Articles
Accounting for non-adherence to assigned antibiotic treatment duration for bloodstream infection (BALANCE): a post-hoc analysis of a randomised clinical trial
Sean W X Ong, Ruxandra Pinto, Robert K Mahar, et al
Lancet Infect Dis 2026; 26: 250-259
https://doi.org/10.1016/S1473-3099(25)00592-4
Summary
Background
The BALANCE trial (NCT03005145) showed non-inferiority of 7-day versus 14-day antibiotic duration for treatment of patients with non-Staphylococcus aureus uncomplicated bloodstream infection (BSI). Non-adherence to assigned duration occurred in 728 (20·3%) of 3581 patients. This post-hoc analysis aimed to identify factors associated with this non-adherence, and evaluate its effect on validity of trial results.
Methods
We identified factors associated with non-adherence using generalised linear mixed models. Adherence was defined as receiving antibiotic duration up to 2 days longer or shorter than assigned duration, measured using a trial-specific adherence variable (assessed daily by research staff) and a separate variable capturing total antibiotic duration. We estimated the effect of antibiotic duration on the primary outcome of 90-day all-cause mortality, accounting for confounding introduced by non-adherence, in adjusted per-protocol and as-treated populations (using two different adherence definitions: first, using an adherence variable as assessed daily by site research staff, and second, using a total antibiotic duration variable recording the total cumulative antibiotic duration received), using inverse probability of treatment weighting (IPTW) and instrumental variable methods. We compared these adjusted estimates to the primary effect estimates in BALANCE and used the same 4% non-inferiority margin to determine non-inferiority of 7-day duration.
Findings
3581 patients with complete outcome data were included. Non-adherence was higher in the 7-day group (432 [24·0%] of 1802 patients) versus 14-day group (296 [16·6%] of 1779). Disease severity, intra-abdominal and skin or soft tissue source of infection, persistent fever, and persistent bacteraemia were associated with treatment prolongation, whereas vascular catheter source and antimicrobial resistance were associated with treatment shortening. Adjusted risk differences for the primary outcome of 90-day all-cause mortality were similar to the primary results in BALANCE: (1) IPTW per-protocol analysis (definition one): –1·61% (95% CI –4·13 to 0·87); IPTW per-protocol analysis (definition two): –1·91% (95% CI –4·61 to 0·93); and IPTW as-treated analysis: –0·75% (95% CI –3·36 to 1·86); and (2) instrumental variable analysis (definition one): –2·68% (95% CI –6·65 to 1·29) and instrumental variable analysis (definition two): –2·80% (95% CI –5·98 to 0·37).
Interpretation
Non-adherence in BALANCE was significantly associated with a range of important prognostic factors, which might have introduced bias. Causal inference methods addressing this bias showed a consistent conclusion of non-inferiority, strengthening the validity of this finding. 7-day antibiotic duration should be the standard of care for patients with uncomplicated BSI.
Funding
The BALANCE trial was funded by the Canadian Institutes of Health Research, Health Research Council of New Zealand, Australian National Medical Research Council, Physicians Services Incorporated Ontario, and Ontario Ministry of Health and Long-term Care Innovation Fund.
