The effectiveness of shoe insoles for the prevention and treatment of low back pain: a systematic review and meta-analysis of randomised controlled trials

An electronic database search of title and abstract was conducted on the 16^th May 2013. The databases searched were MEDLINE (1950-May 2013), EMBASE (1980 – May 2013), the Cumulative Index to Nursing and Allied health Literature (CINAHL) (1982 – May 2013) and The Cochrane Library. Search terms used were back pain, backache, lumbago, shoe insert, shock absorber, insole, footwear, orthoses and orthotic (Additional file 1). No language, publication date or publication status restrictions were used. Hand searches of the reference lists of included trials, clinical guidelines and review articles were also performed.

Only published reports of randomised controlled trials or crossover trials that compared orthoses or insoles with no treatment or placebo treatment were included in this review. Included studies needed to investigate the prevention or treatment of non-specific LBP. For this review, the definition of LBP is pain and discomfort, localised below the costal margin and above the inferior gluteal folds, with or without leg pain. Non-specific LBP is further defined as LBP not attributed to recognisable, known specific pathology (e.g. infection, tumour, osteoporosis, ankylosing spondylitis, fracture, inflammatory process, radicular syndrome or cauda equina syndrome) [21]. Treatment trials were required to report an outcome measure for pain, while prevention trials had to report an incidence rate. Studies were excluded if the individuals involved had LBP caused by specific pathologies or conditions. Trials using insoles to treat limb length discrepancy (LLD) and pelvic obliquity were excluded as there is disagreement regarding whether LLD does predispose to musculoskeletal disorders and what magnitude of LLD is pathological [22]. Clinical trials that were not randomised or quasi-random were excluded.

One reviewer conducted the electronic searches (AS). Titles and abstracts were independently assessed by two reviewers (AS and VC). Disagreements were resolved by consensus and a third reviewer where necessary (MS). A standardised data extraction form was used to collect population characteristics, trial inclusion and exclusion criteria, intervention details, outcome data and overall conclusions from each trial.

Risk of bias in the individual studies was assessed using the Cochrane Collaboration Risk of Bias Tool [23]. Methodological quality was assessed using a modified version of the Quality Index as described by Downs and Black [24]. The final question (Question 27) dealing with statistical power was simplified to a score of 0 or 1, from the original score of 0 to 5. A trial received a score of 1 if a power or sample size calculation was stated, while a 0 was scored if no appropriate power calculation was described. Therefore, our modified index could result in a score between 0 and 28, with a higher score reflecting a superior methodological quality.

All data analyses were performed using STATA version 12 software. A random effects model was used as the underlying assumptions are believed to be better suited to deal with the clinical heterogeneity of the back pain literature [25]. For trials assessing the prevention of LBP, the relative risk (RR) was computed for dichotomous data. With treatment of LBP trials, where different scales were used to measure continuous pain outcomes across trials, standardised mean differences (SMD) were calculated using approximations of the means/standard deviations [26] and Hedges g correction was used to reduce bias [27]. An effect size of greater than or equal to 0.8 was considered to represent a large clinical effect, 0.5 a moderate effect and 0.2 a small effect [28]. Statistical heterogeneity between studies was assessed by use of the I² statistic, and for this review heterogeneity scores were interpreted as low (25%), moderate (50%), and high (75%) [29]. As heterogeneity tests tend to be lower in power, p < 0.1 is used to indicate heterogeneity rather than p = 0.05 [30]. It was predetermined bias would be assessed using a funnel plot, Egger’s test and the Copas Selection model.

The initial database search resulted in a total of 339 citations of which 20 were appropriate for full review (Figure 1). After review, 11 trials were included (Tables 1 and 2) and nine were rejected on the basis of exclusion criteria (Additional file 2).

Five of the studies, involving 293 participants with an age range of 30 to 51 years, investigated the use of insoles for the treatment of LBP (Table 1) [31‐35]. Two of these trials [31, 32] used only female participants (n = 123). The remaining six studies, involving 2379 participants with a relatively younger mean age range of 18 to 36 years, examined insoles in the prevention of LBP (Table 2) [35‐41]. The majority of these trial participants were male military recruits (n = 2281). The materials and design features of the foot orthoses and insoles used as interventions varied widely between the studies (Table 1). The time period for the use of orthoses was also variable, ranging from 5 days to 24 weeks. Four of the studies issued the comparison group with sham insoles [31, 34, 35, 39], another two trials were crossover or wait list designs [32, 33], while the remaining five trials provided no intervention to the control group [36‐38, 40, 41]. Three of the treatment trials used a visual analog scale to measure pain [33‐35], and the other two used an ordinal scale [31, 32]. Three of the prevention trials measured LBP incidence via reported LBP and at least a day off duty [37, 38, 40], two trials used a self-report questionnaire plus a medical assessment [36, 39], while the final trial used a self-report questionnaire only [41].

The modified Downs and Black [24] quality index scores ranged from 57 to 86% (mean = 73%) (Additional file 3). The majority of studies scored highly on reporting of the interventions used and outcome measures. Only three trials [32, 33, 39] provided details of adverse events related to insole or orthotic use. The risk of bias as assessed with the Cochrane Collaboration risk of bias tool (Additional file 4) was generally low or unclear. The greatest potential source of bias was associated with blinding.

As the characteristics of the treatment trials (n = 4) and the prevention trials (n = 6) were similar, meta-analysis of the two separate groups was considered appropriate. One treatment study [32] was excluded from the meta-analysis because we were unable to contact the authors for data that would allow calculation of effect sizes. Statistical analysis to assess the risk of publication bias was not used due to having fewer than 10 trials in the meta-analysis meaning test power is usually too low to distinguish chance from real asymmetry [42].

Four trials with 197 participants were included in this treatment subgroup meta-analysis (Figure 2). The analysis demonstrated a non-significant reduction in LBP between groups (SMD = 0.74, CI 95%: -1.5 to 0.03) in favour of foot orthoses, although a high amount of heterogeneity (I² = 85.4%, p < 0.01) was present. The treatment group generally showed a positive trend, with three of the four trials reporting results that favour the insole intervention over the control treatment [33‐35]. However only two of these trials reported results with statistical significance (Castro-Mendez et al. [34] (ES = -1.91, CI 95%: -2.63 to -1.19, p < 0.01)) and Shabat et al. [35] (ES = -0.64, CI 95%: 1.12 to -0.15, p = 0.01)).

Six trials with 2379 participants were included in this prevention subgroup meta-analysis (Figure 3). The analysis demonstrated a 22% reduction in the risk of developing LBP with the use of foot orthoses or insoles (RR = 0.78, CI 95%: 0.50 to 1.23) compared to the control group, however these results were not statistically significant and a high level of heterogeneity was present (I² = 76.8%, p < 0.01). Of the six prevention trials, only one low quality trial reported results with statistical significance [36]. The results of one trial [39] may have been affected by the high drop out rate, with only 45.5% and 67.5% of the participants in the two intervention groups completing the trial using their orthoses.

LBP remains a considerable health problem in all developed countries [19, 48] and the failure of the current evidence to conclusively identify effective interventions to improve clinical outcomes and reduce associated healthcare costs has led to calls for more targeted trials [49, 50]. Proponents argue that better outcomes may be achieved by classifying patients into subgroups and prescribing treatment relevant to their clinical presentation. Clinical prediction rules (CPRs) are defined as decision making tools for clinicians that utilise information gained from the history and examination, and can be used to guide a therapeutic course of action [51]. Studies using CPRs to guide the choice of physical therapy or exercise plans for LBP treatment have reported better functional outcomes if the participants receive a treatment matched to their subgroup compared to an unmatched treatment [52, 53].

Our analysis of the treatment trials also supports this proposition. Only the Castro-Mendez [34] trial used an abbreviated CPR to guide treatment options and they reported the largest effect size in the reduction of LBP and disability using customised foot orthoses. An inclusion criteria for this trial was foot pronation which has been proposed as a contributing factor to LBP [9, 10]. Foot orthoses has been reported as an effective treatment strategy for pronation [14]. In addition, the majority of these participants were female (female = 43, male = 8) and research has shown a higher prevalence of LBP in females with pronated feet [54]. However, the researchers used pronation as the only factor from the clinical examination in their CPR and further research would be required to develop a robust CPR for the prescription of insoles for patients with LBP. As the other trials reviewed did not attempt to provide a matched treatment to their participants, it is possible that the effect size of the insoles or foot orthoses has been underestimated.

Although this review was designed to be comprehensive with a robust search strategy, it is possible that that not all studies were identified. In addition, only RCTs were considered to have appropriate levels of evidence, so studies with lesser levels of evidence such as case series have been excluded. The strength of evidence in this review is impacted by the small number of trials identified and the low to moderate methodological quality of included trials. Only two of the higher quality studies [34, 38] reported power calculations to enable detection of a clinically important difference between the interventions. Further, the control in four of the trials [31, 34, 35, 39] was a sham insole which would not have provided any functional support to the foot but may have induced a placebo effect. Finally, our analysis only investigated the effect that insoles or foot orthoses had on pain. We did not assess other measures, such as quality of life and psychological health, which are commonly affected in people with LBP, and which may be improved by insoles or foot orthoses.