Low back pain is a musculoskeletal condition that affects many people worldwide and although there are several types of conservative treatments, either physiotherapy and/or pharmacological, the patient does not always obtain satisfactory results after treatment. To improve this situation, many prognostic models have been studied, developed, and validated. However, it is uncertain the available evidence about the prognostic models for predicting the success or failure of patients with low back pain after a conservative treatment.

Identify and evaluate prognostic models’ ability to predict success or failure in patients with low back pain after receiving conservative treatments.

Literature searches were conducted in three different electronic databases (MEDLINE, EMBASE and CINHAL). Prognostic models predicting the success or failure of conservative treatment in adults with low back pain were considered eligible. Studies investigating low back pain related to a severe pathology were excluded. Two independent reviewers performed the study selection and data extraction. The individual performances of the prognostic models were performed descriptively.

Searches initially retrieved 13,013 studies. After analysis considering inclusion criteria, 81 studies were included in this systematic review. Of these, 78 (96.3%) developed and internally validated the prognostic models, and only 4 (3.7%) developed and externally validated the models. Regarding the discrimination of the models studied, the c-statistics or area under the curve (AUC) ranged from 0.44 to 0.96. Regarding the calibration, the calibration slope and intercept ranged from 0.74 to 1.06 and from -0.01 to 0.34, respectively. Regarding the sensitivity and specificity of the prognostic models, there was a variation between 31.0% and 94.5% and from 14.9% to 93.7%, respectively.

Although prognostic models have been developed, discrimination and calibration, as well as specificity and sensitivity, varied significantly among them. In addition, there were few studies investigating the external validation of these models.

Although some prognostic models have been developed, validated, and are able to predict success or failure in patients with low back pain undergoing conservative treatment, necessary to implement such models in clinical practice due to the lack of evidence regarding external validation.