Quantitative versus qualitative cultures of respiratory secretions for clinical outcomes in patients with ventilator‐associated pneumonia

Abstract

Background

Ventilator‐associated pneumonia (VAP) is a common infectious disease in intensive care units (ICUs). The best diagnostic approach to resolve this condition remains uncertain.

Objectives

To evaluate whether quantitative cultures of respiratory secretions and invasive strategies are effective in reducing mortality in immunocompetent patients with VAP, compared with qualitative cultures and non‐invasive strategies. We also considered changes in antibiotic use, length of ICU stay and mechanical ventilation.

Search methods

We searched CENTRAL (2014, Issue 9), MEDLINE (1966 to October week 2, 2014), EMBASE (1974 to October 2014) and LILACS (1982 to October 2014).

Selection criteria

Randomised controlled trials (RCTs) comparing respiratory samples processed quantitatively or qualitatively, obtained by invasive or non‐invasive methods from immunocompetent patients with VAP and which analysed the impact of these methods on antibiotic use and mortality rates.

Data collection and analysis

Two review authors independently reviewed the trials identified in the search results and assessed studies for suitability, methodology and quality. We analysed data using Review Manager software. We pooled the included studies to yield the risk ratio (RR) for mortality and antibiotic change with 95% confidence intervals (CI).

Main results

Of the 5064 references identified from the electronic databases (605 from the updated search in October 2014), five RCTs (1367 patients) met the inclusion criteria. Three studies compared invasive methods using quantitative cultures versus non‐invasive methods using qualitative cultures, and we used them to answer the main objective of this review. The other two studies compared invasive versus non‐invasive methods, both using quantitative cultures. We combined all five studies to compare invasive versus non‐invasive interventions for diagnosing VAP. The studies that compared quantitative and qualitative cultures (1240 patients) showed no statistically significant differences in mortality rates (RR 0.91; 95% CI 0.75 to 1.11). The analysis of all five RCTs showed there was no evidence of reduction in mortality in the invasive group versus the non‐invasive group (RR 0.93; 95% CI 0.78 to 1.11). There were no significant differences between the interventions with respect to the number of days on mechanical ventilation, length of ICU stay or antibiotic change.

Authors' conclusions

There is no evidence that the use of quantitative cultures of respiratory secretions results in reduced mortality, reduced time in ICU and on mechanical ventilation, or higher rates of antibiotic change when compared to qualitative cultures in patients with VAP. We observed similar results when invasive strategies were compared with non‐invasive strategies.

Plain language summary

Quantitative versus qualitative cultures of respiratory secretions in patients with ventilator‐associated pneumonia

Review question 
 We wanted to determine whether quantitative cultures of respiratory secretions are effective in reducing mortality in immunocompetent patients with ventilator‐associated pneumonia compared with qualitative cultures. We also evaluated changes in antibiotic use, length of intensive care unit (ICU) stay and mechanical ventilation.
 
 Background 
 Ventilator‐associated pneumonia (VAP) is a condition which occurs in patients mechanically ventilated for more than 48 hours, which can significantly increase the mortality of ICU patients. The best method for diagnosing VAP and identifying the causative organism (bacteria) is uncertain. Both invasive and non‐invasive techniques are used to obtain samples of respiratory secretions and these can be analysed quantitatively (with a threshold count of the bacterial growth to differentiate between infection and colonisation of the lower airways) or qualitatively (presence or absence of pathogenic germs in the culture). The rationale for using quantitative cultures of respiratory secretions sampled from patients with VAP is to differentiate the infectious organisms (those with a higher concentration) from colonising organisms (those with lower concentration), thereby optimising antibiotic therapy.

Study characteristics 
 After reviewing 5064 articles we found three randomised controlled trials (RCTs) (1240 participants) comparing invasive methods using quantitative cultures, versus the non‐invasive method using qualitative cultures. Two additional RCTs (127 participants) compared invasive versus non‐invasive methods, both using quantitative cultures. We combined all five RCTs (1367 participants) to compare invasive versus non‐invasive interventions for diagnosing VAP.

Key results 
 The cumulative all‐cause mortality was 25.4% (159/626) in the qualitative group and 23.1% (142/614) in the quantitative group over the duration of the trials. There were no statistically significant differences between the use of quantitative cultures versus qualitative (risk ratio (RR) 0.91; 95% confidence interval (CI) 0.75 to 1.11). When we analysed all five studies, a total of 1367 patients were included. The cumulative all‐cause mortality was 26.6% (184/692) in the non‐invasive group and 24.7% (167/675) in the invasive group over the duration of the trials. The invasive versus non‐invasive intervention analysis showed no evidence of mortality reduction (RR 0.93; 95% CI 0.78 to 1.11). The pooled data from trials did not show a significant influence on antibiotic change, but there was significant heterogeneity amongst the studies and the publication bias analysis for the antibiotic change analysis suggests that significant publication bias is likely (Egger's regression (intercept: 1.909; standard error: 0.436; P value (two‐sided): 0.048). The analysis did not show significant differences in days on mechanical ventilation and in the length of ICU stay between either the quantitative versus qualitative culture groups or the invasive versus non‐invasive method groups.

Quality of the evidence 
 The body of evidence supports moderately robust conclusions regarding the objective of our review. We use the word moderately because the sample size was of moderate size, even though this was the largest sample size evaluated to date: five RCTs with a total of 1367 participants. The results were consistent with respect to the mortality outcome, days on ventilation and days in the ICU. However, the results were less consistent with respect to antibiotic change.

Conclusion 
 Evidence from trials included in this review indicates that there is no clinical advantage in the use of quantitative over qualitative cultures, nor in using invasive over non‐invasive diagnostic approaches. The evidence is current to October 2014.

Background

Description of the condition

Ventilator‐associated pneumonia (VAP) is a common infectious disease in intensive care units (ICUs), occurring in 10% to 20% of patients on mechanical ventilation for more than 48 hours. VAP accounts for approximately 90% of all nosocomial pneumonia in ICU patients (ATS/IDSA 2005). The mortality rate of patients with VAP is twice as high as that of patients without VAP (Safdar 2005), and the crude mortality rate ranges from 24% to 57% (Chastre 2002). Many of these patients succumb to an underlying disease rather than to VAP itself. However, it is believed that VAP‐attributable mortality ranges from 15% to 47% (Chastre 2002). VAP significantly increases costs and the length of hospital stay by an average of six days (Safdar 2005).

VAP represents a great challenge to clinical practice and has triggered numerous discussions regarding the best diagnostic approach (Grossman 2000; Shorr 2005). Isolated non‐invasive techniques are sensitive but not specific. Invasive methods used to obtain lower respiratory secretions are theoretically more specific for diagnosis and can potentially lead to a more appropriate use of antibiotics (Chastre 2002). For this reason, various invasive methods are used with the aim of identifying the causative organism and trying to prevent the antibiotic treatment of colonising or contaminant organisms that are not causing the disease. However, there is no evidence or consensus regarding the superiority of one specific invasive technique over another (Grossman 2000). In a recent meta‐analysis, which evaluated the existing randomised and observational clinical studies that have examined the use of invasive strategies and outcomes for VAP, the authors found no changes in mortality (Shorr 2005).

Description of the intervention

Quantitative cultures determining a threshold limit in the bacteria growth are recommended to differentiate infectious organisms (associated with a higher concentration of pathogenic bacteria) from colonising organisms (associated with lower bacterial concentrations), present due to the contamination with bacteria from the oropharynx that occurs in the collection of all respiratory samples (Chastre 2002). Nevertheless, there is no clear evidence or expert consensus to determine if quantitative cultures (that report the number of bacterial colonies) are associated with better clinical outcomes than qualitative ones (that report the presence or absence of pathogenic bacteria) (Grossman 2000). There are no systematic reviews or adequately powered large, controlled, prospective studies that compare the impact of quantitative versus qualitative cultures, independent of the method used to collect respiratory secretions (invasive or non‐invasive), on clinical outcomes.

How the intervention might work

The use of invasive and non‐invasive techniques with or without quantitative cultures of specimens helps to differentiate between colonisation and infection. When culture results are available from these techniques, they allow the precise identification of the offending organisms and their susceptibility patterns. Such data are thought to be valuable for optimal antibiotic selection, improving identification of patients with true VAP and helping decisions on whether or not to treat.

Why it is important to do this review

The assessment of clinical outcomes is the most appropriate way to evaluate the best management of patients with VAP. Studies using other diagnostic criteria as the gold standard to confirm the efficacy of a diagnostic method, such as autopsies for histological confirmation and identification of the causative pathogen, would introduce bias because the non‐surviving patients differ from those who survive.

It is also important to evaluate whether there is any difference in mortality and secondary outcomes such as antibiotic change, ICU stay and duration of mechanical ventilation, when quantitative cultures are used following both invasive and non‐invasive methods of obtaining samples. This Cochrane review examines the evidence to determine how the use of quantitative or qualitative cultures, as well as invasive or non‐invasive techniques, may influence clinical outcomes in patients with VAP (Porzecanski 2006; Shorr 2005; Torres 2004).

Objectives

Primary objectives

To evaluate whether quantitative cultures of respiratory secretions are effective in reducing mortality in immunocompetent patients with VAP, compared with qualitative cultures. We also considered changes in antibiotic use, length of ICU stay and mechanical ventilation.

Secondary objective

To evaluate whether invasive or non‐invasive strategies are effective in influencing the same outcomes, independent of the culture type used.

Methods

Criteria for considering studies for this review

Types of studies

Randomised controlled trials (RCTs) comparing respiratory samples processed quantitatively or qualitatively and obtained by invasive or non‐invasive methods from patients with VAP, where the impact on antibiotic use and mortality rates have been analysed. We excluded studies investigating severe community‐acquired pneumonia.

Types of participants

Immunocompetent patients, 18 years or older, on mechanical ventilation for more than 48 hours with suspected VAP (new or progressive radiological infiltration, fever, tracheobronchial purulent secretion, leucopenia or leucocytosis).

Types of interventions

Invasive and non‐invasive methods using quantitative or qualitative cultures of respiratory secretions from patients with suspected VAP. Invasive methods are those that obtain samples from the lower respiratory tract with the benefit of bypassing the upper airways, which are usually colonised by many micro‐organisms. These invasive methods include bronchoalveolar lavage (BAL), protected specimen brush (PSB) obtained by fibreoptic bronchoscopy, and non‐bronchoscopic techniques (blinded) such as mini BAL, blinded bronchial sampling (BBS) and blinded sampling protected brushes (BPSB). The non‐invasive method used to obtain respiratory samples from patients on mechanical ventilation is endotracheal aspirate (ETA) (Ioanas 2001). Quantitative cultures of respiratory samples are those that determine a threshold count to differentiate between infection and colonisation of the lower airways.

Types of outcome measures

Primary outcomes

1. Twenty‐eight day mortality.

Secondary outcomes

1. Change of antibiotics (any change in the patients' antibiotic prescription based on the results of the diagnostic strategy used).

2. Number of days on mechanical ventilation.

3. Length of ICU stay.

Search methods for identification of studies

Electronic searches

For this update we searched the Cochrane Central Register of Controlled Trials (CENTRAL) (2014, Issue 9) (accessed 21 October 2014), MEDLINE (June 2011 to October week 2, 2014), EMBASE (June 2011 to October 2014) and LILACS (June 2011 to October 2014). Please see Appendix 1 for details of earlier searches.

We used the search strategy described in Appendix 2 to search MEDLINE and CENTRAL. We combined the MEDLINE search with the Cochrane Highly Sensitive Search Strategy for identifying randomised trials in MEDLINE: sensitivity‐ and precision‐maximising version (2008 revision); Ovid format (Lefebvre 2011). We adapted the search strategy to search EMBASE (see Appendix 3) and LILACS (see Appendix 4). There were no language or publication restrictions.

Searching other resources

We searched WHO ICTRP and ClinicalTrials.gov for completed and ongoing trials (17 February 2014). We reviewed bibliographies of all included studies for additional references.

Data collection and analysis

The overall risk of bias is presented graphically in Figure 1 and summarised in Figure 2.

1.

'Risk of bias' graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

2.

'Risk of bias' summary: review authors' judgements about each risk of bias item for each included study.

Selection of studies

Two review authors (DCB, PJZT) independently reviewed and selected potential trials from the search results. We assessed the studies for suitability, methodology and quality. There were no disagreements between these review authors on study selection.

Data extraction and management

We extracted data using a standardised form. Information extracted included:

• age and gender of participants;

• number of participants;

• criteria for suspected diagnosis of VAP (usually defined as a combination of findings: new or progressive radiographic infiltrate; new fever; leukocytosis or leucopenia; purulent sputum; decline in oxygenation);

• type of intervention (invasive versus non‐invasive; quantitative versus qualitative cultures);

• randomisation method;

• disease severity score (i.e. APACHE II);

• previous use of antibiotic;

• appropriateness of initial antibiotic choice (Yes or No);

• duration of mechanical ventilation before the diagnosis of VAP;

• mortality at 28 days;

• antibiotic change (Yes or No);

• ICU stay (days);

• number of days of mechanical ventilation; and

• most common isolated pathogens.

Assessment of risk of bias in included studies

Two review authors (DCB, PJZT) independently assessed the methodological quality of included studies. A third review author (AK) resolved any potential disagreements. We used the Cochrane Collaboration's 'Risk of bias' tool for assessing trial quality (Higgins 2011). In this tool, the risk of bias is evaluated according seven domains: sequence generation; allocation concealment; blinding of participants and personnel; blinding of outcome assessment; incomplete outcome data; selective reporting; and other sources of bias.

Measures of treatment effect

We presented all point estimates as risk ratios (RR). We used forest plots to display results. We combined trials using Review Manager software (RevMan 2014) and analyses were by intention‐to‐treat (ITT). We calculated a fixed‐effect risk ratio (RR) with 95% confidence intervals (CI) for individual studies for dichotomous variables. We calculated the mean difference (MD) with 95% CI for continuous variables. When the trial authors reported standard deviations (SD), we used them directly. In the CCCTG 2006 study, the SDs were not available for these variables, so we analysed them by transforming 95% CI into SD with the formula: SE = (Upper‐Lower limit of 95% CI)/(1.96*2), SD = SE* sqrt (N).

Unit of analysis issues

We only included studies with a randomised, parallel design, therefore participants should be individually randomised to one of two intervention groups (invasive versus non‐invasive; or quantitative versus qualitative culture). We collected and analysed a single measurement for each outcome from each participant.

Dealing with missing data

If we considered missing data to be 'not at random', we would conduct sensitivity analyses to assess how sensitive results were considering the inclusion or not of studies with missing data.

Assessment of heterogeneity

We quantified inconsistency among the pooled estimates by using the I² statistic [(Q ‐ df)/Q] x 100% test, where Q is the Chi² test and df its degrees of freedom. This illustrates the percentage of the variability in effect estimates resulting from heterogeneity rather than sampling error (Higgins 2011). If heterogeneity was found, we used a random‐effects model.

Assessment of reporting biases

We evaluated publication bias analysis by using the Egger's regression method. This method is equivalent to a weighted regression of treatment effect on its standard error, with weights inversely proportional to the variance of the effect size (Sterne 2001). We used the Comprehensive Meta‐Analysis version 2.0 (Biostat, Englewood, New Jersey) software.

Data synthesis

We planned to perform a meta‐analysis if studies performed the same interventions (invasive or non‐invasive; quantitative or qualitative cultures) and reported the same clinical outcomes (mainly survival and antibiotic change). In addition, if studies were suspected to be at risk of bias, or serious publication or reporting biases (or both) were likely, we would not undertake a meta‐analysis and we would describe a systematic synthesis of the studies' findings.

Subgroup analysis and investigation of heterogeneity

We did not plan to perform a subgroups analysis.

Sensitivity analysis

We had planned an additional sensitivity analysis if the eligibility/quality of some studies in the meta‐analysis were dubious: first, including all studies and second, only including those that are definitely known to be eligible. Fortunately, we considered all identified studies to be of the same good quality. We performed sensitivity analyses comparing random‐effects and fixed‐effect models.

Results

Description of studies

Results of the search

We updated our searches in October 2014 and retrieved a total of 605 new records after duplicates were removed from this current search as well as when compared with the search results of the last update (Berton 2012). No new trials were included or excluded in this 2014 update. The updated electronic database searches identified a total of 4903 articles (3931 from the first publication of this review (Berton 2008); 528 from the second publication (Berton 2012); and 605 from this current search update, i.e. October 2014). The review of titles yielded 76 articles that we examined further. Of these, a review of the abstracts, and whenever required, the full text, yielded six articles that appeared to fulfil the inclusion criteria. Five of these met the inclusion criteria and were included in the analysis. We excluded the sixth trial because it had a cross‐over design and was not randomised (Cai 2001).

We identified no additional RCTs following handsearching of the reference lists of identified included studies. All included studies were published in peer‐reviewed literature and their characteristics are described in the Characteristics of included studies table.

Included studies

Three of the five included studies compared invasive methods using quantitative cultures versus the non‐invasive method using qualitative cultures (CCCTG 2006; Fagon 2000; Solé Violán 2000). The other two included studies compared invasive versus non‐invasive methods both using quantitative cultures (Ruiz 2000; Sanchez‐Nieto 1998). We combined them with the three previous studies to compare invasive versus non‐invasive interventions.

Excluded studies

We excluded one study because it was not randomised and had a cross‐over design (Cai 2001).

Risk of bias in included studies

Overall, the risk of bias of the included studies was low.

Allocation

All included studies described the use of computer‐generated randomisation tables of variable and undisclosed block size, providing, in this way, an adequate generation sequence. One study reported the utilisation of a central telephone system (CCCTG 2006), and the other studies, a computer‐generated random number table (Fagon 2000; Ruiz 2000; Sanchez‐Nieto 1998), indicating an adequate concealment allocation.

Blinding

Two studies stated that investigators were aware of the study intervention (CCCTG 2006; Fagon 2000). The other three did not address this issue, but are probably not blinded. However, the review authors judged that the main outcomes (mortality and antibiotic change) are not likely to be influenced by lack of blinding.

Incomplete outcome data

All studies had low incomplete outcome data with almost 100% adequate follow‐up.

Selective reporting

The methodological section of the CCCTG 2006 and Fagon 2000 studies pre‐specified (primary and secondary) outcomes that were of interest in the review and they reported them in the pre‐specified way. The other included studies did not give sufficient information to permit judgement.

Other potential sources of bias

The included studies appear to be free of other sources of bias.

Effects of interventions

Primary outcome

1. 28‐day mortality

The three RCTs that compared quantitative cultures versus qualitative cultures consisted of a total of 1240 patients, including 614 in the invasive quantitative arm and 626 in the non‐invasive qualitative arm (CCCTG 2006; Fagon 2000; Solé Violán 2000).

The cumulative all‐cause mortality was 25.4% (159/626) in the qualitative group and 23.1% (142/614) in the quantitative group over the duration of the trials. There were no statistically significant differences between the use of quantitative cultures versus qualitative (risk ratio (RR) 0.91; 95% confidence interval (CI) 0.75 to 1.11; Analysis 1.1; Figure 3). The intervention effect was statistically homogeneous across the studies with respect to mortality (P value = 0.43; I² statistic = 0%).

1.1. Analysis.

Comparison 1 Quantitative versus qualitative culture, Outcome 1 Mortality.

3.

Forest plot of comparison: 1 Quantitative versus qualitative culture, outcome: 1.1 Mortality.

When we analysed all five studies, a total of 1367 patients were included. The cumulative all‐cause mortality was 26.6% (184/692) in the non‐invasive group and 24.7% (167/675) in the invasive group over the duration of the trials. The invasive versus non‐invasive intervention analysis showed no evidence of mortality reduction (RR 0.93; 95% CI 0.78 to 1.11; Analysis 2.1). The test for heterogeneity among the studies was not significant (P value = 0.32; I² statistic = 14%).

2.1. Analysis.

Comparison 2 Invasive versus non‐invasive method, Outcome 1 Mortality.

The CCCTG 2006 trial excluded patients infected or colonised with Pseudomonas species and methicillin‐resistant Staphylococcus aureus (S. aureus), which are also causes of VAP and associated with a poorer prognosis. Thus, we performed a sensitivity analysis excluding this trial: no difference was found regarding mortality (quantitative versus qualitative: RR 0.82; 95% CI 0.64 to 1.06; invasive versus non‐invasive: RR 0.88; 95% CI 0.71 to 1.09).

Only two studies compared invasive versus non‐invasive interventions, both using quantitative culture and including a total of 127 patients (Ruiz 2000; Sanchez‐Nieto 1998). The combined effect of these trials did not show changes in mortality (RR 1.14; 95% CI 0.54 to 2.41; Analysis 3.1).

3.1. Analysis.

Comparison 3 Invasive quantitative versus non‐invasive quantitative, Outcome 1 Mortality.

The publication bias analysis for the mortality analysis shows the following: Egger's regression (intercept: 0.917; standard error: 1.829; P value (two‐sided): 0.704), which suggests that significant publication bias is unlikely.

Secondary outcomes

1. Change of antibiotics

This endpoint could not be evaluated in the Fagon study because they reported antibiotic‐free days instead of the percentage or absolute value of antibiotic change (Fagon 2000). Antibiotic‐free days means the number of days without antibiotics at day 28 after randomisation, and the number of days for which the patient is alive and not receiving antibiotics. The rate of antibiotic change was evaluated by the inclusion of two trials (CCCTG 2006; Solé Violán 2000). The pooled data from these two trials did not show a significant influence on antibiotic change (RR 1.53; 95% CI 0.54 to 4.39; Analysis 1.2; Figure 4). There was significant heterogeneity among the studies (P value = 0.02, I² statistic = 81%).

1.2. Analysis.

Comparison 1 Quantitative versus qualitative culture, Outcome 2 Antibiotic change.

4.

Forest plot of comparison: 1 Quantitative versus qualitative culture, outcome: 1.2 Antibiotic change.

We found similar results with the summary data from the four studies that reported antibiotic change (CCCTG 2006; Ruiz 2000; Sanchez‐Nieto 1998; Solé Violán 2000), and compared invasive versus non‐invasive interventions (RR 1.67; 95% CI 0.87 to 3.21; Analysis 2.2). The test for detection of heterogeneity was significant (P value = 0.01, I² statistic = 73%).

2.2. Analysis.

Comparison 2 Invasive versus non‐invasive method, Outcome 2 Antibiotic change.

The publication bias analysis for the antibiotic change analysis showed the following: Egger's regression (intercept: 1.909; standard error: 0.436; P value (two‐sided): 0.048), which suggests that significant publication bias is likely.

2. Number of days on mechanical ventilation

The analysis did not show significant differences in days on mechanical ventilation between either the quantitative versus qualitative culture groups (mean difference (MD) 0.58; 95% CI ‐0.51 to 1.68; Analysis 1.3; Figure 5), or the invasive versus non‐invasive method groups (MD 0.61; 95% CI ‐0.47 to 1.68; Analysis 2.3). For this analysis we did not include the data from Fagon 2000, because this information was given as mechanical ventilation‐free days (as explained for antibiotic change). However, no difference was found between these interventions in the Fagon 2000 study (invasive quantitative: 7.8 ± 9.8; non‐invasive qualitative: 7 ± 9.4; P value > 0.2).

1.3. Analysis.

Comparison 1 Quantitative versus qualitative culture, Outcome 3 Duration on mechanical ventilation (days).

5.

Forest plot of comparison: 1 Quantitative versus qualitative culture, outcome: 1.3 Duration on mechanical ventilation (days).

2.3. Analysis.

Comparison 2 Invasive versus non‐invasive method, Outcome 3 Duration on mechanical ventilation (days).

3. Length of intensive care unit (ICU) stay

The length of ICU stay was the same for the quantitative versus qualitative comparison (MD 0.95; 95% CI ‐0.14 to 2.04; Analysis 1.4; Figure 6) and invasive versus non‐invasive analysis (MD 0.94; 95% CI ‐0.13 to 2.01; Analysis 2.4).

1.4. Analysis.

Comparison 1 Quantitative versus qualitative culture, Outcome 4 ICU stay (days).

6.

Forest plot of comparison: 1 Quantitative versus qualitative culture, outcome: 1.4 ICU stay (days).

2.4. Analysis.

Comparison 2 Invasive versus non‐invasive method, Outcome 4 ICU stay (days).

Discussion

Although the current management strategy for patients with suspected ventilator‐associated pneumonia (VAP) recommends obtaining a lower respiratory tract sample for quantitative or semi‐quantitative cultures (ATS/IDSA 2005), this systematic review of the best available evidence did not show that quantitative cultures were associated with a significant reduction in mortality in patients with VAP (Analysis 1.1). Similarly, we found no significant reduction in mortality when invasive methods were compared to non‐invasive methods (Analysis 2.1).

Interestingly, if we consider the lowest mortality rate among the included studies (18% ‐ CCCTG 2006), the pooled sample of included trials gave us statistical power of more than 90% with a significance level of 0.1% to detect an absolute mortality reduction of 9% (a similar mortality reduction observed at day 14 in the study by Fagon 2000; the only study that demonstrated improvement in mortality). However, to confirm the risk ratio reduction of 9% (or absolute risk reduction of 1.6%) suggested by this meta‐analysis estimate, it would be necessary to conduct a trial with more than 7500 patients in each arm in order to achieve a statistical power of 80% and a significance level of 5%. It remains to be seen whether the risks and costs of invasive methods are worth taking in the context of a potential 1.6% reduction in mortality.

The main advantage of using quantitative cultures is that fewer patients may be treated with unnecessary antibiotics, because it is possible to differentiate colonising pathogens from infecting ones (ATS/IDSA 2005). Another possible advantage is that quantitative methods allow antibiotics to be used with more precision, therefore improving appropriate antibiotic use and potentially decreasing mortality. However, the results of Luna 1997 suggest that subsequent changes of inadequate antimicrobial therapy following culture results did not reduce the greater risk of hospital mortality, although the use of adequate empirical treatment from the beginning did. One of the major challenges associated with VAP, therefore, is how to choose the most adequate antibiotic from the beginning of treatment, since multiple‐drug resistant bacteria, polymicrobial cultures and late‐onset VAP are all recognised risk factors for inadequate initial treatment (Teixeira 2007). The impact of these strategies on bacterial resistance would be an important analysis, but unfortunately most studies did not provide enough information to apply the meta‐analytic methodology.

The pooled data from trials that compared invasive quantitative versus non‐invasive qualitative strategies did not show significant changes in antibiotic prescribing (Analysis 1.2). Similar results were observed when invasive versus non‐invasive methods were analysed (Analysis 2.2). Unlike our mortality outcome analysis, the antibiotic change analyses showed statistically significant publication bias (P value = 0.048). An important issue in the three studies, where only a small number of patients were included is that antibiotics were continued in patients with negative cultures (Ruiz 2000; Sanchez‐Nieto 1998; Solé Violán 2000). The other two studies allowed therapy to be discontinued when the culture results were negative, but antimicrobial therapy could be continued based on clinical judgement (CCCTG 2006; Fagon 2000). In the study conducted by Fagon 2000, the outcomes of those patients who had their antibiotics stopped were not reported, and in the Canadian trial (CCCTG 2006), a significant discontinuation or modification of antibiotics occurred in the invasive quantitative arm only for those patients with low or moderate pre‐test likelihood of pneumonia.

The studies by Fagon 2000, Sanchez‐Nieto 1998 and Solé Violán 2000 showed a higher rate of antibiotic change in the invasive method group without mortality reduction. However, the Canadian trial, which included the largest number of patients, showed a high rate of antibiotic change without a difference between the study arms. This could be partially explained by the fact that the research team from this trial monitored the patients and reminded the clinical personnel to adjust antimicrobial therapy as soon as possible, according to culture results. This could result in a more rational antibiotic use and should be considered for future studies.

The lower mortality rate (around 20%) in the studies by CCCTG 2006 and Solé Violán 2000 probably reflects the higher rates of initial appropriate antibiotic therapy (Alvarez‐Lerma 1996; Iregui 2002; Luna 1997). The study by Fagon 2000 had similar rates of initial appropriate antibiotic selection compared to the CCCTG 2006 and Solé Violán 2000 trials, but had a greater mortality rate (see 'Characteristics of included studies' table). This may reflect the significantly higher rate of appropriate initial antibiotic in the non‐invasive/qualitative arms in the two later studies (CCCTG 2006; Solé Violán 2000). The empirical antimicrobial therapy based on the results of direct microscopic examination of respiratory secretions (Fagon 2000), or the empirical broad spectrum antibiotic therapy (CCCTG 2006), guided the initial appropriate treatment choices in a high percentage of cases (above 90%). However, using a strategy similar to Fagon 2000, Ruiz 2000 found only a 73% rate of appropriate initial antibiotic prescribing. It is possible that differences in mortality could reflect the variability of multiple‐drug resistant pathogens, which were lower in the CCCTG 2006 and Solé Violán 2000 studies.

Initial antimicrobial therapy based on local patterns of micro‐organism prevalence and antibiotic susceptibility was used to elaborate empirical strategies with a higher appropriateness rate to treat patients with VAP. This could be further improved by avoiding antibiotics that patients have recently received (ATS/IDSA 2005). Information about micro‐organisms and their antibiotic sensitivities suggests that empirical antibiotic therapy would be appropriate (Ibrahim 2002). Based on these findings, it seems reasonable to assume that the knowledge of local micro‐organism patterns would be more important than the methods used to obtain and cultivate respiratory secretions in reducing the morbidity and mortality of patients with VAP. In fact, results from the CCCTG 2006 trial strongly suggest that in the setting of appropriate initial antibiotic, the method of respiratory culture collection may not affect survival outcomes. Larger randomised prospective trials are required to better answer these questions.

Days on mechanical ventilation and length of intensive care unit (ICU) stay were the same between the intervention groups in the included trials (see 'Characteristics of included studies' table) and in the pooled data. The duration of these outcomes was shorter in the CCCTG 2006 study and this finding should not be attributed to initial antibiotic choice (Dupont 2001), as other studies had the same rate of appropriate antibiotic selection (Fagon 2000; Solé Violán 2000). A possible explanation for patients spending fewer days on mechanical ventilation and in ICU could be related to 'standardisation' of key aspects of the protocol, such as adjustment of antibiotic therapy and discontinuation of mechanical ventilation. This protocol standardisation deserves further evaluation in future trials.

Summary of main results

The studies that compared quantitative and qualitative cultures showed no statistically significant differences in mortality rates. The analysis of all five randomised controlled trials (RCTs) showed there was no evidence of mortality reduction in the invasive group versus the non‐invasive group. There were no significant differences between the interventions with respect to the number of days on mechanical ventilation, length of ICU stay or antibiotic change.

Overall completeness and applicability of evidence

All identified and evaluated studies that were included in our systematic review and meta‐analysis provided sufficient evidence to address the objectives of this review. The relevant participants, interventions and outcomes were all fully investigated by our analysis. The results of our review provide support to optimise the current practice for patients with VAP, i.e. quantitative and qualitative cultures of respiratory secretions are associated with similar morbidity or mortality outcomes in patients with VAP.

Quality of the evidence

The body of evidence supports moderately robust conclusions regarding the objective of our review. We use the word moderately because the sample size was of moderate size, even though this was the largest sample size evaluated to date: five randomised clinical trials with a total of 1240 participants. The results were consistent with respect to the mortality outcome, days on ventilation and days in the ICU. However, the results were less consistent with respect to antibiotic change. The overall results of this review led to a robust internal validity.

Potential biases in the review process

The likelihood that all relevant studies were identified is very high based on our unlimited, exhaustive and replicated literature search. However, due to the number of included randomised trials, publication bias cannot be totally excluded.

Agreements and disagreements with other studies or reviews

No other published systematic review has addressed our objectives with a similar methodological approach. The only other systematic review that has addressed the use of invasive diagnostic methods in patients with VAP was performed by Shorr et al (Shorr 2005). They analysed a similar set of randomised trials plus five observational studies but they did not include the most recent and largest randomised trial on VAP (CCCTG 2006). Similar to our conclusions, they did not find mortality differences between invasive and non‐invasive diagnostic methods. However, different from our conclusion, they found that the invasive approach led to more changes in the antibiotic regimen; this was likely due to several reasons: 1) our antibiotic change analysis (which showed no significant differences between invasive and non‐invasive techniques) was based on randomised studies only, while Shorr et al used additional data from non‐randomised observational studies; 2) our analysis used risk ratios (which are the most appropriate outcome measure to be used in meta‐analysis of randomised studies), while Shorr et al used odds ratio. It is well known that odds ratio leads to an over‐estimation of outcome measures when the rate of the outcome being measured is high (> 10% to 15%); because the rate of outcomes in the antibiotic change analysis was very high (up to 42%), the use of odds ratio in Shorr et al likely produced spuriously inflated estimates; and 3) our review included the largest VAP study (CCCTG 2006) to date, which provided a substantial amount of data on antibiotic change, while Shorr et al did not include this study in their meta‐analysis (it was not published at that time).

Authors' conclusions

Implications for practice.

1. There is no evidence that the use of quantitative cultures of respiratory secretions results in lower mortality compared to qualitative cultures in patients with ventilator‐associated pneumonia (VAP). We observed similar results when invasive and non‐invasive strategies were compared.

2. There is no evidence that the use of quantitative cultures of respiratory secretions results in a higher rate of antibiotic change, or a reduction in intensive care unit (ICU) stay and length of time on mechanical ventilation, compared to qualitative cultures in patients with VAP.

Implications for research.

1. It is unlikely that a multicentre randomised controlled trial with a sample size of fewer than 15,000 patients would settle the question of quantitative versus qualitative cultures because of the potential absolute mortality reduction of 1% to 2% at the most (based on our meta‐analysis and on the most recent, largest clinical trial (CCCTG 2006)). It is unlikely that the risks and costs of invasive methods are worth taking in the context of a 1% to 2% absolute reduction in mortality.

2. We suggest that future VAP studies consider evaluating the best strategy based on specific and local bacteriological patterns to guide empirical selection of appropriate antibiotic therapy.

What's new

Date	Event	Description
21 October 2014	New search has been performed	Searches updated but we did not identify any new trials for inclusion or exclusion.
17 February 2014	New citation required but conclusions have not changed	Our conclusions remain unchanged.

History

Protocol first published: Issue 2, 2007
 Review first published: Issue 4, 2008

Date	Event	Description
23 June 2011	New citation required but conclusions have not changed	Our conclusions remain unchanged.
23 June 2011	New search has been performed	Searches were updated but no new trials were included in this update.
18 February 2008	Amended	Converted to new review format.

Appendices

Appendix 1. Previous searches

For the last update we searched the Cochrane Central Register of Controlled Trials (CENTRAL) (2011, Issue 2), (accessed 23 June 2011), MEDLINE (December 2007 to June Week 4, 2011), EMBASE (December 2007 to June 2011) and LILACS (2007 to June 2011).

Previously we searched the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2007, Issue 4), which contains the Acute Respiratory Infections Group's Specialised Register; MEDLINE (1966 to December 2007); EMBASE (1974 to December 2007); and LILACS (1982 to December 2007).

We used the following MeSH and free‐text terms to search MEDLINE and CENTRAL. The MEDLINE terms were combined with the Cochrane highly sensitive search strategy phases one and two as published in Appendix 5b of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2005). These terms were modified to search EMBASE and LILACS as required.

MEDLINE (PubMed)

#1 Pneumonia OR pneumonias OR pneumoniae
 #2 Lung Inflammation OR Pulmonary Inflammation
 #3 Pneumonitis OR pneumonitides
 #4 Respiratory Tract Infection OR Respiratory Infection
 #5 Tracheobronchitis OR Tracheobronchial
 #6 OR/1‐5
 #7 Mechanical Ventilator OR Mechanical Ventilation OR Pulmonary Ventilator
 #8 Respirator* OR Ventilator* OR Ventilation
 #9 Artificial Respiration
 #10 OR/7‐9
 #11 Culture Media
 #12 Microbiology OR Bacteriology
 #13 Bronchoscopy OR Bronchoscop* OR Bronchoalveolar Lavage OR Bronchioalveolar Lavage OR Bronchopulmonary Lavage OR Lung Lavage OR Bronchial Lavage OR Alveolar Lavage
 #14 Endotracheal aspirate OR Tracheal aspirate
 #15 Protected brush
 #16 Invasive OR Noninvasive
 #17 Qualitative OR Quantitative
 #18 OR/11‐17
 #19 #6 AND #10 AND #18

Appendix 2. MEDLINE (Ovid) search strategy

MEDLINE (Ovid)

1 exp Pneumonia/
 2 pneumon*.tw.
 3 ((lung or pulmonary) adj3 (inflam* or infect*)).tw.
 4 exp Respiratory Tract Infections/
 5 (respiratory adj3 infection*).tw.
 6 tracheobronch*.tw.
 7 or/1‐6
 8 exp Ventilators, Mechanical/
 9 Respiration, Artificial/
 10 (respirat* or ventilat*).tw.
 11 or/8‐10
 12 7 and 11
 13 Pneumonia, Ventilator‐Associated/
 14 vap.tw.
 15 or/12‐14 (66503)
 16 exp Culture Media/
 17 culture media.tw.
 18 microbiology/ or bacteriology/
 19 (microbiol* or bacteriol*).tw.
 20 (invasive or non‐invasive or noninvasive or non invasive).tw.
 21 ((qualitative or quantiative) adj3 (culture* or method* or technique* or analys* or assess*)).tw.
 22 Bronchoscopy/
 23 bronchoscop*.tw.
 24 exp Bronchoalveolar Lavage/
 25 (lavage adj2 (bronchoalveolar or bronchioalveolar or bronchial or lung or alveolar or bronchopulmonary)).tw.
 26 BAL.tw.
 27 (aspirat* adj2 (endotracheal or tracheal)).tw.
 28 (protected brush or protected specimen brush or protected‐specimen brush).tw.
 29 (non‐bronchoscopic or nonbronchoscopic or non bronchoscopic).tw.
 30 mini‐BAL.tw.
 31 ((psb or bbs or bpsb) adj5 (sampl* or techniq* or method*)).tw.
 32 or/16‐31
 33 15 and 32

Appendix 3. EMBASE (Elsevier) search strategy

35. #31 AND #34
 34. #32 OR #33
 33. random*:ab,ti OR placebo*:ab,ti OR factorial*:ab,ti OR crossover*:ab,ti OR 'cross‐over':ab,ti OR 'cross over':ab,ti OR volunteer*:ab,ti OR allocat*:ab,ti OR assign*:ab,ti OR ((singl* OR doubl*) NEAR/2 (blind* OR mask*)):ab,ti
 32. 'randomised controlled trial'/exp OR 'single blind procedure'/exp OR 'double blind procedure'/exp OR 'crossover procedure'/exp
 31. #14 AND #30
 30. #15 OR #16 OR #17 OR #18 OR #19 OR #20 OR #21 OR #22 OR #23 OR #24 OR #25 OR #26 OR #27 OR #28 OR #29
 29. ((psb OR bbs OR bpsb) NEAR/5 (sampl* OR techniq* OR method*)):ab,ti
 28. 'mini‐bal':ab,ti
 27. 'non‐bronchoscopic':ab,ti OR 'non bronchoscopic':ab,ti OR nonbronchoscopic:ab,ti
 26. 'protected brush':ab,ti OR 'protected specimen brush':ab,ti OR 'protected‐specimen brush':ab,ti
 25. (aspirat* NEAR/2 (endotracheal OR tracheal)):ab,ti
 24. bal:ab,ti
 23. (lavage NEAR/2 (lung OR bronchoalveolar OR bronchialalveolar OR bronchial OR alveolar OR
 bronchopulmonary)):ab,ti
 22. 'lung lavage'/de
 21. bronchoscop*:ab,ti
 20. 'bronchoscopy'/exp
 19. ((qualitative OR quantitative) NEAR/3 (culture* OR method* OR technique* OR analys*)):ab,ti
 18. invasive:ab,ti OR 'non invasive':ab,ti OR 'non‐invasive':ab,ti OR noninvasive:ab,ti
 17. 'culture media':ab,ti OR 'culture medium':ab,ti OR microbiol*:ab,ti OR bacteriol*:ab,ti
 16. 'microbiology'/de OR 'bacteriology'/de
 15. 'culture medium'/exp
 14. #11 OR #12 OR #13
 13. vap:ab,ti
 12. 'ventilator associated pneumonia'/exp
 11. #7 AND #10
 10. #8 OR #9
 9. respirat*:ab,ti OR ventilat*:ab,ti
 8. 'artificial ventilation'/exp
 7. #1 OR #2 OR #3 OR #4 OR #5 OR #6
 6. tracheobronch*:ab,ti
 5. (respiratory NEAR/3 infection*):ab,ti
 4. 'respiratory tract infection'/exp
 3. ((lung OR pulmonary) NEAR/3 (inflam* OR infect*)):ab,ti
 2. pneumon*:ab,ti
 1. 'pneumonia'/exp

Appendix 4. LILACS (BIREME) search strategy

(((mh:pneumonia OR pneumon* OR neumonía OR mh:c08.381.677* OR mh:c08.730.610* OR pulmonía OR mh:"Respiratory Tract Infections" OR mh:c01.539.739* OR mh:c08.730* OR "Infecciones de las Vías Respiratorias" OR "Infecciones del Aparato Respiratorio" OR "Infecciones del Tracto Respiratorio" OR "Infecciones Respiratorias" OR "Infecções das Vias Respiratórias" OR "Infecções do Aparelho Respiratório" OR "Infecções do Sistema Respiratório" OR "Infecções do Trato Respiratório" OR "Infecciones del Sistema Respiratorio" OR "Infecções Respiratórias" OR tracheobronch*) AND (mh:"Ventilators, Mechanical" OR "Ventiladores Mecánicos" OR "Ventiladores Mecânicos" OR mh:e07.950* OR mh:"Respiration, Artificial" OR "Respiración Artificial" OR "Respiração Artificial" OR "Ventilación Mecánica" OR "Ventilação Mecânica" OR respirat* OR ventilat*)) OR (mh:"Pneumonia, Ventilator‐Associated" OR vap OR "Neumonía del Ventilador" OR "Pneumonia Associada ao Ventilador" OR "Pneumonia Associada ao uso de Ventiladores Pulmonares" OR "Pneumonia Associada ao uso de Ventiladores Artificiais" OR "Pneumonia Associada a Respirador" OR "Pneumonia Associada a Respirador Mecânico")) AND (mh:"Culture Media" OR "Medios de Cultivo" OR "Meios de Cultura" OR mh:d27.720.470.305* OR mh:e07.206* OR "culture media" OR mh:microbiology OR microbiol* OR mh:bacteriology OR bacteriol* OR invasive OR "non‐invasive" OR noninvasive OR "non invasive" OR qualitative OR quantiative OR mh:bronchoscopy OR broncoscop* OR mh:"Bronchoalveolar Lavage" OR "Lavado Broncoalveolar" OR "Lavagem Broncoalveolar" OR "Bronchial Lavage" OR "Lung Lavage" OR mh:e05.927.100* OR "Lavado Bronquial" OR "Lavado Pulmonar" OR "Lavagem Bronquial" OR "Lavagem Pulmonar" OR bal OR "protected brush" OR "protected specimen brush" OR "protected‐specimen brush" OR "non‐bronchoscopic" OR nonbronchoscopic OR "non bronchoscopic" OR "mini‐bal" OR psb OR bbs OR bpsb) AND db:("LILACS") AND type_of_study:("clinical_trials") AND year_cluster:("2011" OR "2012" OR "2013")

Data and analyses

Comparison 1. Quantitative versus qualitative culture.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Mortality	3	1240	Risk Ratio (M‐H, Fixed, 95% CI)	0.91 [0.75, 1.11]
2 Antibiotic change	2	827	Risk Ratio (M‐H, Random, 95% CI)	1.53 [0.54, 4.39]
3 Duration on mechanical ventilation (days)	2	827	Mean Difference (IV, Fixed, 95% CI)	0.58 [‐0.51, 1.68]
4 ICU stay (days)	3	1240	Mean Difference (IV, Fixed, 95% CI)	0.95 [‐0.14, 2.04]

Comparison 2. Invasive versus non‐invasive method.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Mortality	5	1367	Risk Ratio (M‐H, Fixed, 95% CI)	0.93 [0.78, 1.11]
2 Antibiotic change	4	954	Risk Ratio (M‐H, Random, 95% CI)	1.67 [0.87, 3.21]
3 Duration on mechanical ventilation (days)	4	954	Mean Difference (IV, Fixed, 95% CI)	0.61 [‐0.47, 1.68]
4 ICU stay (days)	5	1367	Mean Difference (IV, Fixed, 95% CI)	0.94 [‐0.13, 2.01]

Comparison 3. Invasive quantitative versus non‐invasive quantitative.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Mortality	2	127	Risk Ratio (M‐H, Random, 95% CI)	1.14 [0.54, 2.41]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

CCCTG 2006.

Methods	Duration: 28 days Parallel design Blinded assessment of outcomes: not met Withdrawals described and were they acceptable: met
Participants	Sample size = 739 APACHE II (severity) ‐ Invasive quantitative = 20.1 ± 6.4 ‐ Non‐invasive qualitative = 19.8 ± 6.2 Duration on mechanical ventilation before study (days): not informed Previous use of antibiotics ‐ Invasive quantitative = 227/365 (62.2%) ‐ Non‐invasive qualitative = 241/374 (64.4%) VAP suspected: new or persistent radiographic infiltrate plus any 2 of the following: fever > 38.3 ºC; leukocytosis or neutropenia, purulent endotracheal secretions, increasing oxygen requirement, potentially pathogenic bacteria from endotracheal secretion
Interventions	Invasive quantitative versus non‐invasive qualitative
Outcomes	Mortality ‐ Invasive quantitative = 69/365 (18.9%) ‐ Non‐invasive qualitative = 69/374 (18.4%) Antibiotic change ‐ Invasive quantitative = 271/365 (74.2%) ‐ Non‐invasive qualitative = 279/474 (74.6%) Intensive care unit (ICU) stay (days) ‐ Invasive quantitative = 12.3 (95% CI 10.9 to 13.8) ‐ Non‐invasive qualitative = 12.2 (95% CI 10.9 to 14.2) ‐ Duration of mechanical ventilation (days) ‐ Invasive quantitative: 8.9 (95% CI 7.4 to 10.7) ‐ Non‐invasive: qualitative: 8.8 (95% CI 7.0 to 10.7) Appropriateness of initial antibiotic choice ‐ Invasive quantitative = 341/365 (93.4%) ‐ Non‐invasive qualitative = 353/374 (94.3%)
Notes	―
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Central telephone system, with a variable, undisclosed block size
Allocation concealment (selection bias)	Low risk	As above
Blinding (performance bias and detection bias) All outcomes	High risk	Physicians were aware of the patient's treatment assignments
Incomplete outcome data (attrition bias) All outcomes	Low risk	1 patient withdrew consent 2 days after randomisation
Selective reporting (reporting bias)	Low risk	Primary and secondary outcomes have been reported in the pre‐specified way and adequately reported
Other bias	Low risk

Fagon 2000.

Methods	Duration: 28 days Parallel design Blinded assessment of outcomes: not met Withdrawals described and were they acceptable: met
Participants	Sample size = 413 SAPS (severity) ‐ Invasive quantitative = 44 ± 15 ‐ Non‐invasive qualitative = 42 ± 14 Duration on mechanical ventilation before study (days) ‐ Invasive quantitative = 10.4 ± 10.2 ‐ Non‐invasive qualitative = 10.7 ± 10.0 Previous use of antibiotic ‐ Invasive quantitative = 105/204 (51.5%) ‐ Non‐invasive qualitative = 103/109 (49.3%) VAP suspected: new or progressive radiographic infiltrate plus 1 of the following: fever > 38.3 ºC, leukocytosis or purulent tracheal secretion
Interventions	Invasive quantitative versus non‐invasive qualitative
Outcomes	Mortality ‐ Invasive quantitative = 63/204 (30.9%) ‐ Non‐invasive qualitative = 81/209 (38.8%) Antibiotic change not informed Intensive care unit (ICU) stay (days) ‐ Invasive quantitative = 26.7 ± 23.9 ‐ Non‐invasive qualitative = 25.1 ± 28.5 Duration of mechanical ventilation (days) not informed Appropriateness of initial antibiotic choice ‐ Invasive quantitative = 203/204 (99%) ‐ Non‐invasive qualitative = 185/209 (88.5%)
Notes	―
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated random number tables were used to assign patients in blocks of 8, with stratification according to treatment centre
Allocation concealment (selection bias)	Low risk	As above
Blinding (performance bias and detection bias) All outcomes	High risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	High risk	All patients were followed up
Selective reporting (reporting bias)	Low risk	Primary and secondary outcomes have been reported in the pre‐specified way and adequately reported
Other bias	Low risk	Not detected

Ruiz 2000.

Methods	Duration: 30 days Parallel design Blinded assessment of outcomes: not met Withdrawals described and were they acceptable: met
Participants	Sample size = 76 APACHE II (severity) ‐ Invasive quantitative = 20 ± 6 ‐ Non‐invasive quantitative = 19 ± 6 Duration on mechanical ventilation before study (days) ‐ Invasive quantitative = 6 ± 4 ‐ Non‐invasive quantitative = 6.2 ± 5 Previous use of antibiotic ‐ Invasive quantitative = 26/33 (70%) ‐ Non‐invasive quantitative = 33/39 (87%) VAP suspected: new or progressive radiographic infiltrate plus 2 of the following: fever > 38.3 ºC or hypothermia < 35 ºC, leukocytosis or leucopenia, purulent tracheal secretion
Interventions	Invasive quantitative versus non‐invasive quantitative
Outcomes	Mortality ‐ Invasive quantitative = 14/37 (38%) ‐ Non‐invasive quantitative = 18/39 (46%) Antibiotic change ‐ Invasive quantitative: 10/37 (28%) ‐ Non‐invasive quantitative: 7/39 (18%) Intensive care unit (ICU) stay (days) ‐ Invasive quantitative = 21 ± 15 ‐ Non‐invasive quantitative = 21 ± 18 Duration of mechanical ventilation (days) ‐ Invasive quantitative = 19 ± 15 ‐ Non‐invasive quantitative = 20 ± 24 Appropriateness of initial antibiotic choice ‐ Invasive quantitative = 27/37 (73%) ‐ Non‐invasive quantitative = 31/39 (79.4%)
Notes	―
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated randomisation table into 1 of the 2 groups
Allocation concealment (selection bias)	Low risk	As above
Blinding (performance bias and detection bias) All outcomes	High risk	Open study
Incomplete outcome data (attrition bias) All outcomes	High risk	All patients were followed up
Selective reporting (reporting bias)	Unclear risk	No data to evaluate
Other bias	Low risk	Not detected

Sanchez‐Nieto 1998.

Methods	Duration: not informed Parallel design Blinded assessment of outcomes: not met Withdrawals described and were they acceptable: met
Participants	Sample size = 51 APACHE II (severity) ‐ Invasive quantitative = 15 ± 5 ‐ Non‐invasive quantitative = 18 ± 5 Duration on mechanical ventilation before study (days) ‐ Invasive quantitative = 11 ± 8 ‐ Non‐invasive quantitative = 11 ± 15 Previous use of antibiotic ‐ Invasive quantitative = 20/24 (83%) ‐ Non‐invasive qualitative = 19/27 (70%) VAP suspected: new or progressive radiographic infiltrate plus 2 of the following: fever > 38.3 ºC or hypothermia, < 35 ºC; leukocytosis or leucopenia, purulent respiratory secretions
Interventions	Invasive quantitative versus non‐invasive quantitative
Outcomes	Mortality ‐ Invasive quantitative = 11/24 (46%) ‐ Non‐invasive qualitative = 7/27 (26%) Antibiotic change ‐ Invasive quantitative = 10/24 (42%) ‐ Non‐invasive qualitative = 4/27 (15%) Intensive care unit (ICU) stay (days) ‐ Invasive quantitative = 28 ± 17 ‐ Non‐invasive qualitative = 26 ± 18 Duration of mechanical ventilation (days) ‐ Invasive quantitative: 23 ± 12 ‐ Non‐invasive quantitative: 20 ± 17 Appropriateness of initial antibiotic choice ‐ Invasive quantitative = 14/24 (58.3%) ‐ Non‐invasive qualitative = 23/27 (85.1%)
Notes	―
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated randomisation table into 1 of the 2 groups
Allocation concealment (selection bias)	Low risk	As above
Blinding (performance bias and detection bias) All outcomes	High risk	Open study
Incomplete outcome data (attrition bias) All outcomes	High risk	All patients were followed up
Selective reporting (reporting bias)	Unclear risk	No data to evaluate
Other bias	Low risk	Not detected

Solé Violán 2000.

Methods	Duration: not informed Parallel design Blinded assessment of outcomes: not met Withdrawals described and were they acceptable: met
Participants	Sample size = 88 APACHE II (severity) ‐ Invasive quantitative = 15.8 ± 0.9 ‐ Non‐invasive qualitative = 15.0 ± 0.9 Duration on mechanical ventilation before study (days) ‐ Invasive quantitative = 7.8 ± 1.1 ‐ Non‐invasive qualitative = 7.3 ± 0.9 Previous use of antibiotic ‐ Invasive quantitative = 16/45 (35.5%) ‐ Non‐invasive qualitative = 19/43 (44.1%) VAP suspected: new radiographic infiltrate plus 2 of the following: fever > 38.5 ºC, leukocytosis or leukopenia, purulent tracheal secretions
Interventions	Invasive quantitative versus non‐invasive qualitative
Outcomes	Mortality ‐ Invasive quantitative = 10/45 (22.2%) ‐ Non‐invasive qualitative = 9/43 (20.9%) Antibiotic change ‐ Invasive quantitative = 15/45 (33.3%) ‐ Non‐invasive qualitative = 5/43 (11.6%) Intensive care unit (ICU) stay (days) ‐ Invasive quantitative = 23.6 ± 3.1 ‐ Non‐invasive qualitative = 22.4 ± 3.1 Duration of mechanical ventilation (days) ‐ Invasive quantitative: 19.9 ± 2.8 ‐ Non‐invasive qualitative: 19.2 ± 3.0 Appropriateness of initial antibiotic choice ‐ Invasive quantitative = 42/43 (93.3%) PS: 2/45 inadequacy (isolated organism treated with only 1 effective antibiotic) ‐ Non‐invasive qualitative = 42/43 (97.6%) PS: 1/43 inadequacy
Notes	―
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated randomisation table into 1 of the 2 groups
Allocation concealment (selection bias)	Low risk	As above
Blinding (performance bias and detection bias) All outcomes	Unclear risk	It is not stated in the paper
Incomplete outcome data (attrition bias) All outcomes	Low risk	3 patients were excluded because of transfer to another institution
Selective reporting (reporting bias)	Unclear risk	No data to evaluate
Other bias	Low risk	Not detected

APACHE: acute physiology and chronic health evaluation
 CI: confidence interval
 ICU: intensive care unit
 PS: positive specimen
 SAPS: simplified acute physiologic score
 VAP: ventilator‐associated pneumonia

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Cai 2001	The study is not clearly randomised and is a cross‐over study

Contributions of authors

Danilo Cortozi Berton (DCB), Andre Kalil (AK) and Paulo José Zimermann Teixeira (PJZT) wrote the draft protocol.
 AK revised the methodology section.
 PJZT and DCB were responsible for identifying studies and extracting data.
 All review authors contributed to the review.

Declarations of interest

Danilo Cortozi Berton: none known.
 Andre Kalil: none known.
 Paulo José Zimermann Teixeira: none known.

New search for studies and content updated (no change to conclusions)