This study aimed to investigate alterations in postural sway associated with decreasing pain intensities. As an observational study with no control group or randomization of patients, it was neither intended nor designed to investigate any effect of manual therapies on non-specific low back pain.
While the reduction in pain followed a course of manual interventions, placebo, analgesic medication or natural history may have elicited similar results with regards to the associated sway alterations. The study design did not intend or allow to assess or quantify any potential additional biomechanical benefit of the therapeutic intervention on postural stability.
During the course of the three measurements, a dropout rate of 25% (13/51) occurred at session three while the groups at measurement 2 remained complete. As the data of these participants was completely removed from the study, no further statistical adjustment such as "intention-to-treat" analysis was deemed necessary. Although for a longitudinal observational study incomplete data sets are not necessarily excluded, their removal was deemed appropriate as individual results are followed over the course of the three sessions. The inclusion of incomplete sway data may adversely affect group means per measurement due to inter-subject variability and thereby alter the interpretation of the results.
Irrespective of the unclear underlying mechanism, the observed decrease in pain intensity over three measurements exceeded two points on an NRS score and is therefore considered a clinically relevant change [30–32]. The study design did not set out to distinguish whether one treatment was associated with a more significant decrease in pain compared to others and therefore this cannot be commented on.
Previously we were able to demonstrate a linear relationship between COP sway and NRS scores in NSLBP patients . The trend observed in this study further strengthens the impression that this close association between postural sway and pain intensity also exists if the original pain NRS-scores change.
The pain reduction occurred following a series of non-specific therapeutic interventions. As mentioned, any contribution of intervention, placebo effects or pain remission due to natural cause remains unclear. As a general trend, both group means and individual COP measurements indicate that a decrease in postural sway was observed if NRS scores also decreased. If this was not the case, the postural sway remained similar (Figures 6 and 7).
While learning effects cannot be excluded as an explanation for altered postural sway at follow-up, this appears less likely as similar effects would be excepted for those patients where no decrease in pain occurred. However, no such effect was observed.
Some studies do not support our observations as no decreasing postural sway was observed following pain reduction [33–35]. This may be attributed to the fact that patients with neurological impairments were enrolled where demyelination probably inhibited full recovery within the follow-up period [34, 35]. Secondly, these studies employed prolonged follow-up periods of 3  and 6 months , while we investigated short-term changes over the course of around 2 weeks.
The chronic low back pain patients without neurological symptoms enrolled by Mehling et al.  experienced only a minimal average pain decrease (VAS 5.15 ± 2.0 at baseline compared to VAS 4.37 ± 2.36 post-intervention). This is not considered a clinically significant improvement and sway data published in an earlier study did also not identify a significant change in postural sway between those pain scores .
In addition, the results may have been affected by high inter-subject variability associated with the small sample sizes or single measurements of short duration .
The results of our study warrant caution in interpretation. First of all, pain perception is multifactorial  and in addition to functional impairments, psychological aspects may play an important role. This was not assessed for and can therefore not conclusions can be reached regarding their implications. It is further possible that both intra- and inter-subject variability in postural sway is masked when calculating means and therefore difficult to interpret.
In addition, the data shows quite wide variations in postural sway velocity likely due to the low sample sizes, particularly at medium pain intensities. When groups consisted of larger patient numbers, generally no significant sway differences were observed compared to other patients experiencing similar pain at baseline. The results from this study suggest that each group should consist of around 10 participants for further analyses. Considering a dropout rate of around 25%, about 14 participants should therefore be enrolled. However, with regards to assessing changes in sway or pain intensity at the follow up recordings, sample size calculations are unable to take this into account as the number of patients that did or did not show alterations in the variable of interest cannot be predicted.
At first sight, these results are quite interesting as a larger inherent variability would have be expected. On the other hand, it is consistent with the subjective nature of pain perception. If a group of individuals receives an identical painful stimulus, a certain variation in pain perception will occur as a result . However, this study suggests that similar postural sway responses occur in those patients reporting the same NRS-score. Secondly, the overlapping 95% CIs for all COP parameters observed between NRS scores particularly at lower NRS scores (Chapters 9 and 10) make results within the same range more probable.
The results further suggest that the presence of pain may be responsible for alterations in postural sway  rather than changes/alterations in proprioceptive information caused by chronic damage to sensory tissues in the neck. Even considering neural plasticity, any reversal of such alterations appears unlikely within the 2-3 day period between measurements.
Further investigations with larger sample sizes are needed to confirm the observed trend for all NRS-scores. Another approach would be to investigate whether this can also be observed when observing natural remission. Also, studies employing analgesics are indicated to further assess the role of direct pain relief compared to the biomechanical, functional approach applied here. Comparing such observations may also indicate whether there are additional benefits associated with manual therapies such as spinal manipulation.
Although the results have to be interpreted with some caution, the COP measurement protocol used in this study may be suitable as an objective outcome measure for clinical monitoring purposes. This in turn also suggests that pain assessment by NRS-11 may be equally objective, thereby limiting the clinical use of COP measures for this specific purpose.
As previously described, it has been demonstrated that elderly fallers show significantly increased postural sway compared to non-fallers [38–40]. There is also evidence that higher COP sway is associated with a higher risk of falling  and sustaining injuries as a consequence, although this is subject to debate [42, 43]. Consequently, if such individuals are additionally suffering from pain, this may further increase the risk of falling in addition to any age-related or pathological changes in postural stability. As this study shows lower sway to be associated with decreasing pain intensities, this underlines the importance of pain control particularly in this population to reduce COP sway and increase postural stability.
There are various limitations to this study. The issues associated with small sample sizes became even more pronounced by the fact that the number of patients per NRS score varied considerably as pain levels changed. Some NRS groups consisted of only n = 2, as seen particularly at higher pain intensities as pain levels decreased over the course of the measurements. This rendered a meaningful statistical analysis difficult. On the other hand, other pain groups grew to n = 14 as a result, which strengthened the conclusions drawn from this data.
In addition, the study design did not allow to determine whether decreasing pain scores alone was responsible for the decreasing postural sway or whether the manual intervention added an additional benefit by increasing biomechanical function. Based on the available literature, however, the latter appears unlikely to exhibit any significant effect (Chapter 5). Furthermore, the cut-off age of 50 years does not allow to extend the results to a geriatric population as the decreased pain perception in this age group  may not lead to similar postural responses. The same accounts for adolescents and children.