Multivariate Prognostic Modeling of Persistent Pain Following Lumbar Discectomy

Background: Persistent postsurgical pain (PPSP) affects between 10% and 50% of surgicalpatients, the development of which is a complex and poorly understood process. To date, moststudies on PPSP have focused on specific surgical procedures where individuals do not sufferfrom chronic pain before the surgical intervention. Individuals who have a chronic nerve injuryare likely to have established peripheral and central sensitization which may increase the risk ofdeveloping PPSP. Concurrent analyses of the possible factors contributing to the development ofPPSP following lumbar discectomy have not been examined.Objective: The aim of this study is to identify risk and protective factors that predict the course ofrecovery following lumbar discectomy and to develop an easily applicable preoperative multivariateprognostic model for the occurrence of PPSP in this patient cohort.Study Design: A prospective study of elective lumbar discectomy with a 3 month follow-up.Setting: University setting in IrelandMethods: All ASA I-II patients, (n = 53, 18-65 years old), undergoing elective lumbar discectomyat a single institute were included and followed for a 3 month period postsurgery. Preoperativepotential predictors were collected: age, gender, pain intensity (McGill score, visual analog scale[VAS], Present Pain Intensity), degree of dysfunction (Roland-Morris Function score), psychologicalstatus (pain catastrophizing, anxiety, and depression scores), health-related quality of life (SF36), quantitative sensory testing (QST), inflammatory biomarkers, and a genetic pain profile.The proposed primary outcome was significant pain reduction (VAS > 70%) 3 months followingsurgery compared to the preoperative pain intensity.Results: A final prediction model was obtained using a multivariate logistic regression incombination with bootstrapping techniques for internal validation. Twenty (37.7%) patientsdeveloped PPSP. Independent predictor factors included age (odds ratio [OR] = 1.0 per year),present pain intensity (OR = 0.6), and degree of dysfunction (OR = 1.2). The concordance indexC (.658) supports a good monotonic association (where perfect prediction is 1) and the Akaike’sinformation criteria indicated a good fit of the model. Inclusion of additional measured parameters(QST, biomarker, or genotyping) did not improve the model.Limitations: Before this internally validated model can be integrated into clinical practice,and used for patient counselling and quality assurance purposes, external validation studies arenecessary.Conclusions: We demonstrated that the occurrence of PPSP can be predicted using a small set ofvariables easily obtained at the preoperative visit. This a prediction rule that could further optimizeperioperative pain treatment and reduce attendant complications by allowing the preoperativeclassification of surgical patients according to their risk of developing PPSP.Key words: Persistent post surgical pain, predictive modeling, prognostic, lumbardiscectomy