Abstract
Introduction
Nearly 2% of youths suffer from chronic migraine or chronic tension-type headache (CTTH). A contributing factor in approximately 20%–50% of these youths is medication overuse, which involves taking analgesics three or more times per week for three months. The objective of this study was to test motivational interviewing (MI) as an approach to promote adherence to recommendations regarding not only analgesic overuse but also other aspects of treatment plans.
Methods
A randomized controlled trial was conducted comparing the experimental condition (standard of care clinic-based treatment with supplemental MI phone calls; n = 24) to a control condition (standard of care clinic-based treatment without these supplemental MI phone calls; n = 23). Four months after enrollment, a research assistant who was blinded to study condition telephoned adolescents inquiring about headache frequency, headache severity, and disability.
Results
Headache frequency was lower in the experimental condition versus control condition for those with relatively lower initial levels of headache frequency. In contrast, we did not find greater improvement in headache severity and disability for the experimental condition relative to the control condition.
Discussion
Our findings provided some preliminary, albeit limited, support for MI as an approach to improve outcomes for adolescent medication-overuse headache (MOH).
Introduction
Nearly 2% of the pediatric population suffers from chronic migraine or chronic tension-type headaches (CTTH) based on criteria from the International Classification of Headache Disorders—3 Beta (ICHD-II) (1,2). These chronic headaches, which occur at least 15 days per month, are associated with significant distress and disability including missed school attendance and family discord. Considering that many adults with these headaches have onset of these refractory headaches during adolescence (3), these headaches may lead to future occupational limitations as well.
Approximately 20%–50% of children and adolescents with these headaches have medication-overuse headaches (MOH), which involve taking analgesics three or more times per week for three months (3,4). A cardinal recommendation for MOH is discontinuation of analgesics for one to two months until the pathophysiologic mechanisms (purportedly repletion of central serotonin) are corrected (5–8). While medication overuse is a well-recognized problem, there is no consensus on how best to assist patients in discontinuing this practice.
Motivational interviewing (MI) represents a promising approach for preventing further medication overuse for MOH. MI involves a clinician using specific nonconfrontational communication strategies to elicit from health care recipients their own personal reasons for adhering to treatment recommendations. A meta-analysis of several randomized trials indicated that MI can improve adherence to diverse health care regimens such as decreasing alcohol use, improving diet, and increasing exercise (9).
We hypothesized that MI might be useful in addressing medication overuse for chronic migraine + MOH or CTTH + MOH for several reasons. First, MI’s strongest scientific support involves reducing alcohol abuse for adults. Therefore, we speculated that MI might be beneficial in reducing another form of substance use—analgesics for MOH. Second, because MI is among the briefest forms of evidence-based behavioral interventions, MI may be more feasible for busy patients to receive relative to alternative psychosocial approaches (e.g. cognitive restructuring) that are sometimes lengthier to deliver. Third, because most treatment providers for pediatric headaches are physicians and nurses and because these nonmental health providers have been successfully trained in MI for other purposes, MI has the potential for widespread dissemination for MOH.
Despite these promising aspects of MI and other authors’ recent advocacy for using MI to promote adherence for headache regimens (10,11), to our knowledge MI has never been empirically tested for any type of headache. The feasibility and efficacy of MI in this population remains unknown. Therefore, in the present investigation we randomly assigned adolescents with MOH to either an experimental condition (standard of care clinic-based treatment with supplemental MI phone calls) or a control condition (standard of care clinic-based treatment without these supplemental MI phone calls). The supplemental MI phones calls were designed to promote adherence to recommendations regarding not only analgesic overuse but also other aspects of headache treatment plans (e.g. prophylactic medication use, adequate hydration). We hypothesized that adolescents in the experimental condition would report less headache frequency, headache severity, disability, and analgesic use at post-treatment relative to the adolescents in the control condition.
Methods
Participants
Our study’s 47 participants were new patients seeking treatment through a pediatric hospital’s headache clinic. Participants were adolescents 12–17 years old who had English as their primary language and who met the International Headache Society criteria (ICHD-2) for chronic headache involving medication overuse. Headaches were either CTTH (n = 20) or chronic migraine (n = 27). Medication overuse must have occurred prior to the initial clinic visit according to adolescent or family report. Study exclusionary criteria were: 1) headaches secondary to any medical cause; 2) chronic pain affecting other body systems; 3) other neurological disorders; 4) lack of regular telephone access; 5) pregnancy; and 6) any cognitive, sensory or motor impairment that prevented valid administration of study measures.
Procedure
Potential participants were approached during the initial clinic visit to the senior author, a neurologist who is the director of the headache clinic, from spring 2009 to summer 2010. Figure 1 presents a flow diagram for our randomized trial. Fifty-two eligible adolescents were approached, and 47 agreed to participate, resulting in a 90.4% participation rate. Reasons for declining participation were unwillingness to receive phone calls (n = 2), time constraints (n = 2), and substantial disinterest in behavioral approaches (n = 1). All adolescents completed a standard of care visit that featured a neurological history and examination, headache education, medication evaluation (abortive medications such as triptans; prophylactic medications such as antiepileptic drugs; and amitriptyline), and behavioral health services (standardized questionnaires, biofeedback, relaxation techniques). The behavioral health services were delivered by a psychologist but not the study’s psychologist. Adolescents were seen for follow-up visits according to standard of care. At the initial clinic visit, guardians provided written consent and adolescents provided written assent for study participation to the senior author and her research and clinical staff.
Flow diagram for the randomized trial.
After the initial clinic visit, the study interventionist, who was not part of the headache clinic, followed a random number table to assign each participant to either the experimental condition (n = 24) or control condition (n = 23). These conditions are described below. Neither the senior author nor the family members were aware of study condition during the initial clinic visit. The random number table was generated by the study’s statistician using an Excel program that generated blocks of 30 in which there was equal allocation to the experimental versus control condition. Four months after enrollment, a research assistant who was blinded to study condition telephoned adolescents inquiring about headache frequency, headache severity, disability, and analgesic use. All participants in the experimental condition and all but one participant in the control condition were reached for this post-treatment assessment, resulting in a final sample size of 46 (24 experimental, 22 control) and a study completion rate of 97.8%. Adolescents completing the post-treatment telephone interview with the research assistant were each sent checks for $50; they were not compensated for received experimental or control phone calls. Study procedures were approved by our hospital’s institutional review board (IRB).
Study conditions
Common features of both conditions
Both conditions shared five common elements. First, they were delivered by the same study interventionist who attempted to reach adolescents up to four times by phone during the four months after study enrollment. Second, at the beginning of the first phone call, the study interventionist asked adolescents two questions on a 1–10 scale (with 10 being highest) regarding their perceived interest in reducing over-the-counter analgesic use and their perceived confidence in reducing over-the-counter analgesic use if they decided to do so. Third, the study interventionist confirmed contact information and the best times to reach participants for future calls. Fourth, the study interventionist assessed adolescents for recent suicidality when an adolescent had screened positive for suicidality during the initial clinic visit. Such assessments occurred during calls for 11 out of the 47 (23.4%) participants. The experimental and control conditions did not differ on likelihood of receiving such a lethality assessment, Chi-square = .18, p = .67. Fifth, the study interventionist did not have any substantive interactions with parents/guardians during any phone calls other than for lethality assessment and safety planning purposes when an adolescent endorsed recent suicidality.
Experimental condition
The study interventionist was a female registered nurse who used MI techniques to promote adherence to the headache clinic’s treatment recommendations. Miller and Rollnick (12) have written a thorough description of MI techniques, but a summary of how MI was applied for the present study is offered here. During each phone call, the study interventionist inquired what (if any) headache treatment recommendations the adolescent wished to discuss. The study interventionist employed reflective listening skills and asked the adolescent to identify advantages and disadvantages of following the particular recommendation from his or her own perspective. Sometimes reviewing disadvantages of making changes/following recommendations is contraindicated within MI because of concerns that patients will talk themselves out of making a change. However, we decided to elicit disadvantages from the adolescent perspective to encourage the experimental condition participants to speak candidly and to help identify barriers in implementing the treatment plan. (Please see Miller and Rollnick (12) for a more complete discussion of this issue.) The study interventionist emphasized the adolescent’s self-control, inquiring if and how the adolescent wished to follow a particular recommendation. The study interventionist periodically asked permission to discuss specific strategies or to provide her own suggestions of ways to improve adherence and results.
Content of experimental phone calls (n = 24).
Several steps were taken in an attempt to maximize the study interventionist’s fidelity to MI principles. To begin with, we selected an interventionist with roughly one year of part-time experience utilizing motivational interviewing strategies over the telephone. In addition, before she interacted with study participants, she had attended a formal two-day training session in motivational interviewing, completed motivational interviewing role-plays with the study’s psychologist involving likely headache treatment plan scenarios, and received didactic instruction in common components of pediatric headache regimens from the study’s neurologist. Furthermore, she met periodically with the study’s psychologist to review particular cases and to discuss MI adherence. Previous research has indicated that ongoing coaching is essential for acquiring MI skills (13). Finally, she completed structured progress notes that reminded her of the key MI strategies she should employ (e.g. reviewing advantages and disadvantages, emphasizing self-control).
Control condition
For these calls, the nurse was instructed not to initiate discussions regarding an adolescent’s treatment plans but instead to engage only in the aforementioned nonspecific activities (e.g. collect updates in contact information). To help prevent contamination across the study’s conditions, she completed structured progress notes that reminded her of the limited activities that were appropriate for the control condition. Total duration of the control calls averaged just 13 minutes (SD = 18), which was more than one hour less than the average total duration for the experimental calls. The total duration of the control calls was significantly shorter relative to the total duration for the experimental calls, t(45) = −9.75, p < .001. However, there were no differences in either (1) the total number of completed calls for the experimental versus control conditions (Ms = 3.3 and 3.7, respectively) or (2) total number of unanswered calls for the experimental versus control conditions (Ms = 7.1 and 6.1, respectively), both p > .05.
Measures
Baseline measures of headache frequency, headache severity, disability, and analgesic use were collected from adolescents as part of the headache clinic’s standard of care. We used the pediatric Migraine Disability Assessment (PedMIDAS) headache disability scale (14) and assessed headache frequency/severity from a diary. Analgesic use was estimated from adolescent report. All of these measures were repeated at four months at the post-treatment assessment, with the PedMIDAS having a 90-day recall period and headache frequency/severity reported using a 90-day diary.
In addition, parents completed a baseline Adolescent Symptom Inventory (ASI) (15) and adolescents completed a baseline Youth Inventory (YI) (15). The ASI and YI assessed behavioral and emotional concerns (e.g. depression, anxiety) and hence facilitated a clinical description of our sample. These questionnaires were not completed at post-treatment.
Statistical analyses
Independent t-tests (continuous variables) and Chi-square tests (categorical variables) were conducted to examine baseline differences in the experimental versus control conditions. For post-treatment scores, a general linear analysis of covariance was used to evaluate the effect of condition on headache severity and disability scores with baseline measures used as the covariates. Disability scores were log transformed to address the normality assumptions of the analysis of variance. A generalized linear model was used to evaluate the effect of condition on headache frequency assuming a Poisson distribution for the count data. The baseline values for the frequencies were log transformed to meet the assumption that the dependent variable was log related to the predictors. The analysis of covariance assumes a parallel relationship between groups for post-scores across the baseline. This assumption was not met for the PedMIDAS scores and frequency so group by baseline interaction terms were included in the models. This led to post-hoc analyses of low and high baseline conditions. The assumptions were met with headache severity. We completed intent-to-treat analyses; we did not conduct “completer” analyses because all experimental group participants completed at least one call.
To identify possible confounding variables across the two study conditions, we found that the two conditions differed on the following baseline variables using a liberal p < .25 criterion: child gender, parent education, baseline analgesic use, baseline caffeine use, being prescribed a second abortive medication at baseline, Adolescent Symptom Inventory (ASI) Inattention, ASI Oppositional Defiant Disorder, ASI Major Depressive Disorder, ASI Dysthymic Disorder, Youth Inventory (YI) Hyperactivity, YI Oppositional Defiant Disorder, and YI Dysthymic Disorder. We then used generalized linear models that included these potentially confounding variables when we found a main effect for experimental condition on a particular outcome variable.
At the beginning of the study, we chose a target sample size of 30 subjects per group based on the assumption that there will be a correlation of 0.5 between pre-treatment and post-treatment outcome scores. Given that assumption, and an alpha of 0.05, a sample size of 30 subjects per group would have allowed us to identify a difference between means of 0.63 standard deviations (SDs) (i.e. a moderate effect size) with a power of 80%. We fell slightly short of our target size sample as funding ended before we reached our initial goal.
Results
Comparisons of baseline characteristics between the experimental and control conditions
Baseline demographic and headache characteristics of the experimental and control conditions.
M: mean; SD: standard deviation. All group differences were nonsignificant (ps ranged from .065 to .900). aAssessed using a 1–9 scale from 1 ($0–4999) to 9 (more than $40,000).
Baseline adolescent emotional and behavioral functioning according to parent report and self-report by condition.
M: mean; SD: standard deviation. aSignificant group difference (p < .05).
Comparison of standard-of-care treatments received between the experimental and control conditions
Regarding medication, all 47 participants were prescribed 25 mg of hydroxyzine as an abortive medication during their initial headache clinic visit. Even though the standard-of-care clinicians were unaware of study condition at the intake clinic visit, one significant difference did emerge: Participants in the experimental condition were more likely to be prescribed an additional abortive medication during this visit relative to the control participants (33.3% versus 8.7%, Chi-square = 4.3, p = .039). Participants in the two conditions did not differ on first type of preventive medications prescribed at the initial visit, Chi-square = 1.37, p = .85. Amitriptyline and topiramate were the most commonly prescribed preventive medications. Just three out of the 47 participants had a second type of preventive medication prescribed.
Regarding biofeedback, all 47 participants received an introductory session of this treatment during their initial clinic visit. Nineteen of the 47 participants received a subsequent biofeedback session; there were no differences by study condition in receiving this second session, Chi-square = .003, p = .958.
Regarding a follow-up appointment with the headache clinic’s neurologist, 39 participants had a follow-up appointment scheduled at the initial appointment, and 38 participants actually kept an appointment. There were no significant differences by study condition and Chi-squares, respectively = .71 and .20, ps = .40 and .66. Days to actual appointment did not differ for the experimental condition (M = 128, SD = 113) versus the control condition (M = 110, SD = 44), t statistic = .633, p = .53.
Comparisons of outcome variables between the experimental and control conditions
Headache outcome variables at four-month follow-up by condition.
M: mean; SD: standard deviation. ap < .001; the experimental and control groups differed on headache frequency only for those with low baseline frequency of headache.
We also conducted a post-hoc analysis to assess clinically significant change by condition. To be consistent with other studies (16,17), we operationalized clinically significant change as a reduction in headache frequency of at least 50% from baseline to follow-up. We found that 71% (17/24) met this criterion in the experimental condition relative to 50% (11/22) in the control condition. This difference was not significant; Pearson Chi-square = 2.09, p = .15.
While the experimental group tended to have lower post-treatment headache severity scores relative to the control group (effect size = .67, root mean square error = 2.15, p = .029), this difference did not hold consistently when we controlled for baseline confounding variables. Therefore, we concluded that there was no consistent effect of experimental condition on headache severity. There were no significant differences in post-treatment headache disability by condition regardless of whether we controlled for parent-reported adolescent mood symptoms (ps = .83 and .33, respectively). We did not run formal inferential statistics for the fourth outcome variable (analgesic use) because of a likely floor effect; participants in both conditions reported more than a 90% reduction in analgesic use at post-treatment relative to baseline reports.
Dose-response effects for the experimental and control conditions
There was a dose-response effect for headache frequency at follow-up after controlling for baseline headache frequency in the experimental group, p < .001. Specifically, more calls in the experimental condition were related to having fewer headaches at follow-up. However, duration of calls in the experimental condition was not related to headache frequency at follow-up (p = .46). The control group was not affected by number or duration of calls in terms of headache frequency at follow-up.
There were no adverse events in either condition.
Discussion
This study found modest support for the efficacy of supplemental MI phone calls on improving outcomes for adolescent headache involving medication overuse. Headache frequency was lower in the experimental condition versus control condition for those with relatively lower initial levels of headache frequency. In contrast, we did not find greater improvement in headache severity and disability for the experimental condition relative to the control condition. Our partially disappointing results may have resulted from a lack of power or from our fairly adherent sample; even control group participants reduced analgesic use from baseline to four months by more than 90%. Nevertheless, our study had numerous strengths, including (a) the first investigation of MI to improve outcomes in a headache population, (b) a highly distressed adolescent sample based on severe initial levels of headache disability and high initial use of analgesics, (c) study participation and study completion rates that were both above 90%, (d) detailed description of baseline demographic and clinical characteristics and appropriate controlling for baseline differences in our study analyses, and (e) the use of an independent evaluator blinded to study condition to assess post-treatment outcomes.
Despite these strengths, six limitations deserve careful attention. First, while we took numerous steps to maximize fidelity to MI principles, we did not employ the gold-standard approach for assessing fidelity—the coding of interactions that are audio-recorded. While such coding can often be complicated, labor intensive, and subject to reliability concerns (18), future investigations with greater resources than our preliminary study should formally assess MI fidelity. Second, our sample may have had limited generalizability. We used an adolescent sample seen in the tertiary care’s specialty headache clinic that featured both a neurologist and child psychologist. Our sample had extremely high levels of analgesic use at baseline. Other samples of adolescents with chronic migraine or CTTH may not have access to these specialized services or may have not been misusing analgesics. Third, objective data, such as school attendance records, emergency department visit records, and pill counts, were not included in the present study but would have added meaningful information regarding the impact of these supplemental phone calls on adherence and outcomes. Fourth, our modest sample sizes may have led to some null results, particularly given that we found a moderate but nonsignificant effect size of 0.49 of condition on post-treatment headache severity. Fifth, our follow-up period was only four months, which was just after the experimental and control calls ended. Lengthier follow-up periods that persist for longer after calls ended would be necessary to see if MI phone calls had enduring effects. Sixth, we did not assess for participants’ baseline headache lifestyles (e.g. hydration, sleep) before presenting to the headache clinic; therefore, by chance the experimental and control conditions could have differed on baseline on this potentially confounding variable.
In summary, our findings provided some preliminary, albeit limited, support for MI as an approach to improve outcomes for adolescent chronic migraine + MOH/CTTH + MOH. We caution readers that while more completed MI calls were associated with lower headache frequency at follow-up, our restricted range of number of completed MI phone calls (from one to four) may have influenced this finding. In addition, our results are consistent with a non-MI study that indicated high completion rates for adolescents with headaches who are offered telephone-based assistance (19). Nevertheless, our brief intervention may be beneficial only for those with lower headache burden with their chronic migraine + MOH/CTTH + MOH. In the future, augmenting our MI calls with other components (e.g. mobile phone applications to monitor adherence and to remind patients of treatment recommendations on a more frequent basis than periodic phone calls) may be necessary to achieve more robust results. Furthermore, the MI phone calls may be more beneficial for other headache populations at risk for greater nonadherence, including those groups with more long-standing analgesic overuse (e.g. adults).
Clinical implications
Motivational interviewing may be a useful approach for improving outcomes in adolescents with medication-overuse headaches. At least one supplemental phone call featuring motivational interviewing was completed by all adolescents assigned to this intervention.
Footnotes
Funding
This work was supported by an intramural grant to the senior author from our hospital’s research institute.
Conflict of interest
None declared.
Acknowledgment
The authors thank Warren Lo, MD, for his assistance regarding the interpretation of the results.