1. Selection of Patients
1.1. Inclusion/Exclusion Criteria
Recommendation: Investigators should clearly specify the eligibility criteria for patient entry into the trial. These criteria should identify a sample of patients who are representative of the patients who will be targeted for treatment in clinical practice.
1.2. Patient Recruitment/Source of Patients
Recommendation: Investigators should clearly specify the sources of patients (eg, the local community, investigators' practices, referrals from other professionals, general practice setting, specialty clinic).
Recommendation: Investigators should clearly specify their methods for patient recruitment (eg, media advertisements, newsletter or e-mail announcements about the trial, news stories or articles, fliers or brochures, direct solicitation to the trial by the investigators or referral sources).
Recommendation: Investigators should clearly specify all inducements for participation in the research (eg, money, in-kind, undergraduate course credit).
Recommendation: Investigators are urged to specify the proportion of patients who were seeking versus not seeking treatment for their headaches.
1.3. Headache Diagnosis
Recommendation: Headache diagnosis should conform to the criteria of the International Headache Society (International Classification of Headache Disorders-II (ICHD-II); IHS, 2004)37 or to other recognized and clinically useful criteria.38
Recommendation: Any outcome study evaluating the effect of an intervention on more than one diagnostic group (eg, migraineurs and tension-type headache patients treated with the same cognitive-behavioral intervention) should separately report outcomes for each diagnostic group. As “episodic” and “chronic” diagnoses are distinct entities, outcomes of patients presenting with chronic versus episodic headache patterns also should be tracked and reported separately. It generally is appropriate to enroll patients who meet criteria for more than one headache diagnosis in outcome research. Results may be pooled across diagnostic groups if the lack of heterogeneity is demonstrated using appropriate statistical tests. When a non-IHS diagnosis is employed, a standard IHS diagnosis also should be provided.
Comments: Research addressing behavioral interventions prior to the release of the original version of the IHS nosology suffers from the lack of uniformity among researchers in the diagnostic criteria employed. While this situation has improved, strict adherence to standard diagnostic criteria in all research evaluating therapies for migraine and tension-type headache (drug and nondrug alike) is essential for the fundamental validity and reliability of research.
Some candidates for research inevitably will meet criteria for more than one headache diagnosis (in separate headache episodes, of course). It may, for example, be acceptable to enroll a migraine patient who also suffers from occasional tension-type headache in a trial evaluating a behavioral treatment for migraine. The investigator should clearly specify whether patients with multiple diagnoses may be enrolled, and if so, such patients should be able to discriminate between types of headache whenever practicable.
1.4. Headache Frequency/Severity
Recommendation: At study enrollment, patients' pattern of headaches should be stable and headaches should be of sufficient frequency and severity to facilitate the demonstration of a change in frequency in response to treatment. In general, patients should report a minimum of four headache days per month, and very often six or more headache days per month should be required.
1.5. Age of Patients at Entry
Recommendation: Patients should be included with a range of ages as appropriate to the intervention.
Recommendation: If patients with a broad range of ages are entered into a single trial, investigators should, when feasible, report outcomes separately by clinically meaningful age groupings.
Comments: Including patients with a broad range of ages may enhance generalizability of trial results, particularly if a consistent effect is observed across age groupings.
1.6. Chronicity of Headache
Recommendation: Headache chronicity should extend for a minimum of 1 year, and the pattern of headache symptoms (eg, headache frequency, severity, location of pain, symptoms associated with head pain) should have been generally stable for a period of 6 months or more.
Comments: Establishing a minimum headache chronicity of 1 year is advisable to help exclude headaches due to other causes, and to ensure that changes observed following implementation of therapy are less likely to arise merely as fluctuation in the course of the disorder or be attributed to “spontaneous remission.”
1.7. Sex of Patients
Recommendation: Women and men alike should be included in headache treatment outcome research in most instances. The proportion of women and men included in each arm of the trial should be clearly reported. Certainly, treatment and comparison groups should not exhibit significant gender differences.
Comments: Given the female preponderance among patients with migraine and with tension-type headache,39,40 the number of men included in intervention studies with small samples size may be low. It is desirable to report outcomes separately by sex, but such reporting may be of limited value when samples are small. If gender differences in treatment response seem likely or if their assessment is important, prespecified numbers of men and women can insure power and stratified randomization by gender can insure balance between treatment groups.
1.8. Hormonal Variables
Recommendation: As hormonal variables are known to exert potent influences on headache—in particular on migraine,41–43 women who are pregnant, planning pregnancy during the trial period, or lactating should be excluded from trials that are not specifically designed to address those circumstances.
Recommendation: Patients' use of exogenous hormones (ie, hormonal contraceptives, postmenopausal hormone therapy) should be reported, tracked and stable throughout all phases of the trial, equated across arms of the trial, and when feasible, examined as a moderator variable.
Recommendation: It is desirable to identify and separately report outcomes for patients who experience headache occurring exclusively with menses (ie, pure menstrual migraine) versus those who experience either menstrually related headache or nonmenstrually related headache.37,44
1.9. Medical and Psychiatric Comorbidities
Recommendation: Patients with substantial medical45 or psychiatric comorbidities46,47 that are deemed likely to interfere with their ability to fully participate should be excluded from trials that are not specifically designed to address those circumstances.
Recommendation: When feasible, known psychiatric and medical comorbidities that may have a bearing on treatment outcome should be tracked and reported throughout all phases of the trial, equated across arms of the trial, and possibly examined as moderator variables.19
Comments: While migraine and tension-type headaches are not in and of themselves psychiatric disorders, epidemiologic and clinical research has revealed a clear association between these headache diagnoses and a number of psychiatric disorders (especially depression and anxiety).46,48–54 While the majority of individuals with headaches in the general population do not experience comorbid psychiatric disorders,55 a sizable number of patients, especially those presenting to specialty clinics do exhibit psychiatric symptoms. Moreover there is some evidence indicating that psychiatric comorbidity is a negative prognostic indicator in children and adults with headache.56
The recommendation to exclude patients with substantial medical or psychiatric comorbidities is not intended to imply that all patients with comorbidities must be excluded; on the contrary, many patients with chronic and stable medical and psychiatric conditions may be appropriate for inclusion in trials of behavioral interventions for headache.19
1.10. Concomitant Medication and Dietary Supplements
Recommendation: Appropriate acute medication therapy (abortive and/or palliative medications) should be allowed for individual headache attacks.
Recommendation: Prophylactic headache medication therapy is discouraged in behavioral trials, particularly within efficacy studies, but may be permitted in studies of clinical effectiveness if the prophylactic medication is stable and ongoing throughout the trial.20
Recommendation: When concomitant medication use (prescription and nonprescription, acute and prophylactic) and/or dietary supplement use (eg, herbal supplements, vitamins, minerals) is permitted during the trial, it is essential that the patient's usage patterns are stable prior to trial entry, and that they have been so for an adequate period of time to ensure that changes observed in headache during the trial are not attributable to recent changes in medication/supplement use (includes increasing as well as decreasing use).
Recommendation: Medication and supplement use should be reported, tracked, and stable throughout all phases of the trial, equated across arms of the trial, and when feasible, examined as moderator variables.
Recommendation: When the trial design calls for discontinuation of medications prior to the initiation of behavioral treatment, an adequate washout period is needed to ensure that medication effects have resolved and a stable baseline of headache activity has been established. Migraine prophylactic medications should be discontinued 3 months prior to the research trial, and tension-type headache prophylactic medications should be discontinued at least 1 month prior to the trial, consistent with the IHS guidelines for controlled trials of drug treatments.1,2
Comments: An allowance for acute headache medication use is made routinely in trials of prophylactic pharmacotherapies for headache, and acute medication use is similarly allowed in trials of behavioral therapies. However, behavioral trials diverge from pharmacologic trials by permitting, in certain cases, ongoing prophylaxis with the patient's established headache medications. This is an important research design consideration, especially in efficacy trials, where the intent is to evaluate the effects of the behavioral intervention independent of any other factor. When concomitant medication is allowed, the efficacy of the behavioral intervention cannot be isolated and evaluated independently from the impact of the medication. The behavioral trial then becomes one of looking at the additive effects of behavioral treatment concurrent to prophylactic medication, thereby threatening the internal validity of the study. Notably, disallowing patients' ongoing prophylactic medication therapy may adversely impact patient recruitment, may prolong the investigation with washout periods, and may exacerbate the patient's headache severity during baseline. Thus, an a priori determination should be made regarding whether to allow patients to remain on their ongoing prophylactic treatments throughout the trial. When prophylactic medication therapy is permitted, the prophylactic medication therapy should be stable, continued throughout the behavioral trial, tracked, and reported. Because ongoing prophylactic treatment may restrict variance of headache outcome in effected patients, stratification of patients on this variable may be desirable to ensure the equivalence of groups.11
Ideally, of course, other active preventative interventions (including medication prophylaxis) are disallowed to ensure that outcome is attributable to the behavioral intervention under study. Once the efficacy of the intervention is established, effectiveness trials can include patients on concomitant preventive regimens to reflect a more “real world” administration of the behavioral therapy.20
Prior to presenting for headache treatment, most patients have attempted a variety of interventions with varying success. Many patients seek additional or “alternative” treatment when prior interventions have proven “less than optimal” (eg, owing to side effects, high cost, insufficient headache relief), yet they nevertheless may deem their present interventions worthwhile and may prefer to continue the therapy. Standard clinical applications of behavioral treatments often involve co-administration with pharmacologic treatments. Therefore, allowing continuation of established medication prophylaxis during an effectiveness trial of behavioral therapy may increase the external validity and generalizability of results.
Nevertheless, permitting concomitant administration of prophylactic medications or other active interventions with the behavioral intervention under investigation may substantially influence outcomes, including potentially restricting headache variance, increasing error variance, and diminishing statistical power to detect important posttreatment change, creating difficulty discerning treatment effects from additive effects of multiple treatments, and possibly reducing patients' ability or motivation to fully participate in all aspects of the behavioral treatment (eg, decreased attention/concentration, decreased physiologic responsiveness to temperature biofeedback).
1.11. Medication-Overuse Headache
Recommendation: Patients suspected of having medication overuse headache should be excluded from trials that are not specifically designed to address those circumstances until the diagnosis can be ruled out.
Comments: The ICHD-II Classification provides specific criteria for medication-overuse headache (ICHD-II 8.2, pp. 94-95), which is recognized as a pattern of high frequency headache arising in the context of frequent and regularly occurring symptomatic headache medication use.37
1.12. Recreational and Psychoactive Drug Use
Recommendation: Use of recreational drugs should be prohibited during the trial, and alcoholic beverage use should be at most modest and with consumption that remains stable through the duration of the trial.
Recommendation: Use of hypnotic medications and other psychoactive drugs should be carefully assessed and prohibited if such medications might conceivably have a bearing on the trial's outcome or might interfere with a patient's ability to fully participate in the trial.
1.13. Patient Characteristics/Demographics
Recommendation: It is important to identify and separately report patient characteristics and demographics of clinical trial participants. Treatment or comparison groups should not exhibit significant demographic differences.
Recommendation: When feasible, demographic variables should be equated across arms of the trial.
Comments: A variety of patient characteristic and demographic variables (eg, age, headache severity) can have a differential influence on treatment response. Thus, it is important for the investigator to consider what particular demographic variables are most relevant for the population being studied. While randomization is designed to equate across groups, there are times in which important variables need to be accounted for in order to ensure equivalence at baseline. While there is no one combination of demographic and patient variables that are relevant to every population, it is vital that investigators thoroughly consider what variables are relevant for the population being studied. When conducting efficacy trials, this is of the utmost importance.20 It is also important in effectiveness trials; however, the nature of the effectiveness trial is to determine whether an efficacious treatment is effective in a certain setting with a certain population.
The demographic variables most commonly accounted for among headache sufferers include: (i) age; (ii) race/ethnicity/sociocultural background; (iii) employment and socioeconomic status; (iv) educational achievement; (v) marital and family status; (vi) disease severity stratification; and (vii) headache diagnostic subgroups (eg, ICHD-II second, third, and fourth digit subtypes).
2. Trial Design
2.1. Range of Research Designs
Recommendation: Case reports, single case experimental designs,57 single-group outcome studies, and other research designs can be appropriately applied in clinical trials evaluating behavioral therapies, and they can prove valuable in illuminating newly developed untested behavioral therapies or substantially new applications of established behavioral therapies.16 However, controlled clinical trials are necessary to establish the effectiveness and utility of behavioral interventions.
2.2. Parallel Group Versus Crossover Designs
Recommendation: Parallel group designs are recommended. True crossover designs cannot feasibly be employed in clinical trials evaluating behavioral interventions.
Comments: Crossover designs are not possible in research on behavioral interventions as the effects of a behavioral therapy cannot be withdrawn (ie, once acquired headache management skills cannot be undone or withdrawn in the same fashion that a medication can be withdrawn) thus essentially ensuring a carryover effect when the patient crosses over from one phase of the trial into the next.16
2.3. Types of Controlled Trials: Design and Implementation of Control Conditions
Recommendation: There is no single “gold standard” design trial for controlled trials. The appropriate trial design will differ depending on the goals of the clinical trial.
Comments: Pragmatic trials that seek only to determine the utility or cost effectiveness of a treatment require a different trial design than explanatory trials that seek to establish the mechanism(s) whereby a treatment produces improvement.58,59 Appropriate control conditions differ for pragmatic and explanatory trials. Control conditions for pragmatic trials seek only to establish the practical utility or cost effectiveness of treatment relative to a benchmark such as pill placebo, treatment as usual, or less cost effective version of the treatment.58,59 On the other hand the goal of explanatory trials is to establish causal relationships identifying “active elements” of treatment. Currently there is no agreed upon “gold standard” control group that allows definitive causal inferences about “active” treatment elements for explanatory trials.16
2.3.1. “Psychological placebo” control conditions
Recommendation: A “pseudotherapy” control condition that is designed to control for nonspecific elements of the therapeutic intervention (eg, demand characteristics, expectations for improvement, influence of self monitoring, number and duration of sessions, clinical contacts) can be an important element of explanatory trials.
Recommendation: Pseudotherapy control conditions are not suited to pragmatic trials as it has not been possible to design a bona fide “psychological placebo” condition that is established as inert in the same fashion that a medication placebo is inert and that has the same appearance and credibility as the active treatment.60
Comments: Simply put, any “psychological placebo” condition will involve interpersonal relationships or other treatment elements that may not be psychologically inert and that are integral components of either the active intervention being investigated or of another active intervention.16,17 Moreover, even the most carefully conceived “psychological placebos” may not be credible to patients and may not evoke the same level of expectancy for improvement as the active condition.60 Nonetheless, appropriate “pseudotherapy” control conditions may narrow possible explanations for treatment effects.
The Declaration of Helsinki (2000, 2002) has concluded categorically that placebo trials should not be undertaken when an established and efficacious therapy is available.61,62 Although cogent arguments have been made in defense of the use of placebo in research,5,63 the implementation of placebo-controlled trials has become increasingly contentious in recent years. While this controversy over appropriateness of placebo-controlled research may not yet have had a major influence on the design and implementation of headache treatment trials, the availability of a variety of efficacious treatments (particularly for acute and prophylactic management of migraine) suggests this issue is relevant for headache researchers.64
At present, however, the application of placebo-controlled trials for examination of headache interventions is widely considered useful and appropriate, and this continues to be the standard design for evaluating pharmacological therapies for recurrent headache.1–3 It is left to the discretion of the investigator to determine the viability and merit of implementing placebo controls when evaluating behavioral therapies for headache.17,18
2.3.2. Dismantling studies
Recommendation: Dismantling studies, where selected components of a behavioral treatment are eliminated and the effects on treatment outcome are examined, provides an alternate methodology for identifying the “active elements” of behavioral treatments.
2.3.3. Medication placebo control conditions
Recommendation: Medication placebo provides an appropriate benchmark control for pragmatic trials, but cannot be considered adequate control for nonspecific aspects of behavior therapies in explanatory trials.17,18
2.3.4. Wait list, usual care, or no treatment controls
Recommendation:“Usual care” may be an appropriate benchmark control condition for pragmatic trials, but is unlikely to be a relevant control condition in explanatory trials. A “no treatment” or wait list control condition is discouraged, at least in part, because it requires withholding treatment. Implementation of a wait list control nevertheless can be justified when other alternative control conditions are not feasible. Wait listed patients should be offered an “active” intervention for headache after the control condition is completed.
Comments: To date, “wait list” or “no treatment” conditions have been the most commonly employed control condition in research evaluating behavioral interventions for headache. Such conditions control for the influences of pre- and posttreatment assessment, symptom self monitoring, and “spontaneous remission” of symptoms.16,17 Patients must be fully informed from the outset of the likelihood that they will be assigned to the control condition, and they should be informed of the length of time treatment will be withheld (and, of course, given the option not to enroll in the trial).
2.3.5. Adequacy of “historical controls”
Recommendation: As a rule, research designs that rely solely upon “historical controls” or published norms are inadequate for controlled research trials because such controls are nonrandomized and nonconcurrent.
Comments: Generally there exist large inter- and intra-individual variations across study populations that render historical controls poorly suited for comparison purposes.
2.3.6. Active-treatment concurrent control (“noninferiority” or “equivalence” trials)
Recommendation: The value of a new intervention can be established through direct comparison to an established intervention in a clinical trial.
Comments: The term “noninferiority” (or “equivalence”) trial is commonly used to refer to a randomized clinical trial in which a new intervention is compared with an established intervention rather than a placebo or untreated control condition.65,66 A clear advantage of noninferiority trials is that they avoid enrolling patients in placebo or inactive control conditions. Simply, the failure to demonstrate a statistically significant difference between two conditions cannot serve to establish “equivalence.” Thus the onus is upon the investigators to examine a sufficiently large number of patients to ensure adequate statistical power to detect clinically meaningful differences in outcomes between the conditions under study. In many instances, noninferiority trials require substantially larger sample sizes than do trials employing placebo or inactive controls.
In testing whether a new intervention is as good as an established one, the “control” or standard treatment must have already been shown to meet certain minimum standards of effectiveness in controlled trials. In addition, the established intervention should have been shown to be effective under similar conditions such as characteristics of study population.
Assay sensitivity67 has not been established for treatments for headache. In a particular sample or trial established drug therapies and pill placebo may not differ in effectiveness; similarly, in a particular sample or trial established behavior therapies and a control condition may not differ in effectiveness. This currently limits the use of equivalence trials because equivalent outcomes with two treatments may reflect the particular responsiveness (or lack of responsiveness) of the trial sample, regression to the mean, or other artifacts.
2.3.7. Ethical issues in design of control conditions
Recommendation: In implementing control conditions, it is desirable to avoid use of deception and to avoid withholding treatment whenever feasible. Use of deception and temporarily withholding treatment may nevertheless be acceptable under some circumstances.
Comments: Research designs that call for implementation of “psychological placebo” or other psychological control conditions can raise ethical issues regarding use of deception. In some past research involving “psychological control” conditions, investigators have not fully informed participants that they might be assigned to a control condition. Rather, in such research, patients initially were led to believe that they were being assigned to one of two or more active-treatment conditions, and the patients were advised of the deception only after the initial phase of the trial was complete. Although use of deception in research can be justified, it is strongly desirable to avoid deception when other research designs are feasible.
Research designs employing psychological control conditions also raise ethical issues regarding the withholding of treatment. As behavioral interventions generally require several weeks or months to accomplish, a control patient may be denied active intervention for a considerable period of time. Given that patients with migraine or tension-type headache have, by definition, a chronic and nonprogressive condition, withholding intervention generally is deemed justifiable when patients are fully informed that they may be randomized into a control condition requiring delay of active treatment. Withholding active treatment becomes a greater concern if patients are not advised that they could be assigned to a control condition. Although it can be justified, it is desirable to avoid withholding treatment when other research designs are feasible.
2.4. Assessment of Treatment Credibility
Recommendation: In explanatory trials, investigators should undertake systematic assessments of the credibility of all treatment and control conditions to ensure that control interventions are deemed equally as credible as the intervention(s). In pragmatic trials, treatment credibility is typically irrelevant.
2.5. Blindedness of Therapist and Patient to Intervention
Recommendation: Double-blinding behavioral treatment and control conditions is rarely, if ever, practical or possible. Single blinding is often practical and is desirable in explanatory trials. Blindness is typically irrelevant to pragmatic trials.
Comments: Behavior therapies involve the teaching and learning of headache management skills. Blinding both the teacher and learner is typically not possible and, where possible, risks undermining effective skill learning.16,17 In rare instances double-blinding biofeedback training procedures may be possible, but this has yet to be demonstrated. Single-blind trials of behavior therapies are feasible,68–72 but their value has been limited by the lack of a true psychological placebo. Nonetheless, single blinding of treatment and control conditions is valuable in controlling for outcome expectancies and thus is desirable in explanatory trials.
2.6. Design Issues for Trials Comparing or Integrating Behavioral and Drug Therapies
Recommendation: Clinical trials comparing the effectiveness of drug versus behavioral treatments require research designs that address the unique characteristics of both forms of intervention. Neither the methodology used to evaluate drug therapies nor the methodology used to evaluate behavioral therapies can be transferred wholesale to trials that compare or combine the two therapy modalities.
Comments: Trials that compare the effectiveness of drug and behavioral interventions present special methodological problems as well as unique scientific opportunities.18,73–75 In such trials, and as appropriate, investigators are urged to implement the recommendations as specified in the IHS guidelines for controlled trials involving headache medications.1–3
2.6.1. Allegiance effects and therapeutic integrity
Recommendation: It is recommended that the health care professionals that administer drug and behavioral therapy have primary allegiance to the therapy modality they are administering and expertise in the implementation of this therapy modality. Inclusion of measures of therapy integrity for each therapy modality also is recommended.
Comments: Allegiance effects, or the tendency of investigators and trial sites to obtain better results with therapies they have primary allegiance to, than with other therapies, are an important methodological issue in drug versus behavioral treatment comparisons.76,77 Effective implementation of both drug and behavior therapy can prove problematic when investigators have greater training, experience, or familiarity with one therapy modality than with the other modality.73,74,77 These methodological confounds can be minimized when drug and behavioral therapy are each administered by health care professionals who have a primary allegiance to the type of therapy they are administering, expertise in its administration, and therapy integrity is monitored.
2.6.2. Outcome and process measurement
Recommendation: It is highly desirable for trials to include a sufficiently wide spectrum of secondary outcome measures to capture possible differential outcomes of drug and behavior therapy.
Comments: In the instance that drug and behavioral therapies are found to exert a similar impact on the primary outcome variable, the impact of these two treatment modalities may differ on other measures of functioning (eg, psychological symptoms, quality of life, efforts to manage headaches).12,18,78,79 Identification of potential differences in the impact of the two therapy modalities requires the inclusion of outcome measures that assess a range of outcomes likely to be impacted by either therapy modality.74,75
2.6.3. Outcome evaluation period
Recommendation: An evaluation period that is adequate to capture the treatment effects of both drug and behavior therapy treatment effects on primary and secondary outcome measures are needed.
Comments: The time course for change following initiation of therapy may differ for behavioral versus drug therapies.80 For example, drug therapies exert their influence more rapidly in some instances and have a delayed onset of effect in other instances.79,81 Thus, a valid comparison of behavioral and drug therapies requires that the evaluation period be of sufficient duration to reveal the likely changes produced by each treatment modality.
2.6.4. Measurement of therapeutic mechanisms
Recommendation: Investigators are encouraged to include measures of the putative therapeutic mechanisms of therapy where feasible.
Comments: Measurement strategies differ in pragmatic trials (designed to establish the effectiveness of therapies) and explanatory trials (designed to examine how therapies work).58 The latter, but not necessarily the former trials require the inclusion of measures of therapeutic mechanisms. However, even pragmatic trials provide can provide useful information about therapeutic mechanisms in behavioral treatments. It can reasonably be assumed drug and behavioral therapies work via different therapeutic mechanisms so information about the differential impact of drug and behavior therapy on the putative therapeutic mechanisms of behavior therapy can be informative. Information obtained from drug-behavior therapy trials might thus be maximized if measures of the putative therapeutic mechanisms of behavioral treatment (eg, self efficacy, headache-related coping) are obtained even in pragmatic trials.18
2.6.5. Patient recruitment
Recommendation: Investigators should employ recruitment procedures that are broad enough to minimize the possibility that trial participants are differentially responsive to drug or behavior therapy. It is desirable to assess variables that might indicate if a trial sample is biased to be more responsive to one treatment modality (eg, reason for refusal of randomization, treatment preference).
Comments: It is important that patients likely to be responsive to each treatment modality are included in the trial sample, unless the goal of the trial is to compare the effectiveness of drug and behavioral therapy in a particular subgroup of patients.74 For example, excluding previous nonresponders to one treatment modality may introduce bias. Typically variables that predict response to drug and behavioral treatment will be unknown. This argues for recruitment of a wide a sample of patients from multiple sources and clear specification of recruitment strategy.
Participants often will express a preference for drug or behavioral therapy, and this preference may undermine adherence, influence dropout rate, and impact treatment response.73,82 In trials evaluating combined drug and behavior therapy, treatment preferences may lead participants to be differentially adherent to the two treatment components. Recruitment strategies that primarily yield participants with one therapy preference (eg, recruitment only from a clinic known primarily for either drug or behavior therapy) thus may introduce bias. Potential participants also may be more likely to refuse randomization when they do not want drug therapy than when they do not want behavior therapy, creating a differential sieve that yields a sample more cooperative with drug than behavior therapy. Assessing reasons for refusal of randomization and participants' treatment preferences can help identify these possible confounds when they cannot be controlled.
2.6.6. Structure of behavioral versus drug therapies
Recommendation: When the structure (ie, the number, frequency and length of treatment sessions, follow-up assessments) of drug versus behavior therapy differ, it is important that decisions about the structure of treatment and evaluation periods be congruent with the goals of the trial.
Comments: As practiced clinically, the number, frequency and length of treatment sessions may differ for drug and behavior therapy. If these factors are artificially equated in comparative trials of drug and behavior therapy, the clinical relevance (external validity) of findings may be compromised; if these factors are not equated, clinical relevance may be enhanced, while the possible role these factors play in observed treatment outcomes (internal validity) may be unclear.18 Choices about the structure of therapies will depend on the goals of the trial. However, in general, external validity is likely to be the guiding concern in pragmatic trials and internal validity the guiding concern in explanatory trials.
The structure of drug and behavior therapy also may differ during any long-term evaluation of treatment effects. Drug therapy and associated clinic visits typically continue throughout the follow-up period; in contrast, all treatment-related contact in behavior therapy typically ends after the relatively brief period required to administer therapy. This leads to a similar trade off between internal and external validity in the design a long-term evaluation that was noted above for the structure of therapy.
2.6.7. Control group issues pertaining to trials comparing drug versus behavioral therapies
Recommendation: A pill placebo condition is desirable reference condition in pragmatic trials comparing drug and behavior therapies. Scientific and ethical pros and cons of using a pill placebo condition warrant consideration by the investigator.60,63,65,67,83–86 Inclusion of a pseudotherapy condition to control for purported “nonspecific” factors in behavior therapy is not necessary in pragmatic trials.
Comments: Pill placebo can provide a useful common reference for both drug and behavior therapy.17,18,59 While pill placebo does not control for all the “nonspecific” elements of behavioral treatment in the same manner as it potentially controls for the nonspecific aspects of drug therapy, this is not of primary concern in a pragmatic trial. Any treatment that is no more effective than pill placebo would probably not be of clinical interest.
Trials that compare drug and behavioral treatments are not likely to be explanatory trials because drug treatment is not a particularly useful comparison treatment in a study examining treatment mechanisms in behavior therapy. This problem of control groups in explanatory trials is addressed above.
2.7. Access to Active Intervention
Recommendation: At the conclusion of the study, every patient entered into the study should be assured of access to the best-proven intervention identified by the study or to an equivalent or more efficacious intervention (whether or not the patient's participation in the additional intervention is evaluated).
2.8. Sample Size Determination and Statistical Power Assessment
Recommendation: The determination of the sample sizes required to adequately address the principal questions of the research should be undertaken a priori, and the method for sample size determination should be reported. Statistical power analyses should be undertaken to establish number of patients required per study arm to reveal effects of interest.10,87,88
Comments: Required sample sizes sometimes can be very large given potential interest of small to modest effect sizes. The onus is upon the investigator to clearly identify and justify what effect size is of interest—often a difficult conceptual undertaking. Prior to initiating the trial, the investigators should be reasonably certain they would have sufficient resources available (including the potential to recruit appropriate patients) to gather data from a sufficient number of patients that in turn will allow meaningful data analyses to be conducted.
2.9. Sample Stratification Based Upon Headache Severity or Other Criteria
Recommendation: In parallel group trials, if a wide range of baseline headache frequency is included in a trial, it is reasonable to stratify randomization based on frequency of attacks (eg, ≤3 or >3 attacks per 4 weeks) occurring during baseline. In some cases where variability in baseline headache frequency is limited, it may prove valuable to stratify on other prognostic factors instead. The exact procedures used for random assignment and restrictions to that assignment should be reported. In addition, to avoid bias, methods should be undertaken and reported to conceal each patient's group assignment so that neither the investigator nor the patient, if possible, could know assignment in advance.
Comments: Randomization alone may not ensure comparability among treatment groups for important prognostic factors, particularly in small trials.11,89 However, prognostic factors currently are not well defined and should be employed judiciously. An empirical or a theoretical rationale needs to be provided when justifying a stratification strategy.
Studies have demonstrated that the degree of improvement following headache treatment may vary depending on baseline headache frequency.90 Therefore, when a wide range of baseline headache frequency is included in a trial, it is reasonable to stratify randomization based on frequency of attacks.1 Other factors may also be important; examples of potential variables for stratification include medication use (acute or prophylactic), gender, and other relevant diagnostic criteria (eg, psychiatric comorbidity).
Finally, it is important to operationally define stratification variable(s). For example, if psychiatric comorbidity is used to stratify, the researcher needs to detail how psychiatric comorbidity was measured/assessed.
2.10. Minimum Length of Baseline and Outcome Assessment Periods
Recommendation: A minimum of 4 weeks for baseline and outcome assessment is strongly recommended. When normally cycling women are included in the trial, it is highly desirable for baseline and outcome assessment period to extend at least 5 weeks to capture headache activity during all phases of the menstrual cycle.
Comments: A minimum of 2 weeks of baseline and outcome assessment may be adequate for patients with daily or near daily headaches, but a 4-week baseline is strongly preferred. Baseline and posttreatment outcome assessment periods longer than 5 weeks might be desirable in some circumstances, but this is usually impractical.
2.11. Follow-Up Outcome Assessments
Recommendation: Investigators are urged to undertake and report follow-up assessments of treatment outcome in order to provide an assessment of the maintenance of treatment gains. The follow-up assessment should include the principal dependent measures employed to assess treatment outcome measures as well as headache frequency measures.
Comments: By convention, follow-up assessments characterized as “short-term” have implemented repeat outcome evaluations between 3 months and 1 year posttreatment. Follow-up assessments conducted at 1 year or more posttreatment have not often been reported to date and are highly desirable. It often is not feasible to track control patients throughout follow-up assessments given that doing so would require withholding treatment for an extended period of time.
2.12. Duration of Treatment Periods
Recommendation: The length of the treatment period necessarily will vary with the nature of the behavioral intervention being evaluated. It is advisable, however, to employ treatment periods of 2 months or longer when evaluating behavioral therapies for headache prevention.
Comments: Employment of relatively long treatment periods will increase the power of the trial by providing more stable estimates of headache frequency during the intervention. In addition, the efficacy of behavioral interventions often accrues gradually (ie, needs some weeks before becoming fully established). Likewise, prophylactic pharmacologic therapies newly implemented within trials examining integrated behavioral/pharmacologic therapies will require time for the efficacy of the pharmacotherapy component to become fully established. Only effects of sufficient duration are clinically relevant.
2.13. Specification of Intervention Setting
Recommendation: Investigators are urged to clearly specify the setting where the actual intervention takes place (eg, headache specialty clinic, psychologist's private practice, primary care physician's office, inpatient unit, university student health center, school classroom, mental health center, patient's home, employee health clinic).
Comments: Clear specification of treatment settings will assist consumers of the research in judging the feasibility of the intervention and generalizability of the findings.
2.14. Qualifications, Training, and Experience of Research Staff
Recommendation: Trials of behavioral interventions should be directed by a principal investigator(s) who has appropriate training and experience in conducting clinical research. The principal investigator assumes responsibility for ensuring the availability of careful supervision for novice researchers and trainees involved with the trial, including guarding against contamination across treatment conditions (ie, blending of treatment elements).
2.15. Qualifications, Training, and Experience of Practitioners
Recommendation: All practitioners should possess the training, qualifications and credentials (ie, licensure, certification) required to administer the interventions under study. Moreover, the training and experience of practitioners of all therapy conditions in the trial should be adequate to ensure evenhandedness of trial.
2.16. Details of the Therapeutic Intervention
Recommendation: Investigators should specify key elements of the intervention in sufficient detail to allow the intervention to be clearly understood and replicated by other researchers as well as to allow the trial to be considered for inclusion in systematic reviews.
Comments: True replication of results is only possible if the essential elements of a therapeutic procedure can be identified and the intervention duplicated across research laboratories. Thus replication of research evaluating behavioral interventions depends heavily upon the original investigators' ability to communicate the details of their intervention and methodology to others. Davidson and colleagues (2003)7 and Whitlock and colleagues (2002)91 outline the intervention detail to be described in behavioral research reports. Beyond the information provided in the primary report, investigators should be prepared to share the specific additional details of the intervention, preferably in the form of a written therapy protocol that articulates all of the essential therapeutic elements.
2.17. Evaluation of Treatment Integrity
Recommendation: It is desirable for investigators to provide an assessment of whether or not the intervention was actually delivered as intended (ie, assessment of treatment fidelity).7
Recommendation: Investigators should provide an assessment of the extent to which patients complied with the procedural requirements of the intervention (ie, assessment of treatment adherence).7
2.18. Evaluation of Predictors and Mechanisms of Outcome: Mediator and Moderator Variables
Recommendation: Investigators are urged to undertake systematic assessment and evaluation of possible predictors and mechanisms of outcome.
Comments: The more specifically the causes and predictors of improvement can be specified, the more effectively interventions ultimately can be tailored to treat patients.60 Behavioral interventions necessarily contain a large number of components that contribute to the efficacy of therapy.21 Identification of the active components of therapy (ie, the components of therapy that contribute to observed therapeutic effects) and their mechanisms of action can only be accomplished through careful research design.60
As detailed by Borkovec and Onken, “Moderator variables involve factors that may correlate with the effectiveness of therapy (for example, an intervention may be effective with one type of gender, ethnicity, therapist, or level of problem severity, but is less effective with another type). Mediator variables involve factors that are causal links between the provision of an intervention and client improvement. Examples of such factors include the nature of the therapeutic relationship, self-healing potential, client and therapist expectancy for improvement from and credibility of an intervention, suppressed emotion, and underlying cognitive beliefs”. (p. 310)60
3. Evaluation of Results
3.1. Selection and Implementation of Dependent Measures
Recommendation: Clearly defined primary and secondary outcome measures should be reported, with a clear distinction being made between the two.
Comments: Primary outcomes are the principal measures hypothesized to be affected by the behavioral therapy and are the main focus of the rationale and objectives of the intervention.12 Secondary outcomes are additional variables that can be influenced by the behavioral therapy but are not the main focus of the investigation.
3.1.1. Measurement reliability and validity
Recommendation: Investigators should exercise great care in their selection of tools for measuring outcomes, opting to use reliable and well-validated measurement instruments.
Comments: When previously validated measurement instruments either are not available or are impractical for a particular investigation, the onus is upon the investigators to clearly establish and provide support for the reliability and validity of the instruments adopted for their research.12 The obligation to establish reliability and validity extends to language translations of well-established and validated instruments.
3.1.2. Use of daily self-report measures of headache
Recommendation: Investigators are urged to employ a daily self-report headache diary as their principal dependent measure for assessing treatment outcome whenever possible.
Comments: Global, retrospective measurement of headache variables yields data that are less likely to be reliable, may not correlate with daily measures, and are considered more readily subject to self-report bias.92 Electronic headache diaries are increasingly employed as a means of improving the integrity of the self-report data. It also may prove worthwhile to employ other collateral measures of outcomes (eg, patient's, physician's, or spouse's judgments of patient's improvement).
3.1.3. Selection and reporting of headache variables
Recommendation: Investigators should report a measure of headache frequency as a dependent variable. Choice of headache frequency measure may vary depending upon whether episodic versus chronic headache disorders are under study (see Table 1).
|Episodic||3–4 to 14 days/mo||attack freq. (#/mo)*|
|HA index |
peak HA severity
|medication use |
|Chronic||15+ days/month||HA days/month*||HA index |
severe HA days/mo
peak HA severity
|(same as above)||(same as above)|
Recommendation: Investigators are urged to report multiple headache variables whenever possible (eg, report measures of headache intensity and duration in addition to frequency).
Recommendation: Investigators should specify their principal dependent measure a priori.
Recommendation: Investigators are urged to report a measure of headache activity (preferably headache frequency) as their principal dependent measure.
Comments: The recommendation for headache frequency reporting is consistent with the IHS guidelines for controlled trials of drug treatments1,2 and will facilitate meta-analyses and other comparisons across studies of various interventions.12,15 The IHS clinical trials guidelines for migraine1 prefer headache frequency to headache index because there is no consistent definition of headache index, and changes in this measure can be more difficult to interpret clinically. The IHS guidelines acknowledge, however, that headache index may more accurately reflect the patient's overall level of suffering,1 and some researchers have argued that for patients with very frequent migraine, chronic tension-type headache, or a mixture of both, clinically important reductions in headache index (most commonly 50% or greater) may be the more appropriate outcome measure.93
3.1.4. Treatment responder rate
Recommendation: The “responder rate” (ie, proportion of patients deemed “clinically improved”) should be reported.
Recommendation: Investigators should employ a criterion of 50% improvement in the primary headache dependent measure as the delimiter for a “treatment responder.” The 50% criterion is congruent with the IHS guidelines' for controlled trials of drug treatments.1,2
Recommendation: It is desirable for investigators to report additional information bearing upon the “clinical significance” of the outcomes associated with the intervention(s) under study.
Comments: By convention, a 25% to 49% change from pre- to posttreatment in the primary headache measure generally is deemed “modestly” or “somewhat” improved, and less than 25% change is deemed “not improved.”
3.1.5. Assessment of functional status, headache-related disability
Recommendation: Investigators are urged to adopt standardized measures of disability, functional status, and/or “quality of life” for assessing outcomes whenever feasible and to employ at least one well-validated headache-specific disability/quality-of-life measure.12
3.1.6. Assessment of additional/secondary dependent measures
Recommendation: Investigators are urged to assess outcomes using multiple measures and across multiple domains.
Comments: Whereas the principal dependent measures in trials evaluating interventions for migraine or tension-type headache generally reflect headache characteristics (eg, frequency, severity, duration), in order to broadly assess the impact of behavioral headache therapy investigators are encouraged to report a variety of outcome measures. Common examples include measures of psychological symptoms, stress, stress coping, cognitive variables, sleep variables, treatment satisfaction, and side effects of treatment. Such measures can prove useful in examining outcome mediator or moderator variables.
3.1.7. Adverse event reporting
Recommendation: Investigators should monitor and report the occurrence of adverse events that occur during the trial and include whether or not an adverse event led to discontinuation of treatment.
Comments: Adverse events seldom occur with behavioral interventions; significant negative reactions are uncommon and serious negative reactions are rare. Nevertheless, there are reports of a small number of patients who have experienced uncomfortable reactions or other problems with behavioral therapies that can result in treatment termination. The severity of any observed adverse events should be rated (eg, mild, moderate, or severe; serious or nonserious) along with the investigators' estimate of the likelihood the adverse event occurred as a consequence of the trial.
3.1.8. Assessment of treatment costs
Recommendation: Monitoring and analyzing costs of therapy and undertaking comparative cost/cost-benefit analyses are desirable. Assessing the impact of therapy on medical utilization/costs of care also can prove worthwhile.
3.2. Consistency in Assessment
Recommendation: Consistency in assessment methodology across phases of trial is essential.
Comments: Investigators should exercise caution to ensure that the assessment methodology employed at baseline is consistently applied throughout the trial, as variability in the assessment methodology (eg, assessing headache using daily self-report forms at baseline and subsequently using telephone interviews to gauge patient's impressions of improvement at posttreatment) will likely render the findings of the research uninterpretable.
3.3. Reporting and Examining Attrition
Recommendation: That attrition rate and reasons for attrition should be clearly enumerated.
Comments: The recommended method to report attrition is through the use of a diagram showing flow (ie, numbers) of participants through each stage, to include the number of patients who: (i) were assigned to each condition, (ii) received the intended treatment, (iii) completed the study protocol, and (iv) were included in analyses of the primary outcome variable (see: http://www.consort-statement.org/revisedstatement.htm).6
Recommendation: Investigators should conduct and report “intention-to-treat” (ITT) analyses for their principal dependent measures (at a minimum), and when appropriate, consider ITT findings as their primary efficacy analyses.
Recommendation: Investigators are urged to conduct and report “completer” analyses.
Comments: As characterized within the Consort Statement, ITT analysis is “…a strategy for analyzing data in which all participants are included in the group to which they were assigned, whether or not they completed the intervention given to the group. Intention-to-treat analysis prevents bias caused by the loss of participants, which may disrupt the baseline equivalence established by random assignment and which may reflect nonadherence to the protocol” (http://www.consort-statement.org). A “completer analysis” is an examination of findings from only those subjects who completed particular phase of or the entire intervention.
4. Issues Pertaining to Research Ethics
4.1. Adherence to Ethical Principles
Recommendation: Investigators evaluating headache therapies must consistently adhere to accepted ethical principles for research with human subjects. (cf. American Psychological Association Ethical Guidelines for Psychologists94; the Belmont Report95; Ethics Subcommittee of the IHS).5
Comments: Included in the imperative to conform to accepted ethical principles is the requirement that investigators consistently strive to minimize any risks associated with the research or the intervention under study. While the risks inherent in behavioral interventions for migraine and tension-type headache (at least those routinely evaluated to date) are admittedly very small, the implementation of behavioral interventions and participation in behavioral research are not risk free (eg, risk of disclosure of confidential information), and thus investigators are obliged to make every effort to minimize participants' exposure to risk.
4.2. Oversight by Institutional Review Board or Equivalent
Recommendation: Investigators evaluating headache therapies should conduct research under the purview of an Institutional Review Board or equivalent ethics committee or oversight body.
4.3. Identification of Support and Potential Conflicts of Interest
Recommendation: In reporting their findings, investigators should clearly specify sources of financial and other important support for the research (and for the investigators) that could raise concerns about conflicts of interest. These include (but are not limited to) grants, loaned or donated equipment, research facilities, and salary (also including lecture fees, honoraria, consultancies from research sponsors) that within reason could be adjudged potentially conflictual. Other potential conflicts of interest similarly should be disclosed.