Risk prediction model for unsuccessful elbow flexion recovery after nerve transfer surgery in patients with brachial plexus injury

OBJECTIVE Nerve transfer surgery has been a mainstay treatment of brachial plexus injury, with varying success rates. Patients undergoing unsuccessful surgery are left with a useless limb for at least 2 years. Preoperative prediction has become a topic of interest to avoid an unsuccessful nerve transfer surgery. This study aimed to find strong predictive factors and develop a prediction model for unsuccessful functional elbow flexion recovery 2 years after a nerve transfer surgery in patients with brachial plexus injury.METHODS This retrospective study reviewed the medical records of 987 patients with traumatic brachial plexus injury who underwent brachial plexus surgery by five hand and microsurgery surgeons at a single tertiary care referral center from December 2001 to July 2018. Four hundred thirty-three patients were eligible for analysis. Patient demographic data, injury factors, surgical details, and postoperative factors were collected. Multivariable logistic regression was used to identify strong prognostic factors for unsuccessful nerve transfer surgery for elbow flexion. A simplified model was developed by rounding the coefficient to the nearest 0.5 score or an integer. Both original and simplified models were validated using the Hosmer-Lemeshow goodness-of-fit test and bootstrapping.RESULTS A full, original prognostic model from a stepwise backward logistic regression consisted of a BMI & GE; 23 kg/m2 (p = 0.015), smoking (p = 0.046), total arm-type injury (p = 0.033), donor nerve (p < 0.001), associated upper-extremity frac- ture (p = 0.013), and associated ipsilateral vascular injury (p = 0.095). The areas under the receiver operating character- istic curve of the original and simplified models were 0.765 and 0.766, respectively. The Hosmer-Lemeshow test showed good agreement of predicted and observed probability of the original (p = 0.49) and simplified (p = 0.19) models. Boot- strapping estimated an average optimism (1.9%) in the original model and minimal optimism (0.1%) in the simplified model. CONCLUSIONS The prediction model for failed elbow flexion recovery after nerve transfer surgery in traumatic brachial plexus injury was developed with good predictive value and internal validity. An alternative treatment, i.e., primary free functioning muscle transfer, should be offered in preoperative counseling in cases of a very high risk of failure.