Transcutaneous electric nerve stimulation (TENS) for acute low back pain: systematic review

2.1 Data sources and search

The electronic databases MEDLINE, EMBASE, CENTRAL, CINAHL, PsycINFO and Cochrane Database of Systematic Reviews (inception to May 2018) were searched for reports of placebo-controlled randomised controlled trials (RCTs) evaluating TENS for acute LBP (Appendix Table 1). Additionally, we screened reference lists of included RCTs and relevant systematic reviews to identify any other relevant studies (Fig. 1).

Fig. 1:

Summary of search.

2.2 Study selection

We included English language RCTs evaluating TENS for acute, non-specific LBP. Two reviewers (CAS and JW) screened titles and abstracts of retrieved studies, and two reviewers drawn from a pool of three reviewers (JB, SG and CAS) independently inspected the full manuscript of potentially eligible RCTs to determine eligibility, with disagreements resolved by consensus.

2.3 Data extraction

Two reviewers drawn from a pool of three reviewers (JW, SG and CAS) independently extracted study and participant characteristics, and outcomes data from included studies. Missing data were obtained by contacting authors or estimated using the methods described in the Cochrane Handbook for Systematic Reviews of Interventions [20].

2.4 Outcomes

Trials were included if they reported on pain, disability or adverse events outcomes. Pain outcomes were converted to a common 0 (no pain) to 100 (worst possible pain) scale. The pain intensity measures used in the retrieved trials were visual analogue scale (VAS) scores (range, 0–100) and numerical rating scale (NRS) scores (range, 0–10). The NRS was converted to the same 0–100 scale as in the VAS, as these two pain measures have been shown to be highly correlated and when transformed, can be used interchangeably [21].

2.5 Data synthesis

We present results as mean differences (MD) expressed in points on a 0–100 pain scale and 95% confidence intervals rather than standardised mean differences (SMD). We considered treatment effects in the range 10–19 points as small; >20 points as moderate and >30 points as large. These values are consistent with the proposed thresholds for clinically important changes in pain response from the literature on chronic pain [22]. Effects <10 points were considered as not clinically meaningful [23].

2.6 Follow-up

Outcomes were grouped with respect to follow up period: immediate term (outcomes measured <2 weeks after administration of intervention), short term (outcomes measured between 2 and 6 weeks after the intervention) and intermediate term (outcomes measured 6–12 weeks after the intervention). We considered immediate term pain relief as the primary time point.

2.7 Quality assessment

Risk of bias was assessed by two reviewers (JB, SG or CAS) using the 11-item PEDro scale [24], [25], [26], which is a valid and reliable method of rating risk of bias of RCTs [24], [25], [26]. Each item (excluding the item for external validity) is scored as either present (1) or absent (0) to give a total score out of 10. Trials scoring <7/10 on the PEDro scale were defined as high risk of bias; those scoring 7 or more were considered at low risk of bias [25].

We used the methods described in the Cochrane Handbook [20] to determine a GRADE rating for the overall quality of the evidence for an intervention. This method is described elsewhere [27], [28], but briefly the quality of evidence was downgraded a level for each of four factors: poor study design [25% or more of trials, weighted by sample size, have a low PEDro score (<7/10)], inconsistency of results (25% or more of the trials, weighted by sample size, have results which are not in the same direction or evidence of clinical heterogeneity), imprecision (small sample size and/or wide 95% CI) and publication bias (assessed using funnel plot analysis/Egger’s regression test where there were 10 or more studies). Where Egger’s regression two-tailed p-value was <0.10, the overall quality of evidence was downgraded by one level [29]. It was not necessary to downgrade for indirectness (when the trial context is not the same as the review question) as this review evaluated administration of specific intervention (TENS). The quality of evidence was defined as “high quality”, “moderate quality”, “low quality”, and “very low quality” [20].

2.8 Data analysis

We planned to conduct meta-analysis using Review Manager (RevMan version 5.3) [30]. Where results could not be pooled (e.g. high heterogeneity, I²>85%), results from individual studies were assessed and reported separately. For adverse events data, we were interested in the proportion of participants experiencing one or more adverse events and where possible, calculating a risk ratio based on this information.

3.1 Pain

One study [13] provided low quality evidence (downgraded for imprecision, risk of bias) of immediate improvements in pain with ~30 min TENS treatment compared with sham TENS; MD −28.0 (CI −32.72, −23.28) in patients with moderate to severe acute LBP being transported to the hospital by paramedics. This is shown in Fig. 2. This effect size is just beneath our threshold for large improvements in pain. No longer term results were reported. Another study evaluating a course of TENS over 5 weeks [32] provided inconclusive evidence regarding the effect of TENS for acute LBP in the immediate term; MD −5.90 (−16.42, 4.62) (low quality evidence; downgraded for imprecision, risk of bias). Pooled effects from two low quality studies [31], [32] similarly provide inconclusive evidence regarding the effectiveness of a course of TENS (over 4–5 weeks) for the short term; MD −2.75 (CI −11.63, 6.13) (low quality evidence; downgraded for imprecision, risk of bias). For the intermediate term, one study [32] provided a lack of evidence of an effect of TENS vs. sham TENS; MD −2.20 (CI −13.97, 9.57) (low quality evidence; downgraded for imprecision, risk of bias).

Fig. 2:

Effects of TENS on pain. Pain is measured on a 0 (no pain) to 100 (maximum) scale. Effects are mean differences in outcome with 95% confidence intervals (CI).

3.2 Adverse events

There were limited data on adverse events. There were no adverse events outcomes reported in two of the studies [13], [32]. In one of the studies [31], an equal proportion of people in the active (2/29) and control (2/29) groups experienced an adverse event, however, the nature of these adverse events was not reported.

4.1 Principal findings

There is insufficient evidence to support or dismiss the use of TENS for acute LBP. Currently there is low quality evidence that treatment with TENS (~30 min) provides immediate improvements in pain compared with sham TENS for patients with moderate to severe acute LBP being transported to hospital by paramedics. This improvement is just below our threshold for a large improvement in pain (>30 points on a 100-point pain scale). There is a lack of evidence of an effect of TENS over a course of 4–5 weeks in a clinical setting (low quality evidence).

4.2 Strengths and weaknesses of the study

The strengths of the study are the use of a comprehensive search strategy, use of a validated scale to measure risk of bias and double data extraction. A limitation of this review is that we only retrieved a few trials, none of which provided data on long term outcomes and there was limited data on safety. A limitation common to all three of the included studies was their small sample size and limitations in study design (all were rated as high risk of bias). None of the studies could truly guarantee patient, therapist and assessor blinding thereby introducing potential performance bias as individuals in the treatment arm would have been more likely to perceive an effect. Furthermore, two of the studies [31], [32] did not report adequate methods for concealed allocation, which may have resulted in selection bias, as particular patients may have been preferred over others for allocation to treatment. Another uncertainty is the unusual results reported in the one positive trial [13]. The study reported that after 30 min treatment with active TENS, mean pain scores almost halved, whereas pain scores for the placebo group increased slightly to 77. This is a very unusual pattern of results for a placebo controlled trial in this field. As such, we would caution readers to consider these findings carefully, as it is based on low quality evidence, and because further studies are needed to determine the true effectiveness of TENS for acute LBP.

4.3 Strengths and weaknesses in relation to other studies

This review provides a comprehensive evaluation of TENS in acute LBP, taking into consideration clinical context and treatment duration. We have also included studies previously excluded in a Cochrane review of TENS for acute pain [33]. Previous systematic reviews evaluating TENS for LBP have focused on the treatment of chronic LBP and it is still inconclusive as a viable treatment option in this context [34], [14], [15], [16]. A 2009 review [19] described moderate effect of TENS for acute LBP, based on the pooled estimate from two studies [13], [31], however there was no quality of evidence assessment for this finding. We considered these studies [13], [31] to be sufficiently heterogeneous to challenge pooling (course of treatment in outpatient clinic vs. single treatment in acute care emergency setting, respectively). As such we presented the findings separately which has implications on the overall findings.

4.4 Implications for clinicians and policy makers

The current NICE guideline [17] advises against the use of TENS for LBP, however, it is unclear if this recommendation relates to acute or chronic LBP. The findings from our review are inconclusive, and support the need for larger and more rigorous trials to resolve this uncertainty. If the effects are truly as large as reported by Bertalanffy et al. [13], it is possible that TENS can be integrated in the care plan for patients with acute LBP being transported to hospital. This approach may also reduce the need for strong medications, e.g. opioid analgesics, upon arrival at the hospital or during emergency transport to the hospital. Evidence from more robust clinical trials are not only needed to establish efficacy and safety but also to inform consideration around cost to the individual, particularly as TENS devices are generally more costly than a standard packet of analgesic medicine. Cost effectiveness studies would therefore also be important, to ascertain whether this intervention is good value.