Placebo responses vary by route of administration in migraine prevention trials. What does this mean?

In this editorial we present a discussion on the interpretation of the results of trials on prophylactic treatments. The aim of prophylactic treatments is twofold, to limit the burden of various diseases and, if possible, change their natural course. However, the way in which we measure the effect size of prophylactic treatments varies significantly across conditions, depending on their morbidity, mortality, symptoms, and quality of life impact, but mainly depending on the way the specific condition manifests, i.e. its symptoms and its natural course. In the case of migraine, a polygenic brain disease that is the second leading cause of disability worldwide, early introduction of migraine prophylactic treatment is crucial to limit attack frequency and delay chronification and involvement of comorbidities, that make it intractable and difficult to treat¹. The International Headache Society (IHS) recommends that testing of emerging prophylactic treatments for migraine should include reduction in migraine attack frequency, among other efficacy parameters, as a primary efficacy endpoint. For the time being, and even though we already have good evidence that several prophylactic treatments significantly reduce the frequency of migraines, the IHS also recommends comparing new drug treatments in clinical trials with a placebo^2–4. The use of placebo (s) allows us to assess the effectiveness of a treatment over the positive expectation we all have that a therapeutic action will do more good than harm, which is called in medicine placebo effect. In contrast, placebo, or placebos in plural, refer to the inert substances or interventions used to induce a positive expectation⁵. The way we evaluate the positive predisposition in a clinical trial is influenced by several parameters and finally we are monitoring the placebo response, which contains, in addition to the placebo effect, random beneficial changes that would occur in a patient without the effect of any treatment, within the framework of the natural history of the disease. Another important cofounder is the return to the average state of the condition, the so-called regression toward the mean. In clinical trials we usually recruit people with an exacerbation of the disease (as the inclusion criteria demand) which is normally expected to regress toward the mean after some time, also without the intervention of any treatment. All those beneficial changes observed after the application of placebo are collectively referred as placebo response and they are identifiable in clinical trials, by assessing the proportion of participants who achieved the primary efficacy endpoint in the placebo-treated arm⁵. Interestingly, placebo responses in trials for migraine prophylaxis have increased over the past 30 years, making the comparison of the novel treatments with placebo more imperative than it used to be⁶. Several factors might influence placebo effect and response including the gender, age, and education of a patient, the price of the drug, the status of the doctor recommending a treatment, the patient’s previous experiences, the route of administation, among several other modifiable, or non-modifiable factors⁵. Previous studies have shown that the placebo responses in anti-migraine preventative medications given parenterally were greater than in studies testing oral medications^7–9. Whether this increased placebo response ultimately narrowed the therapeutic gain (=treatment response minus placebo response) in a study and limited the statistical verification of a therapeutic effect remains unclear, but it is a hypothesis that one should investigate. We knew from studies with botulinumtoxinΑ for example, that placebo response is increased compared to studies testing oral drugs⁷. And the same occurred when placebo responses in trials for the prevention of migraine with monoclonal antibodies targeting the calcitonin gene related peptide (anti-CGRP mAbs) were compared to placebo responses in trials with oral anti-migraine treatments^9,10. Among anti-CGRP mABs three are administered subcutaneously (sc, erenumab, fremanegumab, galcanezumab) and one intravenously (iv, eptinezumab). It is therefore interesting to study whether there is any significant difference in placebo response between treatments administered iv or sc in migraine preventative trials. One recent meta-analysis revealed that an iv route of administration for the placebo was a predictor for higher reduction in monthly migraine days, e.g. higher placebo response¹¹. Whether this has a relevant clinical importance remains unclear, but it may influence indirect comparisons, between anti-CGRP mAbs, and thus impact decision making.

There are several metrics to indirectly compare clinical studies, which all share important limitations. The therapeutic gain or the absolute rate (= treatment response-placebo response) is commonly used for indirect comparisons currently. One may argue that this method may underestimate treatments given iv¹¹. To minimize the influence of the magnitude of placebo response when making cross-trial comparisons, absolute rates for one outcome along with relative rates should be considered as complementary metrics¹². To calculate the relative rate (RR), the rate of the beneficial outcome in the treated group is divided by the same rate in the control group, and the result is expressed as a percent reduction or increase. Neither the relative rates nor the absolute rates method is preferable to the other. It is considered that in case both metrics point toward to the same conclusion, that conclusion is likely to be correct. In case the two methods differ in their results, or even they support opposite conclusions, they cannot be used for safe comparative assumptions. Notably, indirect comparisons cannot replace head-to-head comparisons, in any case¹². For example, in the PROMISE 1 trial with eptinezumab 30 mg, 100 mg and 300 mg administered iv every three months, the placebo response rate for 50% RR (= the proportion of participants who achieved a 50% reduction in monthly migraine days compared to baseline) was 37.4% and the response rates for the treated groups were 50.2%, 49.8% and 56.3% respectively¹³. In the HALO trial with sc administered fremanazumab 225 mg monthly, or 675 mg quarterly, the placebo response rate for 50% RR was 27.9% and the response rates in the treated groups were 47.7% and 44.4%¹⁴.

In PROMISE-1 trial the absolute rates (or therapeutic gains) for the 50% RR in the three treated groups of participants were 12.8%, 12.4% and 18.9% respectively, and the relative rates 1.34 (34%), 1.33 (33%) and 1.51 (51%). In HALO trial, the absolute rates for the same outcome for monthly and quarterly fremanezumab were 19.8% and 16.5% and the relative rates 1.7 (71 morbidity, %) and 1.59 (59%). In both trials treatment lasted three months, but the placebo response rates differ by nearly 10 percentage points (37.4% in PROMISE-1 vs. 27.9% in HALO trial). In this paradigm, both methods slightly favor fremanezumab but not for all doses used, indicating that the indirect comparison cannot be clearly conclusive, since a higher difference in the placebo response could result in different interpretations. Thus, placebo response is crucial for both absolute and relative rates but in an opposite direction. Low placebo responses may exaggerate the relative rate and high placebo responses may exaggerate the absolute rates, and that is why both methods should be used to normalize potential differences in the population included in the trials for comparison. Because absolute and relative rates are metrics and do not represent natural values, statistics are inappropriate, due to systemic and random errors. Furthermore, these differences in the above paradigm do not appear to be clinically relevant.

Other cross-trial methods including net-work meta-analyses, are also available and in use, to help in decision making. But so far, in cross-trial comparisons of migraine prophylactic treatments, the observed difference in the size of placebo response between intravenously and subcutaneously administered anti-CGRP mAbs is not large enough to affect indirect comparisons, nor decision-making in practice consequently. Only a double-dummy trial design could be employed to compare the efficacy of the two distinct administrations of treatment (infusion versus oral ingestion). Furthermore, active placebo trials utilizing noninferiority while equivalence trial designs are preferable to traditional placebo trials to reduce placebo response and its variability. Another way to minimize placebo effects and response is to include an active control group. This method secures that both the experimental and active control groups have similar expectations of improvement; thus, differential improvements can be attributed to the efficacy of the treatment.

The participant-nurse relationship, into patient care when an injectable treatment is given, may also affect placebo effects and responses consequently. The development of non-pharmacological interventions as the initial course of a treatment could be used to minimize these differences in a trial, that includes an active comparator. In conclusion, attempts to exaggerate a trend towards one or another direction for the effect size of a treatment by comparing absolute and relative rates that do not point into the same direction in placebo-controlled trials, should be considered with caution, because differences in placebo responses may reflect differences in the positive expectations of the trial treatment held by trial subjects.

All these aspects must be considered when using the suggestions of the guidelines for this class of drugs clinically (Supplementary Video)^15,16.

Transparency

Declaration of financial/other interests

CID has received IHS research fellowship grant, scholarship from Hellenic society of Neurology; she serves as a member of the Board of Directors in the European Headache Federation. PM has received grants from SpringerNature and from Springer Switzerland, royalties from Springer UK. He serves as Editor in Chief of The Journal of Headache and Pain; Editor in Chief of SN Comprehensive Clinical Medicine; he received grants as Editor of Headache Books Series from Springer, Editor of books Migraine in Medicine, A Machine-Generated Overview of Current Research and Non-Migraine Primary Headaches in Medicine, Publisher Springer. He received grants form Italian Foundation for Study of Headaches to School of Advanced Studies, Unitelma Sapienza University of Rome, for the Advanced Training Course on Headaches in Low Medium Income Countries (LMIC). He serves as a member ex officio of the Board of Directors in the European Headache Federation. DDM has received honoraria, research and travel grants from Allergan/Abbvie, Amgen, Biogen, Cefaly, Genesis Pharma, Electrocore, Eli Lilly, Lundbeck, Merk-Serono, Mertz, Novartis, Pfizer, Roche, Sanofi, Specifar and Teva; he has participated in clinical trials for Amgen, Cefaly, Electrocore, Eli Lilly, Genesis Pharma, Lundbeck, Mertz, Novartis, Specifar and Teva as principal investigator; he is President of the board of the Hellenic Headache Society; and Co-chairman of the management group of the Headache Panel at the European Academy of Neurology. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript.

Acknowledgements

None stated.

Reviewer disclosures

A reviewer on this manuscript has disclosed that they have served as a consultant and have been on the speakers bureau for Amgen, Lilly, Lundbeck, and Teva. Peer reviewers on this manuscript have no other relevant financial relationships or otherwise to disclose.

Additional information

Funding

No funding was received to produce this article.