describes the design and methods of the prospective cohort study.

We assessed the baseline characteristics of patients with chronic non-specific low back pain and described the methods used to investigate the clinical course and to identify potential prognostic factors (including internal validation) in a 12-month follow-up study. All participating patients with chronic non-specific low back pain (n=1,760) were recruited between January 2003 and December 2008 at the Spine & Joint Centre rehabilitation centre (mean age 40.1 years, SD 10.6; 73%

female) and were evaluated by means of (postal) questionnaires and physical examinations at baseline, at 2-months follow-up (after the 2-months therapy program), and again at 5 and 12-months follow-up. At the rehabilitation centre, the multidisciplinary behaviour therapy protocol used a bio-psychosocial approach to stimulate patients to adopt adequate (movement) behaviour aimed at physical and functional recovery. The program consists of 16 sessions of 3 hours each over an 8-week period (in total 48 hours), followed by a 3-month self-management program (e.g. exercises twice a week). The primary outcomes were back pain intensity, disability, work participation, quality of life, and patient’s global perceived effect.

Each model had the same 23 potential prognostic factors. The factors for the domain ‘patient’ characteristics’ were: age, gender, educational level (less than high school vs. high school/university), body mass index (BMI 24.9, 25-29.9, 30 kg/m2₎

and marital status/living with one adult (no/yes). For clinical status the following factors were included: duration of back pain in years, cause of back pain (accident movement; after physical load; during pregnancy or after delivery; unknown; surgery pelvis/back or HNP), previous rehabilitation treatment (none vs. one or more previous rehabilitation treatments), comorbidity (none vs. having one or more co-morbidities), present pain intensity (Visual Analogue Scale (VAS) 0-100 mm; with higher scores indicating more pain), course of pain in the previous 3 months (stable; increased; decreased), degree of present fatigue (VAS 0-100 mm; with higher scores indicating more fatigue) and the Quebec Back Pain Disability Scale (QBPDS; 0-100; with higher scores indicating more disability). Psychological factors were: the Dutch version of the Tampa Scale for Kinesiophobia (TSK-DV, 17-68; with a higher score indicating more pain-related fear), SF-36 MCS and PCS at baseline (0-100; with higher scores indicating better quality of life), and the Symptom Checklist 90 (SCL-90; item 9; psychoneurosis). Domain work-related factors consisted of two prognostic factors, work participation (0-100%; by dividing ‘current work hours’ by ‘former work employment hours’ prior to chronic non-specific low back pain) and sickness benefit (none vs. receiving different types of government welfare benefits). Finally, the domain physical examination consisted of the B200 Isostation (strength of back extension in Newton) and the duration of walking, sitting, standing (0-15, 16-30, 31-60, > 61 min) during daily activities.

In our study, every outcome for recovery was operationalised according to two definitions: 1) 30% improvement during follow-up compared to the baseline score for the outcome back pain intensity, disability and work participation and a 10% improvement in the score on the SF-36 PCS and MCS, and 2) ’absolute recovery’ was defined as a VAS score of 10 mm, a QBPDS score of 20 points, work participation working (0-100%) 90% at follow-up, and global perceived effect on a 5-point scale (GPE) dichotomized into ‘clinically improved’ vs. ‘clinically not improved’.

Multivariable logistic regression analysis included the 23 baseline characteristics with imputed datasets of 5 models and a p-value of 0.157. A sensitivity analysis was performed; the selection of variables was repeated in imputed datasets of 5 and 40 models with p-values of 0.05 and/or 0.157, respectively. We also compared

the results with complete-case analysis, i.e. all patients with missing data were excluded from the analyses. Furthermore, every outcome was also changed regarding the percentage improvement and/or absolute recovery score (cut-off point) during the sensitivity analysis.

The results of our prospective cohort study are described in Chapters 4 to 7. In Chapter 4 we assessed the clinical course and potential prognostic factors (including internal validation of the models) for recovery defined with the outcome back pain intensity at 2, 5 and 12-months follow-up. Patient-reported back pain decreased from 55.5 (SD 23.0) at baseline to 37.0 (SD 23.8), 35.3 (SD 26.1) and 32.3 (SD 26.9) at 2, 5 and 12-months follow-up, respectively. At 12 months, 61% of the patients experienced a 30% improvement and 29% of the patients an absolute recovery (VAS 10 mm) with regard to back pain intensity.

At 5-months follow-up, a 30% improvement resulted in a final model (AUC=0.66) which included 9 prognostic factors, together explaining 11% of the variation in outcome: younger age, female gender, a baseline BMI > 25 kg/m2_{, no previous}

rehabilitation treatment, higher baseline level of back pain intensity, no psycho logical/physical dysfunction, higher baseline scores on the SF-36 (PCS and MCS), and higher work participation at baseline. At 12-months follow-up the following factors were related to a 30% improvement: younger age, female gender, being married/ living with one adult, higher level of education, no comorbidity, higher back pain intensity, higher strength at the extension direction, no fear of movement, and higher scores on the PCS (SF-36), with a 10% explained variance and an AUC of 0.65.

At 5 and 12-months follow-up, for absolute recovery the explained variance was 11% and 18%, respectively, with an AUC of 0.69 and 0.73, respectively. At 5 and 12 months the factors younger age, less back pain intensity at baseline and higher scores on the SF-36 were prognostic factors for back pain intensity. Also added to the 5-month follow-up were the factors: no psychological/physical dysfunction, and more work participation at baseline. At baseline, a BMI 30 kg/m2_{, no comorbidity,}

higher disability score and having stable or more back pain intensity due to chronic non-specific low back pain in the previous 3 months were added to the 12-month follow-up. For every prognostic model, the internal validation showed identical results in experienced variance and AUC as in the developed prognostic models. The sensitivity analyses (e.g. 20 mm VAS) also showed similar results.