A database from an observational prospective cohort study for the evaluation of outcomes from a CNP-specific MBR programme was analysed. Data were collected at an assessment before treatment (T0), at the beginning of treatment (T1), at the end of treatment (T2), at 6-months post-treatment (T3), and at 12-months post-treatment (T4). Due to the explorative nature of this study, we used all available data from time points T1, T2 and T3 and did not perform a sample size calculation. Moreover, we did not use the T4 data due to the high number of missing values for the relevant co-variates. Furthermore, the post-MBR programme effects at T4 were individually compared to patient health statuses from T0 and T1 in a previously published study .
The study was conducted at the day clinic in the Department of Physical Medicine and Rehabilitation at University Hospital, Ludwig Maximilian University, Munich, Germany. It was carried out in compliance with the protocols of the Helsinki Declaration of 2004. All participants provided signed, informed consent prior to study participation. The Ethics Committee at the medical faculty of the Ludwig Maximilian University Munich did not have any objections against the publication.
Patients were referred to an interdisciplinary assessment at our clinic by a family physician or specialist. Depending on the primary diagnosis, patients were allocated to one of four condition-specific assessments for neck pain, lower back pain, osteoporosis, or osteoarthritis of the knee and hip. Assessments were conducted by a specialist in physical and rehabilitation medicine (PRM), a physiotherapist, and an occupational therapist. A psychologist from the treatment team also assessed patients who were suspected of suffering from a mental health disorder.
At the end of each assessment, the treatment team either recommended participation in a three-week MBR programme or another treatment option. The MBR programme was recommended according to predefined inclusion criteria, an appraisal of the results of standardised clinical tests and patient questionnaires, and the general impression of the day clinic treatment team. The predefined inclusion criteria for the neck-pain-specific MBR programme were: CNP lasting at least three months (with or without pain radiation in the upper limbs), previous out-patient physical therapy that did not result in improvement according to the patient, limitations in activities, and sufficient German language skills to follow the instructions of the MBR programme. Previous out-patient treatment was defined as conventional care for at least three months and typically included 3 × 6 sessions of physiotherapy for 20–30 min per session.
The MBR programme was not recommended if patients had severe somatic or mental illnesses that limited their ability to participate (e.G. major depression), acute neck trauma in the previous three months, former whiplash injury with proven structural damage, neurological deficits occurring within the previous three months, chronic neurological deficits that would have prevented participation in exercise interventions, dizziness or vertigo with unclear aetiology, diffuse idiopathic skeletal hyperostosis (DISH), shoulder abduction or flexion less than 90°, or patients undergoing a pension application.
After meeting with the treatment team, the physician explained the recommendation to each patient. In conversation, patients expressed their expectations and goals for treatment and their treatment preferences. Within the framework of participatory decision-making, the recommendation could change. More details regarding the assessment have been described elsewhere .
All consecutive patients who participated in the entire MBR programme answered the North American Spine Society questionnaire (NASS) pain+disability scales [24, 25] at baseline and discharge. A study inclusion flow diagram is presented in Fig. 1.
At the beginning (T1) and the end (T2) of the MBR programme, patients completed a set of questionnaires and underwent standardised clinical tests. Each patient was assessed by the same trained physiotherapist or occupational therapist both times. Data was collected using the NASS questionnaire [24, 25], the mental health scale from the Short Form 36 (SF-36) questionnaire [26, 27], and a standardised co-morbidity questionnaire . Socio-demographic information was also collected, and cervical spine range-of-motion (ROM) measurements were obtained using a cervical ROM instrument (CROM) . At the 6-month follow-up (T3), the same questionnaires were sent to the patients by mail. Pre-addressed, stamped envelopes were provided to all participants.
The clinic provides condition-specific, 3-week MBR programmes for patients with CNP, chronic lower back pain, osteoporosis and osteoarthritis of the knee and hip. Each programme alternates one after the other.
Patients completed a three-week, neck-pain-specific MBR programme that included a total of nine treatment days and 44 treatment hours. The programme fulfilled the German procedure classification (Operationen- und Prozedurenschlüssel (OPS)) 8–563.1 criteria of the German health care system, which requires at least 15 treatment units (a minimum of 30 min) of physical therapy or psychological therapy per week . The OPS code, in combination with staying at least 6 h per day, is required for day clinic reimbursement by statutory health insurance in Germany.
The treatment team consisted of a specialist in PRM, physiotherapists, occupational therapists, psychologists, medical massage therapists, and a swimming trainer. Most treatments were provided to groups; although, all participants had two individual physiotherapy lessons that occurred at the beginning and end of the programme. During the initial, individual physiotherapy lesson, patients were trained in deep neck muscle strengthening exercises using biofeedback . During the final individual lesson, patients were instructed in how to perform individual home exercises. Group treatments consisted of up to five participants in practical lessons and up to 10 participants in educational lessons and pool therapy. The MBR included land-based group exercises, gym training, pool exercises, occupational training, psychological lessons (including relaxation strategies), instructions for self-help techniques, patient education by a PRM specialist, and interactive group discussions at the end of each week with the entire treatment team. The physician provided daily ward rounds for the group, as well as individual appointments on demand. Details of the intervention have been described elsewhere .
North American Spine Society questionnaire
The NASS is a condition-specific instrument with specific modules for lower back and neck pain [24, 25]. Its original version includes two scales that measure pain+disability and neurogenic symptoms that were derived from a principal factor analysis in the primary validation study. The cervical spine NASS pain+disability scale includes 11 items and the neurogenic symptom scale eight items. All items range from 1 (best health) to 6 (worst health). The scales were scored by calculating the arithmetic mean of the answers.
The German language version of the cervical spine NASS has demonstrated good criterion and discriminant validity, and sensitivity to change in validation studies [32, 33]. It has also shown good psychometric properties in a validation study conducted on patients who underwent intensive out-patient rehabilitation at health resorts . Validated German versions of more commonly used cervical spine instruments, such as the Neck Disability Index (NDI), were not available at the time of the study; therefore, the cervical spine NASS was used.
Subsequent validation studies of the cervical spine NASS have shown a better fit in factor analyses and good responsiveness for separated scales for pain (two items) and disability (eight items) [32, 35]. In this study, we adhered to the original combined pain+disability scale as a primary outcome because it was defined a priori as a primary outcome in the evaluation of the neck-pain-specific MBR programme.
Short form 36 mental health
The SF-36 is the most widely used generic instrument for measuring health-related quality of life [26, 27]. Using 36 items, the following eight scales were determined: physical functioning, role-physical, bodily pain, general health, vitality, social functioning, role-emotional, and mental health. All of the scales ranged from 0 (worst health) to 100 (best health). The mental health scale, consisting of five items, was included in our analysis as it covers the construct of affective health, especially depression, with high validity . The SF-36 mental health scale shows high rates of completeness, high reliability, and high sensitivity to change in the rehabilitation setting of CNP patients , and has shown associations with the course of pain for different chronic conditions, including neck pain .
Cervical range-of-motion instrument
A CROM instrument (Performance Attainment Assoc., St. Paul, MN, USA) was used to measure the cervical spine ROM in degrees. The instrument consists of a mounting device for the head, two gravity-dependent goniometers, and a compass that measures rotation in 2° increments. A validation study showed good intra- and inter-tester reliability , and reliability and validity has been confirmed in different populations . The minimal detectable change in each direction was between 3.6° and 9.3° [37, 38]. The total active cervical ROM was the sum of 6 directions and showed higher intra- and inter-observer reliability (ICC = 0.99 and 0.95, respectively) compared to each separate cervical measure . The standard errors of the mean for the intra- and inter-observer studies were 6.6 and 17.7°, respectively . We used the total active cervical ROM rather than the separate measures due to limitations in the number of co-variates used in the analysis.
Socio-demographic data and co-morbidities
Information concerning co-morbidities was collected using the standardized Self-Administered Comorbidity Questionnaire (SCQ) , and socio-demographic data was gathered through specific questions.
Descriptive statistics and treatment effects
Descriptive statistics were calculated for the baseline characteristics. Effect sizes (ESs) for the primary outcome (NASS pain+disability) and the secondary outcomes (NASS pain, NASS disability, SF-36 mental health, and the total active cervical ROM) were determined by dividing the mean change between baseline and discharge (T1 and T2), and between baseline and the 6-month follow-up (T1 and T3), by the standard deviation of the baseline score . An ES above 0.30 is generally considered to be clinically meaningful unless instrument-specific studies have provided more reliable results for the minimal clinically important effects . For the NASS pain+disability scale, no specific minimal clinically important differences have been previously quantified. Significance of changes were tested using t-tests for dependent samples of normally distributed data or with Wilcoxon signed-rank tests for non-normally distributed data.
In the primary exploratory linear regression models, the dependent variables were ∆ discharge – baseline, and ∆ 6-month follow-up – baseline of the NASS pain+disability scale. For additional linear regression modelling, the dependent variables were the ∆ discharge – baseline of the separated NASS pain and disability scales.
Independent variables were selected from candidate variables in the database and based on previous research concerning risk factors and prognostic factors in neck pain patients [12,13,14,15,16,17,18,19,20, 22, 42, 43], as well as on clinical experience. We further aimed to cover both physical and psychological health and were specifically interested in the change in the ROM co-variable, which was a treatment aim of the MBR programme  and is associated with pain and disability in CNP patients .
The analyses were adjusted for the baseline variables of the change scores and important socio-demographic characteristics. The total number of co-variates was limited to 10, as 10 cases per co-variate were needed for the finite models and sufficiently valid estimates of the regression coefficient .
The independent variables were sex, age, living with a partner, education level, number of co-morbidities, SF-36 mental health baseline value and mean change, the total active cervical ROM, and the change in the total active cervical ROM. All models were adjusted for the baseline score of the corresponding NASS scale, of which the change in score was the dependent variable. To adjust for any confounding, all listed co-variates were kept in the models, irrespective of whether their correlation was statistically significant. Multivariable partial correlations were determined and adjusted for all other potentially confounding co-variates . The overall explained variances (%) were calculated to quantify the fit of the regression models.
For missing values of single co-variates, the mean imputation method was used (i.e. missing values were replaced with the mean of the valid values within each independent variable). This method was provided by the linear regression module of the statistical software program SPSS 25.0. Imputation by linear regression would have been inappropriate as it was the same strategy as the evaluation of prognostic factors and, therefore, would have increased the number of valid cases, but not the outcome of the prognostic parameter estimates. We assumed missing values were due to random processes, as the main reasons for missing data were incomplete distributions of the questionnaire regarding socio-demographic characteristics (12 patients = 10.7%) and incomplete clinical tests at T1 or T2 (10 patients = 8.9%) and not due to refusals by the patients to fill-in questionnaires or undergo clinical tests. This assumption was further supported by 97% completeness of the SF-36 mental health scale at baseline and 98% at follow-up, despite patients typically viewing questions about mental health as more sensitive compared to those concerning physical health.
All statistical analyses were calculated using SPSS 25.0 for Windows (IBM Corp., Armonk, NY, USA). ESs were calculated using Microsoft Excel 2010.