• BACKGROUND
    • Radial nerve injury after operative treatment of humeral shaft fracture usually improves, but a subgroup of patients does not recover. Understanding the probability of recovery over time is important in deciding on watchful waiting, nerve exploration, or tendon or nerve transfer. Bayesian analysis is helpful in estimating such probabilities. This type of analysis is predicated on estimating the probability that an event will occur and subsequently updating that estimate as new information becomes available-for instance, when nerve recovery does not occur over time.
  • QUESTIONS/PURPOSES
    • Using Bayesian methods applied to a previously studied cohort, we asked: (1) Over time, up to the end of 18 months, what is the probability of radial nerve motor recovery after surgical fixation of humerus fractures? (2) What variables are associated with the timing of radial nerve recovery?
  • METHODS
    • Between January 2002 and November 2014, we treated 375 patients with open reduction and internal fixation (ORIF) for a traumatic diaphyseal humeral fracture at six urban hospitals (two Level 1 trauma centers, two Level 2 trauma centers, and two community hospitals). From this previously studied cohort, we lost access to one hospital's data, leaving us with 295 records to screen for eligibility. We considered patients with an isolated preoperative or postoperative radial nerve palsy, defined as an inability to extend the wrist against gravity (Medical Research Council grade < 3), as potentially eligible. Based on that, 24% (72 of 295) were eligible; of those, 2.8% (2 of 72) were excluded because a nerve disruption was repaired during ORIF. Another 24% (17) had incomplete data sets, leaving 74% (53) for analysis here. Patients with incomplete data sets did not differ from analyzed patients with respect to demographics or injury characteristics. The median (IQR) age was 43 years (25 to 61), and 49% (26 of 53) of patients were male. Most palsies presented preoperatively (83% [44]). Motor recovery was defined as the ability to extend the wrist against gravity (Medical Research Council grade ≥ 3). We conceptualized the probability of radial nerve recovery over time as two conditional probabilities: (1) the probability that the nerve injury is recoverable (neuropraxia, or recoverable axonotmesis), and (2) the probability that it did not recover at a certain point in time. We used a Bayesian network analysis to model these two probabilities. We based our estimate of the probability of a recoverable nerve injury on the largest systematic review, in which 90% (438 of 488 [95% confidence interval (CI) 87% to 92%]) recovered. To reflect uncertainty, we repeated the analysis for the upper and lower limits of its CI. To estimate the probability that recovery did not happen yet at a certain point in time, we used the timing of recovery from the patients in our cohort. We used Cox proportional hazards analysis to examine associations between time to recovery and demographic and injury variables, including age, sex, fracture type and location, vascular injury, type of fixation, and timing of palsy (preoperative versus postoperative).
  • RESULTS
    • If a nerve has not recovered by 7 months, the probability of nerve recovery by 18 months was still better than chance, at 56% (range 48% to 62%). If the nerve had not recovered by 1 year, then the probability of recovery was 17% (range 13% to 21%). No variables (such as age, fracture location, vascular injury, or fixation type) were associated with timing of radial nerve recovery.
  • CONCLUSION
    • Providers can use our findings to counsel patients on the expected probability of nerve recovery, which might reduce anxiety while patients await recovery. These probabilities can aid in the decision whether and when nerve reconstruction, nerve transfers, or tendon transfers may be beneficial. Because the probability of recovery remains relatively high for at least 7 months after injury, early surgery is unlikely to be beneficial in patients with radial nerve motor injury after surgical fixation of humerus fractures. Future studies can provide more specific recovery probabilities by including findings on electrodiagnostic studies and patients treated nonoperatively.
  • LEVEL OF EVIDENCE
    • Level III, therapeutic study.