Abstract
Dear Sir We thank Drs Delzell and Haag for their letter (1) supporting the cautious interpretation of the data in (2). We are puzzled by their letter, as their comments seem to fully agree with our manuscript. Their primary criticism appears to be directed to less conservative statements made in another manuscript whose authors overlap with the present one (3).
We refer the interested reader to the introduction and discussion (2), in which we discuss at length the limitations of our and other evidence regarding medication type and medication overuse as risk factors for headache progression/chronic daily headache (CDH). To clarify our position, we agree that the available evidence on the association between nonprescription caffeine-containing combination analgesics and headache progression is not sufficient to support “definitive causal inferences”. Nor would we advocate making “definitive causal inferences” regarding opiate- or barbiturate-containing analgesics and headache progression, although for these categories of treatment the epidemiologic data are more consistent. Although Drs Delzell and Haag’s comment relates to our findings regarding caffeine-containing combination analgesics—not a specific focus of our study as they note—the limitations of observational studies in this particular area apply to other forms of medication and the related issue of whether treatment frequency is related to risk of CDH (e.g. causes “medication overuse headache”).
The absence of definitive evidence raises two key questions. First, is it possible to gather evidence to support definitive inferences in this area? And second, in the absence of definitive information, how should clinicians advise and treat their patients? We consider these issues one at a time.
For studies of medication “X” as a risk factor for migraine progression, the optimal study design is probably a randomized trial. Eligible patients would be randomized to long-term acute treatment with medication X or to an acute treatment alternative. Various limits on treatment frequency might be added in both arms. The endpoints might be the incidence of new-onset chronic migraine/CDH, the proportion of patients whose headaches increase in frequency by a pre-specified amount or a simple measured change in headache or migraine days per month.
Because CDH is an event with an expected rate of about 3% per year in unselected episodic headache or migraine patients (4,5) and because we would want to detect small differences in our non-superiority design, such studies would need to be enormous. Even if we used a design focused on a group at high risk for progression, hundreds of patients per arm and at least one year of follow-up would be required. The choice of acute treatment for the control group is problematic, as medication X and the alternative treatment must be of equivalent efficacy and tolerability. Compliance over such a long follow-up would also be problematic. Although other designs, particularly randomized withdrawal studies, are possible, they answer a related but different question. Finally, as a practical matter, such studies would need to be conducted without industry funding for obvious reasons. Whether such a trial would be ethical is a topic for a different discussion.
In the absence of definitive evidence, practicing clinicians need to make choices and provide advice to their patients about acute treatment. Treatment guidelines generally recommend restricting the use of any acute treatments, including caffeine-containing combination products, to two or at most three days per week (or 10 days per month) to reduce the risk of medication overuse headache. The quality of evidence for this recommendation is low. It is not clear how many patients are helped (by prevention of CDH) versus how many are hurt (by pain under-treatment) by this advice, as this has not been studied. Other treatments, including preventive and behavioral methods, can be used to reduce headache frequency and the need for acute medications.
In conclusion, it is essential to obtain more definitive information on the role of specific acute treatments, and limits on use, as risk factors for the progression of headache in general and migraine in particular. Evidence from observational studies can be used to inform the design of clinical trials but is inherently limited for the reasons stated in our study and elsewhere. Despite the extreme difficulty of designing and conducting interventional studies, we believe such studies are needed, as CDH is not a rare condition, headache prophylactic strategies and medication withdrawal therapies do not work for a significant number of patients, and it may be that more patients are harmed than helped by current treatment guidelines.