How effective are physiotherapy interventions in treating people with sciatica? A systematic review and meta-analysis

Purpose Physiotherapy interventions are prescribed as first-line treatment for people with sciatica; however, their effectiveness remains controversial. The purpose of this systematic review was to establish the short-, medium- and long-term effectiveness of physiotherapy interventions compared to control interventions for people with clinically diagnosed sciatica. Methods This systematic review was registered on PROSPERO CRD42018103900. Cochrane Central Register of Controlled Trials (CENTRAL), CINAHL (EBSCO), Embase, PEDro, PubMed, Scopus and grey literature were searched from inception to January 2021 without language restrictions. Inclusion criteria were randomised controlled trials evaluating physiotherapy interventions compared to a control intervention in people with clinical or imaging diagnosis of sciatica. Primary outcome measures were pain and disability. Study selection and data extraction were performed by two independent reviewers with consensus reached by discussion or third-party arbitration if required. Risk of bias was assessed independently by two reviewers using the Cochrane Risk of Bias tool with third-party consensus if required. Meta-analyses and sensitivity analyses were performed with random effects models using Revman v5.4. Subgroup analyses were undertaken to examine the effectiveness of physiotherapy interventions compared to minimal (e.g. advice only) or substantial control interventions (e.g. surgery). Results Three thousand nine hundred and fifty eight records were identified, of which 18 trials were included, with a total number of 2699 participants. All trials had a high or unclear risk of bias. Meta-analysis of trials for the outcome of pain showed no difference in the short (SMD − 0.34 [95%CI − 1.05, 0.37] p = 0.34, I2 = 98%), medium (SMD 0.15 [95%CI − 0.09, 0.38], p = 0.22, I2 = 80%) or long term (SMD 0.09 [95%CI − 0.18, 0.36], p = 0.51, I2 = 82%). For disability there was no difference in the short (SMD − 0.00 [95%CI − 0.36, 0.35], p = 0.98, I2 = 92%, medium (SMD 0.25 [95%CI − 0.04, 0.55] p = 0.09, I2 = 87%), or long term (SMD 0.26 [95%CI − 0.16, 0.68] p = 0.22, I2 = 92%) between physiotherapy and control interventions. Subgroup analysis of studies comparing physiotherapy with minimal intervention favoured physiotherapy for pain at the long-term time points. Large confidence intervals and high heterogeneity indicate substantial uncertainly surrounding these estimates. Many trials evaluating physiotherapy intervention compared to substantial intervention did not use contemporary physiotherapy interventions. Conclusion Based on currently available, mostly high risk of bias and highly heterogeneous data, there is inadequate evidence to make clinical recommendations on the effectiveness of physiotherapy interventions for people with clinically diagnosed sciatica. Future studies should aim to reduce clinical heterogeneity and to use contemporary physiotherapy interventions. Supplementary Information The online version contains supplementary material available at 10.1007/s00586-022-07356-y.


Introduction
'Sciatica' is a broad term describing spinally referred pain of neural origin that radiates into the leg. The reported prevalence of sciatica varies widely (1.2-43%) [1], probably due to different diagnostic criteria, reflecting a heterogeneous patient population. Sciatica is a significant burden to healthcare and the economy, as a neuropathic component in low back pain it is not only linked to poorer quality of life, but also increases the already high costs of back pain by a further 67% [2]. Although prognosis is good for most patients, up to 45% continue to have symptoms for 12 months or longer [3].
Physiotherapy interventions such as exercise, manual therapy and psychological therapy are recommended in clinical guidelines for people with sciatica [4]. However, the available systematic reviews examining the effectiveness of physiotherapy interventions are at least ten years old. For example, study selection in the most recent systematic review comparing surgery versus conservative care ended in 2009 [5]. Their results could not be meta-analysed due to poor reporting and clinical heterogeneity. Similarly, a network-meta-analysis concluded its search in 2009 [6], finding no support for the effectiveness of exercise or traction while manipulation may be beneficial. However, the latter was based on a single study only. Prior to this, reviews specifically focusing on conservative management of sciatica were published in 2010 [7] and 2007 [8] and were unable to make strong conclusions on the superiority of any treatment. More recent reviews published in 2015 and 2016 were limited to a subset of physiotherapy interventions (e.g. physical activity versus surgery [9] and exercise versus advice to stay active [10]). A recent review [11] looked at a range of physiotherapy interventions, however the review did not include a meta-analysis.
Of note, sciatica is a heterogeneous condition with no agreed diagnostic criteria [12]. Most reviews to date make no reference to the clinical diagnosis of included study participants rendering it unclear whether patients had confirmed nerve involvement. The objective of this systematic review was therefore to assess the up-to-date evidence on the effectiveness of physiotherapy interventions compared with control interventions in people with clinically diagnosed sciatica.

Registration
The protocol was prospectively registered on PROS-PERO (CRD42018103900). We are reporting our findings according to the updated guidance for the PRISMA guidance [13].

Search strategy
We searched the following databases from inception to 29th January 2021: Cochrane Central Register of Controlled Trials (CENTRAL), CINAHL (EBSCO), Embase, PEDro, Pub-Med and Scopus. We also searched grey literature including trial registries (OpenGrey and clinicaltrials.gov). The search strategy was developed in consultation with a medical librarian and included keywords relating to sciatica, physiotherapy and randomised controlled trials (Supplemental Table 1).

Study eligibility
Included studies were randomised controlled trials evaluating physiotherapy interventions compared to a control intervention in people with 'sciatica'. Trials were eligible if study participants were diagnosed with spinally referred leg pain of neural origin. This diagnosis required at least one of the following: positive sensory, myotomal or reflex tests on neurological examination; positive neurodynamic test (e.g. straight leg raise, slump); imaging confirming spinal nerve compromise correlating with symptoms; presence of neuropathic pain determined with neuropathic pain questionnaires; electrodiagnostic testing or quantitative sensory testing suggesting nerve root involvement. Studies which either did not specify how the sciatica diagnosis was made or were simply using pain referral into the leg without other clinical tests confirming a neural component were excluded.
No restrictions were made on sciatica symptom duration or intensity. Eligible trials must evaluate physiotherapy interventions such as exercise, manual therapy, physiotherapy-led education, or a combination of these. The control intervention needed to be a non-physiotherapy intervention (e.g. surgery, GP care, other non-physiotherapy care). The control intervention could also be placebo, sham or no intervention. No restrictions were made on language.
Trials that included participants with serious pathology (e.g. cancer, fracture, cauda equina), pregnant women or participants aged below 18 were excluded. Studies evaluating post-surgical physiotherapy were excluded. As recent reviews address the effectiveness of acupuncture for people with sciatica [14,15], and acupuncture is not core physiotherapy practice in many countries, trials evaluating acupuncture were excluded.

Study selection
Two reviewers (LD, GJ) screened studies independently. In a first step, titles and abstracts were screened, followed by 1 3 full texts. Discrepancies were resolved by discussion and arbitration by a third reviewer (AS) if required.

Quality assessment
Two reviewers (LD, LK) independently used the Cochrane Risk of Bias tool to assess study quality and risk of bias [16]. The tool was piloted on three excluded studies to test agreement of decision-making. Disagreements between reviewers were resolved by a third reviewer where required (GJ).

Data extraction
Two reviewers (LD, LK) independently extracted data using a standardised form; consensus was used to resolve any discrepancies. The following information was extracted: author, year, country, characteristics of participants (e.g. age, duration, severity of symptoms), diagnostic criteria, physiotherapy and control intervention (type, frequency and duration). Outcomes were extracted at baseline and follow-up time points. Primary outcomes of interest were pain (e.g. numerical pain rating scale) and disability (e.g. Oswestry disability index). Secondary outcomes were global perceived effect, quality of life, change in neurological function, psychological parameters, adverse events, and dropout rates. Means, standard deviations and sample sizes were extracted for each outcome. If alternative summary statistics were provided, we transformed the data using recommended calculations [17]. If available, outcomes were extracted for different time points, and grouped according to time after randomisation as: short term (< 3 months); medium-term (> 3 months but < 12 months) or long-term (≥ 12 months). If multiple terms were reported within one period, the outcome closest to 7 weeks, 6 months and 12 months was used. When more than one body part was used to assess pain (e.g. leg and back pain), the highest score at baseline was used to reflect patients' dominant symptoms. When more than one outcome measure was used within a trial for a specific outcome of interest, the outcome measure described by the trial authors as their primary measure was used.

Data synthesis and analysis
If data were available for the same outcome measure from at least two trials, meta-analysis was performed using Revman v5.4. We calculated standardised mean differences (SMD) and 95% confidence intervals (CI). Random effects models with inverse variance weighting were used to account for the variability of included studies. Heterogeneity was calculated with I 2 statistics and interpreted as follows: 'might not be important' (0-40%), 'moderate' (30-60%), 'substantial '(50-90%), and 'considerable' (75-100%) [16]. We performed separate overall meta-analyses comparing physiotherapy interventions with control interventions for our primary outcomes of pain and disability.
We planned to perform a subgroup analysis according to type of physiotherapy interventions. However, this was impossible as interventions were too heterogeneous to pool. We performed a post hoc subgroup analysis comparing the effect of physiotherapy interventions according to the type of control intervention (minimal vs. substantial). Minimal intervention included advice/education only, GP care, or sham treatment. Substantial intervention included surgery, disc and epidural injections. Due to high risk of bias, we performed a post hoc sensitivity analysis, removing those studies where at least two parameters of risk of bias were rated as high. Results that could not be included in the metaanalysis were narratively described.

Overall meta-analysis on physiotherapy versus control intervention
For pain, 13 trials were included in the overall meta-analysis comparing physiotherapy versus all control interventions at short term, eight trials at medium term and nine trials at long-term time points. There was no difference in effectiveness of physiotherapy versus control interventions at short term (

Subgroup analysis on physiotherapy versus minimal intervention
For pain, six studies comparing physiotherapy with a minimal intervention were included in the subgroup analysis at short term, [18][19][20][21][22][23] three at medium [19,21,23] and two at long term [19,23]. There were no group differences at short (SMD − 0.94 [95%CI − 2.11, 0.23] p = 0.11 I 2 = 99%, One study [24] could not be meta-analysed due to insufficient data. Nonetheless, the results were broadly consistent with the meta-analysis. Santilli et al. [24] reported number of participants with reduction in radiating pain. At medium term, 48 participants (100%) of the physiotherapy group (spinal manipulation) reported reduction in radiating pain compared with 39 (81%) of those in the sham group. At long-term follow-up, 48 patients (100%) of the physiotherapy group continued to report reductions in radiating pain compared with 40 participants (83%) in the sham group.
For disability, six trials were meta-analysed comparing physiotherapy with minimal intervention at short term, [18][19][20][21][22][23] three at medium [19,21,23] and two trials at long term [19,23]. No group differences were observed at short (SMD − 0.34 [95%CI − 0.70, − 0.01] p = 0.06, I 2 = 87%,    Fig. 8). The Santilli [24] study did not report a measure of disability at any time point. Overall, these findings suggest that physiotherapy interventions are slightly more effective than minimal treatment for pain in the long term but not at short or medium term.

Subgroup analysis on physiotherapy versus substantial intervention
Eleven trials compared physiotherapy with substantial control intervention. Nine [26][27][28][29][30][31][32][33]35] were included in the subgroup analysis for pain. There was no difference between physiotherapy and substantial intervention for Two trials reported results that were not possible to incorporate in either meta-analysis [25,34]. Amundsen [25] reported improvements in both the physiotherapy and surgical arms, however groups were not statistically compared. Weber [34] reported slightly higher rates of improvement in surgical compared to physiotherapy interventions at one year.

Sensitivity analysis
Four studies with high risk of bias in at least 2 parameters [28][29][30]33] were removed from the meta-analysis. The sensitivity analyses revealed consistent results for all comparisons apart from the subgroup comparison of physiotherapy versus substantial control intervention (Supplemental Figs. 1-6). With the removal of high risk of bias studies, the effect on pain at medium term and on disability at short term favouring substantial interventions was no longer present (Supplemental Figs. 2 and 4).

Discussion
This systematic review, including 18 studies and 2699 participants with a clinical diagnosis of sciatica suggests that physiotherapy interventions are only better than minimal interventions in reducing pain at long-term time points. Physiotherapy interventions are less effective than substantial interventions (e.g. surgery) in reducing pain at medium term and disability at short-and medium-term time points. However, heterogeneity was considerable in most metaanalyses, and confidence intervals were large, indicating substantial uncertainly surrounding the precision of these estimates. The favourable results for substantial intervention for pain in medium term and disability in short term did not persist following sensitivity analyses removing studies with high risk of bias. The currently available literature therefore provides insufficient evidence to support strong recommendations for physiotherapy interventions in the treatment of people with sciatica.
This systematic review reflects a wider collective inability to show significant benefit of non-surgical treatments for people with sciatica. Pharmacological options fail to demonstrate effects beyond placebo [37], including non-steroidal anti-inflammatories [38], anti-convulsants [39], anti-depressants [40] or opioids [4,41]. Epidural cortisone injections have small effect sizes and short-term benefits [42]. These findings are disappointing given the clear need for effective conservative interventions voiced by patients [43].
Apart from the possibility that physiotherapy is indeed not effective for patients with sciatica, there are multiple possible reasons for the lack of evidence. The physiotherapy interventions used in the 11 trials comparing physiotherapy with substantial interventions are not all considered contemporary in line with current clinical guidelines [4]. This is a reflection of a lack of recent physiotherapy trials, with only four of the 11 studies published in the last decade [26,28,30,31]. Current clinical guidelines recommend group exercise and continuation of normal activities; however, bedrest was a component of the conservative treatment arm in two trials [28,34]. The UK NICE Guidelines [4] find no evidence supporting the use of corsets or belts, but these were a core component in another trial [25] conducted before publication of these guidelines. The physiotherapy interventions are highly heterogeneous and remain unclear in several studies. The Bailey study [26] leaves physiotherapy interventions at the discretion of the treating clinician, and the Peul study [33] refers people to physiotherapy only if they are fearful of movement, leaving uncertainty about how many participants in those trials had active physiotherapy treatment. It could also be argued that patients deemed suitable for surgery are likely to represent a specific subgroup that may be less amenable to physiotherapeutic interventions (e.g. with intractable pain or neurological deficit). Indeed, two trials comparing physiotherapy interventions with surgery included patients who had already failed conservative treatment [28,29], raising serious concerns that physiotherapy interventions could possibly succeed in such a population.
A further challenge to progress in treatment is the diagnosis of sciatica itself [44]. There is no agreed definition for sciatica, reflected in the wide range of definitions used in clinical trials [12], including our review. The broad term 'sciatica' comprises radiculopathy, radicular pain, or somatic referred pain. The differing patient populations bring clinical heterogeneity to most meta-analyses. Unfortunately, the high heterogeneity among studies reduces the confidence in our results. Together with previous systematic reviews with inconclusive findings, our results question the value of continuing to perform clinical trials in heterogeneous groups of patients. Although subgrouping according to risk stratification showed promise in the management of people with non-specific low back pain [45], this has failed in patients with sciatica [46]. Subgrouping using a mechanismbased approach shows promising signals in patients with neuropathic pain of different aetiologies [47], but has yet to be examined in sciatica.
The risk of bias analysis highlights areas of improvement for future trials. Performance bias is the area with the highest risk of bias. Although recent studies have shown that blinding of participants is possible [48], it is not easy to eradicate this bias where the intervention is a physical one such as surgery or physiotherapy. The main area that could easily be addressed is detection bias. Blinding outcome assessment would have reduced overall risk of bias in four studies.

Strengths and limitations
The main strength of this review was the strict inclusion criteria based on clinical diagnosis confirming spinally referred leg pain of neural origin. A consequence of the tight inclusion criteria is the exclusion of 45 studies due to inadequate information on diagnosis of sciatica. As a result, our data reflect outcomes in patients with true nerve involvement. Insufficient reporting and low number of studies prevented a subgroup analysis according to type of physiotherapy intervention. Future trials with physiotherapy intervention should adhere to the TIDieR framework to fully describe the complexity of the intervention [49].

Conclusion
In summary, in patients with clinically diagnosed sciatica, physiotherapy interventions trialed to date provide inadequate evidence to make specific recommendations on their effectiveness in reducing pain or disability. The lack of convincing evidence may be due to several factors including incomplete trial reporting, clinical, methodological, and statistical heterogeneity, and trials lacking high methodological quality. Rather than continuing to perform trials in the heterogeneous population of 'sciatica', future studies should focus on reducing clinical heterogeneity, using contemporary physiotherapy interventions and high methodological quality to hopefully end the roadblock of discovery on the most effective physiotherapy interventions for these patient populations.

Acknowledgements
The authors would like to recognise the contribution of Neal Thurley, Outreach Librarian at the Bodleian Library, University of Oxford, in assisting with the search strategy for this systematic review. The authors would also like to thank Hubert van Griensven, Lecturer, University of Hertfordshire for assisting in translation from Dutch to English of one of the included studies. Author contributions All authors contributed to the study conception and design. LD performed literature search, LD and GJ reviewed articles for inclusion/exclusion with third-party arbitration by AS if required. Data collection was performed by LD and LAK and data analysis was performed by LD and AS. The first draft of the manuscript was written by LD and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.