Subjects

We aimed at enrolling around 80 participants for both symptomatic and control group. Previous sample size calculations for a group of controls and symptomatic patients with an NRS-11 score of 5.0 ± 2.1 using an Altman Nomogram [14] suggested recruitment of around 50 symptomatic and healthy participants each. We decided to exceed this number in order to compensate for potential dropouts.

All new patients entering a private chiropractic clinic in Wolfsburg, Germany were asked on the phone whether they would take part in this study. The healthy controls were friends and partners of already enrolled participants and were initially approached by them regarding the possibility of participation. If they displayed interest they were asked to contact the clinic for further details. After verbal and written information had been given, the subjects consented to participate in this study, which was approved by the Murdoch University Human Research Ethics Committee (Approval 2010/173).

The cut-off age for both controls and symptomatic individuals was 50 years, as after that age related impairments to postural stability could not be excluded [9–11].

Inclusion criteria for the symptomatic participants were NSLBP of any duration and the presence of pain ≥ 2 on the NRS-11 scale on the day of the postural sway recordings. Participants were excluded if the pain went below the gluteal fold, there were positive nerve root findings, serious spinal deformities, any condition that might affect balance (e.g. whiplash associated disorder or vestibular pathologies) or previous significant injuries such as traumatic damage to the spine or spinal surgery. No pain medication was allowed within 24 hours prior to the recordings. Participants were also excluded if they were unable to perform the postural sway recording either due to pain or other reasons. We aimed at enrolling around 10 patients for the 9 pain intensity groups (NRS 2-10).

For the purpose of this study, healthy was defined as the absence of any self-reported neurological or musculoskeletal impairments, pain or disability for a minimum of 6 months prior to the time of evaluation. Specifically, individuals with a history of low back pain within 6 months or previous injury to the neck or lower extremities, any known balance problems or the usage of medication associated with pain suppression or altered sensory perception were excluded. The physical examination of the control group must also have ruled out any back or extremity complaints or significant biomechanical impairments that might influence the measurements.

Procedures

Prior to the COP measurements, a physical examination was conducted on all participants by two experienced and trained chiropractors (TB and AS) who were otherwise not involved in the study. This procedure aimed to assess whether the volunteers met the criteria for their respective group and met the physical demands of the study. The NSLBP participants were further asked to describe their pain intensity at the time of recording by means of an NRS-11, a rating scale ranging from 0 (no pain) to 10 (worst possible pain) [15].

The experimental setup was based on an earlier literature review where a best practice setup for obtaining reliable COP data was published [16]. Accordingly, trials were conducted with eyes closed as the data obtained shows higher reliability than with eyes open. We further considered that the loss of visual input would prove an additional challenge to the balance system. In this way deficits in proprioception may be more easily detected and the discriminative value of the measurement between healthy controls and symptomatic NSLBP participants enhanced.

The system used for this study was a Metitur Good Balance GB300^® CE (Metitur Oy, Finland). Signals were sampled at 100 Hz, amplified and converted from analogue to digital. High frequency noise was reduced by a low-pass filter with a cut-off frequency of 10 Hz.

Mean velocity (mVel) was chosen as the main COP parameter as this has consistently shown to be both reliable [16] and discriminative for NSLBP [1]. It is described by taking the total distance of the COP path travelled in the respective direction and dividing it by the sampling duration. In addition, the 90% circle diameter was included to offer a broader spectrum of analysis. This parameter refers to the diameter of a circle containing 90% of the COP path travelled over a given time.

The participants were asked to remove their shoes and stand upright on the forceplate with their eyes closed, the head erect and their arms hanging loosely by their sides. The foot position was narrow stance with toes and heels touching. For the duration of the recording, the participants were further instructed to "stand as still as possible" [17].

Three successive trials of 90 seconds duration each were conducted with a preceding 5 sec adaption period that was not recorded. Rest periods of 60 sec were provided between each trial during which the participants were allowed to sit down while maintaining their original foot position on the forceplate. All participants were asked afterwards whether pain or discomfort may have influenced their balance performance.

All tests were conducted in a quiet room with normal temperature. The forceplate was calibrated prior to the recordings and further underwent an automatic calibration check before each trial.

Data analysis

Subjects

Age groups

Reliability

Relationship between pain intensity and postural sway

As a general trend, a steady linear increase in mVel AP/ML and 95% circle diameter direction can be observed. Levene's Tests showed no homogeneity of variance (p ≤ 0.018) while Shapiro-Wilk test indicated a normal distribution of the independent and dependent variables (p ≥ 0.11).

Compared to healthy controls, a significant difference (p ≤ 0.01) in mVel was present in NSLBP participants beginning at an NRS score of 3 in ML direction. In AP direction, statistical significance (p ≤ 0.05) was also reached at a pain intensity of 3 with an increase in significance from 5 to 8 (p ≤ 0.001) (Figure 1).

Compared to healthy controls, a significant difference in 90% circle diameter was only present at NRS scores of 6, 7 and 8 (p ≤ 0.001) (Figure 2).

Finally, the relative differences between pain scores for the parameter 90% circle diameter are demonstrated in Table 5. The same trend as seen with mean sway velocity can be observed. However, at pain intensities 2 and 3, significant differences between pain scores are present at larger intervals (3 NRS scores compared to 1-2 at mVel ML/AP).

Regression analysis

The SPSS Curve Estimation function showed that a linear relationship was the most suitable line of fit (p ≤ 0.001). Hence, linear regression was used for further analyses of the data. No co-linearity between the variables was determined.

The univariate regression analysis included the variables gender, age, weight, height, previous pain duration and pain intensity. With the exception of previous pain duration, all other independent variables exhibited a significant effect on mVel AP/ML and 90% circle diameter and were consequently included in the multivariate analysis. This further analysis showed that only pain intensity exhibited a significant effect on the selected COP parameters.

For mean velocity and pain intensity, the regression analysis was a reasonably good fit, describing 53.0% of the variance in mVel ML and 40.0% in mVel AP (R²adj = 51.0% and R²adj = 38.4% respectively), the overall relationship was highly significant in both ML and AP direction (F = 40.8, p < 0.001 and F = 24.9, p < 0.001 respectively). Mean sway velocity increased by 1.53 mm/s for every extra pain level in ML, and by 1.27 mm/s for every extra pain level in AP direction.

The regression analysis for the parameter 90% circle diameter and pain intensity was a poor fit, describing just 18.7% of the variance in circle diameter (R²adj = 16.5%). The overall relationship, however, was highly significant (F = 8.6, p < 0.001). The 90% circle diameter of the COP excursion increased by 0.6 mm for every extra pain level.

Clinical considerations

The COP measurement protocol used in this study may in future be suitable as an objective outcome measure for clinical monitoring purposes. However, the results are unidirectional in that increasing pain was associated with increasing postural sway. We have not established that decreasing pain leads to a decreasing postural sway.

Secondly, given the linear relationship between pain intensity and, for example, mVel, a clinically significant decrease of two points on a pain NRS [20] is equivalent to a reduction in mean sway velocity of 3.6 mm/s in ML and of 3.0 mm/s in AP direction. These changes lie between 1 and 2 standard deviations from the mean. It remains to be seen if such a reduction is also clinically significant.

In addition, this study indicates that any future sample size calculations for COP measurements involving pain sufferers should be considered in the light of the respective perceived intensity. Depending on the research purpose, the inclusion criteria may focus on those with NRS-scores of 5 or higher to reach significance compared to controls more readily.

The results may also cast a new light on the interpretation of studies that reported no significant differences in postural sway between symptomatic individuals and healthy controls. In those instances (e.g. Brumange et al. [22] and Mok et al. [23]), these observations may be attributable to the low perceived pain intensities of the NSLBP participants enrolled.

There is evidence that higher COP sway is associated with a higher risk of falling in the elderly [24] and sustaining injuries as a consequence, although this is subject to debate [25, 26]. Our results did not include geriatric participants and therefore cannot be generalized to that population, however our data may nevertheless underline the importance of suitable pain control in elderly pain sufferers to avoid falls.

In addition, as pain interference appears a likely underlying mechanism, the focus of a rehabilitative approach in pain sufferers with increased COP excursions should be on pain reduction rather than proprioceptive training.

Future studies may also show a role for COP measurements as part of a suite of other procedures to identify malingerers. Even if the individual is aware that pain is associated with greater COP excursions, a study with pseudo-malingerers showed that imitating pain related sway pattern is difficult at best and the average results for sway velocity and sway area greatly exceeded those expected from a real pain sufferer [27].

Strengths and Limitations

The major strength of this study is in its best practice experimental setup which ensured reliable data collection. Our inclusion and exclusion criteria further prohibited our overall results from being affected by demographic or anthropometric factors.

In this cross-sectional study the subjective nature of pain perception and therefore pain rating may have influenced the results. In addition, pain perception between younger and older NSLBP participants varies and a decrease in pain perception in geriatric individuals has been described [28]. Although this does not affect our sample groups with a cut-off age of 50 yrs, it nevertheless prohibits our results to be generalized to elderly patients.

While significant differences in postural sway compared to healthy controls could be demonstrated in our patient population, the overall number of participants per NRS score was still comparably small. Our results are therefore prone to be affected by extreme COP measures. Other sample groups with identical NRS scores may therefore show varying results. However, we expect the linear trend to be preserved. Similar studies with an identical experimental setup and larger sample sizes should be conducted to confirm our results.