Physical and mechanical therapies for lower limb symptoms in children with Hypermobility Spectrum Disorder and Hypermobile Ehlers-Danlos Syndrome: a systematic review

All RCTs and quasi-RCTs of physical and mechanical therapies for lower limb problems in children with hypermobility were eligible for inclusion. Trials focusing on prevention of hypermobility-related lower limb problems were excluded. A trial could be included regardless of whether or not participants were blinded to their intervention.

Trials of children age 0–17 years with hypermobility and lower limb symptoms were included. Children could have been diagnosed with JHS or Ehlers-Danlos Syndrome (hypermobility type), or ‘hypermobility’ described as being symptomatic. From our knowledge of the literature, it was expected that JHS would be diagnosed using the Brighton criteria, and Ehlers-Danlos Syndrome (hypermobility type) using the Villefranche criteria, though these diagnostic criteria were not essential for inclusion. Other hypermobility described as being symptomatic was eligible for inclusion. Trials of asymptomatic hypermobility and other heritable disorders of connective tissue (e.g. Osteogenesis Imperfecta, Marfan Syndrome) were excluded. Lower limb symptoms included any pathology of the gluteal and femoral regions, and any pathology distal to that, including pathology of the ankles and feet [20].

All settings were eligible for inclusion.

Any non-invasive mechanical or physical therapy for lower limb problems was eligible for inclusion. Mechanical interventions include braces, splints, footwear and orthotic devices. Physical interventions include massage, stretching, and strengthening exercises. All other mechanical and physical interventions were also eligible for inclusion. Trials of drug therapies, surgery and invasive interventions (e.g. dry needling) were ineligible. Trials were excluded if the effect of the mechanical or physical intervention could not be isolated (e.g. if the mechanical or physical therapy was provided in combination with other therapies).

Quantifiable and validated measures of health-related quality of life (for example, PedsQL Measurement Model) [21].

Any validated quantifiable measure of pain (for example: the Paediatric Pain Questionnaire) [22]; disability or functional ability, or both; hypermobility-related injuries (for example ankle sprains); participant satisfaction with intervention; fatigue and adverse events. Tools validated for use in children were included even when not specifically validated for hypermobility.

Electronic searches were performed on 31st of October 2017. Search terms for OVID MEDLINE (from January 1966 to 31/10/2017) were:

This was adapted to suit the following databases: Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library, Issue 10 2017); EMBASE (January 1980 to October 2017); CINAHL (January 1982 to October 2017); and PUBMED (January 1966 to October 2017). No language or publication restrictions were applied and authorship/ results were not masked. Reference lists of included studies were screened for other eligible studies. Researchers and corresponding authors of eligible studies were contacted by email to help identify other published or unpublished potentially eligible articles.

Two authors (BP and either AC or FH) assessed titles and abstracts of studies retrieved by the electronic searches and retrieved full-text versions of all potentially eligible studies. These were assessed by two authors (BP and either AC or FH) for inclusion. Disagreements between BP and AC were mediated by FH. There were no disagreements between BP and FH. Study authors were contacted as required to determine study eligibility.

Data were extracted using standardised pilot-tested forms by one reviewer (BP) and were checked by another reviewer (AC). Extracted data included: study design, inclusion and exclusion criteria, recruitment procedures, setting, interventions, outcome measures, follow up duration, power calculation, baseline characteristics, outcome data, and funding source. Corresponding authors were contacted for more information as required.

The risk of bias of included studies was independently assessed by BP, AC and FH using the following criteria described in the Cochrane Handbook for Systematic Reviews of Interventions [23]:

Each criterion was assigned a judgement of ‘high risk’, ‘low risk’ or ‘unclear risk’. All disagreements in the assessment of risk of bias were resolved by discussion by BP, AC and FH.

Data for different types of interventions and different lower limb problems were analysed separately in Review Manager 5 [24] following Cochrane Handbook guidelines [23]. Data were entered by one reviewer (FH) and checked by another (BP). Where data were not appropriate for analysis in Review Manager 5, additional data were requested from corresponding authors. In the event that corresponding authors did not provide appropriate data as requested by email, the results were reported as they appeared in the original trial publication. Continuous data that were analysed in Review Manager 5 were analysed using a fixed effect model and the inverse variance method with mean difference effect measure with 95% confidence intervals (CI). Analyses of dichotomous data were to report odds ratios (OR) with 95% CI, though this was not required. If meta-analysis was performed, sensitivity analysis was to exclude papers that did not conceal allocation sequence, and outliers if the reason for the errant result was evident, though this was not required.

Electronic searching retrieved 520 titles and abstracts following removal of duplicates. Of these, 24 were identified as being potentially relevant [3, 15‐18, 25‐43]. Upon reading full texts, 21 articles were excluded (14 were not RCTs or quasi-RCTs [3, 15‐18, 25‐33], 4 included an inappropriate age range [34‐37], 2 did not address a specific lower limb complaint [38, 39], and 1 reported non quantifiable measures [40]). Authors of the remaining three trials were contacted to determine eligibility. Of these, one was excluded as research was ongoing [41].

Two studies were eligible for inclusion [42, 43]. Reference lists of both included trials were screened for other potentially eligible trials and known researchers in the field were contacted by email to identity other potentially relevant studies. No additional studies were identified. Fig. 1 presents the PRISMA flow diagram.

Both included studies were parallel-group randomised controlled trials published in peer-reviewed journals between 2010 and 2012 [42, 43]. Table 1 summarises characteristics of included studies.

A total of 86 participants (48 males and 38 females) were included in the two trials [42, 43]. All participants in both trials were diagnosed using the Brighton criteria. The mean age of participants in the trial by Kemp et al. (2010) was 10.88 years (SD = 2.5) and in the trial by Pacey et al. (2013) was 12.04 years (SD = 2.93).

Both trials assessed arthralgia associated with JHS. The trial by Kemp et al. (2010) included participants presenting with arthralgia for at least the previous three months. The trial by Pacey et al. (2013) included participants presenting with JHS-associated knee pain.

Risk of bias assessment is presented in Table 2.

Pain severity was assessed in both trials. The trial by Pacey et al. (2013) employed the 100 mm Visual Analogue Scale (VAS) on which children indicated their level of pain as both an average and maximum rating of pain intensity over the past week. The trial by Kemp et al. (2010) also used the VAS scale to quantify parent-report of child’s pain, which was self-reported by children aged eleven and older. Children below eleven years of age in the Kemp et al. (2010) trial reported pain using the Wong-Baker Faces adaptation of the VAS scale. Physical functioning was also assessed by both authors using the Child Health Assessment Questionnaire (Minimally important difference (MID) for improvement: − 0.188) [44]. The trial by Pacey et al. (2013) assessed functional ability, by measuring the number of flights of stairs participants could climb in two minutes (MID unknown), whereas this outcome measure was assessed by Kemp et al. (2010) via a six-minute shuttle walking test (MID: 54–80 m) [45].

In total, 86 participants were allocated to undergo treatment via physical therapy. In the trial by Pacey et al. (2013), 12 participants were allocated to the hypermobile training group and 14 were allocated to the neutral training group. Four participants withdrew from this trial prior to random allocation, including one who was later diagnosed with osteogenesis imperfecta, thus rendering them ineligible for participation. In the Kemp (2010) trial, 30 participants were allocated to the targeted physical therapy group and 27 participants were allocated to the generalized physical therapy group.

In both studies, clinically important baseline differences precluded the use of unadjusted follow-up data in Review Manager 5. Mean change from baseline scores were reported by Kemp et al. (2010) and were analysed in Review Manager 5 for this review. Results are presented in Table 3. The contact author of the trial by Pacey et al. (2013) were asked to provide change-from-baseline standard deviation data for all outcomes included in this review, however, none were provided. Hence, no re-analysis of data were performed for the trial by Pacey et al. (2013). The results from the trial by Pacey et al. (2013) reported in this systematic review reflects the original trial report.

These results are presented in Table 4.

The trial by Pacey et al. (2013) measured the change in outcomes during the baseline control period prior to group allocation. During this period, there were statistically significant improvements in only the parent-reported limitations in emotion and behaviour domain of the CHQ (p = 0.02), which resulted in improvement in the Psychosocial Summary of the CHQ (p = 0.03). The size of the change was not reported. There were no statistically significant changes for any other outcome measured prior to intervention.

No statistically significant differences were found between targeted and generalised physiotherapy after three or five months. In the trial by Pacey et al. (2013), comparison of exercising knees into the neutral or hypermobile range for eight weeks, no statistically significant differences were found between groups for the outcomes of pain, CHAQ or function measured by number of flights of stairs climbed per two minutes. There were statistically significant differences between neutral knee range and hypermobile knee range training groups for CHQ. The CHQ Physical summary improved more in the neutral knee range training group (mean difference − 7.77; 95% CI: 0.055 to 14.99), however, there were large between-group differences at baseline, which may have influenced these results. The CHQ Psychosocial summary improved more in the hypermobile range training group (mean difference 9.06; 95% CI: 2.66 to 15.47). Pacey et al. (2013) report that the following three individual domains of CHQ significantly favoured the exercise into the hypermobile range group: self-esteem (p = 0.03), behaviour (p = 0.019) and mental health (p = 0.001). Between-group change from baseline standard deviation data for these domains were not reported. None of the individual domains significantly favoured the group exercising into the neutral range of knee extension. The trial by Pacey et al. (2013) noted that self-esteem and mental health domains were significantly lower than Australian normative values at baseline (all p < 0.05) [46]. Only the hypermobile training group equalled Australian norms post treatment (self-esteem: hypermobile p = 0.84, neutral p < 0.05; mental health: hypermobile p = 0.53, neutral p < 0.05). Differences between pre- or post- training mean scores of either group with Australian norms were not seen in behaviour domains (all p > 0.05).