Do the results of the OXYGEN trial change if analyzed as "as-treated?": A secondary analysis of the OXYGEN trial

Objective: To determine if the results of the OXYGEN trial changed using an "as-treated" approach instead of the original "intention-to-treat" approach. The multi-center randomized controlled OXYGEN trial aimed to determine the effectiveness of high FiO2 in decreasing infection rates for high-risk tibial plateau, tibial pilon, and calcaneus fractures. Methods: A secondary analysis of a multi-center randomized controlled trial conducted at 29 US trauma centers was performed. A total of 1231 patients aged 18-80 years with tibial plateau, tibial pilon, or calcaneus fractures thought to be at elevated risk of infection were enrolled. Patients were randomly assigned to receive inspired oxygen at a concentration of 80 % FiO2 (treatment) or 30 % FiO2 (control). Adherence was defined using two different criteria. Criterion 1 required at least 80 % of the surgery time <= 40 % FiO2 for the control group or >= 70 % FiO2 for the treatment group. Criterion 2 required at least 80 % of surgery time within 20-40 % (control) or 70-90 % FiO2 (treatment). The primary outcome was surgical site infection (SSI) within 182 days of definitive fracture fixation. Secondary outcomes were deep and superficial surgical site infections within 90, 182, and 365 days of definitive fixation. Results: Under Criterion 1, the primary outcome occurred in 7 % (38/523) and 10 % (49/471) of patients in the treatment and control groups, respectively (p = 0.10). Deep infection occurred in 30 (6 %) treatment and 30 (6 %) control patients (p = 0.75). Superficial infection occurred in 9 (2 %) treatment and 20 (4 %) control patients (RR, 0.41; p = 0.03). Using Criterion 2, the primary outcome occurred in 7 % (36/498) of treatment and 10 % (48/468) of control patients (p = 0.12). Deep infection occurred in 28 (6 %) treatment and 29 (6 %) control patients (p = 0.81). Superficial infection occurred in 9 (2 %) treatment and 20 (4.3 %) control patients (RR = 0.43; p = 0.03). Conclusions: When re-analyzing based on which patients actually received high or control levels of perioperative oxygen fraction, the results are somewhat consistent with the original "intent-to-treat" analysis. Specifically, high perioperative oxygen lowered the risk of superficial SSI but did not affect deep infections.