Feasibility and safety of a 6-month exercise program to increase bone and muscle strength in children with juvenile idiopathic arthritis

Between September 2014 and February 2015, we recruited children aged 8 to 16 years with a diagnosis of JIA [21] who were willing and able to participate, irrespective of their subtype of JIA or disease activity status, and who lived in the Metro Vancouver area and could attend monthly group exercise sessions. Exclusion criteria were: 1. receiving bisphosphonate treatment (past or planned), 2. participation in high performance sports, training or competition > 3 h/week, 3. participation > 1 resistance training session per week for the past 4 months, and 4. pregnant or planning pregnancy. The Clinical Research Ethics Boards at the University of British Columbia approved all procedures (certificate H14–01572). Parents or guardians provided written informed consent and children provided assent.

Following recruitment, participants attended baseline measurement at the Centre for Hip Health and Mobility in Vancouver. To account for rolling recruitment the intervention was conducted in three waves (February, March and April, 2015). Participants had follow-up assessments 3, 6 and 12 months after their baseline visit. At each assessment participants completed questionnaires, muscle testing and clinical assessment by a pediatric rheumatologist. Bone outcomes were assessed at baseline, 6- and 12-months.

Clinical assessment included count of joint and enthesitis sites with active inflammation by a pediatric rheumatologist, standing and sitting height using stretch stature (Seca Model 242, Hanover, MD) to the nearest 0.1 cm, body weight calibrated electronic scale (Seca Model 840, Hanover, MD) to the nearest 0.1 kg and arm (ulna) and lower leg (tibia) length (cm) to the nearest millimeter using a standard steel anthropometric tape. All measures were performed in duplicate by trained research assistants.

We estimated age at peak height velocity (APHV) as an indicator of somatic maturation using the Moore equation [22], and subsequently calculated maturity offset as a continuous measure of years from APHV. Self-report questionnaires included pain intensity in last week (10 mm visual analog scale where higher scores indicate more pain), function (Child Health Activity Questionnaire (CHAQ), 0–3 where higher scores indicate greater functional disability) [23], physical activity questionnaire for children (PAQ-C) and adolescents (PAQ-A) (score 1–5 with higher scores indicating higher PA) [24, 25], fatigue (PedsQL Multidimensional Fatigue Scale, 0–100 where higher scores indicate less fatigue) [26], Juvenile Arthritis Quality of Life Questionnaire (JAQQ, scores 1–7 with lower scores indicated better quality of life) [27], Child and Youth Physical Self-perception Profile (CYPSP, 4 sub-domains of sport competence, body attractiveness, physical condition, and physical strength as well as general physical self-worth; score 0–4 with 2.5 midpoint and higher scores indicating higher self-perception) [28], current medications, fracture history, dietary calcium intake [29] and self-assessed stage of sexual maturation as per the method of Tanner [30].

We used both DXA and HR-pQCT to assess bone outcomes at baseline, 6- and 12-months. Image acquisition and analysis protocols are described in detail elsewhere [31]. Lumbar spine bone mineral content (BMC, g) and areal bone mineral density (aBMD, g/cm²) were assessed using DXA (Hologic QDR 4500 W) with standard positioning. A trained technician acquired and analyzed all scans and performed daily quality control procedures. We calculated lumbar spine BMC z-scores using published reference data [32]. Second, we used HR-pQCT (XtremeCT; Scanco Medical AG, Brüttisellen, Switzerland) to assess BMD, bone microarchitecture and bone strength at the non-dominant distal radius (7% site) and distal tibia (8% site). Outcomes of interest included total BMD (Tt. BMD, mg HA/cm³), trabecular bone volume ratio (BV/TV), cortical thickness (Ct.Th, mm; mean cortical volume divided by the outer bone surface) and failure load (F.Load, N). We calculated age-, sex- and ethnicity-specific z-scores for each HR-pQCT outcome using published centile values [31].

Muscle testing included grip strength, peak power (single two-legged countermovement vertical jump) and force (multiple one-legged hopping) on Leonardo Mechanograph® Ground Reaction Force Plate, (GRFP; Novotec Medical GmbH, Germany), and isokinetic strength of non-dominant elbow flexors / extensors, knee flexors / extensors, hip abductors/ adductors using a Biodex dynamometer (System 4). We highlight grip strength, peak power, knee extensor peak torque, average power and total joint work per participant as clinically significant muscle measures.

The home-based exercise program included: jumping exercises (3 sessions/week, jumps associated with ground reaction forces of between 3 and 5 times body weight [33, 34]); handgrip exercises (3 sessions/week completed with hand grippers) and resistance training (2 sessions/week completed with Therabands® without handles). Before starting the program, our kinesiologist (KD) conducted a home visit to individually tailor the program to each participant’s abilities. The kinesiologist provided each participant with pictures of their exercises, a schedule for exercise progression, Therabands® (color coded for resistance) and hand grippers. She also demonstrated modified exercises should specific exercises cause joint pain and recommended one rest day between resistance and jumping exercises.

The progressive exercise program was implemented over 26 weeks. Participants completed six 4-week blocks of jumping exercises (exercise load increased from 1 set of 5 jumps to 3 sets of 10 jumps; jumping difficulty progressed from simple jumping movements (e.g., tuck jumps/‘motorcycle jumps’) to hopping, skipping and multidirectional plyometric movements (e.g. jumping lunge/‘leaping lizards’) Additional file 1: Figure S1). We estimated ground reaction forces associated with these exercises were approximately 3 to 5 times body weight [35]. For resistance exercises the load increased from 1 set of 8 repetitions for 4 exercises to 3 sets of 15 repetitions for 7 exercises. Difficulty increased from simple single-joint exercise (e.g. biceps curl) to complex multiple- exercises (e.g. lunges). Participants were instructed to take a 20-s rest break between sets and exercises. Resistance exercises included upper extremity, core and lower extremity muscle groups. Handgrip exercises progressed from 2 sets of 5 repetitions to 3 sets of 12 repetitions. Children were instructed to squeeze a rubber cylinder with either 5 short, hard squeezes or hold a moderate to maximal squeeze for 5 s. We implemented a 2-week program of exercise maintenance at the end of the six 4-week blocks. The resistance training exercises required 10–30 min per session; jumping and handgrip exercises required 2–10 min per session.

Participants also attended one group-based exercise session each month. These kinesiologist-led sessions were held at three locations in Metro Vancouver as per the geographic distribution of participants. At each session, the kinesiologist reviewed participants’ progress, introduced new exercises, conducted interactive strength and agility activities and provided children with the opportunity to meet fellow participants. The kinesiologist modified the exercise program for participants who reported pain during or after completion of prescribed exercises.

Participants received usual medical care throughout the study and had regular rheumatology clinic visits every three to 4 months. If clinically indicated, clinic visits included assessment by a physiotherapist who could prescribe additional exercises to target specific problem areas.

Participants completed a weekly exercise log, in print or online. (Additional file 1: Figure S2). Parents/guardians were asked to assist with scheduling and monitoring home exercise sessions, as needed. The kinesiologist contacted the family every 2 weeks to address questions about exercises and to monitor compliance. We recorded the proportion of prescribed exercise completed (total repetitions reported / total repetitions prescribed) and the proportion of group exercise sessions attended (sessions attended/total sessions). We also asked participants to use their log book to record completion of any additional exercises.

Adverse events included JIA flare (defined as any increase in active joint count or Physician Global Assessment of disease activity that resulted in a change in treatment), post-exercise pain lasting > 24 h, and falls or injuries that occurred during exercise. To determine feasibility, we recorded participant recruitment, adherence to exercises, attendance at group exercise sessions, completion of questionnaires and attendance at measurement session. As a companion study, we conducted one-on-one interviews with child-caregiver dyads to identify facilitator and barriers to adherence [36].

We recruited a convenience sample of 24 participants as this allowed 8 subjects per group for each recruitment wave.

We performed all analyses using Stata, version 10.1 (StataCorp, College Station, TX). We report means and standard deviations for anthropometry, clinical, bone and muscle outcomes. We report median and interquartile range (IQR) for categorical outcomes, including active joint count. We calculated the average (and range of) adherence for each 4-week exercise block and the proportion of participants attaining at least 80% adherence for each 4-week block. We used repeated-measures analysis of variance to compare clinical outcomes over time. To examine change in bone and muscle outcomes while accounting for the time-varying covariate change in height, we fit mixed effects models (xtmixed in Stata) with time as a fixed effect and maturity offset at baseline as an additional covariate. We used the margins command in Stata for pairwise comparisons between time points. We considered p values < 0.05 as statistically significant.

Participants completed a median of 46.9% (5.4, 66.7 IQR) of prescribed exercises. Adherence was highest during the first 4-week block of exercises (median 72.3%). It decreased but remained stable over the next 16 weeks (approximately 50%) and decreased during the final 4 weeks (median 38%). Attendance at monthly group sessions was variable across the intervention period, with a median of 66.7% (16.7, 100 IQR) sessions attended. Only 2 participants (15.4%) achieved the goal of 80% adherence, 3 (23.1%) achieved 60% adherence and 5 (38.5%) achieved 50% adherence. Completion of logbooks (paper or online) was 53.8% (19.2, 91.7 IQR) and declined across 24-weeks.

Pain scores were variable and generally low (27 ± 26 points) on 100-point scale (Table 2, Additional file 2: Table S2). Ten incidents of pain lasting > 24 h were reported by 8 participants. The incremental change in pain from pre- to post-exercise sessions was small (5.0, 0–20 Median, IQR on 100-point scale). Two participants reported falls or injuries; one participant ‘rolled their ankle’ during the jumping exercises and one participant reported a knee injury in two consecutive jumping exercises sessions. Seven participants had their exercise programs modified for pain, injury or flare, and 3 withdrew because of pain. There were 4 JIA flares during the intervention (3 mild, 1 moderate). Of those participants with mild flares, one participant (#16) was previously in clinical remission and started NSAID therapy and two participants (#9, #19) were previously in clinical remission on stable DMARD therapy but started NSAID therapy during the intervention. One participant (#8) had a moderate flare that required addition of prednisone and biologic therapy to stable DMARD therapy.

Total fatigue scores were low at baseline (mean 57.9, SD 18.2) and improved by 8.1 points and 12.0 points, on average, at 3 and 6 months, respectively, with some improvement maintained at 12 months (9.4 points) (Table 2, Fig. 2). Total fatigue scores were lower than previously reported scores of a general pediatric rheumatology population aged 2 to 18 years (mean 73.8, SD 21.9) [26]. We did not observe improvements in any other clinical outcomes.

We provide bone outcomes at baseline, 6- and 12-months in Table 3. We excluded one participant’s baseline radius HR-QCT scan because of significant motion artifact (motion grade of 5 on a 5-point scale). We identified one potential outlier for HR-pQCT outcomes at 12-months. Distal radius Z-scores for this participant at 12-months were more than 2 standard deviations higher than their z-scores at baseline or 6-months and review of the participant’s scan revealed callus formation of unknown origin (no history of fracture). We removed the 12-month HR-pQCT scan for this participant from our analysis.

At baseline, all participants had a LS BMC z-score within the normal range as compared with international reference data [32]. In contrast, 4 participants had z-scores < − 2 for HR-pQCT outcomes at the distal tibia (Tt.BMD, BV/TV, Ct.Th and F.Load). Three of these participants also had z-scores < − 2 for F.Load and BV/TV at the distal radius.

LS BMC z-score did not change over the course of the intervention or follow-up period (Table 3, Fig. 2). At the distal tibia, Ct.Th z-score appeared to decrease during the intervention; however, this trend was not maintained after 12-months. Total BMD at the distal tibia increased by 5% after 12 months, but Tt.BMD z-score did not change significantly. Z-scores for distal tibia BV/TV and F.Load did not change significantly. At the distal radius, Tt.BMD z-score decreased during follow-up after adjusting for change in height and maturity. Z-scores for other bone outcomes at the distal radius did not change significantly over time.

We provide key muscle outcomes at baseline, 3-, 6- and 12-months in Table 4. Briefly, after adjusting for height and maturity offset, grip strength and knee extensor peak torque, average power and total joint work did not improve over the course of the intervention (Table 4, Fig. 2). Absolute values for peak force of the right and left legs increased between baseline and 12-months, but z-scores for mechanography outcomes did not change significantly during the intervention.

Our exercise program is the first to include impact exercise and resistance training in both home- and group-based sessions. As noted by others [6], a ‘one size fits all’ approach to exercise prescription is likely not appropriate for children with JIA. Thus, our kinesiologist tailored the program to each participant’s abilities and modified the program when necessary. Modifications were made for pain, injury or flare in seven participants. Adverse events and post-exercise pain were rare, which suggests the exercise program is safe for children with JIA.

The six-month intervention was necessarily longer (to promote osteogenesis) than previous studies that employed resistance training and weight-bearing exercise to promote bone strength [19, 38]. The longer duration may have contributed to high participant attrition and relatively low adherence. Some participants expressed lack of motivation, which also emerged as a recurring theme in our companion study where we used qualitative methods to better understand barriers to participating in exercise among children with JIA [36]. Previous studies also reported low adherence to exercise programs and physical therapy among children with JIA [39‐41]. However, even partial adherence to treatment plans may significantly decrease pain and improve function and quality of life [39].

Our study demonstrates a clear need to better understand barriers and facilitators to participation in and adherence to exercise programs in children with JIA. In future, the exercise intervention could be adapted to include more group- or in person-contact with the exercise instructor, incentives for participation, or incorporate interactive technologies (e.g. Fitbits), social media (e.g. Strava) or internet-based interventions to monitor and encourage adherence, particularly when participants are not in close geographic proximity [42‐44]. Armbrust and colleagues recently reported promising results for their internet-based PA intervention, “Rheumates @ Work” (R@W) [45]. In contrast to the present study, R@W did not include a prescriptive exercise component. Instead, each week of the interactive, cognitive-behavioral program targeted a different theme such as health education and barriers and benefits of being physically active through online films, animations, puzzles and/or assignments. Acceptance and satisfaction with R@W were very high (93.8% and 85%, respectively) among children aged 8 to 13 years [44] and this translated into improvements in children’s PA and exercise capacity [45]. Further, low program costs suggest this might be a feasible program for others to implement, perhaps in combination with recommended activities to promote muscle and bone strength, rather than a prescriptive exercise component.

Participants had low disease activity for the duration of the study, but almost half reported some functional difficulties (CHAQ above 0). PA levels were low to average and did not change during the study. Fatigue was common but diminished after 6 months of exercise; this decrease was sustained at 12 months. This is promising as fatigue may be a barrier to participation in PA and organized sport among children with JIA [46]. We did not observe any other improvements in clinical outcomes, which may in part relate to low adherence.

Ours is the first study to utilize HR-pQCT to examine bone structure and strength in children with JIA. Of note, whereas DXA-derived lumbar spine BMC z-scores were within the normal range for all participants, several participants had z-scores for bone strength, structure and density lower than two standard deviations at both the distal tibia and radius. Similar to previous studies that employed pQCT imaging [15‐17], this finding suggests that bone strength and structure may be compromised in children with JIA. We aim to investigate this further in the larger LEAP study.

The intervention did not positively influence children’s bone mass, structure or strength z-scores. Although short bouts of jumping activity are known to benefit bone structure at the weight-bearing tibia in school-aged children without JIA [34, 47], low adherence and attrition (implementation factors) likely contributed to our inability to adequately assess effect of the intervention itself (design factors) on bone mass, structure and strength in children with JIA. Further, 6 months may have been insufficient to accurately assess exercise-related adaptations in these bone outcomes. The decline in Tt.BMD z-score at the distal radius may also suggest that modifications to the exercise program are required to provide a more osteogenic stimulus (e.g., dynamic loads of greater magnitude) to the upper limbs.

We extend previous studies of children with JIA that measured strength for single muscle groups, by incorporating jumping mechanography to assess peak force and peak power [48]. Muscle function measured by mechanography is not well described in children with arthritis. However, these measures may be more relevant to children’s daily activities than traditional single muscle or muscle group strength testing. Children with JIA performed jumping mechanography without difficulty and demonstrated modest improvements in muscle function over the course of the study. However, changes in z-scores for mechanography outcomes as well as grip strength and knee extensor torque, power and total work were likely a function of normal growth and development as the increases over 3-, 6- and 12-months were no longer significant after we accounted for height and maturity.

We acknowledge several limitations of our study. Intervention efficacy was hampered by relatively high attrition and low adherence. It will be important in future to identify factors that negatively impact implementation and overcome them, where possible. Also, we did not have a control group with which to compare change in outcomes and accurately tease out the distinct influences of growth and maturation separate from the intervention on bone and muscle outcomes. Finally, we did not assess serum levels of vitamin D nor did we provide participants with calcium (or vitamin D) supplements despite the fact that most participants did not meet the dietary reference intake for calcium of 1300 mg/day [49]. Supplementation may be warranted in future exercise trials. Further, the influence of dietary calcium intake on bone outcomes in children with JIA will be explored in the larger LEAP study.