences of other adverse events, including changes in blood pressure, hot flushes, drowsiness and severe asthenia (1/14 versus 2/15; RR 2.00; 95% CI 0.20 to 19.78; p = 0.55). However, both effect estimates had wide confidence intervals and were

ASAA
Fragoso YD, Lopes GCM, Santos GM, Carturan P, Martimbianco ALC Pizotifen for the treatment of migraine. A systematic review and meta-analysis highly imprecise.

Patients' overall evaluation
The participants in two studies 30,31 reported that the effects from both treatments were positive, but there was no significant difference between the groups after 4 months of treatment (RR 0.90; 95% CI 0.61 to 1.32; 2 RCTs; I2 = 0%; 101 participants; p = 0.40) ( Figure 5).
In two studies 26,27 , other adverse events relating to the treatments were reported. These included nausea, Comparison 3: Pizotifen versus naproxen -Frequency, intensity and duration of episodes Two studies 25,26 assessed these outcomes, but it was not possible to group their data in a meta-analysis because the outcomes were measured using different methods. In one study 25 (67 participants), no significant differences between the groups were found regarding frequency (MD -0.10; 95% CI -0.68 to 0.48) or severity of episodes (MD -0.20; 95% CI -0.93 to 0.53), after 3 months of treatment. There were no significant differences between the groups regarding the duration of attacks or use of rescue medication (no numerical data provided).
In another study 27  -Adverse events vomiting and gastrointestinal effects. However, there were no differences between the groups observed. -Adverse events The participants in the study by Behan et al. 24 presented weight gains. However, no numerical data were provided, and the authors reported that only in one case, was it sufficient to cause the patient's withdrawal from the study.
No adverse events were seen in the clonidine group.

Certainty of the evidence
Based on the GRADE approach, the certainty the evidence regarding pizotifen versus placebo was classified as 'very low'. It indicated that we had little confidence in the effect estimate. The evidence was downgraded due to methodological limitations, inconsistency (substantial heterogeneity between studies) and imprecision (wide confidence intervals and small sample size). The findings are summarized in a table of assessment details, which is presented as Supplementary Material 3.

Discussion
Migraine is a burden for patients and society. The socioeconomic costs of migraine are immense 32 , with average direct costs of €2427/patient/year in the USA and Canada 33 and €1222 to €1482/patient/year in Europe. 34,35 These societies will invest in newer drugs and will reimburse the costs of anti-CGRP MAbs so that migraineurs will have lower financial burdens and less absenteeism and presenteeism. The same cannot be said for developing countries, where treating migraine at the cost of circa €1600/patient/year is unrealistic. Poorer countries require efficient but inexpensive drugs that provide effective migraine treatments. For this situation, pizotifen may be an alternative. This systematic review has shown that pizotifen is superior to placebo and not inferior to naproxen, flunarizine, valproate or clonidine. The ensuing weight gain is not tolerated by all patients 36 , but it may be acceptable in some cases and situations. An additional benefit of pizotifen is a very safe profile in pregnancy cases (category B in the Food and Drug Administration -FDA). All other prophylactic migraine treatments are at least category C by the FDA. 37 The adverse events reported in pizotifen trials are essentially restricted to weight gain. The profile of adverse events of most prophylactic drugs for migraine leads to high rates of discontinuation. 38 Even for erunumab, an anti-CGRP MAb, the discontinuation rate over six months was 27.7% in real life. 39 Trials on pizotifen were conducted decades ago, with low numbers of participants and unclear risks of bias. The patients with migraine included in most pizotifen studies were not selected to take the International Headache Society criteria into consideration criteria. 40 It is unlikely that an inexpensive drug-like pizotifen will be tested in modern trials. In an ideal scenario, a new trial should compare pizotifen's efficacy and safety compared to another drug.
Despite the limitations mentioned above, we believe that these meta-analyses' results are sufficient for pizotifen to be considered in treatments for migraine for patients who are not preoccupied with weight gain.

Conclusion
Pizotifen was superior to placebo and not inferior to naproxen, flunarizine, valproate or clonidine, for treating migraine. The adverse events from the use of pizotifen were restricted to weight gain. The clinical trials on pizotifen are now a few decades old, and the body of evidence was classified as 'very low' due to methodological limitations.
A new trial in the 21 st century could render good evidence for this inexpensive drug's safety and efficacy.

Conflicts of interest/competing interests: none.
Authors contribution: all authors contributed to the study concept and design and participated in searching and selecting articles. ALCM, carried out meta-analyses; YDF, paper draft and final version.