OVIVA — journal-club digest

Personal study digest for a new ID fellow. Findings are from the cited trial; the appraisal, Stats Corner, and wider-literature notes are reviewer synthesis. Not a substitute for the paper.

Bottom line

Oral antibiotics were noninferior to IV for the first 6 weeks of bone & joint infection — with shorter hospital stays and far fewer catheter complications.

The question: for complex bone & joint infection, can appropriately chosen oral therapy replace 6 weeks of IV?

PICO

Population Adults >18 with bone/joint infection who would ordinarily get 6 wks IV — native osteomyelitis, native joint infection needing excision arthroplasty, prosthetic joint infection, orthopedic fixation-device infection, or vertebral osteomyelitis; 26 UK centers. Randomized ≤7 days after surgery (or start of antibiotics). Excluded: concurrent S. aureus bacteremia or endocarditis, infection needing prolonged IV (e.g. CNS), septic shock, no suitable oral option.

Intervention Oral antibiotics for the first 6 weeks — agent chosen by an infection specialist (susceptibility, bioavailability, tissue penetration). Follow-on oral beyond 6 wks allowed.

Comparison Intravenous antibiotics for the first 6 weeks. Adjunctive oral agents (e.g. rifampin) permitted, reflecting usual practice.

Outcome Primary: definite treatment failure within 1 year (clinical / microbiologic / histologic, blinded adjudication). Key secondary: serious adverse events, IV-catheter complications, early discontinuation, C. difficile, length of stay, function/QoL.

Why this trial mattered

The rule that osteomyelitis needs “prolonged (4–6 week) parenteral” antibiotics traces to an influential 1970 NEJM review, not to a trial — a preference for IV built on the belief that parenteral is inherently better. The only prior randomized signal was a small, underpowered meta-analysis (180 chronic-osteomyelitis patients) that found no IV advantage but could not settle the question. Meanwhile IV carries real costs: catheters, thrombosis, line infections, longer stays. OVIVA was the pragmatic noninferiority RCT built to test the pharmacologic premise that appropriately selected oral agents reach adequate concentrations at the site — and can carry the first 6 weeks.

Design at a glance

Design RCT, open-label, noninferiority Randomized 1,054 (527 / 527) Analyzed 1,015 mITT · 909 per-protocol Primary definite failure at 1 yr NI margin 7.5 pp (was 5 pp) Analysis ITT + imputation, 90% CI Ran 2010–2015 · 26 UK sites

What they found

Definite treatment failure (1 yr): IV 14.6% (74/506) vs. oral 13.2% (67/509). Difference (oral − IV) −1.4 percentage points (90% CI −4.9 to 2.2; 95% CI −5.6 to 2.9) — the upper 90% bound (2.2) sits well below the 7.5-pp margin, so noninferiority is met (it clears the original 5-pp margin too).
Robustness: mITT, per-protocol, and a worst-case sensitivity analysis all agreed; an exploratory Bayesian analysis put the probability that oral is ≥5 pp inferior at 0.1% (≥1 pp inferior, 12.7%). No subgroup favored either route; no difference in time to failure (P=0.57).
IV-associated harms: catheter complications 9.4% (49/523) vs. 1.0% (5/523), P<0.001; early discontinuation of the assigned strategy 18.9% vs. 12.8%, P=0.006; median hospital stay 14 vs. 11 days, P<0.001.
No safety penalty for oral: serious adverse events 27.7% vs. 26.2% (P=0.58); C. difficile 1.7% vs. 1.0% (P=0.30).
Duration myth busted: total antibiotic course did not differ — median 78 (IV) vs. 71 (oral) days (P=0.63); 76.7% continued beyond 6 weeks in both arms.

Risk difference in percentage points (oral − IV failure). Thick bar = 90% CI (the interval tested against the margin); thin whisker = 95% CI. Both estimates — and their whole intervals — sit left of the 7.5-pp margin, and both cross 0, so oral is noninferior but not superior.

Bottom line & practice change

For most patients with bone & joint infection who have a suitable oral option and infection-specialist input, oral therapy can carry the first 6 weeks: equal 1-year treatment failure, shorter hospital stays, far fewer line complications, and no excess C. difficile or serious adverse events. Four caveats travel with that: the agents were expert-selected (this is “appropriately chosen oral,” not “any oral”); follow-up was only 1 year, so late osteomyelitis relapse is not captured; the trial excluded concurrent S. aureus bacteremia and endocarditis, so it does not license oral step-down there; and patients who cannot absorb or have no good oral agent are out of scope.

How good is this evidence?

Strengths

Large pragmatic multicenter RCT (1,054 patients, 26 centers) answering a decades-old dogma.
Objective primary endpoint, adjudicated by an endpoint committee unaware of assignment.
Broad real-world eligibility (all sites, organisms, procedures) → generalizable.
Noninferiority held across mITT, per-protocol, and worst-case analyses, and every subgroup.
High retention (96.3% endpoint data) and good measured adherence.

Weaknesses

Open-label — blinding was impractical/unethical (no matched IV placebo).
NI margin widened mid-trial from 5 to 7.5 pp after the control failure rate ran high.
1-year follow-up only; late relapse of bone infection can occur beyond that.
Agents/durations not standardized — expert-chosen per patient.
Few resistant organisms (e.g. MRSA); poor absorbers underrepresented.

Bias / threats

Open-label can nudge soft/subjective outcomes (SAE attribution, discontinuations) — not the objective primary.
Margin change + imputation of 3.7% missing endpoints are analytic degrees of freedom (sensitivity analyses agreed).
Selection: no-oral-option, septic shock, and IV-only pathogens were excluded.
Result presumes an infection specialist to pick the oral regimen.

Note the asymmetry: the randomized, blindly-adjudicated primary failure result is robust; the mid-trial margin change and the open-label soft endpoints deserve more caution.

Stats Corner: noninferiority on a risk-difference scale (and a moving margin)

OVIVA is a noninferiority trial, not a superiority trial: the question is not “is oral better?” but “is oral no worse than IV by more than a line we drew in advance?” That line — the noninferiority margin — was set on the absolute risk-difference scale at 7.5 percentage points. Noninferiority is declared when the upper end of the 90% confidence interval for (oral − IV) failure stays below +7.5 (a 90% two-sided interval because the test is one-sided at α=0.05). Here the difference was −1.4 pp with a 90% CI of −4.9 to 2.2: the upper bound (2.2) is well left of 7.5, so oral is noninferior. Notice the interval also crosses 0 — so the trial does not claim oral is better; noninferiority and superiority are separate questions. Two teaching points. (1) The margin moved. It began at 5 pp (assuming a 5% control-failure rate) and was widened to 7.5 pp mid-trial when a planned interim showed the real control rate was ≈12.5%. That is defensible — a fixed absolute margin becomes very strict once events are common, and holding 5 pp would have ballooned the sample size from 1,050 to ~1,668 — but widening a margin makes noninferiority easier to reach, so it is a genuine degree of freedom to keep in view. (2) In noninferiority, per-protocol matters as much as ITT. Intention-to-treat blurs the arms toward “no difference,” which flatters an NI claim, so a trustworthy NI result needs ITT and per-protocol to agree — in OVIVA they did (with a worst-case analysis on top).

Where it sits in the literature

Framed with the accompanying NEJM editorial (Boucher), which covered OVIVA and POET together — kept separate from the trial’s own results.

The editorialist accepts that noninferiority was shown but, in 2019, calls it premature to recommend a widespread early switch to oral therapy for bone & joint infection — reserving targeted oral for selected patients who have the infrastructure for close monitoring. Her cautions: the population and regimens were heterogeneous and included few resistant organisms, which biases toward noninferiority; the open-label design was mitigated (not erased) by blinded endpoint adjudication; and the elaborate, expert-driven antibiotic selection presumes a robust ID workforce. She also notes oral carried as many serious adverse events as IV — so outpatients on oral still need monitoring — and suggests OPAT guidelines could add selected oral regimens. Wider context: OVIVA overturned a belief anchored to that 1970 review and a small underpowered meta-analysis, and — published the same week as POET (partial oral for left-sided endocarditis, Iversen, NEJM 2019;380:415-424) — helped move osteoarticular practice toward earlier oral step-down. Keep the scope tight: OVIVA excluded concurrent S. aureus bacteremia and endocarditis, so oral step-down there rests on different trials (e.g. SABATO in low-risk S. aureus bacteremia, Lancet Infect Dis 2024; and POET).

Sources: Li H-K, Rombach I, Zambellas R, et al. Oral versus Intravenous Antibiotics for Bone and Joint Infection (OVIVA). N Engl J Med 2019;380:425-436 (doi:10.1056/NEJMoa1710926; ISRCTN91566927) — article, supplementary appendix, and protocol on file. Editorial: Boucher HW. Partial Oral Therapy for Osteomyelitis and Endocarditis — Is It Time? N Engl J Med 2019;380:487-489 (doi:10.1056/NEJMe1817264). Wider literature: POET (Iversen K, et al. N Engl J Med 2019;380:415-424); SABATO (Lancet Infect Dis 2024;24:523-534); prior osteomyelitis meta-analysis cited in the trial’s introduction.