Assessment of variability in traction interventions for patients with low back pain: a systematic review

Lumbar traction is a commonly used method to treat patients with low back pain (LBP) with or without sciatica. In the UK and the US, lumbar traction is used by 41 and 77% of outpatient rehabilitation providers respectively [1, 2]. Despite this common use of lumbar traction in the clinical setting, several systematic reviews have concluded that lumbar traction has little or no value on the clinical outcomes of pain intensity and functional status. The reviews also suggest that traction does not appear to lead to quicker return to work among people with LBP with or without sciatica [3‐5]. These conclusions present a clear discordance between evidence-based recommendations and how lumbar traction is regarded in current clinical practice [1, 2, 6].

The earliest systematic review, conducted in 1995, included 17 randomized controlled trials (RCTs) that assessed traction on neck and low back pain [7]. Of the 17 RCTs, only 3 (2 lumbar, 1 cervical) had good quality. This systematic review concluded that traction efficacy was unclear, and called for more proper design and better methodological quality in future traction trials.

An update of the above systematic review, published in 2006, included 24 RCTs that assessed the effectiveness of traction in the management of LBP [4]. The RCTs were selected if they examined any type of traction on acute, subacute, or chronic LBP with or without sciatica. Of the 24 RCTs, only 5 were considered of high quality, and suggested that there was strong evidence that traction was not effective in the management of patients with mixed duration of LBP with or without sciatica. However, there was moderate evidence that autotraction was effective in the management of patients with mixed duration of LBP with or without sciatica.

The most recent update of the above systematic reviews, published in 2013, included 32 RCTs that assessed the effectiveness of traction in management of LBP using the same selection criteria that were used previously [5]. Of the 32 RCTs, 16 studies were considered to have low-risk of bias. The overall conclusion of this systematic review suggested that traction, alone or in combination with other interventions, has little or no impact on the clinical outcomes of pain and function on people with mixed duration of LBP with or without sciatica. However, this systematic review suggested that large, high-quality studies, were still required to make definitive conclusion about traction effectiveness.

Interestingly, rehabilitation providers are reported to be aware of the recommendations from systematic reviews against traction, yet 64% of them disagree with these recommendations and 25% remain undecided [1]. One explanation for this large amount of disagreement may be that rehabilitation providers regularly report the empirical observation that some patients are dramatic responders to lumbar traction [1]. This clinical observation may be driven by the ability of rehabilitation providers to somehow recognize certain clinical patterns that allow them to match patients’ symptomatic presentations to specific traction strategies [2]. This pattern recognition of a traction subgroup has been recommended within the treatment-based classification system, which is commonly utilized by physical therapists [8‐10].

Another explanation for this divergence between the continued use of lumbar traction by rehabilitation providers and the recommendations against it from systematic reviews may be related to the variability in the delivery methods of traction among the RCTs included in these systematic reviews [5]. The variability in delivery of traction interventions can stem from using different types of traction, different traction parameters, and different patient populations [5]. When RCTs with different traction methods are pooled together, the overall treatment effect size is diluted.

A number of RCTs suggest that traction can be an effective intervention in the management of patients with LBP. Fritz et al. found that mechanical traction in combination with extension exercises can result in significant improvement in disability and fear-avoidance beliefs after two weeks of treatment compared to extension exercises alone for patients with acute LBP and nerve root compression symptoms [11]. Also, Prasad et al. found that using inversion traction plus physical therapy in patients awaiting surgery for disc herniation helped 77% of them avoid surgery compared to physical therapy alone that helped only 22% avoid surgery [12]. Additionally, Kim et al. found that when prescribing the inversion traction for patients with chronic LBP, the tilt degree of the traction table matters [13]. Kim et al. found the a tilt degree of 60 resulted in improve levels of pain, spine flexibility and trunk extensors strength compared to tilt degrees of 30 or 0 (supine position) [13]. Further, Simmerman et al. found that aquatic traction resulted in significant pain reduction and centralization of symptoms compared to land-based exercises in patients with chronic LBP associated with nerve root compression symptoms [14]. Finally, Diab and Moustafa found that traction in combination with stretching and infrared radiation resulted in significant improvement of pain and disability levels compared to stretching and infrared in patients with chronic LBP at 6 months [15]. Collectively, these individual RCTs point to the potential effectiveness of traction in the treatment of patients with LBP with or without sciatica.

Given the possibility that the treatment effect size could be diluted when heterogeneous studies are pooled together, the purpose of this review is to map the evidence regarding the diversity in traction delivery methods. Specifically, this systematic review will explore the various traction intervention protocols by reporting on traction types, traction parameters, dosage and patients’ characteristics. Assessing the diversity of traction delivery methods will help determine the appropriateness of conducting meta-analysis.

This is a systematic review of RCTs that have included some type of lumbar traction as a treatment intervention. The data collection of this systemic review started in September 2016. The first step was to perform a generalized search of the literature using the key words “lumbar traction” in the following databases: Cochrane Central Register of Controlled Trials, MEDLINE, EMBASE, and CINAHL. The search returned several systematic reviews that directly addressed the topic of lumbar traction. The most recent Cochrane systematic review included RCTs of lumbar traction published from inception until August 2012 [5]. From this Cochrane review [5], all of the RCTs contained within the evidence tables were extracted and compared with those found in the evidence tables of the other systematic reviews [3, 7]. This step was necessary to confirm that the recent Cochrane review included all of the RCTs (or more) contained in the evidence tables of the older systematic reviews.

The next step was to update the systematic review by searching for additional RCTs published between September 2012 and December 2016. To identify new RCTs, the following keywords were used to search the same databases: “traction” OR “traction therapy” OR “traction physical therapy” OR “decompression” OR “unloading”; OR “lower back” OR “low back pain” OR lumbar pain OR sciatica OR radiculopathy OR lumbago OR backache. After these additional RCTs were retrieved, two authors (MAlr and MAlm) examined the titles and abstracts to select studies that would potentially be worthy of full text review. After that, the two authors extracted and synthesized the data about the specific traction protocols used in each RCT by reporting on the traction type, traction parameters and patient population. The authors used consensus to agree on which trials would warrant a review of the full text article for potential inclusion in this systematic review. When disagreement occurred, the third author (MS) was consulted to resolve the disagreement.

For a study to be included in this systematic review, it had to be an RCT of patients 18 years of age or older; with acute, subacute or chronic LBP; with or without sciatica. Also, the studies had to include at least one type of traction: manual, auto-traction, gravitational, aquatic and mechanical traction. The traction may or may not have been combined with other interventions such as manual therapy or exercise, with the requirement that traction was the primary intervention. Additionally, any type of comparison group was allowed including placebo, sham or active intervention. Finally, the RCTs must have included at least one clinically relevant outcome measure such as numeric pain scale, self-reported function, global measure of improvement, or return to work. These inclusion criteria are similar to those reported in the most recent Cochrane systematic review [5].

The last step was to create an evidence table that combined the RCTs extracted from the most recent Cochrane systematic review with the new RCTs identified through the updated search (Table 1). Because this review focused on extracting details about the specific traction protocols used in each RCT, there was no need to collect data on quality and risk of bias. This format was used to allow the reader to quickly visualize the similarities and differences in traction protocols across the RCTs.

The most recent Cochrane systematic review included 32 individual lumbar traction RCTs published from inception through August 2012 [3, 5, 7]. Our updated search identified an additional 14 newer studies that were published between September 2012 and December 2016. Of these newer studies, 5 RCTs were combined in Table 1 with the previously identified 32 RCTs for a grand total of 37 RCTs [12, 13, 15‐17]. This search process is summarized in (Fig. 1) [18].

In the columns of Table 1, the primary author and date of each RCT are organized in chronological order. In the rows of Table 1, the qualitative factors and traction parameters of each RCT were included. One study included the results of two RCTs in a single publication [19], so both of those two RCTs were included in Table 1, each reported in a separate column.

Table 2 uses descriptive statistics to summarize categories of traction parameters and patient characteristics from the included RCTs. Of the 37 RCTs, 59.5% used some type of mechanical traction while the remaining studies used auto-traction (16.2%), manual (10.8%), inversion (8.1%), or aquatic traction (5.4%). In 27% of the trials, traction was used in combination with some other type of rehabilitation intervention, such as exercise or physical agents.

The amount of force used during the traction treatment varied widely across these 37 studies, ranging from 2.3 kg in one trial to 100% of body weight in another. In 35% of the RCTs, the amount of traction force was determined by using some arbitrary percentage of the patient’s body weight that varied from 20 to 100%. However, another 37.8% of studies used an arbitrary pre-determined amount of weight ranging anywhere from 2.3 kg to 60 kg as the traction force. The traction rhythm was evenly distributed between continuous and intermittent types of application, with each type of application used 43.2% of the time. In the remaining of the studies, the traction force and rhythm were not clearly described.

The traction session time and treatment frequencies were very difficult to categorize. The traction sessions lasted from 3 to 4 min in duration in some trials, to more than 30 min in other trials. The frequency of application of the traction treatments used in these trials varied from as few as one single session, to as many as 20 traction sessions applied over 6 to 10 weeks.

Other traction parameters were also found to vary widely across these 37 traction RCTs. With respect to patient positioning during the application of traction, 29.7% of the trials had the patient lie in a supine position, 16.2% used prone positioning, 5% applied traction in standing, 3% used a side-lying position, and in 43.2% of the studies there was no clear description of patient positioning.

Regarding the onset and duration of symptoms, 54% of the studies included patients with a mixed acuity of LBP, 16% included only patients with chronic LBP, 8% included only subacute cases of LBP, 8% included only acute cases, and 14% of the studies did not provide any clear description of symptom acuity. With respect to leg symptoms (e.g. sciatica), 59% of the RCTs included only patients with sciatica, 24.3% included a mixture of patients with and without sciatica, 10% included only patients without sciatica, and the remaining studies did not contain any report of leg symptom or sciatica status. In 24% of the included studies, the patients had a specific LBP diagnosis such as herniated disc, while the other 76% of studies did not report any specific pain generator or anatomical cause of the LBP (non-specific LBP).

The results of this systematic review show that there are widespread variations in most of the traction protocols used in the RCTs found in the traction literature. When examining Table 1, each RCT appears to have a distinct combination of traction type and traction parameters applied to different populations of patients with LBP, with or without sciatica. There is no main theme or pattern that emerges about the traction parameters used in the studies included and rated in the previously published systematic reviews. This variability in traction delivery protocols represents a primary gap in the traction literature and might be a key factor that underpins the negative conclusions about traction in the most recent systematic review [5].

RCTs of lumber traction have employed a mixture of traction types, traction parameters and patient populations. The large variability in the delivery of traction intervention provides evidence that the RCTs included in systematic reviews were extremely heterogeneous. This suggests that negative conclusions about the overall clinical effectiveness of lumbar traction should be interpreted with caution. More research about the effectiveness of traction is still necessary, and future trials should consider two important points: (1) discovering optimal dosage and traction parameters to inform the development of standardized traction protocols, and (2) discovering the important baseline variables predictive of successful traction response. By standardizing the traction dosage and parameters, improving patient selection criteria, and response to axial force, more clinically meaningful traction research could be conducted.