A two-group randomized controlled clinical trial was undertaken.
Participants in this study were patients with persistent LBP who attended a specialized outpatient spine clinic (Spine Center of Southern Denmark) after referral from the primary care sector. Criteria for referral were: (1) back problems with or without radiculopathy, (2) a maximum of two years' duration of the current episode, and (3) previous appropriate but unsuccessful treatment in the primary care sector.
In this clinical setting, from August 2007 to December 2008, MRI was routinely performed [22
] on all patients (with no contraindications for MRI) meeting the following criteria: (1) LBP or leg pain of at least 3 on an 11-point Numerical Rating Scale, (2) duration of current symptoms from 2 to 12 months, and (3) age above 18 years. These criteria were chosen on the basis of previous projects where there was a high prevalence of MCs.
All patients with an MRI showing MCs Type I, II or III with a distribution exceeding the endplate, were clinically examined by a researcher (RKJ) to determine if they met the inclusion criteria. The researcher was blinded to subsequent patient allocation to intervention groups. Eligible patients were then informed about the results of the MRI, showing that MCs were a possible reason for their LBP and invited to participate in the trial. As directed by the Ethics Committee for the Region of Southern Denmark, both written and oral information about the trial was provided to each patient before admission to the project. This information also contained a layperson's explanation of the rationale for both treatment approaches. A team consisting of a nurse, a medical practitioner, a physiotherapist and a chiropractor attended participants during the trial if any complications occurred or if any pain medication required modification.
Inclusion and exclusion criteria
Patient inclusion criteria were: (1) LBP of at least 3 on an 11-point Numerical Rating Scale, (2) LBP greater than any leg pain present, (3) duration of current symptoms from 2 to 12 months, (4) age from 18 to 60 years, (5) ability to read and speak Danish, (6) a willingness to participate in the study, and (7) a minimum of one MC (Type I, II or III) that extended beyond the endplate into the vertebral body.
Patients were excluded if they: (1) were unable to participate in the project because of other physical or mental disorders, (2) had a competing LBP etiology such as disc herniation with symptomatic root compression, or (3) had undergone previous spinal surgery with no pain relief after the operation.
MRI was performed with a 0.2 T MRI-system (Magnetom Open Viva; Siemens AG, Erlangen, Germany). A body spine surface coil was used with the patient in the supine position. The imaging protocol consisted of sagittal and axial T1- and T2-weighted sequences.
The MRI evaluation, at baseline, was performed by a musculoskeletal radiologist (JSS) with extensive MRI experience and trained in using the standardized evaluation protocols [23
] used in this study. MCs were evaluated according to the Nordic Modic Consensus Group classification [23
], which had shown substantial to almost perfect reproducibility with Kappa values for intra-observer reproducibility (k = 0.77 to 1.0) and inter-observer reproducibility (k = 0.73 to 0.91).
Patients were included if the 'maximum height' of the MC extended beyond the endplate into the vertebral body on the sagittal image (Figure ). This criterion was chosen to enhance the probability that this MRI finding was clinically relevant and because it had been shown to have high Kappa values for intra-observer reproducibility (k = 0.83) and inter-observer reproducibility (k = 0.80) [23
Figure 1 Classification of size. Classification of the size of a Modic change (MC) based on its depth of extension into the vertebral body height: A: Endplate only, B: > endplate to 25%, C: 25 to 50%, and D: > 50%. Only patients with B, C or D (more ...)
Patients wishing to participate in the project were allocated to one of two groups, the new treatment approach (rest group) or a comparison treatment (exercise group).
The rest group was instructed to avoid hard physical activity and to rest twice daily for one hour, by lying down. To imitate the session structure in the exercise group and thereby the potential effect of being in a group [25
], patients met in a 'café-like' environment which provided the opportunity to exchange personal experiences of pain and physical incapacity. They met in groups of up to a maximum of 10 people once every second week for 10 weeks for a session of 45 minutes where they engaged in unstructured talk with a physiotherapist present. To remind the patients not to be physically active and to attempt to support the spine, they were given an orthopedic flexible lumbar belt from 'Camp Scandinavia' and were instructed to use it as needed for up to a maximum of four hours per day. Patients were instructed to use a diary to record separately how many hours a day they were resting and wearing the lumbar belt. After 10 weeks, patients were instructed to increase their physical activity gradually until a self-determined acceptable level for the patient was achieved.
The exercise group received exercises for the stabilizing muscles in the low back and abdomen together with dynamic exercises, exercises for postural instability and light physical fitness training. The patients exercised in groups of up to a maximum of 10 people for one hour once a week for 10 weeks guided by a physiotherapist. They were encouraged to do the same exercises at home three times a week and to maintain a 'normal' level of activity. The patients had the opportunity to socially interact with each other during the sessions. After 10 weeks of intervention, the patients were given directions on how to continue their exercise at home. This intervention represented the normal 'active' treatment as recommended in current clinical guidelines [1
Two physiotherapists were assigned to each of the two intervention groups for the duration of the trial, although only one attended each session. These four physiotherapists had 12, 15, 20 and 37 years of experience, respectively, working with back pain patients. For both interventions, two nurses, one for each group, participated in two of the sessions to discuss pain medication, introduce pain coping strategies and to tackle 'unhelpful' beliefs about LBP. The first session in both groups started with 20 minutes of information about anatomy and MCs as well as repetition of the rationale for treating MCs with either exercise or rest. Only the rationale that was relevant for the group to which the patient was allocated was explained.
The patients' attendance was recorded at each session and if they were absent, they were contacted by a secretary to determine the reason for their non-attendance and to help to remove any obstacles to participation. When the patient had more than 20% absence, she or he was counted as a dropout.
A detailed protocol of procedures and content of the sessions was made for each intervention group for use by the attending nurses and physiotherapists, to ensure a uniform and consistent approach to the interventions. There was no overlap of personnel between the two intervention groups. The interventions ran from August 2007 to April 2009 with an intake of patients every fifth week to ensure that patients waited a maximum of five weeks from their initial examination to commencement of their involvement in the intervention.
A booklet of questionnaires was distributed to patients to complete in their own time at baseline, 10 weeks from baseline (post-treatment) and 12 months from the end of intervention (one-year follow-up).
Variables collected at baseline
Information on the presence and type of MCs for 11 endplates from upper L1 to upper S1 was extracted using the evaluation protocol (Nordic Modic Consensus Group classification [23
Questionnaire information was collected at baseline (before allocation) containing the following background variables: age, sex, body mass index (BMI), smoking status, employment status, type of occupation, education level, whether currently on sick leave and days on sick leave during the previous year. The outcome measures of pain, disability and general health described below were also contained in the baseline questionnaire booklet. In addition, information on any ongoing complaint at the National Social Appeals Board, the Patient Complaints Board of the National Health Service or a financial compensation case was collected together with information on any previous treatment administered in relation to the present episode of LBP. Finally, patients were asked how they would expect the treatment approaches (rest and exercise, respectively) to affect their back pain. A five-point Likert scale for each treatment with response options ranging from 1 (much better) to 5 (much worse) was used as measurement.
Variables collected post-treatment (10 weeks) and at one-year follow-up
Details on the primary and secondary outcome measures are described in Additional File 1
Primary outcome measure
The numerical rating scale [27
] (NRS) measures current back pain on a 0 to10 scale.
Secondary outcome measures
The Roland Morris Disability Questionnaire [28
] (RMQ) is a 23-item disability questionnaire with a 0 to 23 scale, measuring activity limitation.
] (EQ-5D) is a standardized instrument measuring health status-related quality of life consisting of a health status index (EQindex
0 to 1 scale) and a visual analogue scale (EQVAS
0 to 100 scale).
The global assessment transition questionnaire measures the patients' perceptions of the overall change in their back pain since the beginning of the study on a 7-point Likert scale [31
Beck Depression Inventory [32
] (BDI) is a 21-question inventory measuring the presence and severity of depressed mood with a 0 to 63 scale.
SMS-Track (pain and sick leave)
In order to obtain detailed information about each patient's clinical course, automated text messages via short message service tracking (SMS-Track) [33
] were used for data collection. SMS-Track is a system for data capture. Each week, for 52 weeks starting at baseline, a programmed database sent a text message to patients' cell phones containing two questions: 'How many days have you had low back problems during the last week?' and 'How many days have you been on sick leave due to your back problem during the last week?' Patients were instructed to answer from '0' to '7' according to the number of days per week that were relevant for each question. If the patient did not answer the question within five days, a text message reminder was automatically sent. If the patient had three or more missing answers, a secretary called the patient and asked if the patient had any problems with the technique and if so, tried to solve the problem. The sick leave data were transformed into a five-day week instead of a seven-day week, by recoding answers with six or seven days as five days.
Post-treatment and at the one-year follow-up, any adverse events associated with the treatments, together with a description of any care-seeking for the current back problem (that is, having consulted a general practitioner (GP), specialist, chiropractor, physiotherapist, other care provider or been hospitalised) was also measured.
A power calculation was made using a data file (unpublished observations) from the Spine Center of Southern Denmark, which showed that patients with MCs on a numerical rating scale (0 to 10) reported an average pain of 6.4 (SD 1.8). The aim was to have 90% power to observe a significant difference (alpha level) of 5%. To ensure a mean difference of 30% or more between the groups and with an estimated dropout rate of 20%, we needed to include 38 patients in each group.
On receiving the completed baseline questionnaire, the project secretary allocated each patient into one of the two intervention groups by means of computerized minimization software ('Minim', an MS-DOS program [34
]). Minimization [35
] is a covariate adaptive randomization method [36
] which aims to ensure that treatment arms are balanced with respect to certain predefined factors as well as for the number of patients in each group [37
]. The technique has been shown to be a highly effective allocation method recommended in the conduct of controlled trials particularly when trial sample sizes are small [37
]. Variables equally distributed through minimization were age (3 groups), sex (2 groups), heavy smoking, that is, ≥ 20 cigarettes a day (2 groups) and self-reported hard physical work (2 groups). The variables were chosen for the following reasons: age because activity level and care-seeking might be influenced by age [38
], sex because women more often report some kind of consequences due to spinal problems than men [38
], workload and smoking because hard physical work in combination with heavy smoking is associated with MCs [39
]. A concern about using minimization is the risk of selection bias due to the potential for assignment being predicted if the person conducting the minimization has knowledge of the characteristics of earlier patients [37
]. This was managed by ensuring that minimization was consecutive in order of patient registration and by the task of minimization being shared between two people. The researchers were masked to group assignment.
The MRI evaluator (JSS) was blinded to all patient information except name, sex and age. This was achieved by sending the images to an external radiologist with no access to other study data. The researchers analyzing the data (RKJ and NW) were blinded to patient identification as the study participants were given a random identification number when assigned to the project. All researchers were blinded to identification of the groups until all the analyses were done. The key for identification of patients and group was concealed by the project secretary.
Data management and analysis
Comparisons were made of baseline data between the two intervention groups, and between compliers and non-compliers in each group using t-tests or a non-parametric equivalent for continuous variables and a chi-square test or Fisher exact test for categorical variables.
All data were analyzed using the intention-to-treat (ITT) principle, whereby all patients who returned a questionnaire were included regardless of their participation in the intervention. In addition, an efficacy subset analysis was carried out for the primary and secondary outcome measures, selecting the subset of patients who received their treatment and who did not drop out for any reason.
For the whole study sample, the mean difference between baseline and each of the two follow-ups for primary and secondary outcome measures were compared using paired t-tests. A comparison between groups was performed using multiple linear regression with robust variance estimations. Regression models were adjusted for baseline score and for the following covariates: age, sex, hard physical work and heavy smoking. The covariates were included because it has been suggested that factors used for minimization in the allocation process should be adjusted for in the analyses [37
The number of patients in each group who achieved a minimal clinically important difference (MCID) was counted for the pain and disability outcomes using the raw change score from baseline to follow-up. A potential group difference was analyzed using a chi-square test. MCID values for secondary sector patients from a study by Lauridsen et al.
] were used because these MCID values were derived for patients undergoing standard treatment in the same clinical setting as the current study. The MCID values were 'quartile-specific', which means that the value of MCID is dependent on the quarter of the scale in which the baseline score is located. For the pain scale (0 to 10), the MCID is 1 if the baseline score lies in the first quarter of the scale (0 to 2.5), 2 if it lies in the second quarter (> 2.5 to 5), 4 if it lies in the third quarter (> 5 to 7.5), and 1 if it lies in the fourth quarter (> 7.5 to 10). For the RMQ scale (0 to 23), the MCID is 4 if the baseline score is in the first quarter of the scale (0 to 5.75) and 2 for the second and third quarter (> 5.75 to 17.25). No quartile-specific value was available for the fourth quarter, and so an overall score with the value 2 was used [40
Items on the 7-point global assessment transition questionnaire were dichotomized into not better (that is, reports of 'much worse', 'worse', 'a little worse', 'about the same' and 'a little better') and better (that is, 'better' and 'much better') and the proportion of patients who improved in each group was compared using a chi-square test.
When analyzing the SMS-Track data, only patients with a total response rate of more than 80% were included. Generalized Estimating Equations (GEE) were used to assess any difference between the groups over time with age, sex, hard physical work and heavy smoking as covariates. Further, we counted the total number of days with back problems over the 52 weeks as a proportion of the maximum number of days possible. The same analysis was performed for the number of days on sick leave, assuming each working week consisted of 5 days. The groups were compared using a chi-square test. The SMS-Track data were also presented visually. Adverse events and care-seeking were reported in raw numbers and percentages.
Finally, an analysis was performed investigating if the patients' expectations of the treatment they received would influence the outcome [41
]. The five-point Likert scale for each treatment option was dichotomized into not better (that is, reports of 'much worse', 'worse' or 'unchanged'), and better (that is, 'better' and 'much better'). A dichotomous 'treatment expectation' variable was constructed containing patients who received their preferred treatment and those who did not. A comparison of the distribution in the two intervention groups was analyzed using Fisher's exact test. Thereafter, a regression model was created with intervention group and the 'treatment expectation' variable as the independent variables, and the dependent variables (outcome variables) of pain, disability and general health. The model was adjusted for baseline score, age, sex, hard physical work and heavy smoking.
Significance (alpha level) was set at 5%. All analyses were performed using the Stata statistical software, version 11.
The study was approved by the Ethics Committee for the Region of Southern Denmark, approval # S-VF-20060111 and registered in ClinicalTrials.gov Identifier # NCT00454792.