Back to presentation overview

Title: Prognostic factors for developing chronic low back pain.

Authors and affiliation: Martijn W Heymans,  PT, PhD 1,2, Stef van Buuren, PhD3, Dirk L Knol, PhD4, Willem van Mechelen, MD, PhD1,2  Henrica CW de Vet, PhD1

1 EMGO Institute, VU University Medical Center, Amsterdam, The Netherland
2 Body@Work, Research Center Physical Activity, Work and Health, TNO-Vumc / Department of Public and Occupational health, VU University Medical Center, Amsterdam, The Netherlands
3 TNO Quality of Life, Leiden / Department of Methodology and Statistics, University of Utrecht, The Netherlands
4 Clinical Epidemiology and Biostatistics, VU University Medical Center, Amsterdam, The Netherlands

Email address: mw.heymans@vumc.nl

Introduction: Low back pain (LBP) is a common and expensive musculoskeletal problem. Most patients getting LBP will recover soon. However, a small group progresses to develop chronic LBP. For the purpose of prevention or early therapeutic intervention, it is important to identify patients who are at high risk of long-term LBP.

Objective: For this purpose, we developed a clinical prediction rule which can be used during daily care by general and occupational physicians. This enables family and occupational physicians to take evidence based action on the level of the individual worker.

Methods: We conducted a prognostic study in a cohort of 628 patients with (sub)acute LBP by merging data of three RCTs recently conducted in our institute. We explored all relevant prognostic indicators, considering demographical (e.g. age, sexe), LBP related (e.g. pain and duration of complaints), psychological (e.g. kinesiophobia) and work related indicators (both physical and psychosocial, e.g. lifting, social support). We also included early changes in pain intensity and function because recent studies demonstrated their strong predictive value. Outcome was pain for more than 6 months. To select prognostic factors and for the development of the prediction model, we used bootstrapping combined with backward selection procedures. Missing data as a result of merging the data sets were multiply imputed. Bootstrapping was used to validate the prediction model.

Results: The factors showing a significant influence in the multivariate prediction model for chronic LBP were the level of functional status at 3 months, the change in pain intensity during the first 3 months, pain intensity at baseline, physical activity, working with vibration tools, whole body vibration, long-term sitting, job demands, passive pain coping, duration of complaints and treatment status before study onset. The c-index for this prediction model was 0.75. The effect of optimism as indicated by the slope was 0.80. The calibration plot showed that low predictions were predicted slightly too low by the model and high predictions slightly too high.