VAP is the most frequent type of infection in ICU patients in Europe and Latin America (almost half of all nosocomial infections)  and ranks second in US ICUs . The attributable mortality is higher in medical than in surgical patients, and rates vary according to the case mix and aetiological agent .
Inadequate or delayed antimicrobial treatment in VAP is an established independent predictor of death . According to published data, changing an initial empirical treatment based on subsequent culture results may have either a beneficial effect (in terms of mortality, less antibiotic use, less days on antibiotics)  or no effect in more severely ill patients . For this reason, efforts must be directed at choosing adequate empirical treatment as early as possible, which may be accomplished with a high degree of suspicion and adequate guidelines based on local antibacterial susceptibilities. In addition, adhering to ideal pharmacological principles (choosing continuous as opposed to intermittent administration, adjustment for renal and hepatic failures), reducing dosages when appropriate, and shortening the duration of treatment are presently standard of care for VAP.
In order to avoid any delay in instituting antibiotic treatment, reliable diagnostic methods should be employed. Despite their variable sensitivity and specificity , clinical/radiological findings may currently be considered the best option, although rapid tests, such as the percentage of infected leucocytes on bronchial specimens, are promising in that they can provide rapid confirmation . Culture results for bronchial or tracheal samples may be available late in the course of an episode of VAP and should not be used to decide whether to treat, especially in patients who are severely ill. On the other hand, culture results should be used to adjust (narrow or extend antibiotic spectrum) or withdraw empirical treatment – a practice that has been shown to be beneficial, with no increase in mortality, and that directs medical staff to seek other unsuspected foci of infection .
Although bronchoscopic samples increase the degree of confidence that a diagnosis of VAP is correct , endotracheal aspirates, despite their lack of consistency as a diagnostic tool , are widely employed in the management of VAP. Recent small trials have consistently shown that there is no advantage of using bronchoscopic methods over relying on tracheal aspirate cultures when mortality is an end-point [6, 20, 21]. Reduced costs and similar outcomes were reported using either quantitative or qualitative tracheal aspirates for guiding or deciding to interrupt antibiotic treatment for VAP . This may be due to the high correlation between tracheal aspirates (both quantitative and qualitative) and bronchoscopic cultures when presence of VAP is highly probable [21, 22]. However, the above-mentioned studies did not determine the value of quantification of micro-organisms in tracheal aspirate samples as compared with qualitative assessment.
Quantification of micro-organisms in biological samples for the purpose of diagnosing infectious conditions is widely used, particularly for nosocomial infections. Regarding respiratory infections, bronchoscopic samples have established cutoff values (104 cfu/ml for bronchoalveolar lavage [BAL] fluid and 103 cfu/ml for protected brush specimen [PBS]) for improving diagnostic performance. On the other hand, use of these cutoff values has yielded conflicting results, and previous antibiotic treatment has great impact on these values. Souweine and coworkers  showed that the standard cutoff values of BAL and PSB would have to be lowered to 103 cfu/ml and 102 cfu/ml to retain diagnostic accuracy where antibiotics were previously administered, mainly when they are given in the preceding 24 hours.
Only a small number of studies have evaluated the role of quantitative endotracheal cultures in the diagnosis of VAP. Albert and coworkers , studying 20 ventilated patients and using clinical/radiological parameters, found the threshold of 105 cfu/ml to have a sensitivity of 81%, specificity of 65%, positive predictive value of 55% and negative predictive value of 55%. In that study different cutoff values were not tested to evaluate the real usefulness of quantification. Jourdain and coworkers  studied a group of 57 patients with presumed VAP, 19 (33%) of whom were confirmed by PSB sample with more than 103 cfu/ml. Using quantification in this population, those investigators showed that the sensitivity of the test reduced considerably from 86% to 43% whereas specificity increased from 52% to 95% when a cutoff of 103 cfu/ml was compared with one of 107 cfu/ml. No data regarding previous use of antibiotics were available to explain the decreased sensitivity.
We conducted a prospective follow up of severely ill patients with a high rate of antimicrobial use prior to diagnosis of VAP. Not surprisingly, the most frequent agents recovered were multidrug-resistant agents, such as methicillin-resistant S. aureus, P. aeruginosa and Acinetobacter spp.
We found different levels of sensitivity (81%, 65%, 26%) and specificity (23%, 48%, 78%) for qualitative and quantitative (cutoffs 105 cfu/ml and 106 cfu/ml) findings, respectively, as was expected. However, the positive (18%, 21%, 20%) and negative (86%, 87%, 83%) predictive values obtained were very similar.
Our data reveal sensitivity values for tracheal aspirates similar to those observed in the above-mentioned studies, although specificity values were lower. According to our data, use of the cutoff value 105 cfu/ml reduced the sensitivity of the test to levels too low to be useful in clinical practice, bearing in mind the proposed role of tracheal aspirates to guide antibiotic withdrawal or modification. Moreover, quantification did not improve predictive values for the purposes of diagnosing VAP at the time when a suspected case was evaluated.
Patient characteristics may have an impact on the accuracy of diagnostic tests. Although there is broad correlation between the number of bacterial colonies in biological samples and the occurrence of infection as opposed to colonization, the exact bacterial count cannot be predicted in highly ill patients, for whom a lower inoculum may be sufficient for disease development. This has been observed for catheter-related infections in severely ill patients in a surgical ICU , in which true catheter-related bacteraemia was reported with fewer than 15 cfu on catheter tips. In our patient population there was a significant proportion of patients with renal failure, diabetes, cancer and sepsis – conditions that are known to be associated with immunosuppression.
These decreased sensitivity values may also be explained by antimicrobial use. More than 95% of the patients studied were receiving antibiotics when the sample was collected for analysis, and the majority of them were broad-spectrum antibiotics (almost 50% had received glycopeptides and 35% carbapenems). About 80% had received them for longer than 72 hours. Decreased accuracy of quantification with samples obtained by bronchoscopy was reported by Soweine and coworkers . BAL and PSB had significantly less sensitivity when the procedure was performed within 24 hours of antibiotic use than when antibiotics had not been given for longer than 72 hours. The impact of antibiotic use may be greater for tracheal aspirates, irrespective of the timing of administration; this may be due to the higher concentration of the antibiotic in upper tract secretions, although this point requires further investigation.
Our study has a number of limitations. While we attempted to achieve a high degree of certainty in clinical/radiological parameters, with the participation of three experienced ICU physicians (with a high degree of correlation between them), no 'gold standard' technique was employed, such as bronchoscopic samples (although it remains controversial whether bronchoscopy samples can be regarded as the gold standard for VAP). Because of the low specificity of clinical judgement, we must consider the fact that we are studying a population in which VAP rate is over-estimated. This is supported by the rate of 18.4% of VAP diagnoses with a negative tracheal aspirate finding and a 13.3% rate of fungal isolates, which only rarely can be considered true causative agents. Thus, it is possible that we have false-positive rate of at least 31.7%, although technical problems with specimen collection cannot be ruled out. The virtual absence of a gold standard for VAP makes study designs that address the issue of diagnostic tests difficult. In accordance with our study design, we evaluated all patients with mechanical ventilation every week, irrespective of clinical suspicion of VAP. This strategy may have beneficial effects because we included in the same population patients who were likely and those who were unlikely to have definite VAP, but increasing the possibility of false-positive cases.
Other study designs use populations selected because clinical/radiological judgement suggest the presence of VAP. In these studies, the control cases (no VAP) are defined as having negative bronchoscopic cultures, based on predetermined cutoff values. In these situations, problems with the lesser sensitivity of bronchoscopic samples in patients on antibiotics, and even the intrinsically low sensitivity of this diagnostic strategy when compared with histological criteria , increase the likelihood of including false-negative control individuals. In other words, with our study design we might have overestimated VAP, as compared with underestimating it with conventional study designs. For this reason we think that there is no ideal design for such studies, and studies that rely solely upon clinical/radiological parameters should not systematically be discarded. Furthermore, the use of bronchoscopy in our hospital is unreliable, as it may be in a large number of general ICUs.
Tracheal aspirates have a definite role to play in the management of VAP, but only when correlated with clinical findings . The use of quantitative results may be associated with under-diagnosis of VAP, leading to inappropriate changes to antibiotic regimens and, in some cases, antibiotic delay or withdrawal.