Immediate effects of a lumbar spine manipulation on pain sensitivity and postural control in individuals with nonspecific low back pain: a randomized controlled trial

The present study is reported in accordance with the TIDieR recommendations [19]. This HVLA lumbar manipulation in subjects with LBP study is a randomized controlled clinical trial involving a control group (CG) and intervention group (IG) with a 1:1 allocation rate. It was conducted between January and March of 2015 with the approval of the Ethics Committee of the Universidade Federal do Rio Grande do Sul (number 834.848/CAEE 36001414.2.0000.5347) and registered at ClinicalTrials.​gov (NCT02312778), according to the CONSORT 2010 [20].

The participants were recruited through social media and newspapers as well as in physiotherapy clinics. Manual therapy studies use the clinical prediction rule to select and classify homogenous groups of participants that may benefit from the use of spinal manipulation as a therapeutic intervention [4, 21‐24]. The spinal manipulation clinical prediction rule has five criteria: pain duration of less than 16 days, no symptoms distal to the knee, score less than 19 in the fears and beliefs questionnaire, spinal stiffness, and internal rotation of the hip greater than 35 degrees [4].

The inclusion criteria for the study were men and women between the ages of 20 and 60 years with daily or almost daily lumbar pain in the previous 3 months [25] and who met at least four of the five clinical prediction criteria for HVLA lumbar manipulation [4, 23]. The exclusion criteria were the presence of radiating lower back pain, neurological alteration in the lower limbs (sensitivity, muscle force, and/or patellar or Achilles reflexes), previous surgery, medical diagnosis of spondylolisthesis, spinal stenosis, inflammatory disease, cancer, lower limb musculoskeletal degenerative diseases, pregnancy, pathologies and/or medications that may affect balance, osteopenia and/or osteoporosis, and women over 50 who had not had a bone densitometry exam.

Prior to intervention, the most hypomobile vertebra within the L1 to L5 vertebral segment was identified in all the members of both groups, using the clinical posterior-anterior vertebral pressure test applied with the subjects in ventral decubitus. Testing was performed on the spinous processes of the vertebrae. The osteopath placed the hypothenar eminence of the hand over the spinous process of the vertebra to be tested and applied gentle but firm pressure on the spinous process. By doing so, the stiffness of each vertebra was judged as either normal, hypomobile, or hypermobile. Thus, the most hypomobile vertebra was identified by comparing the mobility of the vertebrae immediately above and below [4]. It is important to highlight that, while assessing for segmental hypomobility is common in manual therapy practice, this test has only fair inter- and intra-rater reliability (kappa <.40) [26].

Each subject from both the CG and IG received a single intervention. All interventions were conducted on an examination table, with the subject in right lateral decubitus (Fig 1) because assessment on that side presents better intra- and inter-reproducibility [27, 28]. All interventions were conducted by an osteopath with 3 years of experience, who had been trained to identify vertebral mobility and perform the spinal manipulation (HVLA).

The participants allocated to the CG received simulated manipulation with no intended therapeutic effect. The participants allocated to this group were positioned in the right lateral decubitus position with the left leg flexed at the hip and knee and the left foot resting in the right popliteal fossa without stretching the paravertebral tissues (Fig 2). The participant remained in the position for approximately 20 s without receiving the HVLA thrust, which is the average time required to carry out HVLA lumbar manipulation [29]. For the IG, the HVLA lumbar manipulation was conducted according to Gibbons and Tehan [30], by locating the hypomobile vertebra when performing the thrust (Fig 3). During the manipulation it was unnecessary to produce an audible ‘pop’ [31].

All assessments were performed by the same assessor, following the same sequence, namely: numerical pain scale (subjective pain intensity), algometer (pressure pain threshold) and test on force platform (postural outcomes).

A self-reported 10-point-scale [32, 33] was used to assess the subjective pain intensity at rest. Pressure pain threshold (PPT) was assessed using a 10 kgf analogic pressure algometer (Wagner Instruments, Greenwich, CT-USA). The PPT was assessed on three sites: the lumbar spinal processes of the most hypomobile vertebra identified during the posterior-anterior vertebral pressure test; and bilaterally on the spinal erector muscle group, located 5 cm each side of the lumbar spinal processes [34]. The assessment was carried out with the participants lying in the prone position (Fig. 4), in accordance with the procedures used by Oliveira [34]. In short, the assessor pressed the algometer at a rate of approximately 0.5 kgf/s. Participants were asked to say “pain” when the sensation of pressure or discomfort became a clear sensation of pain. Three measurements were collected for each site at 30-s intervals. The average of three measures was used for data analysis. If the participant did not report pain at a force equivalent to 10 kgf, the test was interrupted and this value was considered the PPT. Prior to the assessment, the assessor performed 2 demonstrations of the procedure on the extensor muscles of the dominant forearm to ensure that the participant understood the test. While no information was found regarding the amount of difference between two assessments that should be considered clinically important using the analogic pressure algometer, van der Roer et al. [35] recommended that when using a self-reported 10-point-scale a difference of 2.5 points is clinically important.

Considering the possibility that the excitatory threshold of the mechanoreceptors is influenced by spinal manipulation [1], a secondary outcome, postural control was assessed in two ways: based on the COP and the Center of Projected Gravity (COPG). A BTS P-6000 force platform (BTS Bioengineering, Milan-Italy) with a sampling frequency of 500 Hz was used to evaluate the COP.

To evaluate the COPG, the SMTS DX high-definition motion capture system (BTS Bioengineering, Milan-Italy), consisting of 10 infra-red cameras with a sampling frequency of 500 Hz and a spatial resolution of 4 megapixels, was used. First, the barycenter of four reflective markers placed on the anterior and posterior superior iliac spines were used to estimate the center of gravity. This method was adapted from the sacral method, which is used for walking [36, 37]. Then, the COPG was calculated by projecting the three-dimensional coordinates of the center of gravity onto the floor. The COP and COPG signals were smoothed using a fourth-order low-pass Butterworth filter with a cut-off frequency of 4 Hz [38]. The variables for postural control were [11]:

The collection environment was controlled in order to avoid sound or visual stimuli that could affect the evaluation of the COP. The researcher reviewed the procedural instructions with each participant. In accordance with the guidelines from Zok et al. [41], the participants were instructed to “stand as still as possible and stare at the fixed target in front of you” while in the orthostatic position on the force platform, with their feet approximately shoulder-width apart and their arms resting at the side of the body. The subjects from both groups were asked to adopt a semi-static upright posture, which was held for 30 s, eyes open with the gaze fixed on an x-shaped marker placed 3 m in front of the platform at a height of 1.75 m from the ground.

The procedure was repeated three times, generating three curves both before and after the intervention for each group. The initial analyses were performed based on Curve 2, arbitrarily. If any problems were encountered when collecting Curve 2 (i.e. unwanted participant movements, noising signal, etc) then Curve 1 was used. If necessary, Curve 3 was used as the last option. If such problems persisted, the participant would be excluded.

When calculating the sample size, the subjective pain intensity was taken into account based on data from de Oliveira et al. [34], who also investigated the immediate effects of manipulative therapy in patients with low back pain. Using the G-Power software version 3.1.7 (Universität Kiel, Germany), the t-test family (means: difference between two dependent means – matched pairs), means and standard deviations pre- (6.07 ± 2.12) and post- (4.16 ± 2.34) intervention, and assuming a correlation between groups of 0.5, an effect size of 0.85 was determined. Thus, assuming an alpha of .05 and power (1-β) of 0.80, a total of 10 subjects per group was necessary. Assuming a loss of 20%, we decided on 12 subjects per group.

The group randomization was simple and was generated using Random Allocation Software (Informer Technologies, Inc.) by a researcher uninvolved in the subsequent research stages. This same researcher distributed the labels in sealed opaque envelopes, respecting the numbering on each label.

This study was participant-blinded and assessor-blinded. The participants did not know to which group they had been allocated. They were informed there would be two different interventions and that they would be offered an alternative intervention after the end of the study if they wished. Nobody asked for an alternative intervention. However, the participant-blinding was not formally assessed. The evaluators rating the variables were also blinded because they did not know which intervention the participants had received, since they were not present at the time of the individual interventions.

The first stage consisted of recruiting individuals and having them sign the consent form. Then, one of the researchers completed an assessment form containing items such as the participant’s age, body mass, height (Parisian point equivalent to 0.66 cm), level of physical activity (considered when executed with guidance from a health professional and at least three time a week), and study eligibility criteria, including the clinical prediction rule.

In the second stage, the primary and secondary outcomes were evaluated prior to beginning the intervention. The third stage, consisting of the post-randomization intervention, was always performed by the same osteopath without the presence of any of the other researchers. The post-intervention evaluation was performed during the fourth stage. There was no interval between stages. The time between pre-intervention evaluation and intervention, and between intervention and post-intervention evaluation was just enough for the respective researchers to leave and enter the room.

The data normality was assessed using the Kolmogorov–Smirnov test. If the data distribution was normal, parametric tests were applied to analyze the data. For the variables with non-parametric distribution, transformations were performed (log [x], sqrt [x], 1/x, etc.). SPSS 22.0 software (IBM) was used with a significance value of 5%.

The descriptive variables were presented as the mean and standard deviation with a confidence interval of 95%. For all outcomes, multiple mixed two-way ANOVAs were performed, with the group (control or intervention) as the independent factor and time (pre- or post-intervention) as the repeated factor. Given that the statistical test for the primary outcome was different from that in the original pre-register, the G-Power software was used to calculate the posthoc power (1 – β). Effect size (r) was calculated using the square root of the partial-squared eta (sum of squares of the effect divided by the total sum of the squares of the evaluated effects plus the sum of the squares of the error of the evaluated effect) and was classified as small when r ≤ 0.10, average when 0.10 < r 0.30, and large when r = 0.50 [42, 43].

This study has several limitations. The first is related to the study design, which include only a single intervention and no follow-ups. Secondly, the limited sampling duration for the standing balance trials could be considered a potential study limitation. Thirdly, the participant-blinding was not formally assessed, which means there is no guarantee the participants were unaware the nature of the intervention they received (placebo or real). Another limitation refers to the sample size calculation, which only considered the subjective pain intensity. This fact may have under-powered our study for secondary outcome measures. Probably the main limitation is related to the sample size calculation, which we had assumed within-group comparisons. Thus, the results regarding the between-group comparisons are underpowered. In other words, our results regarding the lack of difference between groups need to be interpreted with caution (type II error possibility).