Objectives To conduct a pilot study examining whether adding biofeedback-assisted relaxation training (BART) to medication treatment results in better clinical outcomes than medication treatment alone for children with functional dyspepsia (FD) associated with duodenal eosinophilia, a subgroup of children with recurrent abdominal pain. Methods Twenty children were randomly assigned to receive a standardized medication treatment or medication plus 10 sessions of BART. Children and parents completed psychosocial functioning and quality of life measures at baseline, posttreatment, and 6 months. Children rated pain daily via PDA. Physicians provided biweekly assessments of clinical improvement. Results Children receiving medication plus BART demonstrated better outcomes on pain intensity, duration of pain episodes, and clinical improvement than children receiving medication alone. Conclusions BART is a promising adjunctive treatment for pediatric FD associated with duodenal eosinophilia. Electronic daily diaries appear to be a useful approach to assessing changes in self-reported pain ratings in this population.
biofeedback and relaxation, chronic and recurrent pain, gastroenterology, longitudinal research, randomized controlled trial
Chronic or recurrent abdominal pain, historically referred to as “RAP,” affects an estimated 10–20% of school-aged children and adolescents (see Banez & Cunningham, 2003, for a review). RAP exacts significant psychosocial and financial costs to children, their families, and the healthcare system as pain, other somatic symptoms, anxiety, functional disability, and increased health care use often persist into adulthood (Campo et al., 2001). Over the past two decades, there has been an increasing attention paid to this common pain condition and efforts to better define it through development of the Rome criteria (Rasquin et al., 2006). Referred to collectively as functional gastrointestinal disorders (FGIDs), these diagnostic entities include functional dyspepsia (FD), irritable bowel syndrome, functional abdominal pain syndrome, and abdominal migraine. These diagnoses are considered “functional” because they are made in the absence of structural or biochemical abnormalities that explain the symptoms. Most of the children who had been classified earlier as having RAP meet criteria for at least one FGID, with FD and/or irritable bowel syndrome being most common (Schurman et al., 2005; Walker et al., 2004).
Despite an improved classification scheme and increased appreciation for potential long-term costs, there are currently no universally accepted clinical practice guidelines for pediatric RAP broadly, or the FGIDs more specifically (Di Lorenzo et al., 2005). However, a biopsychosocial framework for treatment recently has been gaining acceptance. Within this framework, pain results from the varying contributions of, and interactions among, biological, psychological, and social factors (Drossman, 2006). Histologic inflammation is one biological factor implicated in the development and maintenance of FGIDs that may interact with psychological factors, such as anxiety. For example, duodenal eosinophilia (a higher density of one type of white blood cell indicating inflammation in the lining of the first part of the small intestine) appears relatively common, found in 71% of children undergoing endoscopy with mucosal biopsy for FD (Friesen, Sandridge, Andre, Roberts, & Abdel-Rahman, 2006). Duodenal eosinophilia also is one of the few conditions associated with FD where medical therapy has been shown effective (Friesen et al., 2004). Finally, significant correlations have been found between mucosal eosinophils and anxiety scores in children with FD (Friesen et al., 2005). Thus, children with FD associated with duodenal eosinophilia appear uniquely poised to benefit from multifaceted treatment, including medications aimed at histologic inflammation and psychological therapies targeting anxiety.
Cognitive-behavioral therapy (CBT) has been the most studied adjunctive psychological treatment for RAP. While deemed “probably efficacious” by one review (Janicke & Finney, 1999) and determined to meet the criteria for “level B” evidence of effectiveness in another (DiLorenzo et al., 2005), a more recent Cochrane review (Huertas-Ceballos, Logan, Bennett, & Macarthur, 2008) identified only six randomized controlled trials of CBT interventions for RAP (Robins, Smith, Glutting, & Bishop, 2005; Sanders, Shepherd, Cleghorn, & Woolford, 1994) and deemed the current evidence for effectiveness as “relatively weak.” These authors called for more research on psychosocial interventions with pediatric RAP, including—but not limited to—CBT, and encouraged examining these in terms of match with a specific clinical picture or subgroup.
Biofeedback (BF) is a technique wherein individuals are trained to relieve physical and emotional symptoms by using signals from their bodies that are displayed visually or aurally. Relaxation training, a common component of CBT that can be an effective intervention even alone (Blanchard, Greene, Scharff, & Schwarz-McMorris, 1993), can be added to BF to yield biofeedback-assisted relaxation training (BART). As such, BF/BART addresses psychological contributors to RAP by providing children with concrete methods of coping with pain and stress that can also have direct physical benefits via decreasing sympathetic arousal. Together, these psychological and physical benefits may decrease the experience of pain (Scharff, 1997).
While previous research has shown that BF/BART can be effective in treating pediatric pain conditions such as headache (Scharff, Marcus, & Masek, 2002; Trautmann, Lackshewitz, & Kroner-Herwig, 2006), there have been few empirical studies of its efficacy for RAP or specific FGIDs. Although initial findings are encouraging (Humphreys & Gevirtz, 2000), the biopsychosocial model suggests that combining BF/BART with medication targeted at biological factors implicated in pain maintenance may yield stronger effects. To date, no empirical investigations could be identified that examine the efficacy of BF/BART in treating pediatric RAP or specific FGIDs when added to medication management. Furthermore, there have been a limited number of studies examining structured biofeedback protocols (Brent, Lobato, & LeLeiko, 2008), which are important both for replication and for providing a foundation for future dismantling studies.
The limited research on specific treatments for pediatric RAP and FGIDs may be due, at least in part, to the challenges associated with studying pediatric pain treatments. These challenges include identifying and measuring outcomes in an appropriate, accurate, and sensitive fashion. Prior studies generally have used a 50% reduction in pain frequency, intensity, and/or duration from pre- to post-treatment as a marker of clinically significant improvement (Hicks, von Baeyer, & McGrath, 2006; Osterhaus, Lange, Linssen, & Passchier, 1997). However, pain is a highly dynamic process and single time-point (e.g., pre-/post-test) measures are rarely sensitive or stable enough to accurately measure pain changes in response to treatment (Bolger, Davis, & Rafaeli, 2003). In addition, paper “diaries,” often used for prospectively tracking pain outcomes, can be subject to problems with poor data quality and unrecognized retrospective reporting that may greatly influence findings (see Gendreau, Hufford, & Stone, 2003, for a review of this issue). Fortunately, newer technologies have been developed that allow time-date stamping of diary entries to minimize recall bias (Affleck, Zautra, Tennen, & Armeli, 1999). In addition, statistical techniques for modeling patterns of change over time that use information from multiple data points (rather than single time-point pre-/post-measures) have emerged. Together, these have the potential to improve sensitivity and stability of outcome measurement in pain populations and propel research in this area forward.
The current pilot study used this type of electronic daily process approach to pain measurement, involving repeated measures nested within participant, to explore whether BART confers an additive benefit beyond medication treatment alone for children with FD associated with duodenal eosinophilia. Focus on this well-defined group of patients addresses concerns that RAP is not a single condition, but a symptom that may have multiple etiologies with differential responses to treatment that may be masked in trials involving a broader abdominal pain sample (Huertas-Ceballos et al., 2008). Consistent with a biopsychosocial model, we hypothesized that children who received multifaceted therapy with BART as an adjunct to standardized medication treatment would evidence better clinical outcomes across a variety of areas (lower pain ratings, greater clinical improvement, better psychosocial functioning, and higher quality of life) than children who received standardized medication treatment alone. Conducting this work as a pilot study allowed refinement of the measures and procedures prior to initiation of a costly large-scale randomized clinical trial, as well as preliminary documentation of a treatment group effect needed to justify the next step in this line of research.
Twenty children were recruited for this pilot study from a multidisciplinary abdominal pain center serving children aged 8–17 years at a large Midwestern children’s hospital over the course of approximately 1 year. All children had been referred for abdominal pain of at least 3 months duration, failed a trial of acid suppression therapy, and been identified as having duodenal eosinophilia by mucosal biopsy prior to study entry (see Figure 1 for the participant flow diagram and inclusion and exclusion criteria). The final sample was predominantly female (65%) and had a mean age of 12.2 years (SD = 2.8). Parents identified children’s race as Caucasian (80%), Hispanic (10%), African-American (5%), and “other” (5%). Parent questionnaires were completed most frequently by mothers (85%).
Measures—Primary Clinical Indicators
Children completed a daily pain diary on PDA throughout the intervention phase. The PDA recorded data entry dates, as well as the time entries were started and finished. Data not completed within the 24 hr window of a particular calendar day were considered missing. Children were asked to complete the pain diary only once per day, preferably at bedtime to best capture all information from that calendar day. In the rare instance that more than one entry was recorded on the same day, the latest entry that day was selected to represent the day’s pain ratings as it was thought to contain the most complete information about that calendar day. Children were provided with a copy of the Faces Pain Scale-Revised (Hicks, von Baeyer, Spafford, von Korlaar, & Goodenough, 2001), with faces labeled from 1 to 6 (with 6 representing the worst pain possible) to correspond to the radio button options provided on the PDA. Children were asked to rate highest level of pain, lowest level of pain, and current pain using this scale each day. Information on total duration of pain (summed across episodes) also was collected daily, using the following scale: 0 = no pain today; 1 = 1–5 min; 2 = 6–60 min; 3 = 1–2 hr; 4 = >2 hr.
A Global Response Assessment (GRA), completed by a single physician blinded to treatment group, measured clinical response to treatment on a 5-point scale based on that provider’s assessment of both pain and symptom-related impairment in functioning after reviewing information provided by patients/families during bi-weekly medical visits, as follows: 1 = “worse;” 2 = “same;” 3 = “better, but still interferes;” 4 = “minimal pain, no interference;” and, 5 = “complete resolution of the problem.” The GRA methodology has been validated as an endpoint for measuring improvement in functional bowel disease (Gordon et al., 2003).
Measures—Secondary Clinical Indicators
The Behavior Assessment System for Children (BASC; Reynolds & Kamphaus, 1992) is a widely used measure of children’s emotional, behavioral, and social functioning with separate forms for parent- and youth self-report, both of which were used in this study. Items on the BASC are rated on a 4-point likert-type scale (0 = Never to 3 = Almost always). Subscale scores are reported as t-scores, with scores above 60 classified as “at-risk” and above 70 classified as “clinically significant.” The BASC has demonstrated good reliability and validity. Alpha coefficients for subscale scores range from .64 to .89 for self report and range from .69 to .89 for parent-report (Reynolds & Kamphaus, 1992). This study focused on the internalizing problems subscales (i.e., Somatization, Anxiety, and Depression) due to prior research suggesting a higher incidence of internalizing than externalizing symptoms in children with RAP (Banez & Cunningham, 2003; Schurman et al., 2008).
Quality of Life/Functional Disability
The Pediatric Quality of Life Scale—Version 4.0 (PedsQL; Varni, 1999) has been widely used in pediatric research. Separate forms are available for child/adolescent self-report and parent-report, both of which were used in this study. The measure is comprised of 23 items rated on a 5-point likert-type scale (0 = Never to 4 = Almost Always). Scores range from 0 to 100, with higher scores indicating higher levels of quality of life. The PedsQL has demonstrated good reliability and validity. Alpha coefficients for total and subscale scores range from .80 to .88 for self-report and .86 to .90 for parent-report (Varni, Seid, & Kurtin, 2001). The four narrow-band scales (i.e., Physical, Emotional, Social, and School Functioning) were selected to examine intervention effects on different areas of functioning and functional disability.
Biofeedback readings were recorded with a single I-330 C2 Physiograph (J & J Engineering, Poulsbo, WA), utilizing Windows USE 3 Physiolab software; software was run on a Dell Latitudes Pentium 4 laptop computer. The ground sensor cable was attached to a SE 35 EDG 8 mm snap style silver/silver chloride disc imbedded in the Velcro fastener. After a saline-based gel was applied to the disc to ensure skin contact on the palmar surface of the fourth digit, dominant hand, the fasteners were wrapped around the middle phalanges of the finger. The following modalities were monitored on all patients for every session (except where noted): (a) general muscle tension was evaluated using surface electromyograph (sEMG) measured in microvolts (mcv) by wide band placement of pre-gelled silver/silver chloride Red Dot patch sensors a finger’s width above each eyebrow on the frontalis and attached to an MV-1L sEMG cable via CL 50 alligator clips; (b) peripheral skin temperature (TEMP) was measured in degrees Fahrenheit (F) by an RV-5 TEMP/EDR sensor attached to the pad of the participant’s middle finger of the dominant hand; (c) electrodermal response (EDR), specifically skin conductance, was measured in micromhos (mmho) by an RV-5 TEMP/EDR sensor attached to the dominant index finger by a SE 35 EDG 8 mm snap style silver/silver chloride disc imbedded in a Velcro fastener with saline-based gel on the disc; and (d) breathing rate and effort were monitored during sessions 2, 3, 4, 6, 8, and 10 by a pneumograph (PNG) strain-gauge, strap-style sensor with a single RV 1 respiratory cable placed around the abdomen below the rib cage.
Children and their parents independently completed the BASC as part of the routine initial multidisciplinary evaluation for abdominal pain. Two to four weeks later, following endoscopy with biopsy and prior to initiation of any treatment, children and their parents returned for a baseline study visit. After providing informed permission/assent, children and their parents independently completed the PedsQL. Children were randomized using a previously determined computer-generated random number sequence to one of two treatment groups (10 participants per group): (a) standardized medical care (SMC; brief biweekly medical visits and a stepwise progression of prescription medication beginning with a combination of montelukast and ranitidine, addition of hydroxizine at 2 weeks if needed, and then addition of budesonide at 4 weeks if needed); or, (b) SMC plus 10 sessions of BART over a 6-week period. The allocation sequence was concealed in a series of numbered envelopes by the PI prior to study recruitment and opened by other research personnel at the time of randomization in order to limit possible subversion of allocation. Although it was not possible to blind families, the physician providing SMC was blinded to group assignment. All families were asked to avoid beginning any new mental health treatment during the 6-week study intervention period, although they were allowed to continue with any preexisting services.
Following randomization, children received a PDA, written instructions about caring for the PDA, and a hands-on instruction on how to use the device. Children were asked to complete the PDA diary once daily, preferably at bedtime, beginning that evening. PDA data was downloaded at medical visits or BART sessions. Children who missed several days of data entry were reminded verbally during these visits/sessions to complete the PDA measures every day. No other prompts, reminders, or incentives were provided. GRAs were completed at each bi-weekly medical visit for children in both treatment groups. All children and their parents returned to complete an immediate posttreatment evaluation (including the BASC and PedsQL, as well as a stress profile) at the conclusion of the intervention phase and again 6 months after study entry.
Children assigned to the SMC+BART group had their first BART session at the conclusion of the baseline study visit, with follow-up scheduled at a frequency of 1–2 sessions per week. BART sessions were conducted by one of two registered nurses with professional biofeedback certification. Sessions were approximately 50 min in length and followed a manualized protocol that included instruction in relaxation methods such as abdominal breathing (session 1), progressive muscle relaxation (sessions 2–4), imagery (sessions 5–7), and autogenic handwarming (session 8). Sessions 9–10 focused on integration and maintenance of these skills. Sessions generally were structured as follows: (a) review of homework/practice log; (b) discuss progress and current symptoms; (c) hook up to equipment and obtain brief baseline readings; (d) engage in specified relaxation training activities; (e) discuss session and progress; and, (f) review written home practice directions. Participants were provided with a practice CD and a temperature trainer for use at home in conjunction with assigned homework, which focused on mastery of relaxation skills taught in session and use of these skills routinely throughout the day. Study procedures were approved by the hospital’s Institutional Review Board.
All analyses were performed using SPSS Version 17. Mixed linear regression models evaluated the primary clinical indicators measured multiple times over the intervention period (i.e., pain intensity, pain duration, clinical improvement), as well as the secondary clinical indicators measured at pretreatment, posttreatment, and at 6 months (i.e., psychosocial functioning, quality of life), to establish group equivalence at baseline, to determine the main effect of time, and to detect possible interaction effects related to treatment group. Mixed linear regression models allow flexible analysis of repeated measures data (particularly in situations involving different numbers of repeated measurements, different time intervals between measurements, or both) by utilizing all data available to generate the predictive model. This analytic approach was robust to missing data and allowed use of available data from all 20 randomized participants without excluding the two drop outs (see Figure 1) from analyses. A z-test for proportions was used to compare averaged rates of diary compliance between the two groups to evaluate the possibility of systematic bias in missing data prior to the application of mixed linear regression models. Using the general rule that this type of study will have sufficient power to detect medium effect sizes with at least 30 measurements within 20 participants (Hox, 1998; Kreft, 1996), we anticipated sufficient power to detect medium effects at an α-level of .05 for our primary clinical indicators (42 measurements within 20 participants). Less power was anticipated for analysis of secondary clinical indicators, but was considered acceptable given the pilot nature of this project.
Compliance with Diary Entries
No significant difference was observed between the BART and SMC groups on rates of compliance with PDA daily diary entries (67% vs. 58%, respectively). Best compliance (above 80%) was noted during the first two weeks of study participation for both groups, with remaining data points spread across the remainder of the 6-week intervention period.
Analysis of PDA daily diaries indicated a significant main effect of time for highest [F(543) = 32.50, p < .001], lowest [F(544) = 18.12, p < .001], and current [F(543) = 39.56, p < .001] level of pain intensity, as well as for pain duration [F(543) = 14.33, p < .001]. However, several significant interaction effects also emerged to qualify these main effects, as follows. A significant interaction of time with treatment group was found for highest level of pain intensity [F(543) = 3.96, p < .05; see Figure 2]. Although children receiving SMC alone demonstrated significant declines in highest level of daily pain [β = −.02, t(544) = −2.51, p < .05], children receiving SMC + BART exhibited a relatively steeper rate of decline [β = −.03, t(543) = −1.99, p < .05]. Self-report of lowest [F(544) = 3.12, p = .08; βSMC = −.01, t(546) = −1.68, p = .09; βSMC + BART = −.02, t(544) = −1.77, p = .08] and current [F(543) = 3.30, p = .07; βSMC = −.02, t(544) = −3.02, p < .005; βSMC + BART = −.03, t(543) = −1.82, p = .07] level of pain intensity largely mirrored these results, although the interaction of time with treatment group reached only the level of a trend for each. Finally, a significant interaction of time with treatment group was found for pain duration [F(543) = 17.63, p < .001; see Figure 3]. However, in contrast to findings for pain intensity, children receiving SMC alone reported no significant decline in pain duration [β = .001, t(545) = .279, p = .78], while children receiving SMC + BART reported significant declines in this area [β = −.03, t(543) = −4.20, p < .001].
Visual representation of significant time × group interaction for self-reported highest level of pain intensity
Visual representation of significant time × group interaction for self-reported highest level of pain intensity
Visual representation of significant time × group interaction for self-reported pain duration
Visual representation of significant time × group interaction for self-reported pain duration
Analyses of physician-reported clinical response to treatment indicated a significant interaction with treatment group [F(48) = 4.12, p < .05; see Figure 4]. Specifically, children receiving SMC + BART [β = .58, t(47) = 2.03, p < .05] exhibited a steeper rate of recovery than children receiving only SMC [β = .30, t(47) = 3.02, p < .005].
Take a Virtual Tour