Electronic behavioral interventions for headache: a systematic review

Background There is increasing interest in using electronic behavioral interventions as well as mobile technologies such as smartphones for improving the care of chronic disabling diseases such as migraines. However, less is known about the current clinical evidence for the feasibility and effectiveness of such behavioral interventions. Objective To review the published literature of behavioral interventions for primary headache disorders delivered by electronic means suitable for use outside of the clinician’s office. Methods An electronic database search of PubMed, PsycINFO, and Embase was conducted through December 11, 2015. All eligible studies were systematically reviewed to examine the modality in which treatment was delivered (computer, smartphone, watch and other), types of behavioral intervention delivered (cognitive behavioral therapy [CBT], biofeedback, relaxation, other), the headache type being treated, duration of treatment, adherence, and outcomes obtained by the trials to examine the overall feasibility of electronic behavioral interventions for headache. Results Our search produced 291 results from which 23 eligible articles were identified. Fourteen studies used the internet via the computer, 2 used Personal Digital Assistants, 2 used CD ROM and 5 used other types of devices. None used smartphones or wearable devices. Four were pilot studies (N ≤ 10) which assessed feasibility. For the behavioral intervention, CBT was used in 11 (48 %) of the studies, relaxation was used in 8 (35 %) of the studies, and biofeedback was used in 5 (22 %) of the studies. The majority of studies (14/23, 61 %) used more than one type of behavioral modality. The duration of therapy ranged from 4–8 weeks for CBT with a mean of 5.9 weeks. The duration of other behavioral interventions ranged from 4 days to 60 months. Outcomes measured varied widely across the individual studies. Conclusions Despite the move toward individualized medicine and mHealth, the current literature shows that most studies using electronic behavioral intervention for the treatment of headache did not use mobile devices. The studies examining mobile devices showed that the behavioral interventions that employed them were acceptable to patients. Data are limited on the dose required, long term efficacy, and issues related to the security and privacy of this health data. This study was registered at the PROSPERO International Prospective Register of Systematic Reviews (CRD42015032284) (Prospero, 2015).


Introduction
Experts in behavioral headache medicine have identified ten areas of critical need for behavioral headache research. One important unmet need is the development of an innovative technology-based treatment platform for headache self-management [1]. Behavioral headache treatments (e.g., progressive muscle relaxation (PMR), biofeedback, and cognitive-behavioral therapy (CBT)) are Level A Evidence-Based migraine treatments [2] that are essentially free of side effects [3]. Behavioral treatments have enduring benefits [4] and may be less costly than pharmacologic interventions [5]. Questions remain whether these evidence-based treatments can effectively be delivered electronically so that patients can do them on their own outside of the clinical setting. This is an important and timely topic because in 2014, 64 million Americans had smartphones [6]. The FDA states that per industry estimates, 500 million smartphone users worldwide will be using a health care application (app) by 2015 [7], and by 2018, 50 percent of the more than 3.4 billion smartphone and tablet users will have downloaded mobile health apps [8]. Most published literature regarding health apps has focused on preventing and managing chronic disease [9,10], monitoring app acceptability and utility [11][12][13][14], and qualitative studies of user experience and desired functions [15][16][17][18]. However, it is unclear the extent to which apps are effective at facilitating behavior change [19].
The creation of electronic headache apps has proliferated, with over 40 headache apps in the U.S. Google Play Store (Date accessed 12/23/15) and over 70 headache apps in the U.S. Apple iTunes store (Date accessed 1/8/ 16). These apps advertise various purposes-the ability to track headache frequency with electronic headache diary functionality, the ability to detect headache triggers, and the ability to treat headaches with behavioral treatments. Prior studies have demonstrated that smartphone apps with electronic headache diaries are a reliable method for data collection preferred over paper headache diaries by patients because electronic diaries are more discreet in the work place [20]. Electronic headache diary data collection also results in fewer secondary data errors [21], less administrative burden [22,23], high participant acceptance [23], and potential cost savings [24]. Additionally, this format allows the use of reminders and timely follow-up of non-compliant participants via real time investigator data monitoring capabilities. Thus, while we know that electronic headache diaries are useful for collecting data, it is unknown whether behavioral interventions, oftentimes delivered along with the electronic diaries, are feasible and effective.
The purpose of this systematic review is to describe the current body of literature on electronic behavioral interventions for primary headaches that can be used outside of the clinician's office. Specifically, we reviewed the literature to first examine the modalities in which such treatments were delivered (computer, smartphone, watch and other), types of behavioral intervention delivered (CBT, biofeedback, relaxation, other), the headache type being treated, duration of the treatment, adherence, and outcomes obtained by the trials. We also examine the overall feasibility of smartphone interventions for primary headache disorders. Understanding the existing literature on the electronic behavioral interventions will increase our understanding of what has been learnt so far, where the potential of this technology may be best realized in future, and what may be applied to the growing field of mHealth.

Review
Methods A health sciences librarian (AP) conducted searches in PubMed.gov, Embase (via Ovid), and PsycINFO (via Ovid) from January 1, 2000 through December 11, 2015. The search strategy combined three concepts: (1) headaches (e.g. "migraines"), (2) electronic or computerized formats (e.g. "internet"), and (3) behavioral interventions (e.g. "behavior therapy"). Terms were searched both as keywords and Subject Heading terms. The search was restricted to studies published in or after 2000. A manual search among references of selected articles and reviews was also performed (MM). The full search strategies are available in the Appendix. After deduplication, 291 unique records were identified.
Studies were considered eligible for inclusion if they were randomized controlled trials, prospective nonrandomized trials, or observational studies using a behavioral intervention for a primary headache disorder. Secondary headache disorders, including dental disorders, were not eligible. Studies involving telephonic interventions were not included.
Two study investigators (MM, JT) included or excluded articles based on the predefined eligibility criteria using a two-step procedure. In the first step, the investigators independently reviewed titles and abstracts. Twenty-two articles were resolved by consensus, and 256 articles were excluded. Main reasons for exclusion based on title and abstract were because the papers did not pertain to a primary headache disorder (N = 99), were review articles (N = 78), or were studies about medications (not behavioral interventions) (N = 25).
In the second step, the same investigators independently reviewed the full-text version of the remaining 32 eligible articles. Any differences in selection from the two independent searches were resolved by consensus between the two investigators. Of the 32 remaining articles for full-text screening, 23 met inclusion criteria. Two articles were resolved by consensus.
A standardized form was used for data extraction, including the following items: first author, year of the study, country where the study was performed, type of behavioral intervention, mode of electronic delivery, study design, recruitment and setting, headache type, participants, outcomes measured, and results.
This study was registered at the PROSPERO International Prospective Register of Systematic Reviews (CRD42015032284) [25]. Reporting of this study conforms to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [26] and the flow diagram can be found in Fig. 1. Table 1, 23 studies met our search criteria, 8 were performed in pediatric populations and the remainder performed in adults. The majority (14/23, 61 %) used the internet to deliver the behavioral electronic intervention. Only 2/23 (9 %) used CD ROMs and 2/23 (9 %) used Personal Digital Assistants (PDAs). A PDA is a handheld digital device which is similar to its successor of the smartphone but often lacks either the touch screen, connectivity, or ability to download and run native applications that a smartphone can. Five used alternate means but none used smartphone technology. About half (13/23, 57 %) were randomized controlled trials. For the behavioral intervention, CBT was used in 11/23 (48 %) (one study used family CBT), relaxation was used in 8/23 (35 %) (half of these studies specified that PMR was used for the relaxation therapy), and biofeedback was used in 5/23 (22 %) of the studies. The majority (14/23, 61 %) used more than one type of behavioral modality. Descriptions of the procedures for teaching and/or practicing the CBT were varied. Descriptions ranged from "a CD ROM program designed to give children additional strategies to help manage their headache pain" to references of full CBT procedure modules. The table in the Appendix illustrates the varied descriptions provided in the papers. The duration of therapy ranged from 4-8 weeks for CBT with a mean of 5.9 weeks. The duration of other behavioral interventions ranged from 4 days to 60 months.   • Decrease in stress (DASS) in tx group compared to controls from baseline to post-intervention (p = 0.0324) + from baseline to 3 month follow-up (p = 0.0045). Table 1 Description of studies meeting eligibility for the systematic review (Continued)
• CPCI-42:Increase in relaxation (baseline to post-intervention, 3month assessment + 6 month assessment), task persistence (baseline to postintervention + 3month assessment), exercising (baseline to post-intervention) + use of social support (baseline to postintervention) in tx group compared to controls.
• Increases in selfefficacy in tx group compared to controls (baseline to postintervention, 3-month assessment + 6 month assessment) CBT (familybased) RCT  • There was a statistically significant reduction in HA frequency from baseline to post-tx + baseline to 3-month follow up in both tx conditions, however there was NS difference in HA frequency between tx + control groups.
• There was a significant reduction in activity limitations, emotional functioning + parent response to pain behavior from baseline to post tx in both groups, but no significant difference between tx + control groups 28.9 % Overall (29.5 % in tx group vs 28.2 % in control group • There was a statistically significant reduction in HA pain intensity from baseline to post-tx + baseline to  Table 1 Description of studies meeting eligibility for the systematic review (Continued) • Frequency of HA decreased significantly from pretx to post-tx in CBT group but not in control (EDU) group.
• Pain catastophizing was significantly decreased from baseline to post tx in CBT group but not in control.
• NS difference between the groups in satisfaction or "patient-therapistalliance/assistance"     • Raw values of the "headache" subtest of the GBB improved in both groups but NS between groups.

No Adherence Data
• NS in FPI-R, STAI or SDS As indicated in Table 2, the outcomes measured from the various studies ranged considerably. About one third of studies (7/23, 30 %) used measures of headache occurrence as outcomes such as a 50 % reduction in headache frequency, headache intensity or headache index as an outcomes measure. The most frequent outcome assessed was acceptability of/satisfaction with the intervention, with 9/23 (39 %) studies assessing this. Five studies (22 %) used medications or a medication index for assessment. 6/23 (26 %) studies used measures of depression, anxiety and/or stress. Only 2/23 (9 %) studies evaluated cost. Three (13 %) used self-efficacy and/or locus of control.
Many of the studies were considered "pilot" studies and thus had a small number of subjects, limiting the ability to detect differences between treatment groups and controls. 16/23 (70 %) studies had a N ≥ 30. Of the 16, one study by Bromberg et al. [27] had technical issues leading to the loss of data. Of the remaining 15 studies, six had statistically significant positive findings for their respective primary outcomes (frequency and/or intensity and/or disability and/or medications taken) in the electronic behavioral intervention group compared to controls. No studies addressed power analysis.
Looking at the 9/23 (39 %) studies that offered CBT interventions with N ≥ 30, there was a positive outcome in headache frequency, severity, intensity and/or migraine medications taken in 4/9 (44 %) studies. Of the 3/23 (13 %) studies which used biofeedback as part of the behavioral intervention, only two (67 %) had N ≥ 30 and one (33 %) of these studies had a decrease in peak intensity. Of the 8/23 (35 %) studies using relaxation, 7 (88 %) had N ≥ 30 and of these, only 2/7 (29 %) had positive outcomes (frequency and peak intensity).
One study by Bromberg and colleagues evaluated whether there was a dose response by test determining whether more time spent on the intervention website resulted in a greater treatment effect. The researcher divided the treatment group into low-dose and high-dose groups. Participants who used the site the most reported greater increases in feelings of self-efficacy from baseline to post-intervention (t = 5.41, P < .0010), 3 months (t = 4.53, P < .0010) and 6 months (t = 4.64, P < .0010) but outcome variables aside from self-efficacy were not affected [27]. This study also performed a completer analysis (i.e. whether there was a different intervention effect for participants who completed all study assessments compared with those who did not). For all but 2 outcomes (self-efficacy and headache locus of control), no difference in treatment effect was noted for completers compared with noncompleters.

Discussion
Of the 23 papers on digitally delivered behavioral interventions for headache, all were published after the year 2000 and most (55 %) were published after the year 2010. Overall results suggest moderate to high rates of acceptability and feasibility of using technology to treat headaches, although efficacy data are more limited. Patients were comfortable using a variety of delivery modalities including CD-ROMs in two studies, the internet in 14 studies, personal digital assistants in two studies, custom biofeedback devices in three studies, and a personal audio player in one study. Although the pilot nature of many of these studies makes it difficult to directly compare outcomes, results of these 23 studies can help guide future efforts to develop more effective, engaging, and evidence-based digital interventions for headache.
Focusing on outcomes, results of the 23 studies suggest that patients are willing to use digitallydelivered behavioral headache interventions. The studies used varied headache related outcome measures: headache index, headache duration, headache frequency, headache intensity, medications, self-efficacy, quality of life, sleep, anxiety, depression, distress, and disability. Also, the studies used many diverse outcome measures regarding the use of the technology being studied: satisfaction, feasibility, tolerability, adherence, and engagement among others. The variability in reporting instruments used to record headache symptoms limits direct comparison between studies. However, in an effort to broadly understand the overall impact on headache, we considered which interventions led to a 50 % or greater reduction in headache symptoms seen only in the intervention and not in the control group. Only a minority of studies were able to demonstrate this statistically significant reduction in symptoms. The combination of internet based interventions with CBT had the most positive effect on headache reduction. Given the widespread acceptance of the effectiveness of CBT for headache symptoms and the increasing access as well as comfort that many have with the internet, the successful combination is not surprising. Internet-delivered CBT has also been an area of intense research in other fields such as psychiatry in exploring applications for treating depression [28][29][30] and in pain medicine for reducing discomfort. [31][32][33] Along these lines, given the high prevalence of psychiatric comorbidity and migraine [34], future studies can be done to explore whether these electronic behavioral interventions help pain outcomes and the psychiatric symptoms.
The results of many of the reviewed studies must be understood in terms of several common themes. Dropout rates for the digital interventions ranged widely. Adherence is a common concern for headache medication [35,36], with estimates that anywhere from 25 % to 94 % of adults adhere to headache treatment regimens [35]. Thus, adherence is also a concern for digital interventions. Recent studies of computer based CBT delivered for purposes outside of headache have suggested that the average user may only complete a single session of computer based CBT [37] and our review suggests the same issue may occur when using computer CBT for headache. These adherence rates in the review may actually be higher than in clinical practice as some studies tied compensation to adherence and others offered personal check-ins or other study staff interventions to encourage adherence. Given the chronic nature of many headache conditions, long term adherence or adherence with booster sessions may be important if digital interventions are to have long term efficacy. The average duration of studies was only 6 weeks, rendering it difficult to fully understand how engaged patients will be with digital tools that may be used over a period of months instead of weeks. Future studies can be done to better assess adherence. Future studies can evaluate the need for booster sessions to help maintain skills. Engagement of patients, families, and treating clinicians in the development, refinement, and post study iterative improvement in the intervention would likely increase adherence and potency of the treatment. Studies can also be done to determine whether the psychiatric comorbidities such as anxiety and depression play a role in adherence with electronic behavioral treatments for headache. In addition, whether these evidence-based treatments can effectively be delivered electronically outside of the clinical setting remains to be seen.
Beyond concerns regarding adherence, understanding the impact of these digitally delivered behavioral interventions for headache is difficult because of the numerous confounding variables in many of the individual studies. But rather than being just a cause for caution, these confounding variables also suggest important research agenda items in seeking to understand the mechanism and efficacy of these interventions. For example, numerous studies involved human contact, be it in the form of a research assistant checking in to a therapist offering personalized feedback, although the impact of such additional human interaction and feedback was hardly explored. In our review, we found one study conducted to explicitly determine the role of such human contact; the study by Andersson and colleagues examined whether telephone contact by a therapist improved adherence and headache related disability, stress, and coping outcomes when used with an internet-based cognitive behavioral self-help program. However, therapist-initiated telephone calls did not influence the results [38]. Also, many of the interventions employed technologies that prompted use between once a day to several times per day but again the impact and optimal number of prompts was examined in only one study, in which the investigators noted that users displayed some annoyance about the number of digital prompts they received [39]. Along similar lines, no studies looked at the right "dose" of behavioral interventions, e.g., how many CBT sessions are necessary, the most effective length of each CBT session, and if and when should there be refresher lessons. Of the reviewed papers, the average number of sessions or lessons was six and was static and fixed for each study. Finally, few studies employed blinded controls (e.g., sham CBT) making it hard to understand the true placebo effect of receiving what may be perceived as 'high tech treatment.' However, per the Behavioral Clinical Headache Trials Guidelines, the placebo is not a realistic expectation of behavioral trials anyway [40].
In this era of increasing interest in mobile health, smartphone apps, and wearables, it was interesting to note the paucity of literature exploring the mobile delivery of behavioral interventions for headache. This is especially interesting as smartphones can fulfill many of the delivery modalities explored in the literature (cd-rom, internet, personal digital assistants, audio) and also offer all the interventions offered in the literature (CBT, self-management, relaxation, and biofeedback). With ownership of smartphones in the general population at 64 % in 2015 [41] and expected to increase in 2016, it seems likely that smartphones will may offer an accessible and practical tool for behavioral interventions for treatment of headache. However, further data is needed on those who do not have access or ability to use a smartphone in order to understand the true potential. Still, the lack of any outcomes literature on this topic reflects a critical opportunity of researchers.
Despite the paucity of data on smartphone interventions, a quick search of the Apple or Android app marketplaces shows that this lack of a clinical evidence base has not stopped industry or consumer interest in headache intervention apps. Recent reviews have called into question the quality of these headache apps on the commercial marketplaces [42]. Assuming that the same complexities of efficacy and adherence that were noted in the 14 studies of internet delivered CBT for headache may also be applicable for smartphone delivered interventions-such concerns about the utility of commercially based headache apps seem valid although more research is needed here.
Several limitations must be considered in light of our methodology. As a literature review, we only report on published clinical studies and ignore industry and private efforts to use technology to deliver behavioral interventions for headache. While it is likely that industry efforts have achieved results beyond those noted in our review-validating such technologies in clinical studies remains critical. Also, we did not evaluate the quality of CBT, biofeedback or PMR delivered in the reviewed papers and it is possible that differences in the delivery of the behavioral intervention itself may have impacted outcomes. For example, one study used not individual but family based CBT [43]. In addition, we did not control for study methodology with some studies offering subjects more encouragement and assistance which may have also impacted adherence rates. Quantifying such support is difficult as many studies were vague or did not report on exact amount or quantity of assistance provided in using the technology. Finally, we acknowledge that there are no established data on the effect sizes of the electronic behavioral interventions for headache. We used N ≥ 30 to collectively describe studies that were not just very small pilot studies.

Conclusion
Through this systematic review we have explored the literature on digitally delivered behavioral interventions for headache. While results suggest feasibility, data on efficacy and adherence is harder to interpret. The lack of studies on smartphones is notable and presents an important research opportunity going forward.  Table 3 Various descriptions of cognitive behavioral therapy (CBT) conducted in the studies Connelly, 2005[44] A CD-ROM program was provided to give additional strategies to help manage head pain.
Rapoff, 2014 [45] "…lessons on how to use various empirically supported cognitive-behavioral treatments to self-manage recurrent headaches…focused on problem-solving and stress management, and targeted pain behavior and parental response to pain." Bromberg, 2012 [27] "Lessons-interactive instruction for learning practical pain self-management skills and strategies, and how to apply them to solve problems. Tools-visual and graphic interactive learning experiences that allow users to actively manipulate information to construct knowledge and learn to solve problems. Self-assessments-structured sets of questions that help users to reflect on and learn what skills and knowledge they have, where their relative strengths and deficits are, and how to identify what behaviors to target for change." Law, 2015 [43] "Psychological therapy included face-to-face cognitive behavioral therapy for pain management and/or biofeedback. Kopfschmerztherapie mit Kindern und Jugendlichen. Ein Trainingsprogramm.Gottingen: Hogrefe. "The first module presented education on headaches, the second unit focused on stress management (perception of own stress symptoms, coping with stress). In the following modules the participants acquired various skills including "cognitive restructuring (identification of dysfunctional cognitions regarding headache and stress and identifying functional cognitions)." Sorbi, 2015 [48] F o c u s e d o n "graded tasks, specific goal setting and review of behavioral goals; behavior self-monitoring with teaching the use of prompts and cues, feedback on performance and barrier identification; behavioral modeling and social comparison, a focus on time-or stress-management, and relapse prevention. Adapted from several studies and later on turned into a module: Sorbi MJ and Swaen SJ. Protocollaire behandeling van patienten met migraine en spanningshoofdpijn:Ontspannings training en cognitieve training (Training according to protocol in migraine and tension-type headache: Relaxation training and cognitive training