Design
A cross-sectional study was conducted. The study was approved by the Swiss Ethics Commission of Northwest and Central Switzerland (EKNZ) (reference number 2015–243). All participants gave their informed written consent prior to study start and all procedures conformed to the Declaration of Helsinki.
Participants
A convenience sample of 60 participants, 30 subjects with nonspecific CLBP and 30 healthy controls, were recruited from three outpatient physiotherapy clinics in Central Switzerland. The private clinics were selected according to the quantity of patients with relevant cases. The CLBP and control groups were matched for gender and age, but no further subgrouping was made. Inclusion criteria were: 1) age over 18 years; 2) proficient in the written and spoken German language; 3) no current pregnancy, or pregnancy in the past 6 months; 4) at least four points on the Roland Morris Disability Questionnaire (RMDQ) - indicating at least moderate disability due to LBP; and 5) the presence of CLBP - defined as at least 3 months of unilateral or bilateral nonspecific LBP. Exclusion criteria were: 1) clinical bedside signs of nerve root pain, or evidence of specific spinal pathology (e.g., malignancy, fracture, infection, inflammatory joint or bone disease; and 2) surgery on the lower back in the past 6 months. Healthy controls were excluded when they had any history of LBP in the past 6 months, or a period of LBP of more than 1 month in the past.
Questionnaires
Basic demographic data, − gender, age, weight, height, body mass Index (BMI), affected side, pain duration and pain intensity - were obtained from all participants. The Roland Morris Disability Questionnaire (RMDQ) [22, 23] was used to screen for eligibility, with a score of at least 4 points on this scale being necessary for inclusion. Subsequently, participants completed the Fear Avoidance Belief Questionnaire (FABQ) [24], Fremantle Back Awareness Questionnaire (FreBaQ) [23] and Hospital Anxiety and Depression Scale (HADS) [25,26,27]. These questionnaires had been previously validated in the German language.
Subjects of the CLBP group were asked additionally to report the locality of their pain (bilateral, left-sided, or right-sided), its duration and mean intensity, using a numeric rating scale (NRS). This scale ranges from 0 (no pain) to 100 (worst pain) [28].
Assessments
Examiners, blinded to the participants’ condition, recorded the results of the BPA, TPD and MCT. The examiner who produced all the back photos made no assessments and was blinded to the subjects’ group. The two examiners undertaking the physical testing were also blinded to the subject grouping. The latter were both experienced physiotherapists with a musculoskeletal physiotherapy specialisation. However, the tests were not explicitly trained.
Back-photo assessment
BPA had previously been used to assess limbs in a population with complex regional pain syndrome [8]. The lower back was the focus of this study. The lumbar vertebra four was marked on the skin of the subject and a photograph was taken from the middle part of the gluteal area to the occipital part of the skull to depict the participant’s back in a standing position. The photograph was then modified at level L4 in steps of 3% enlargement and shrinkage, using the GNU Image Manipulation Program (GIMP 2.8.14 for OS X). We chose to use 9 different sizes, instead of 7 used by an earlier study [8] and therefore increased the sizes only by 3% instead of 5%. The maximal limit was set at ±12%. This resulted in eight modified photos, plus the original photo for each participant. The photos were allocated numbers from 1 to 9 in order of the extent of modification. Numbers 1 to 4 were allocated to the shrunken photos, with 1 representing the photo with the maximal shrinkage of − 12% (Fig. 1). Number 5 was given to the original, unmodified photo. Numbers 6 to 9 were allocated to the enlarged photos, with 9 representing the photo with the maximal enlargement of + 12%. The photos were arranged on a sheet of paper in a randomised sequence, with the same sequence being used for all participants. The participants were then requested by the examiner to identify the original photo of their back. The BPA outcome value was in the range from 0 to 4, reflecting the number of steps of modification between the original photo and the photo selected by the subject, irrespective of direction.
Two-point discrimination
TPD is a reliable intra-rater measure to detect altered tactile acuity [29]. TPD measurements were taken using a plastic calliper, according to an established protocol [15, 17, 30] (Fig. 2). The participant lay prone and unable to see the calliper. An examiner measured both the horizontal and vertical TPD bilaterally on the participant’s lower back at level L4 [29]. The calliper tip distance ranged from 100 mm to 5 mm, with the test started at the maximum spread. For every correct detection, the spread distance was decreased by 10 mm. Conversely, for every incorrect detection, the spread distance was increased by 5 mm. This procedure was repeated three times in descending and ascending order and the average of the smallest distance between the calliper tips at which the participant was still able to discriminate between the two separate points was recorded as the TPD value [17].
Movement control test
MCI of the lumbar spine was evaluated using a battery of six tests designed to ascertain the movement control of the back (Fig. 3). The MCT battery has been shown to be a reliable tool in detecting impaired lumbopelvic control [20, 21, 31]. Before the assessment start, the examiner explained the six specific movement tasks of the MCT to the subject. The following protocol was used for each specific movement. At the beginning, the subject was verbally instructed by the examiner to perform a specific movement. A correctly performed movement was rated as a negative outcome for this task and the examiner continued directly to the next specific movement. When the specific movement was not performed correctly, the examiner provided verbal corrections. If the movement was still performed incorrectly, then the examiner demonstrated the correct movement. When, with this support, the subject was able to perform the specific movement correctly, then the outcome for this task was rated negative and the examiner went on to verbally instruct the next specific movement. When the subject was unable to perform the specific movement correctly, even with support, then the outcome for this task was rated positive and the examiner went on to verbally instruct the next specific movement. The tasks outcomes were recorded and a final score calculated ranging from 0 (all movements performed correctly) to 6 (no movement performed correctly). The subjects’ scores were noted as the outcome values for MCT. A detailed description of the MCT test battery and the definition of ratings can be found elsewhere [20].
Statistical analysis
All statistical analyses were performed using R (version 3.2.3) [32]. Normality of data was determined by inspection of histograms. Demographics, questionnaire data, BPA, TPD and MCT were analysed with descriptive statistics. Spearman correlations were used for independent variables (BPA, TPD, MCT, FreBAQ, FABQ, HADS, Age, Body height, Body weight, BMI). Multiple logistic regressions were applied with conditional likelihood to determine associations between the main outcomes (BPA, TPD and MCT) with the presence of CLBP. Concerning the goodness-of-fit test criteria for applying a multiple logistic regression, it was not applicable to run a Hosmer-Lemesbow test, since this is only suited to unconditional logistic regression. The Hosmer-Lemesbow test is computed with the expected probabilities of an event. These are unknown for conditional logistic regression because the model omits the unobserved intercept for each individual. As a compromise, a likelihood-ratio test was added, as well as concordance statistics. The concordance is defined as the probability that the prediction goes in the same direction as the actual data. CI > 0.5 implies a good predictive ability. Checks for outliers/influential data and for collinearity were performed. There was no evidence of outliers/influential data and there was low collinearity. To check for multicollinearity, variance inflation factors (VIf) were computed, with cut-off value 5 [33]. The log odds of the presence of CLBP were modelled with six covariates. These covariates were BPA, TPD (left and right horizontal TPD, left and right vertical TPD) and MCT. The objective was to quantify the effect of each covariate on the outcome CLBP, reported in odds ratios. The questionnaires revealed no associations with the main outcomes and were not investigated further.