Delivery and adherence of biomarker-guided antibiotic treatment protocols in the ADAPT-Sepsis clinical trial: an in trial descriptive process evaluation of daily procalcitonin- and C-reactive protein-guided protocols

Introduction. Biomarker-guided antibiotic protocols using daily procalcitonin or C-reactive protein have been proposed to optimise antibiotic duration in sepsis. This pre-planned process evaluation of the ADAPT-Sepsis trial examined delivery and clinician adherence to procalcitonin- and C-reactive protein-guided protocols. Methods. An embedded descriptive process evaluation was conducted within the multicentre ADAPT-Sepsis trial including 2760 adults across 41 UK critical care units. Patients were assigned to procalcitonin-guided, C-reactive protein-guided or standard care. Algorithm-generated recommendations (strong stop, suggest stop or standard care) were relayed to research staff and clinicians who remained blinded to group allocation. Adherence was defined as stopping biomarker sampling after advice. Advice timings, frequencies, adherence rates, and reasons for non-adherence were recorded and analysed. Results. Among 2760 patients, strong stop advice occurred earlier in those allocated to the procalcitonin-guided group compared with the C-reactive protein-guided group (mean difference -1.7 days, 95%CI -2.5 to -0.9, p < 0.001). C-reactive protein guidance generated a higher proportion of standard care advice (mean difference -9.1%, 95%CI -10.9% to -7.3%, p < 0.001). Overall advice adherence was greater in patients allocated to the C-reactive protein group than those allocated to the procalcitonin group (mean difference -6.8%, 95%CI -9.0 to -4.5, p < 0.001), driven by the higher proportion of standard care advice and later stronger stop advice with higher adherence (mean difference -14.3%, 95%CI -21.8 to -6.8, p < 0.001). Reasons for advice non-adherence included fixed-duration microbiology guidance, illness severity, continuing fever and elevation of routine inflammatory markers, which did not differ between groups (p = 0.54). Discussion. A C-reactive protein‑guided discontinuation protocol achieved higher clinician adherence probably because its recommendations aligned more closely with established prescribing norms. In contrast, procalcitonin guidance produced earlier and more frequent stop signals, offering greater potential to reduce antibiotic exposure but requiring clinicians to accept earlier discontinuation than usual practice.