Skip to main content

Foundational Trials

PRORATA

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Use of procalcitonin to reduce patients' exposure to antibiotics in intensive care units (PRORATA trial): a multicentre randomised controlled trial
  • Acronym: PRORATA
  • Year: 2010
  • Journal published in: The Lancet
  • Citation: Bouadma L, Luyt C-E, Tubach F, et al. Use of procalcitonin to reduce patients' exposure to antibiotics in intensive care units (PRORATA trial): a multicentre randomised controlled trial. Lancet. 2010;375(9713):463-474.

Context & Rationale

  • Background
    • ICU patients have high empiric antibiotic exposure, with downstream risks including antimicrobial resistance, drug toxicity, Clostridioides difficile infection, and selection of multidrug-resistant organisms.
    • Clinical signs, radiology, and routine biomarkers (e.g., C-reactive protein) are often non-specific in critical illness; de-escalation/discontinuation decisions are frequently delayed.
    • Procalcitonin (PCT) rises in many bacterial infections and typically falls with source control and effective therapy, offering a biologically plausible “stop signal”.
    • Before PRORATA, evidence for PCT-guided stewardship in ICU was limited by small trials, heterogeneity in algorithms/assays, and uncertainty about safety (mortality/non-inferiority).
  • Research Question/Hypothesis
    • Whether a PCT-guided antibiotic strategy can reduce antibiotic exposure (increase antibiotic-free days) in ICU patients with suspected bacterial infection.
    • Whether this strategy is non-inferior for mortality (28-day and 60-day), using a pre-specified non-inferiority margin.
  • Why This Matters
    • Provides a pragmatic stewardship framework that tests a biomarker-informed discontinuation strategy in a high-risk population where over-treatment is common.
    • Directly confronts the “stewardship–safety trade-off” by pairing superiority for antibiotic exposure with non-inferiority for mortality.
    • Informs later trial designs, guideline recommendations, and implementation programmes around biomarker-supported antibiotic discontinuation.

Design & Methods

  • Research Question: In critically ill adults with suspected bacterial infection, does a PCT-guided antibiotic algorithm reduce antibiotic exposure while remaining non-inferior for mortality versus standard care?
  • Study Type: Multicentre, randomised, controlled, parallel-group, open-label, investigator-initiated trial in seven French ICUs; allocation stratified by centre with variable block sizes; blinded expert adjudication of infectious outcomes.
  • Population:
    • Setting: Adult ICUs (France); patients enrolled at ICU admission or during ICU stay when a new infectious episode was suspected.
    • Inclusion criteria: Age ≥18 years; suspected bacterial infection prompting consideration of systemic antibiotics; expected ICU stay >3 days.
    • Key exclusions: Expected ICU stay <3 days; Simplified Acute Physiology Score (SAPS) II >65; receipt of antibiotics for >24 hours before eligibility assessment; anticipated need for long-term antibiotic therapy; pregnancy; do-not-resuscitate order at inclusion; bone marrow transplant; AIDS; cystic fibrosis; medullary thyroid carcinoma.
  • Intervention:
    • PCT-guided strategy: PCT measured at each suspected infectious episode up to day 28 and daily during antibiotic therapy; results communicated to clinicians within ~2 hours of sampling.
    • Initiation guidance (PCT at infection suspicion): <0.25 µg/L strongly discouraged; 0.25 to <0.5 discouraged; 0.5 to <1 encouraged; ≥1 strongly encouraged.
    • Continuation/discontinuation guidance (during therapy): stopping strongly encouraged if PCT <0.25; stopping encouraged if PCT 0.25 to <0.5 or if PCT decreased by ≥80% from peak; continuation encouraged if PCT ≥0.5 with a decrease <80% from peak; change of antibiotics strongly encouraged if PCT increased compared with peak and remained ≥0.5.
    • Clinical override: The algorithm could be overridden for syndromes requiring immediate treatment (e.g., septic shock, meningitis) and where clinical judgement indicated ongoing instability despite low PCT.
  • Comparison:
    • Standard-care antibiotic management based on contemporary guidelines and clinician judgement.
    • PCT values were not available to treating clinicians in the control group.
  • Blinding: Open-label for treating clinicians and patients; an expert committee blinded to allocation adjudicated relapse and superinfection and confirmed infectious episodes.
  • Statistics: Power calculations: (1) 133 patients per group to detect a 3-day increase in antibiotic-free days (assumed mean 12, SD 7.5) with 90% power at a 5% significance level; (2) 300 patients per group to exclude a 10% absolute mortality difference (assumed mortality 35%) with 80% power at a 10% alpha risk (non-inferiority assessed using a two-sided 90% CI); planned enrolment 630. Primary analysis: intention-to-treat.
  • Follow-Up Period: Antibiotic exposure and infectious outcomes through day 28; mortality assessed at day 28 and day 60.

Key Results

This trial was not stopped early. Recruitment proceeded to the planned sample size; 630 were randomised and 621 were analysed in the intention-to-treat population (307 PCT-guided; 314 control).

Outcome PCT-guided strategy Standard care Effect p value / 95% CI Notes
All-cause mortality (day 28) 65/307 (21.2%) 64/314 (20.4%) Absolute difference 0.8% 90% CI -4.6 to 6.2; P=Not reported Non-inferiority margin 10%; point estimate favoured control.
All-cause mortality (day 60) 92/307 (30.0%) 82/314 (26.1%) Absolute difference 3.8% 90% CI -2.1 to 9.7; P=Not reported Non-inferiority margin 10%; confidence interval allows clinically important harm up to 9.7%.
Days without antibiotics (day 1–28) 14.3 (SD 9.1) 11.6 (SD 8.2) Mean difference 2.7 days 95% CI 1.4 to 4.1; P<0.0001 Co-primary endpoint (superiority); favoured PCT-guided.
Antibiotic exposure (days of therapy, day 1–28) 10.3 (SD 7.3) 13.3 (SD 8.7) Mean difference -3.0 days 95% CI -4.1 to -1.8; P<0.0001 Pragmatic reduction in total antibiotic days per patient.
Relapse (day 1–28) 20/307 (6.5%) 16/314 (5.1%) Absolute difference 1.4% 95% CI -2.3 to 5.1; P=0.45 Adjudicated; no evidence of excess relapse.
Superinfection (day 1–28) 106/307 (34.5%) 97/314 (30.9%) Absolute difference 3.6% 95% CI -3.8 to 11.0; P=0.29 High event rates; imprecision permits moderate harm or benefit.
Multidrug-resistant bacteria (day 1–28) 55/307 (17.9%) 52/314 (16.6%) Absolute difference 1.3% 95% CI -4.6 to 7.2; P=0.67 No reduction detected despite lower antibiotic exposure.
SOFA score (day 28) 1.5 (SD 3.0) 0.9 (SD 2.4) Mean difference 0.6 95% CI 0.0 to 1.1; P=0.0370 Statistically higher in PCT-guided group; clinical significance uncertain.
ICU length of stay from inclusion 15.9 (SD 16.1) days 14.4 (SD 14.1) days Mean difference 1.5 days 95% CI -0.9 to 3.9; P=0.23 No evidence of shorter LOS despite fewer antibiotics.
Duration of first antibiotic episode (overall) 6.1 (SD 6.0) days 9.9 (SD 7.1) days Mean difference -3.8 days 95% CI -4.8 to -2.7; P<0.0001 Largest between-group separation occurred in episode duration (stopping decisions).
Duration of first episode: ventilator-associated pneumonia 7.3 (SD 5.3) days 9.4 (SD 5.7) days Mean difference -2.1 days 95% CI -4.0 to -0.3; P=0.0210 Syndrome-level reduction consistent with stewardship direction of effect.
    • PCT guidance increased antibiotic-free days by 2.7 (95% CI 1.4 to 4.1) and reduced total antibiotic exposure by 3.0 days (95% CI -4.1 to -1.8), without clear differences in relapse, superinfection, multidrug resistance, or length of stay.
    • Mortality estimates were imprecise: day-60 mortality was numerically higher in the PCT group (30.0% vs 26.1%) with a 90% CI allowing harm up to 9.7%.
    • Clinicians frequently overrode algorithm advice in both arms, yet substantial separation in antibiotic exposure was maintained (particularly for duration of the first antibiotic episode).

Internal Validity

    • Randomisation and allocation: Computer-generated allocation; stratified by centre; variable block sizes; investigators masked to allocation before randomisation but not afterwards.
    • Dropout/exclusions: 630 randomised; 9 excluded from analysis; intention-to-treat population 621 (307 vs 314); loss to follow-up 1 per group for 60-day mortality.
    • Performance/detection bias: Open-label antibiotic decisions introduce performance bias (clinicians could stop earlier when prompted by PCT); mortality is objective; relapse/superinfection were adjudicated by a blinded expert committee, limiting detection bias for these endpoints.
    • Protocol adherence: Recommendations were overridden frequently.
      • PCT-guided arm: algorithm advice not followed in 219 episodes; examples included antibiotic initiation despite PCT <0.5 µg/L in 65 patients because infection could not be ruled out, and continuation beyond “stop” advice in clinically unstable patients.
      • Control arm: guideline recommendations not followed in 146 episodes; reasons included postponed initiation (n=15), premature stopping (n=46), or continuation beyond guidance (n=85); overall non-adherence reported as 45%.
    • Baseline characteristics: Broadly balanced (examples): age 64.4 vs 64.7 years; SAPS II 44.7 vs 42.6; SOFA day 1 7.5 vs 7.2; septic shock at inclusion 21.8% vs 18.5%; mechanically ventilated at inclusion 78.5% vs 74.8%.
    • Heterogeneity: Mixed infection syndromes (respiratory predominating) increase clinical heterogeneity; randomisation should balance measured/unmeasured confounders, but syndrome-level effect modification is plausible (and episode-duration reductions varied by syndrome).
    • Timing: Rapid turnaround (reported within ~2 hours) supports real-world decision-making; however, initial antibiotic initiation was similar between groups (reflecting that biomarker guidance mainly affected stopping rather than early empiric start decisions).
    • Dose/intensity of intervention: The “dose” was the frequency and fidelity of PCT-informed decisions; high override rates diluted the biological separation, yet antibiotic exposure differed meaningfully.
    • Separation of the variable of interest:
      • Days without antibiotics (day 1–28): 14.3 (SD 9.1) vs 11.6 (SD 8.2).
      • Antibiotic exposure (days of therapy, day 1–28): 10.3 (SD 7.3) vs 13.3 (SD 8.7).
      • Days of antibiotic exposure per 1000 inpatient days: 653 vs 812.
      • Duration of first antibiotic episode (overall): 6.1 (SD 6.0) vs 9.9 (SD 7.1).
    • Key delivery aspects: The intervention worked as a discontinuation tool more than an initiation tool; the control group received explicit guidance on recommended treatment duration (which may have reduced the contrast versus “usual” care).
    • Outcome assessment: Mortality objective; relapse and superinfection adjudicated; multidrug-resistant bacteria outcomes depend on surveillance practices and may be underpowered for modest differences.
    • Statistical rigour: Pre-specified co-primary endpoints (antibiotic-free days superiority; mortality non-inferiority with 90% CI); trial achieved planned sample size; interpretation of non-inferiority depends on acceptance of the 10% margin and the use of 90% CI.

Conclusion on Internal Validity: Overall, internal validity is moderate: randomisation and adjudication support causal inference for antibiotic exposure outcomes, but open-label care and substantial algorithm overriding introduce meaningful performance and implementation bias, and mortality non-inferiority remains imprecisely estimated.

External Validity

    • Population representativeness: Adult ICU patients with suspected infection typical of many mixed ICUs; however, exclusions (SAPS II >65, major immunosuppression such as bone marrow transplant or AIDS, pregnancy) limit applicability to the sickest and specific high-risk immunocompromised groups.
    • Setting: Seven ICUs within one health system context (France); local antimicrobial practices and baseline durations influence effect size and may differ across regions.
    • Feasibility requirements: Requires timely PCT testing with rapid reporting and a structured stewardship workflow; resource-limited settings or those without reliable biomarker turnaround may not replicate delivery fidelity.
    • Applicability across syndromes: Syndrome-level reductions were largest where “standard” durations were longest (e.g., community-acquired pneumonia, urinary tract infection), suggesting local baseline practice patterns modulate benefit.

Conclusion on External Validity: Generalisability is moderate: findings translate well to adult mixed ICUs with access to rapid PCT assays and stewardship support, but are less applicable to profoundly immunocompromised patients, very high-severity cohorts excluded by design, and settings where antibiotic durations are already short.

Strengths & Limitations

  • Strengths:
    • Pragmatic ICU trial with clinically relevant stewardship outcomes and mortality safety assessment.
    • Structured, explicit PCT algorithm with rapid turnaround embedded in real-world decision-making.
    • Blinded expert adjudication of relapse and superinfection mitigated outcome-classification bias.
    • Meaningful achieved separation in antibiotic exposure (e.g., -3.0 antibiotic days and +2.7 antibiotic-free days over 28 days).
  • Limitations:
    • Open-label design and non-binding recommendations: antibiotic decisions were susceptible to behavioural and co-intervention effects.
    • High protocol deviation/override rates in both groups dilute interpretability of “algorithm efficacy” versus “decision support under clinician discretion”.
    • Non-inferiority framework: a 10% absolute mortality margin may be considered clinically wide, and mortality estimates remained imprecise (day-60 upper 90% CI approaching margin).
    • Single-country, seven-centre setting with explicit control-group guidance; effect size may differ where baseline antibiotic durations and stewardship maturity differ.

Interpretation & Why It Matters

  • Clinical and methodological implications
    • PCT can function as an “off-ramp” for antibiotics in ICU care: PRORATA’s effect concentrated in shortening episode duration rather than changing initial prescribing.
    • The trial highlights the implementation reality of stewardship tools: clinician overrides were common yet reductions in antibiotic exposure persisted, suggesting decision support can shift default durations even without perfect adherence.
    • Safety interpretation requires nuance: while mortality differences were compatible with non-inferiority under the pre-specified framework, the day-60 point estimate and higher day-28 SOFA underscore the need for careful clinical context, especially in persistently unstable patients.
    • PRORATA helped establish the now-common design pairing stewardship endpoints with patient-centred safety outcomes and influenced subsequent larger ICU trials and guideline language around biomarker-aided discontinuation.

Controversies & Subsequent Evidence

    • Open-label stewardship effects and algorithm adherence: The accompanying Lancet commentary emphasised that non-blinded care and frequent clinician overriding complicate attribution of benefit solely to biomarker biology and may dilute or distort safety signals.1
    • Non-inferiority margin and mortality interpretation: A 10% absolute margin can be judged permissive for a stewardship intervention; PRORATA’s 60-day mortality CI approached this margin and the point estimate favoured control, keeping residual concern about harm in some subgroups or implementation contexts.1
    • Subsequent trials: Larger multicentre ICU trials of PCT-guided discontinuation (e.g., SAPS) broadly reproduced reductions in antibiotic duration and did not demonstrate excess mortality, supporting safety when embedded in structured clinical review and stewardship workflows.2
    • Meta-analytic evidence: Systematic reviews and meta-analyses focused on critically ill/septic populations generally report reduced antibiotic exposure with no increase in mortality; some analyses suggest potential survival benefit, but heterogeneity and implementation fidelity remain key drivers of effect estimates.36
    • Guideline incorporation: International guidelines now frame PCT as an adjunct to clinical assessment for discontinuation decisions (not a stand-alone trigger for initiation), reflecting the totality of ICU trial and meta-analytic evidence post-PRORATA.45

Summary

    • PRORATA tested a pragmatic PCT-guided algorithm to guide antibiotic decisions in ICU patients with suspected bacterial infection.
    • The intervention increased antibiotic-free days (14.3 vs 11.6; mean difference 2.7 days) and reduced total antibiotic exposure (10.3 vs 13.3 days; mean difference -3.0 days) over 28 days.
    • Mortality was similar at 28 days and numerically higher at 60 days (30.0% vs 26.1%), with imprecision compatible with clinically important harm up to 9.7% by 90% CI.
    • Relapse, superinfection, multidrug-resistant bacteria, and length of stay were not meaningfully different, but SOFA at day 28 was slightly higher in the PCT group (1.5 vs 0.9).
    • Clinician override rates were high in both groups, underscoring that biomarker-guided stewardship is an implementation intervention as much as a biological one.

Further Reading

Other Trials

Systematic Review & Meta Analysis

Observational Studies

Guidelines

Notes

  • PRORATA is best interpreted as an implementation trial of biomarker-informed discontinuation: the achieved antibiotic reduction occurred despite frequent clinician overrides, and mortality estimates remain imprecise enough to mandate context-aware adoption.

Overall Takeaway

PRORATA was a landmark ICU stewardship trial demonstrating that a structured PCT-guided algorithm can meaningfully reduce antibiotic exposure in critically ill adults, primarily by shortening treatment duration rather than preventing initiation. While mortality was broadly similar under the trial’s non-inferiority framework, residual uncertainty remains because the day-60 point estimate favoured standard care and confidence intervals permitted clinically important harm, reinforcing that PCT should augment—rather than replace—clinical judgement and reassessment.

Overall Summary

  • PCT-guided decision support reduced antibiotic exposure by about 3 days over 28 days without clear signals for increased relapse or superinfection, but mortality safety remained imprecise and implementation fidelity (override rates) was a defining feature.

Bibliography