Methods

The PRORATA study. A multi‐centre study based in France. Unblinded, randomised controlled trial comparing strategy utilising serum procalcitonin measurement to guide initiation and cessation of antibiotics for ICU patients with suspected bacterial infection, with antibiotic administration according to an agreed guideline

Participants

630 adult ICU patients (8 medical, surgical or medical‐surgical ICUs; 6 hospitals). Mean age was 62; 66% male; SAPS II 47; SOFA score 7.8) with suspected bacterial infection (either at admission or during ICU stay). 89% patients were medical.16% enrolled patients were defined as being "Immunocompromised" (including patients with AIDS, solid organ transplant, haematological malignancy, prior chemotherapy or radiotherapy, and long‐term corticosteroid or other immunosuppressant therapy); identifying and excluding data from these patients for the purpose of the systematic review was not possible. There appear not to have been significant differences in baseline characteristics between treatment groups. Of 621 patients entered into the analysis, 214 (34%) had VAP or "HAP": 141 (23%) patients were diagnosed with VAP, and 73 (12%) patients had HAP not requiring mechanical ventilation. For all patients with microbiologically‐confirmed infection, initial antibiotic therapy was appropriate in 92% cases. Diagnostic criteria for VAP and HAP were not provided in the full‐text article

Exclusions: pregnancy, expected ICU stay < 3 days, bone marrow transplant or chemotherapy‐induced neutropenia, infections requiring long‐term course antibiotic therapy (e.g. infective endocarditis), poor chance of survival (SAPS II > 65), DNR order

Interventions

Patients randomised to intervention group (311 randomised; 4 withdrew consent; 307 included in analysis; 75 patients with VAP; 29 with HAP not requiring mechanical ventilation) had serum procalcitonin level measured at inclusion, at each infectious episode until day 28, and for every morning that antibiotics were administered. The guidance for withholding or stopping antibiotic therapy was: i. antibiotics to be withheld when procalcitonin was < 0.5 µg/L on day of study entry; ii. antibiotics to be discontinued when procalcitonin level had fallen to < 0.5 µg/L or to less than 20% of the peak procalcitonin concentration

For patients randomised to control group (319 randomised; 4 withdrew; 1 randomised twice; 314 included in analysis; 66 patients with VAP; 44 patients with HAP not requiring mechanical ventilation), physicians were encouraged to administer antibiotics according to agreed recommendations (including duration of therapy)

Outcomes

Most outcome data were presented for all patients, without specific reference to patients with VAP or HAP (non‐ventilated)

For all patients, the following outcomes were presented:

• Mortality (28‐day and 60‐day)

• Number of days without antibiotics (28‐day)

• Relapse or superinfection (28‐day)

• Mechanical ventilation‐free days (28‐day)

• SOFA scores (Days 1, 7, 14, 21 and 28)

• Duration of stay in ICU

• Duration of stay in hospital

• Days of exposure to each antibiotic per 1000 patient days

• Number of days of continuous antibiotic treatment (28‐day)

• Duration of antibiotic treatment according to infection site

• Percentage of emerging multi‐drug resistant bacteria isolated from routine microbiological assessment (28‐day)

For patients specifically with VAP, the following outcomes were presented:

• Duration of first episode of antibiotic treatment

• Number of days without antibiotics (28‐day)

• Mortality (28‐day)

Notes

Procalcitonin measurements were used to guide whether to initiate antibiotic therapy as well as when to discontinue therapy. A very low proportion of recruited patients were surgical, which may limit generalisability of results. There was a high incidence of protocol non‐adherence; the algorithm for antimicrobial therapy administration was not followed (overall) in 162 patients (53%) in the intervention group, nor for 141 patients in the control group (45%). For all patients, a non‐significant increase in 60‐day mortality was observed in the procalcitonin group (92/307 (30%) versus 82/ 314 (26.1%)] in the control group). Furthermore, the study may have been under‐powered to determine non‐inferiority of PCT‐guided therapy in terms of death, since it was based upon a 35% absolute mortality in the control group and a 10% between‐group mortality difference. For all patients, in the procalcitonin group there were non‐significantly greater relapse rates (20/ 307 (6.5%) versus 16/ 314 (5.1%) in control group; P = 0.45) and superinfection rates (106/ 307 (34.5%) versus 97/ 314 (30.9%) in the control group; P = 0.29)

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

"Independent, centralised, computer‐generated randomisation sequence... was used to randomly assign patients."

Allocation concealment (selection bias)

Low risk

"Investigators were masked to assignment before... randomisation."

Blinding (performance bias and detection bias)
All outcomes

High risk

Unblinded study

Incomplete outcome data (attrition bias)
All outcomes

Low risk

2 patients lost to follow‐up, 1 from intervention and 1 from control group. This is unlikely to have a clinically relevant impact

Selective reporting (reporting bias)

Unclear risk

Unclear

Other bias

High risk

The algorithm for antimicrobial therapy administration was not followed (overall) in 162 patients (53%) for the intervention group, nor for 141 patients in the control group (45%)

Methods

The PNEUMA study, a multi‐centre study based in France. Unblinded, randomised controlled trial comparing fixed durations (8‐day versus 15‐day) of antibiotic therapy for VAP. Randomisation occurred 3 days after bronchoscopy confirming diagnosis of VAP

Participants

401 adult patients. 51 French ICUs. 72% male; mean age 61; episodes due to NF‐GNB 32.5%, MRSA 11.2%; mean SAPS II score 45; mean SOFA score 7.4 at admission. VAP was diagnosed according to the following criteria: new and persistent radiographic infiltrate, plus 1 of: purulent tracheal secretions, temperature of 38.4 °C or higher, or leukocyte count > 10,000/µL; and positive quantitative culture of 10⁴ cfu/mL from BAL or 10³ cfu/mL from PSB. Duration of mechanical ventilation prior to VAP: 13.6 days. No significant differences between groups at baseline, with the exception of significantly higher proportion of men (76.6%) in 8‐day regimen versus men in 15‐day regimen (67.6%; P = 0.046). All patients received appropriate initial antibiotics. Exclusions included early onset pneumonia (within 5 days of commencing mechanical ventilation) in patients who had received no antimicrobial therapy in the 15 days prior to diagnosis of pneumonia, and immunocompromised state, characterised by: neutropenia, AIDS, long‐term corticosteroids or other immunosuppressant therapy

Interventions

197 patients received fixed 8‐day course of antibiotics (chosen by treating physician); 204 patients received a 15‐day course

Outcomes

The following outcome measures were reported:

• Death from any cause (28‐day and 60‐day)

• Microbiologically‐confirmed pulmonary infection recurrence (28‐day)

• Antibiotic‐free days (28‐day)

• Mechanical ventilation‐free days (28‐day)

• Number of organ failure‐free days (28‐day)

• Evolution of parameters comprising the SOFA score and ODIN score (Day 1 to 28)

• Clinical features relevant to pulmonary infection (fever, leukocyte count, PaO₂/ FiO₂ ratio, radiologic score; Day 1 to 28)

• Duration of stay ITU

• Rate of unfavourable outcome (death, recurrence, prescription of new antibiotic therapy)

• In‐hospital mortality

• Percentage of emerging multiresistant bacteria during ITU in bronchoscopic samples collected to investigate possible recurrence

Notes

Repeat bronchoscopy was performed on the basis of fever, purulent secretions, new or progressive pulmonary infiltrates, or deterioration in respiratory or haemodynamic parameters. It was not performed routinely, e.g. on completion of 8‐day course of therapy, and consequently data regarding persistent colonisation with NF‐GNB is not available

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

"Randomisation was performed...and stratified... according to a computer‐generated random‐number table."

Allocation concealment (selection bias)

Low risk

"Randomisation performed centrally, using an interactive voice system... randomisation was not communicated to the investigators until day 8... On that day, investigators had to telephone the randomisation centre to receive the treatment assignment by fax."

Blinding (performance bias and detection bias)
All outcomes

High risk

"...Patients, medical and nursing staffs, and pharmacists remained blinded until [day 8]." However, importantly, no attempt was made to blind from day 8, i.e. the point from which thereafter allocation might make a significant difference

Incomplete outcome data (attrition bias)
All outcomes

Low risk

Following randomisation, 0 patients were lost to follow‐up. 1 patient excluded from analysis following withdrawal of consent

Selective reporting (reporting bias)

Unclear risk

Study protocol not available for examination. However, outcome measures are those expected and appropriately presented within the report

Other bias

Low risk

There was a significantly higher proportion of men in the 8‐day group (76.6%) compared with the 15‐day group (67.6%; P = 0.046). However, there were not significant differences in illness severity scores, prior duration of mechanical ventilation, prior antibiotic administration, micro‐organisms responsible for VAP, and antibiotic regimes to treat VAP

Data regarding proportions of patients who contrary to protocol did not receive a full 8‐day or 15‐day course of antibiotics are not provided. However, absolute antibiotic‐treatment days and 28‐day antibiotic‐free day data indicates significantly less antibiotic exposure as a consequence of allocation to short or prolonged‐course therapy

Methods

Single‐centre study based in Tunisia. RCT comparing fixed durations (7‐day versus 10‐day) of antibiotic therapy for VAP

Participants

Medical ICU. 30 adult patients (63% male; mean age 63 years; NF‐GNB 72%; SAPS II 42.4). VAP (onset more than 48 hours after mechanical ventilation in ICU) diagnosed was suspected on the basis of: new and persistent radiographic infiltrate, purulent secretions, fever or deteriorating gas exchange or white cell count and confirmed on quantitative analysis of culture of endotracheal aspirate (> 10⁴ cfu/ml) or protected distal respiratory specimens (>10³ cfu/ml). Mean onset of VAP after institution of mechanical ventilation: 10 days. Initial antibiotics were appropriate in 94% of cases. No significant differences in baseline characteristics between the 2 groups

Exclusions include: second episode of pneumonia during single hospitalisation, terminal illness, failure to isolate bacterial growth

30 patients randomised from 39 patients with clinical features of VAP: 9 not enrolled because of terminal illness or failure to isolate bacteria

Interventions

14 patients randomised to receive 7‐day course of antibiotics; 16 patients to receive 16‐day course. Choice of antibiotic: on microbiology advice, taking into account whether early (up to including 5 days after commencing mechanical ventilation) or late‐onset VAP, and whether risk factors for multi‐resistant bacteria present, and modified according to culture/sensitivity results

94% patients received appropriate initial antibiotic therapy

Outcomes

The following outcome measures were presented:

• Mortality (14‐day and 28‐day)

• Antibiotic‐free days (28‐day)

• Microbiological resolution (timescale not specified)

• CPIS resolution (days 1, 10, 14)

• Mechanical ventilation‐free days (28‐day)

• Recurrence: relapse or super‐infection (timescale not specified)

• Duration of ITU stay

Notes

Data regarding protocol violations and patients lost to follow‐up not available. Unable to make contact trial with authors to request supplementary information

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

Use of random number table

Allocation concealment (selection bias)

Unclear risk

Inadequately reported

Blinding (performance bias and detection bias)
All outcomes

Unclear risk

Inadequately reported

Incomplete outcome data (attrition bias)
All outcomes

Unclear risk

Inadequately reported

Selective reporting (reporting bias)

Unclear risk

Inadequately reported

Other bias

Low risk

No evidence of other source of bias

Methods

2 centre study, based in Uruguay, conducted May 2003 to December 2006 (during the period December 2005 to April 2006 recruitment was interrupted while eligible patients were enrolled in the Pontet 2007 study). Randomised controlled study comparing fixed short (8‐day) and long (12‐day) courses of antibiotic therapy for VAP.

Participants

77 patients (medical, surgical and neurosurgical; mean age 59 years, 53% male, 63.6% NF‐GNB, 9.1% MRSA; median APACHE II score 21, MODS score 5, SOFA score 6) with VAP. VAP was diagnosed on the basis of: new and persistent radiographic infiltrates, 2 of temperature ≥38.5 ºC or <36 ºC, leukocytes ≥12,000/ mm³ or ≤ 4 x10³/ mm³, and BAL culture ≥ 10⁴ cfu/ml, or positive ETA plus CPIS > 6, or micro‐organism in ≥ 2 blood cultures with identical sensitivity to tracheal secretions and in absence of other possible infection, or positive culture of pleural fluid. Mean time after commencing mechanical ventilation before onset of VAP: 9.3 days. 68% patients had received a prior course of antibiotics. In 97% cases, initial antibiotic therapy for VAP was appropriate. There were no significant differences in baseline characteristics between treatment groups
Exclusion criteria: failure to meet clinical and microbiological criteria, decision not to institute antibiotic therapy, duration of therapy < 6 days, death before 6th day of treatment, onset of VAP within 48 hours of admission from other centre

Interventions

77 patients randomised to 8‐day treatment or 12‐day course on Day 8 of antibiotic therapy. Antibiotic choice was that of the attending physician; in 75/77 (97%) cases, initial antibiotic therapy was appropriate. The most commonly used antibiotics were: cefoperazone‐sulbactam, carbapenem and other third‐generation cephalosporins. In 51% cases, antibiotic combinations were used

Outcomes

The following outcomes were studied:

• ITU mortality

• VAP‐related mortality

• Clinical resolution of VAP

• Non‐resolution of VAP ("Therapeutic failure")

• Recurrence of VAP

• Duration of mechanical ventilation

Notes

0 patients were lost to follow‐up. No patient received a shorter duration of antibiotic therapy than allocated; data incomplete for patients who had antibiotics continued beyond the allocated duration

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

Block randomisation using random number table

Allocation concealment (selection bias)

Low risk

Sealed, numbered envelopes opened sequentially

Blinding (performance bias and detection bias)
All outcomes

High risk

Unblinded study

Incomplete outcome data (attrition bias)
All outcomes

Low risk

No patients lost to follow‐up

Selective reporting (reporting bias)

Unclear risk

Unclear. Data incomplete for patients who had antibiotics continued beyond the allocated duration

Other bias

Low risk

Study appears to be free of other sources of bias

Methods

Single‐centre, medical ICU from USA. Randomised controlled study comparing strategy involving discontinuation policy to decrease antibiotic administration for VAP with standard‐therapy

Participants

302 patients receiving antibiotic treatment for VAP were randomised. VAP was diagnosed on the basis of: new persistent radiographic infiltrates together with one of: positive pleural culture (same organism as in sputum/tracheal aspirate), radiographic cavitation, histopathologic evidence of pneumonia, or 2 of: fever, leukocytosis and purulent tracheal aspirate or sputum. Mean age 60 years; 50% male; 29% COPD; APACHE II score 23; mean CPIS 7.1; Pseudomonas or Acinetobacter infection 11%; MRSA 20%. 25% patients had received prior antibiotics. Proportion of patients receiving appropriate initial antibiotic therapy was 94% overall. 18.6% patients were defined as "Immunosuppressed". No significant differences in baseline characteristics between groups

Exclusions: transfers to host institution because of lack of capacity at external institutions

Interventions

154 patients randomised to discontinuation group; one of investigators recommended discontinuation of antibiotics to treating physician on weekdays if: non‐infectious cause for radiographic infiltrates identified, signs and symptoms of active infection had resolved (in terms of temperature, white cell count, radiographic appearance, sputum characteristics and PaO2/ FiO2 ratio). 148 patients were randomised to the control group. Recommendations regarding choice of antibiotic therapy were made to both groups

Outcomes

Reported outcome measures:

• Duration of antibiotic therapy for VAP

• ITU mortality

• Hospital mortality

• Duration of ITU stay

• Duration of hospital stay

• Duration of mechanical ventilation

• VAP recurrence during same ITU stay

Notes

Outcome data was missing for 4 patients in the intervention group, and for 8 patients in the control group. Among intervention group, recommendations to discontinue antibiotics were given to the physicians of 142 patients (94.7%); among 88.7% of these patients antibiotics were discontinued within 48 hours of the recommendation. Trial author contacted, but no further information available

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Unclear risk

No details provided

Allocation concealment (selection bias)

Unclear risk

No details provided

Blinding (performance bias and detection bias)
All outcomes

High risk

Intervention involved investigators contacting the physician team

Incomplete outcome data (attrition bias)
All outcomes

High risk

Outcome data missing in 12/302 (4%) cases; no information given regarding missing outcome data

Selective reporting (reporting bias)

Unclear risk

Unclear

Other bias

Low risk

There do not appear to be other major potential sources of bias

Methods

4‐centre Uruguayan study, conducted December 2005 to April 2006; 2 of these centres also recruited to the study Medina 2007, but outside this time period. Unblinded, randomised controlled study comparing PCT‐guided strategy for antibiotic cessation versus standard therapy for VAP

Participants

81 adult patients (medical, surgical, cardiothoracic and neurosurgical; mean age 54 years, 63% male; NF‐GNB 23.4%; MRSA 6.1%; Day 1 MODS 6.2; Day 1 SOFA score 4.8) with suspected VAP. VAP was diagnosed on the basis of: new and persistent radiographic infiltrates, 2 of temperature ≥ 38.5 ºC or < 36 ºC, leukocytes ≥12,000/ mm³ or ≤ 4 x10³/ mm³, and BAL culture ≥ 10⁴ cfu/ml, or positive ETA plus CPIS > 6, or micro‐organism in ≥ 2 blood cultures with identical sensitivity to tracheal secretions and in absence of other possible infection, or positive culture of pleural fluid. Patients were excluded on the basis of: AIDS, leukaemia or immunosuppression. A further 14 patients were excluded after randomisation; from control group, 5 patients were excluded because of short duration of treatment (3 patients) or negative BAL (3 patients); from the PCT group, 9 patients were excluded post‐randomisation because Day 1 or 2 PCT was < 0.5 ng/mL. There were no significant baseline differences between groups in terms of demographics, underlying disease or prior antibiotic administration

Interventions

Intervention: for the group receiving PCT‐guided therapy; if at day 7 PCT was < 0.5 ng/ml, antibiotic discontinuation was encouraged. Duration of therapy in control group was according to pre‐existing guidelines in place at each ICU. Choice of antibiotic therapy was that of the treating physician. Most commonly used antibiotics were: cephalosporins, ampicillin‐sulbactam and amikacin; antibiotic combinations were used in 54.5% cases

Outcomes

Data are presented as a per‐protocol analysis. The following outcome measures were reported:

• ITU mortality

• VAP‐related ITU mortality

• Duration of mechanical ventilation

• Duration of ITU stay

• Non‐resolution (therapeutic failure)

• Relapse (Day 29)

• Super‐infection (Day 29)

• Clinical resolution

• CPIS (Day 1, 7)

• MODS (Day 1, 7)

• SOFA score (Day 1, 7)

Notes

All patients in the procalcitonin group completed at least 7 days of antibiotic therapy. At Day 7, antibiotics were discontinued for all patients with procalcitonin < 0.5 ng/ml

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

Block randomisation using random number table

Allocation concealment (selection bias)

Low risk

Sealed, numbered envelopes opened in sequence

Blinding (performance bias and detection bias)
All outcomes

High risk

Treating team were made aware of assignment on Day 7, when PCT measured and communicated to PCT group

Incomplete outcome data (attrition bias)
All outcomes

Low risk

No patients lost to follow‐up

Selective reporting (reporting bias)

Unclear risk

Unclear

Other bias

High risk

Potential source of bias as this is a per‐protocol analysis; exclusion of 9 patients with low PCT measurements in the PCT group may exclude a higher proportion of relatively well patients compared with the control group

Methods

Single‐centre study from the USA. Randomised controlled trial comparing strategy of 3 days' ciprofloxacin monotherapy versus standard therapy (duration and antibiotic choice at the discretion of physician) for patients with pulmonary infiltrates but low‐probability pneumonia

Participants

Study entry criteria: new‐onset pulmonary infiltrate associated with possible nosocomial pneumonia. Modified Clinical Pulmonary Infection Score (CPIS) < 7 on Day 1 (suggesting low probability pneumonia)

81 adult medical and surgical ICU patients; 47 (58%) receiving mechanical ventilation; mean age 66.7 years; no data on sex. APACHE III score 41.8; mean CPIS 4.9. Prior mean duration of ICU stay 8.8 days and duration of mechanical ventilation 7.6 days. Chronic obstructive airways disease present in 27% cases. 5% patients were transplant recipients. With the exception of abnormal respiration (92% experimental group versus 71% standard‐therapy group, P = 0.016), there were no significant differences in baseline characteristics between the 2 groups

Exclusions: HIV, chemotherapy‐induced neutropenia, concurrent antibiotic administration (other than surgical prophylaxis), flouroquinolone allergy

Interventions

Randomised at Day 1 of episode of possible pneumonia. Intervention group (N = 39): 3 days' ciprofloxacin monotherapy. At Day 3 if CPIS < 7, antibiotics would be discontinued; if CPIS > 6, therapy would be continued, with choice of agent and duration of therapy at the discretion of treating physician, and incorporating microbiology results

Standard therapy (N = 42): choice and duration of antibiotic therapy at choice of treating physician

Outcomes

The following outcomes were reported:

• Mortality (Day 3, 14, 30)

• Duration of ITU stay

• Emergence of antimicrobial resistance or superinfection (Day 28)

• Duration of antibiotic therapy

• Antimicrobial therapy cost

Notes

Pathogens associated with possible HAP were incompletely presented

For patients allocated to the short‐course therapy, 0 patients with CPIS < 7 at Day 3 had antibiotics continued beyond 72 hours

A significant decrease in duration of therapy in "Standard therapy" group was observed with time (P = 0.0001), thought to be a result of unblinded nature of study. The study was terminated by institutional review board as it was deemed "unethical to continue study."

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Unclear risk

Randomisation process not described

Allocation concealment (selection bias)

Unclear risk

Randomisation/concealment not described

Blinding (performance bias and detection bias)
All outcomes

High risk

This was an unblinded study

Incomplete outcome data (attrition bias)
All outcomes

Low risk

No missing outcome data

Selective reporting (reporting bias)

High risk

The components of the composite measure of "emergence of antimicrobial resistance or superinfection" are not reported

Other bias

High risk

Stopped early by institutional review board. Results for patients managed in "experimental" group appear to have influenced management of patients subsequently allocated to "standard therapy group"

Methods

The ProVAP study, an international, multi‐centre study involving 7 ICUs in USA and Switzerland. Unblinded, randomised controlled trial comparing strategy utilising serum procalcitonin measurement to guide discontinuation of antibiotics for ICU patients with VAP, with antibiotic administration according to an agreed guideline

Participants

101 patients with VAP; 36% NF‐GNB; 10% MRSA; mean age 56; 75% male; 27% surgical patients; 19% with chronic obstructive airways disease. Mean SAPS II score 43; ODIN score 2.1; SOFA score 7.8. Mean duration of mechanical ventilation before onset of VAP 6 days. 75% received antibiotics during the 14 days before VAP. 89% patients had CPIS ≥ 6.ODIN score was slightly higher in the control group than the procalcitonin group (2.3 versus 1.9; P = 0.042); otherwise, there were no significant differences in baseline characteristics. VAP diagnosed according to clinical criteria: new or persistent infiltrates on chest radiography with at least 2 of purulent tracheal secretions, temperature > 38 ºC, leukocyte count > 11,000/ µL or < 3000/ µL; microbiological confirmation was not an essential criterion, and indeed was absent in 27% patients. Exclusions: treatment with immunosuppressants or long‐term corticosteroid treatment; or underlying immunosuppressant disorder; co‐existing extrapulmonary infection

Interventions

101 patients were randomised on day of diagnosis of VAP (Day 0). Serum procalcitonin levels were measured from Day 0 to Day 10 for all patients (including patients in the control group). For the 51 patients in the intervention group, discontinuation of antibiotic therapy was encouraged according to an algorithm 72 hours (Day 2) after randomisation if PCT was < 0.5 µg/L or had decreased to 20% or less of level on Day 0

For the 50 patients randomised to the control group, procalcitonin levels were withheld, and duration of therapy was determined by treating physician

In both intervention and control groups, choice of antibiotic therapy was that of the attending physician

Outcomes

The following outcome measures were reported:

• Mortality (Day 28)

• In‐hospital mortality

• Duration of antibiotic therapy

• Antibiotic‐free days (Day 28)

• Mechanical ventilation‐free days (Day 28)

• Duration of hospital stay

• Number of ITU‐free days alive (Day 28)

• Clinical features associated with respiratory infection

• PaO₂/ FiO₂

• SOFA score

• ODIN score

• CPIS

Notes

It is unclear how many patients were treated with antibiotics beyond the point at which discontinuation was advocated by the protocol

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Unclear risk

"Arbitrary allocation to... treatment assignments based on sealed, opaque envelopes."

Allocation concealment (selection bias)

Low risk

"Treating physicians were not aware of envelope contents before randomisation."

Blinding (performance bias and detection bias)
All outcomes

High risk

Not blinded

Incomplete outcome data (attrition bias)
All outcomes

Low risk

No patients lost to follow‐up

Selective reporting (reporting bias)

High risk

Complete data on numbers of patients for whom physicians violate (PCT) protocol not presented

Other bias

Low risk

There appear not to be other major potential sources of bias