Oral versus Intravenous Antibiotics for Bone and Joint Infection

Abstract

Background: The management of complex orthopedic infections usually includes a prolonged course of intravenous antibiotic agents. We investigated whether oral antibiotic therapy is noninferior to intravenous antibiotic therapy for this indication.

Methods: We enrolled adults who were being treated for bone or joint infection at 26 U.K. centers. Within 7 days after surgery (or, if the infection was being managed without surgery, within 7 days after the start of antibiotic treatment), participants were randomly assigned to receive either intravenous or oral antibiotics to complete the first 6 weeks of therapy. Follow-on oral antibiotics were permitted in both groups. The primary end point was definitive treatment failure within 1 year after randomization. In the analysis of the risk of the primary end point, the noninferiority margin was 7.5 percentage points.

Results: Among the 1054 participants (527 in each group), end-point data were available for 1015 (96.3%). Treatment failure occurred in 74 of 506 participants (14.6%) in the intravenous group and 67 of 509 participants (13.2%) in the oral group. Missing end-point data (39 participants, 3.7%) were imputed. The intention-to-treat analysis showed a difference in the risk of definitive treatment failure (oral group vs. intravenous group) of -1.4 percentage points (90% confidence interval [CI], -4.9 to 2.2; 95% CI, -5.6 to 2.9), indicating noninferiority. Complete-case, per-protocol, and sensitivity analyses supported this result. The between-group difference in the incidence of serious adverse events was not significant (146 of 527 participants [27.7%] in the intravenous group and 138 of 527 [26.2%] in the oral group; P=0.58). Catheter complications, analyzed as a secondary end point, were more common in the intravenous group (9.4% vs. 1.0%).

Conclusions: Oral antibiotic therapy was noninferior to intravenous antibiotic therapy when used during the first 6 weeks for complex orthopedic infection, as assessed by treatment failure at 1 year. (Funded by the National Institute for Health Research; OVIVA Current Controlled Trials number, ISRCTN91566927 .).

Figure 1. Enrollment, Randomization, and Follow-up.

Figure 2. Route and Duration of Antibiotic Therapy.

Panel A shows the percentage of participants receiving intravenous antibiotics from the start of the treatment episode (i.e., the date of definitive surgery or, if surgery was not performed, the start of planned curative antibiotic therapy) through day 60. Participants who had been randomly assigned to receive oral therapy and received intravenous therapy were doing so because they were prescribed intravenous antibiotics for up to 5 days for an intercurrent infection unrelated to the incident orthopedic infection (permitted by the protocol); were unable or unwilling to take oral therapy for any reason (secondary end point); were, subsequent to randomization, considered to have no suitable oral options for antibiotic therapy on the basis of emerging susceptibility results (secondary end point); or had had a potential treatment failure (primary end point). Most participants who had been randomly assigned to receive intravenous therapy but were receiving oral therapy over the same period were doing so because of a failure of intravenous access (secondary end point). Panel B shows the percentage of participants receiving any antibiotic through the final follow-up. The vertical line indicates 6 weeks after the start of treatment (i.e., the end of the intervention period).

Figure 3. Differences in Risk According to the Analysis Performed.

The point estimates for the differences in failure rates are shown with 90% (thick lines) and 95% (thin lines) two-sided confidence intervals. The noninferiority margin is indicated by the vertical dashed line. The use of two-sided 90% confidence intervals was prespecified in the trial protocol in accordance with the sample-size calculation. Because two-sided 95% confidence intervals are also now commonly included in noninferiority trials, they are shown here to assess the sensitivity of the results to a change in significance level. In the intention-to-treat population, missing data were imputed with the use of multiple imputation by chained equations. The modified intention-to-treat population included only the participants with complete end-point data. The worst-case sensitivity analysis shows the results based on the worst-case assumption that, for participants with missing data, all participants who were randomly assigned to receive oral therapy and no participants who were randomly assigned to receive intravenous therapy had definitive treatment failures, thus introducing the worst possible bias against the oral strategy.