Perron M.1,2, Gendron C.3,4, Langevin P.1,5, Roos M.1, Roy J.-S.1,6
1Laval University, Rehabilitation Department, Quebec, Canada, 2CIUSSS de la Capitale-Nationale, DEAU, Quebec, Canada, 3Canadian Forces Health Services Group, Valcartier Health Center, Quebec, Canada, 4Laval University, Kinesiology Department, Quebec, Canada, 5Physio Interactive, Quebec, Canada, 6CIRRIS, Quebec, Canada
Background: Low back pain (LBP) is a diagnosis that encompasses heterogeneous patients unlikely to respond to a unique treatment approach. Identifying sub-groups of LBP patients with shared characteristics may help to plan interventions and to improve treatment outcomes.
Purpose: To create a clinical prediction rule (CPR) that will accurately predict the success following participation in a supervised multi-station exercise program in soldiers with LBP.
Methods: Soldiers with non-specific LBP were consecutively recruited to participate in a 6-week supervised program comprising 7 stations, each consisting of exercises of increasing difficulty. The exercises were grouped together as follows: hip strengthening and control, squats, core stability, abdominal planks, abdominal and back extensor strengthening and lifting techniques. Progression to higher levels of difficulty was determined according to irritability level, the most limited plane of motion and quality of exercise execution. Advancement in the exercises culminated in occupational task simulation (gradual exposure and task oriented approach). Demographic, impairment and disability data were collected during a baseline testing session. The modified Oswestry Disability Index (ODI) was administered at baseline and following the 6-week program. An improvement of 50% in the initial ODI score was considered the reference standard to determine success following participation in the exercise program. Univariate associations with success were tested for each baseline variable using chi-square or paired t-tests. Variables that showed between-group (success/failure) differences (p 0.01) were considered potential predictors. After determining the sampling adequacy of the potential predictors using factor analysis (Kaiser-Meyer-Olkin statistic > 0.6), variables were entered into a logistic regression. Finally, continuous predictors were dichotomized and the sensitivity, specificity and positive and negative likelihood ratios were determined for the model and for each predictor.
Results: Of the 104 included participants, 20 (18.3%) did not fill out the ODI questionnaire at the end of the program and were excluded. Five variables contributed to prediction of success following participation in the exercise program: no pain in lying down (p = 0.017), no use of antidepressants (p = 0.061), FABQ work score 22.5 (p = 0.061), fewer than 5 physiotherapy sessions before entering the program
(p = 0.144) and less than 6 months work restriction (p = 0.161). This model yielded a sensitivity of 0.775 (95% C.I. 0.61-0.89), specificity of 0.80 (95% C.I. 0.65-0.90), LR+ of 3.88 (95% C.I. 2.1-7.1), and LR- of 0.28 (95% C.I. 0.16-0.50). A 77.5% success rate following the program can be predicted if more than three predictors are present, while an 80% probability of failure can be expected if only three or fewer predictors are present.
Conclusion(s): This study led to the identification of predictors of success for an exercise program that can be easily objectified in the clinic and are mostly not subject to evaluator bias.
Implications: The use of such predictors may guide clinicians in identifying patients most likely to respond positively to the exercise program. However, care should be taken to avoid misinterpretation until further studies confirm the validity of this CPR with different populations.
Funding acknowledgements: The authors wish to thank the physiotherapists of the VHC for data collection. Project funded by OPPQ and REPAR
Topic: Musculoskeletal: spine
Ethics approval: The research protocol was approved by the research ethics committee of the CIUSSS de la Capitale-Nationale (IRDPQ).
All authors, affiliations and abstracts have been published as submitted.
