A mixed treatment comparison of selected osteopathic techniques used to treat acute nonspecific low back pain: a proof of concept and plan for further research

Methods

A thorough literature search was performed in April 2020 according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [5]. There were no outside sources of funding for this research project and institutional review board approval was not required.

Studies were identified by searching the MEDLINE/PubMed, OVID, Cochrane Central, PEDro, and OSTMED.Dr databases; scanning the reference lists of articles; and using the Canadian Agency for Drugs and Technologies in Health grey literature checklist [6]. Each database was searched from inception to April 1, 2020. The following search terms were used: acute low back pain, acute low back pain plus physical therapy, acute low back pain plus spinal manipulation, and acute low back pain plus osteopathic manipulation.

Articles studying outcomes of patients with ANLPB were selected to generate an MTC assessing pain related outcomes. To be eligible for the MTC, each study had to have been published in English; have been an RCT with human subjects; have had a population of working age (16–70 years); and had to have had an objective outcomes measure of pain assessed within one month from the date of onset.

Eligibility assessment was performed in an unblinded manner by the single author (J.W.P.) using the National Institute for Health and Care Excellence (NICE) methodology checklist for reviewing RCTs [7]. The NICE methodology checklist is a tool used to qualitatively assess RCTs for risk of four types of bias: selection bias, performance bias, attrition bias, and detection bias [7]. As there was only one reviewer, the author adapted the NICE checklist to add a quantifiable measure for determining suitability. Each category of bias was initially assigned a numerical value of one. One point was added for each ‘no’ or ‘unclear’ answer. The remaining value was the quality measure for the given type of potential bias (4=high risk; 3=moderate-high risk; 2=low-moderate risk, 1=low risk). The quality measures were summed and then divided by 4. The average of the quality measures was then rounded to the nearest integer. Studies with an average value of three or four were eliminated from the NMA due to unacceptable risk of bias.

The author developed a data extraction sheet based on the NICE methodology checklist for reviewing RCT. Information was extracted from each study that qualified for full text assessment, including characteristics of trial participants (age and sex); type of intervention and comparison (placebo, nothing, or another intervention); type of outcomes measure; presented results; and bias assessment. The outcomes measure chosen for this NMA was the Visual Analogue Scale (VAS) instruments.

The NMA was performed using the GeMTC user interface for automated NMA utilizing a Bayesian hierarchical model of random effects [8], [9]. Again, the primary outcome measure was mean VAS score. Heterogeneity priors were determined automatically by the software [10].

The network geometry is presented visually as a figure (Figure 1), with the treatment nodes representing the various treatments and comparisons from each of the included studies. The solid lines between some of the nodes indicate direct comparisons made between the intervention and control groups from the included RCTs. Indirect comparisons were made between nodes not connected by a solid line. Potential scale reduction factor (PSRF) was calculated to assess for adequate convergence within the network; PSRF compares the variation within each chain to the variation between chains. The PSRF starts with a high value, and slowly approaches 1.0 as the chains become more similar. Residual deviance (D_res) was used to assess model fit. Model fit was deemed adequate when D_res was equal to or less than the number of independent data points. Empirical examination for detecting across-studies (publication) bias was performed using a GeMTC produced comparison adjusted funnel plot of study effect.

Figure 1:

Presentation of the mixed treatment comparison network structure. The circles represent the interventions. The solid lines indicate direct comparisons between interventions. The dashed lines indicate indirect comparisons of interventions. OMT, osteopathic manipulative treatment; Tx, treatment; SCS, strain-counterstrain; ME, muscle energy; HVLA, high-velocity low-amplitude.

Conventional care was selected as the baseline against which all other treatments were compared. Differences between treatments were considered significant (at the 5% level) if their confidence intervals did not overlap the no-effect line. In addition to relative effects, the Bayesian analysis also generated a rank probabilities table.

Results

The literature search identified 805 studies meeting search criteria. After removing duplicates, 483 studies remained to be screened for eligibility (Figure 2). The screening excluded 469 studies. The full text of the remaining 14 studies [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24] were assessed by the single author (J.W.P.) for risk of bias. The results of the risk of bias assessment are presented in Table 1. Seven studies [12], [15], [18], [20], [22], [23], [24] failed to present the numerical values, confidence intervals, or standard deviations of their findings. High overall risk of bias was identified for two of the remaining studies [11], [14]. High risk of performance bias was determined for one study [14], and high risk of selection bias was suspected for one study [20]. This left five studies [13], [16], [17], [19], [21] to be used for the MTC model. The characteristics of each, including interventions, are presented in Table 2. Each of the groups in the included studies had small sizes, ranging from 30 to 70 subjects. The OMT groups were particularly small, with 30–44 subjects. High-velocity low-amplitude technique (HVLA), strain-counterstrain technique (SCS), muscle energy technique (MET), and a mixed OMT protocol were included in the model. It is also worth noting that the MET and SCS interventions were limited to treating the erector spinae muscle group.

Figure 2:

Preferred Reporting Items for Systematic Reviews and Metaanalyses (PRISMA) literature search flow diagram.

The network geometry of the resulting model consisted of six nodes, five direct comparisons, and no closed loops (Figure 1). The exercise node had a sample size of 118 patients from two studies [16], [17]. The counterstrain node had 74 subjects from two studies [16], [19]. The mixed treatment protocol node had 33 subjects from a single trial [13]. The HVLA node had 37 subjects from only one study [21], and the muscle energy node contained 30 subjects, also from a single study [19]. As such, the model relied heavily on indirect comparisons. The PSRF compares the variation within each chain to the variation between chains. The PSRF was below 1.05 for each of the comparisons made in this model, indicating adequate convergence of the model (Table 3). The overall D_res was 10, with 10 data points; this is consistent with adequate model fit. The per-arm residual deviance for each of the included studies determined that there were no important outliers (Table 4).

Table 5 presents comparisons of the studied interventions with mean difference and 95% confidence intervals (CI). Each cell in the league table represents the effect of column-defining intervention relative to the row defining intervention. The 95% CI for each comparison crossed zero, indicating that there were no statistically significant mean differences. The relative effects plot (Figure 3) graphically presents the VAS mean difference and 95% CI for each active treatment relative to conventional care, suggesting that there was no statistically significant decrease of reported pain when exercise or OMT were added to conventional treatment. However, the high-velocity low-amplitude (HVLA) arm and OMT mixed treatment protocol arm trended toward demonstrating an added benefit. Inspection of rank probabilities (Table 6) demonstrated that HVLA and OMT mixed treatment protocols plus conventional care were ranked superior to conventional care alone for improving ANLBP (Table 6).

Figure 3:

Relative effects plot. OMT, osteopathic manipulative treatment; Tx, treatment; SCS, strain-counterstrain; ME, muscle energy; HVLA, high-velocity low-amplitude.

Heterogeneity was not a consideration for this model because each direct comparison in the network had only a single study; inconsistency was not assessed because the model had no closed loops [25]. The comparison adjusted funnel plot of study effect detected no indication of across-studies bias (Figure 4).

Figure 4:

Funnel plot for detecting across-studies bias. OMT, osteopathic manipulative treatment; Tx, treatment; SCS, strain-counterstrain; ME, muscle energy; HVLA, high-velocity low-amplitude.

Discussion

The findings of this systematic review and meta-analysis are not conclusive, but they suggest that the HVLA and OMT mixed treatment protocol arms in prior studies trended toward being superior to erecter spinae muscle energy (MET) and strain-counterstrain (SCS) techniques for reducing ANLBP in the first month after onset. These findings also suggest that the addition of prescribed exercises offered no therapeutic advantage over conventional treatment alone.

The results of this MTC should not be used to infer that MET and SCS are ineffective techniques for ANLBP. However, the findings do suggest that MET and SCS directed at erector spinae dysfunction trended toward being inferior to conventional care alone. This may be related to the effectiveness of the studied techniques, or erector spinae dysfunction may not be a major contributor to ANLBP.

These findings agree with two previous Cochrane reviews, one examining the use of muscle energy technique for low back pain [26] and one exploring the effect of prescribed exercises for low back pain [27]. A 2016 Cochrane review [26] determined that there was low‐quality evidence of no clinically relevant difference from adding MET to other interventions for management ANLBP with the outcome of pain. There was low‐quality evidence of an effect in favor of MET for functional status. However, the authors also noted that the quality of research related to testing the effectiveness of MET was poor [26]. The 12 studies in the review were generally small (n=20–72) and at high risk of bias due to methodological inadequacies; there was not sufficient evidence to reliably determine whether MET was likely to be effective in practice [26]. The authors of that review [26] also stated that there is a need for larger, higher‐quality studies with more robust methodology. Ideally, all studies should clearly describe all methods, have larger sample sizes, use robust methods of statistical analysis, demonstrate baseline equivalence of patient characteristics between groups, and use treatment protocols that can be generalized to clinical practice. A 2005 Cochrane review [27] of 61 randomized controlled trials with a total of 6,390 participants concluded that exercise therapy was as effective as either no treatment or other conservative treatments for ANLBP. However, the authors also noted that their review largely reflected limitations of the literature, including low quality studies with heterogeneous outcomes measures, inconsistent and poor reporting, and possibility of publication bias [27].

The model presented in this study is subject to the limitations noted by the authors of the previous Cochrane reviews [26], [27]. Heterogenous study methodologies and outcome measures were identified during the literature review. This resulted in a limited number of studies meeting inclusion criteria. Because of this, exercise was included as a variable to allow the inclusion of muscle energy technique and counterstrain in the final model. Fortunately, the use of exercise for acute LBP has been researched extensively, thus providing a point of reference from which to compare the results of the exercise arm of this MTC to evaluate the utility of this methodology. The sample sizes for each of the studies included in the model were small and likely contributed to the failure to demonstrate any statistically significant results (Table 2). Several of the included studies had moderate to high risk of performance bias because of failing to blind participants to their allocation group and lack of clarity as to whether comparison groups received the same care and support apart from the intervention studied (Table 1). A few of the included studies had moderate to high risk for detection bias for failing to blind investigators to the participants’ intervention and to important confounding factors.

A limitation that is unique to the present review was having a single author (J.W.P.) who performed the literature search and quality assessment. This may have led to the introduction of selection bias. An attempt was made to mitigate this issue by modifying the NICE checklist to generate a more objective quantitative measure of bias risk.

While this study failed to provide any definitive evidence upon which to make clinical recommendations, it did demonstrate the utility of performing NMA for MTCs of osteopathic modalities used to treat ANLBP. However, to take full advantage of this statistical technique, future studies should be designed with consideration for the shortcomings in osteopathic research identified by this study and noted in the previous Cochrane reviews [26], [27].

To address these deficiencies, the author proposes a coordinated series of RCTs with an identical design other than the studied intervention. Individual studies would be performed at multiple sites coordinated through an interinstitutional research network. The intervention would be a single OMT technique plus conventional care or a mixed treatment protocol plus conventional care. The chosen intervention would be used to treat all identified dysfunctions. The control could be a different OMT technique plus conventional care for all identified dysfunctions, sham OMT plus conventional care, or only conventional care. The choice of the intervention and the control would be left to the institution performing the study. Inclusion criteria for participants in the series suggested by the author here would be adults who meet the case criteria for ANLBP. ANLPB should be uniformly defined. A suggested definition is pain from the lower costal margin to the gluteal folds without defined pathology excluding somatic dysfunction of less than 30 days duration. The author suggests exclusion criteria as follows: contraindications to OMT, pregnancy, neurodegenerative disease, prior surgery to the lumbosacral spine, history of chronic low back pain, and positive “red flags” for serious spinal pathology. The author also suggests that each institution would be obligated to assure an adequate sample size to more precisely address the research question.

In the suggested series of future research, the author also recommends that physical assessment should be standardized for each of the studies. The structural examination would use a standardized protocol to assess for postural imbalance, psoas dysfunction, quadratus lumborum dysfunction, piriformis dysfunction, sacral dysfunction, pelvic dysfunction, and lumbar segmental dysfunction. Each dysfunction would be treated using only the intervention modality, active control modality, or sham technique as defined by randomized group allocation. Further, the author recommends several measures to mitigate bias. A validated method of randomization must be utilized to assign group allocation; group allocation should be concealed from investigators and participants. Groups should be comparable at baseline, including major confounding factors, such as age, sex, body mass index, smoking status, employment status, fear-avoidant behaviors, and catastrophization. Comparison groups should receive the same care apart from the intervention being studied. Guideline-based conventional care should be the same for all studies. The accepted definition of conventional care is to explain the usual course of ANLBP, advise remaining active, and provide a non-steroidal anti-inflammatory drug or acetaminophen if indicated [3]. The number and the characteristics of dropouts for each group should be noted. Future investigators should, in the authors’ opinion, assure an appropriate length of follow up, with intervention and control groups followed for an equal length of time; for example, outcomes measures may be obtained at baseline, two weeks, one month, and three months. Pain beyond three months would be considered chronic. This future study series could use self-reported pain and function ability as a definition of outcome; VAS and the Roland–Morris disability questionnaire [28] would be valid and reliable methods to determine the outcomes.

In this future study series suggested by the author, statisticians should be kept blind to the participants’ group allocation and to other important confounders. Statistical assessment would be performed to assess the difference in the means of the outcome measures. The results should be reported with the corresponding standard deviations or confidence intervals. Finally, each study would be submitted for publication no matter the outcome, to avoid publication bias in a future MTC. The results of each study would then be used to generate a MTC of osteopathic techniques for the treatment of ANLBP. The resulting model would be at extremely low risk for within-study bias, reporting bias, indirectness, imprecision, heterogeneity and incoherence. Data gathered relating to specific dysfunctions and group characteristics could also be used for network metaregression to determine how a covariate might interact with the treatment effects.

There are many benefits to be gained by using this future study design recommended by the author, beyond efficiency. The final product would provide level 1A evidence [29] comparing a variety of OMT techniques used for treating ANLBP. The acquired knowledge could then be applied to the development of clinical practice guidelines for the treatment of ANLBP. A study plan like this one would provide structure and training opportunities for institutions with fledgling clinical research programs. This approach to research is also aligned with the AOA’s strategic plans for research [30].

Conclusions

This study failed to provide evidence to recommend for or against any of the included OMT techniques or prescribed exercises to conventional care for ANLBP. This may reflect the limited number of previously-published studies meeting inclusion criteria due to methodological heterogeneity, failure to adequately report results, and unacceptable risk of biases. However, the model’s findings do align with some prior literature, suggesting that NMA for MTC of osteopathic techniques may be a useful method for examining some of osteopathic medicine’s more complex questions if the shortcomings in osteopathic research are addressed in future investigations.