Background and Objectives: There is a need to validate and suggest easy and less costly diagnostic method for diagnosis of ventilator-associated pneumonia in developing country. The study was performed to compare available methods for the diagnosis and to characterize the organisms causing VAP.

Design and Methods: All patients on mechanical ventilation for more than 48 hours and simplified CPIS ≥ 6 were enrolled prospectively. Four diagnostic procedures, endotracheal aspiration (ETA), blind bronchial sampling (BBS), blind bronchoalveolar lavage (blind BAL) and bronchoscopic BAL (BAL) were performed in same sequence within 12 hours. The bacterial density ≥ 104 cfu/ mL BAL samples were taken as reference standard.

Introduction Ventilator associated pneumonia (VAP) is defined as nosocomial pneumonia developing in a patient 48 hours after the initiation of mechanical ventilatory support (by endotracheal tube (ETT) or tracheostomy tube) (1). Despite major advances in the techniques for the management of ventilator dependent patients, VAP continues to complicate the course of 8-28% of the patients receiving mechanical ventilation (MV) (1-3). Rates of pneumonia are considerably higher among patients hospitalized in intensive care units (ICUs) compared with those in the hospital wards. The risk of pneumonia is increased 3 to 10 folds for the intubated patient receiving mechanical ventilation (1, 2, 4-6). The mortality with VAP is considerably high, varying from 24 to 50% and can reach as high as 76% in some specific settings or when lung infection is caused by high risk pathogens (1).

For many years, VAP was diagnosed by clinical criteria such as fever, leukocytosis and purulent tracheobronchial secretions supported by the radiological evidence of new or persistent pulmonary infiltrates (7). However these criteria are non-specific (8). Studies have shown that clinical criteria have imperfect diagnostic reliability in ventilated patients, and therefore, additional procedures such as cultures of the lower respiratory tract are required for the accurate diagnosis and treatment of VAP (9, 10). However, clinical criteria do remain crucial for defining those patients who may require respiratory sampling (1).

Given the invasive nature and high incidence of complications with techniques such as lung biopsy and percutaneous needle aspiration, these have now been replaced by safer methods such as endotracheal aspiration (ETA), bronchoscopic bronchoalveolar lavage (BAL), protected sampling brushing (PSB) and non-bronchoscopic methods such as blind BAL and blind bronchial aspirates (BBA) as approaches for the definitive diagnosis of VAP (8, 12). The ETA is the most widely used sampling technique in ventilated patients (13). This technique is known to have a high sensitivity but also has a high false positive rate and correlates poorly to actual pathogen due to respiratory tract colonization (14). Invasive bronchoscopic sampling techniques like bronchoscopic BAL and PSB are currently considered as the reliable sampling techniques in adult ICU to recover organism from the lower respiratory tract (15) and have high sensitivity and specificity (16). The usefulness of these techniques in routine clinical practice is, however, hampered by the potential risks of bronchoscopy, the unavailability of equipment and trained personnel on a 24-hour basis in many critical care facilities and the associated cost.

The microscopic identification of organisms causing pneumonia requires a simple, safe, effective and inexpensive method with good sensitivity and specificity with results at least equal to bronchoscopic BAL. Non-bronchoscopic bronchoalveolar lavage and BBS have been reported to have good efficacy in the microbiological diagnosis of VAP (17- 19). Though these techniques have not been standardized in children, they appear promising (20).

The methods of diagnosing VAP are debatable and there is no accepted 'gold standard'. No study has shown the superiority of a specific method. The methods proposed have different sensitivities and specificities (11). So, additional validation studies are needed. Other indications of the need for further studies of the non-bronchoscopically directed techniques are the absence of standardized diagnostic thresholds for quantitative cultures, and to develop cheaper, reliable and easy to use diagnostic method involving inexpensive easily available equipment.

Numerous studies have been conducted in the adult medical and surgical ICUs (1, 5, 12, 13, 16-18) establishing the above-mentioned techniques as reliable methods for the diagnosis of VAP, but corresponding studies in children are very few (21, 22) and none from the developing world.

Therefore this prospective study was conducted to know the causative microorganisms and to evaluate and compare the four procedures namely endotracheal aspirates (ETA), blind bronchial sampling (BBS),blind bronchoalveolar lavage (blind BAL), and bronchoscopic bronchoalveolar lavage (BAL) for the diagnosis of VAP in the tertiary care PICU of a developing country.

Materials and Methods

Hospital, ICU characteristics and Facilities. This prospective study was conducted in the 6-bedded PICU of a tertiary care multidisciplinary teaching hospital. This 600-bedded hospital has separate adult medical and surgical, pediatric, neonatal, cardiac and post transplant ICUs.The PICU is equipped with 6 ventilators, central oxygen, air and vacuum supply. Each bed has multiparameter monitor for continuous hemodynamic and respiratory monitoring .Nurse on duty maintain vital sign and intake-output record on hourly basis. Besides standard treatment modalities and ventilation, bedside renal replacement therapy, electroencephalography, echocardiography and flexible bronchoscopy facilities are available. Extracorporeal membrane oxygenation and nitric oxide therapies are not yet available. The PICU is staffed with full time, experienced pediatric intensivists, critical care fellows and residents. The nurse / patient ratio of 1:2 is provided round the clock.

Study Population and data collection. During 9 months period (January to September 2003) of study all patients on mechanical ventilation for more than 48 hours by endotracheal tube or tracheostomy were evaluated for the development of VAP. Data pertaining to the clinical suspicion of pneumonia included daily temperature, total leucocytes counts, PaO2 / FiO2 ratio, nature of tracheobronchial secretions, and chest X-ray. Each patient was assigned a score based on the simplified clinical pulmonary infection score (CPIS) criteria (23). Total points in this composite score vary from 1 to 10 points. Only those patients with a CPIS score of 6 or more were included in the study.

For all patients included as study case following data was recorded: age, gender, clinical presentation, dates of admission and discharge. Time period of ICU stay prior to initiation of ventilation, duration of mechanical ventilation, duration of ICU and hospital stays were also recorded. Date of suspicion of VAP and number of episodes of VAP were also noted. Chest radiographs at the time of admission, on initiation of ventilation and at the time of clinical suspicion of VAP were also recorded.

Specimen collection. With each episode of clinically suspected VAP, the patient was subjected to four different sampling techniques within 12 hours of clinical diagnosis of VAP with an average of an hour between each procedure. All procedures including bronchoscopy were performed by same investigator. A set sequence of sampling was followed in each case with the ETA first, followed by BBS, then blind BAL, and BAL at the last. All patients were premedicated with midazolam and fentanyl prior to performing these sampling procedures unless they were already sedated and paralyzed.

Endotracheal Aspiration. Sterile suction catheters of 53 cm length (model GS 2006 Romsons, Agra, India) of sizes appropriate for the different sized endotracheal tubes were used. The catheter connected to mucus trap unit (model GS 51800, Romson, Agra, India) was advanced through the ETT till some resistance was encountered and after withdrawing about 1-2 cm, suctioning was performed and the aspirate collected in a sterile mucus trap. During this procedure the patient was temporarily disconnected from the ventilator with mean duration of 10 seconds.

Blind Bronchial Sampling. Blind bronchial sampling was performed using sterile catheter 2 sizes smaller than ETA catheter. For 4.0 size ETT, 6 Fr size suction catheter was used. Catheter was introduced and advanced into the ETT blindly to the length around 5 cm greater than the length of ETT tube at the lip level. To increase the chances of placing the suction tube in the affected lung, the patient's head was turned to the opposite side in cases where infiltrates were unilateral. No saline was injected before or during the procedure. The procedure was repeated 2-3 times to obtain sample.

Blind Bronchoalveolar lavage. Patients were preoxygenated for 5-10 minutes with 100% oxygen and sterile, disposable swivel adapter was inserted between ETT tube and ventilator circuit. Thus, all the patients received ventilation during the procedure. The catheter used for performing the blind BAL was a balloon tipped wedge pressure catheter (Model A1- 07121, Arrow International, PA 19605, USA). This was 60cm in length, 4Fr, double lumen with inflatable balloon of 0.6 mL capacity at the distal end. The length of the catheter was primed and filled with saline before insertion. The catheter was introduced into the ETT tube via swivel adapter's small opening and advanced much further than the length of the ETT from the lip level till resistance met. To increase the chances of obtaining lavage from affected lung, head turning maneuver was used as described previously. The distal balloon was inflated with 1.0mL of normal saline since fluid is less compressible than air. Attempt was also made to wedge the catheter properly by further pushing it if possible. The volume of saline aliquot injected through catheter varied with the weight of the child. We used 3 mL for babies less than 5 Kg, 5 mL for children between 5 - 10 Kg, 7.5 mL for 11- 20 Kg and 10 mL for patients above 20 Kg. An aliquot of sterile saline was injected over 10 seconds and immediately reaspirated into the syringe attached to the 3-way of the catheter. Four such separate aliquots were used without withdrawing the catheter and using 4 different syringes. The first syringe aspirate was discarded while other 3 samples were collected in a sterile mucus trap.

Bronchoscopic bronchoalveolar lavage. Before starting BAL, all patients were pre-oxygenated with 100% oxygen for 5-10 minutes. The site for bronchoscopic BAL was chosen according to the x-ay appearance. BAL was thus performed in an area of localized pulmonary infiltration if present. In case the chest x-ray showed diffuse disease or no specific area with infiltrate, BAL was performed in the area with most inflammation or with purulent secretions. If no inflammation or purulent secretions were seen during bronchoscopy, BAL was performed in the right lower or middle lobe. All bronchoscopic lavages were obtained with Olympus BF type XP40 bronchoscope (Olympus Optical Co. Japan) with outer diameter of 2.8mm and a suction channel of 1.2mm size. Endotracheal suction was done just prior to introduction of bronchoscope. Lignocaine spray was avoided as local anesthetic during the procedure. BAL was performed via the endotracheal tube using a swivel adapter. In infants with 4.5 or less endotracheal tube size, an appropriate size laryngeal airway mask was used for bronchoscopic lavage (24). Under visual control, the bronchoscope was advanced in the direction of the chosen segment until a wedged position was achieved. Lavage was carried out using 4 aliquots of sterile saline (depending on weight of the patient)

All patients were monitored during and after all the four procedures with multiparameter monitor and any significant hypoxic episode (saturations dropping below 80) or cardiac arrhythmia were recorded as complications of the procedures. In 3 infants transient hypoxia was noticed during BBS procedure. This was corrected rapidly with bagging and the procedure was completed on second attempt using sterile disposable swivel adaptor providing continued ventilation.

Microbiological methods. All the samples were transported to the laboratory within 15 minutes and cultured within an hour of collection. After receipt in the laboratory, the samples were first vortexed for 60 seconds after which, gram stained preparations were performed and studied for the presence of squamous cells, polymorphonuclear cells and the type of microorganism present. The presence of polymorphonuclear cells (PMN) in the aspirated fluid was semi-quantified using techniques published previously (25). Simultaneously, quantitative cultures using the calibrated loop method were performed on common media such as blood agar, chocolate agar and Mckonky's agar using standard techniques (26). Organisms were identified using automated Vitek - 1 system (bioMerieux, France). Microbiological examination for unusual organisms such as Mycoplasma, Chlamydia, Pneumocystis carinii and viruses did not form a part of this study.

The institutional review board approval was obtained for the study and informed consent for all diagnostic procedures was taken from the parents.

Statistical analysis. In accord with previous studies (1, 14, 15, 21, 22) of quantitative bacteriology of BAL cultures, a bacterial density of >104cfu/ml was considered as 'POSITIVE' for VAP and those episodes were referred to as 'DEFINITE VAP' episodes. The organisms isolated on blood culture were compared against the organisms isolated from the cultures of various tracheobronchial techniques using the chi square test. Analysis was also done for any relation between the semi-quantitation of PMN on gram stain and cultures of the lower tracheobronchial secretions. Taking BAL colony counts of ≥ 104 cfu/ml as the reference standard, the other three methodologies - ETA, BBS and blind BAL were analyzed and their receiver operating curves (ROC) were plotted and the area under the curve were also obtained. Following this, the sensitivities, specificities, positive and negative predictive values (PPV, NPV) and accuracies at various cut off values of colony counts for all the above mentioned techniques were calculated. Concordance between different diagnostic methods was analyzed using kappa score. P value less than 0.05 was considered significant.

DISCUSSION To our knowledge this is the first prospective study from a developing country comparing the four methods ( ETA, BBS, blind BAL, bronchoscopic BAL) for the diagnosis of VAP. Taking bronchoscopic BAL as the reference standard, the operative characteristics of the first three techniques at different threshold values of colony counts were worked out along with the microbiology of VAP.

Pseudomonas aeruginosa was the most common organism to be isolated from the forty episodes of VAP.This finding is in agreement with other studies, which have stated the incidence of Pseudomonas VAP to be the highest among all pathogens. (14, 27, 28). A systematic analysis reported that gram negative bacteria represented 58% of the recovered organisms (1). Of these, most of the episodes were caused predominantly by Pseudomonas aeruginosa (24.4%), Enterobacteriaceae (14.1%), Acinetobacter (7.9%) while Staphylococcus aureus was reported in 20.4% and Candida sps for 0.9% of the episodes of VAP.

In the present study simplified CPIS was used because it is simple and can be used repeatedly. This includes readily available parameters indicating clinical, laboratory and oxygenation status of patient. This score has been extensively studied in adults. This score has the sensitivity ranging from 72 - 77% and the specificity varying from 58 - 85% for the diagnosis of VAP (29, 30). The use of this score has been reported once in a retrospective pediatric study in 40 cases (31). This study showed CPIS has a PPV of 93% and was also found to be an early predictor of poor prognosis. In present study the CPIS score was also evaluated against the reference standard at values of 7, 8, 9 and 10. It was found that the most accurate value of CPIS to diagnose DEFINITE VAP was between 7 and 8 with the area under the ROC being 0.812 (p value=0.001). The mean CPIS in patients with definite VAP was 8.4 while in no definite VAP group it was 6.7 (p 0.007). In a study by Pugin et al (18), the cut off of 6 for CPIS was found to show a good correlation (r 0.84, p < 0.0001) between this clinical score and quantitative bacteriology of BAL samples.

There was not a single instance when the BAL culture was positive with a corresponding negative endotracheal aspirate. ETA cultures yielded microorganisms in 37 episodes as compared to 29 by the reference standard and therefore only 29 episodes (78%) could be called definite VAP on endotracheal aspirates It has been shown that ETA has high sensitivity but low specificity for the diagnosis of VAP (22). So a sterile ETA assures the absence of VAP.In a study by Torres et al (32) of the 51 isolates on qualitative ETA only 29 (57%) correlated with same organisms growth on PSB (103 cfu/mL) and BAL (>103 cfu/mL). In another study (22) endotracheal secretions culture was positive in 70% of 30 pediatric cases. The concordance between reference standard (expert's opinion) and positive ETA was 57% (kappa 0.24). This high false positive rate in our as well as the studies cited above, could be explained by the bacterial colonization of the proximal airways observed in most patients in the ICU.

In the present study threshold value for ETA of 105 cfu/mL appeared to be most accurate with kappa statistic of 0.631. This cut off value was similar to that obtained by El Ebiary et al (33) where protected specimen brush (PSB) (>103 cfu/mL) and BAL (>103 cfu/mL) were used as the reference standard.

Blind bronchial sampling showed 88% sensitivity and 82% specificity at 104 cfu/mL (kappa 0.68). Thus quantitative cultures of BBS appear to be simple, non invasive and useful tool in the diagnosis of VAP. The high accuracy of BBS is not unusual as VAP is usually secondary to the aspiration of colonized oro-pharyngeal secretions into the dependent areas of the lung (especially the right), which can easily be reached with a catheter passed blindly through the endotracheal tube (19). Papazian et al (30) showed the sensitivity of BBS (58%) was greater than that of PSB (42%) but less than that of BAL (93%) and the area under ROC of BBS was greater than that of PSB (p<0.05). The author concluded that BBS was preferable to PSB for the diagnosis of VAP.

The blind BAL performed with wedge pressure catheter was most accurate at ≥ 103 cfu /ml (kappa 0.78) in the present study. Pugin et al (18) compared blind BAL results with bronchoscopic BAL. They calculated a sensitivity of 73% and a specificity of 96% and PPV of 92% for blind non-bronchoscopic BAL. Gaussorgues etal (34) used 7 Fr right heart catheterization cuffed catheter for blind BAL in 13 adults and compared with postmortem lung tissue histology and culture. When lung histology and cultures were negative for pneumonia and organisms, the blind BAL were also negative for organisms on culture. Among the 10 positive blind- BAL cultures, lung biopsy showed histologic pneumonia in 9 cases. Fourteen organisms were cultured from lung tissue while blind BAL correctly identified the causative microbes in 13 cases. Alpert etal (35) used 4 Fr balloon tipped wedged pressure catheter to perform blind BAL in 20 pediatric cases aged 1 month to 6.5 years. Clinically significant information was obtained in 17 (85%) cases and no patients required an open lung biopsy. Other innovative method for blind BAL using feeding tube in small children has been described by Koumbourlis etal (36).

Strengths of the study. This prospective study included consecutive patients selected with strict criteria including simplified CPIS and not on the attending physician suspicion, so reducing the selection bias. All the diagnostic procedures were done in the same sequence in a predefined time period and by the same investigators. As it appears, BBS and blind BAL can be used for the diagnosis of VAP. In this study performance of blind BAL was best, followed closely by BBS when compared with bronchoscopic BAL as a reference standard for the diagnosis of VAP. The imported balloon tipped pressure catheter used in this study for blind BAL costs INR 1600.00 whereas suction catheter made in India used for BBS costs less than INR 10.00. This huge cost difference is definitely a major consideration in the PICU of developing countries while investigating VAP.

Limitations of study. Lung tissue specimen is the best tool for the diagnosis of VAP (37) but not perfect as a gold standard (22). Lung tissue sample results may be negative if the patient is already on antibiotics and histological changes may be disseminated in the different segments of the lungs and not in a homogenous fashion (39). Studies have shown that sampling limited to one specimen from the lungs cannot exclude VAP (40, 41). Moreover the autopsy is difficult to use as gold standard in the pediatric population because of the mortality attributed to VAP is as low as 8% (42). In the present study the mortality due to VAP was 10%. In this study the bronchoscopic BAL culture (≥104cfu/mL) was used as reference standard. Compared with lung specimen cultures and histology, PSB (≥ 103 cfu/mL) and bronchoscopic BAL (≥ 104 cfu/mL) showed strong correlation in identifying the causative organisms (38). Moreover BAL appears to have an acceptable level of reproducibility of 75% (28). High concordance has been found for the presence (93%), type (86%) and quantity (78%) of bacteria in two protected BAL samples taken 2 hours apart (20).

CONCLUSIONS The most common organism responsible for VAP in this study was Pseudomonas aeruginosa. Negative ETA culture rules out the diagnosis of VAP. Quantitative cultures of BBS and blind BAL were useful in the diagnosis of VAP at threshold of ≥ 104 cfu / mL and ≥ 103 cfu / mL respectively. The role of BBS should be reevaluated in the future studies since it is cheap, easy to perform, well tolerated by ventilated children and can be repeated easily.

Table 1- Baseline Characteristics of the study population (n= 30)

No of VAP episodes 40 Age median (range) 6.5 years (1 mo -12 yrs) Males 20 (66.5%) PRISM score (range) 13 ± 6 (0 - 34) Primary system involvement Central nervous system 9 (30) Respiratory system 6 (20) Septicemia 5 (16.6) Diabetic ketoacidosis 2 (6.6) Postoperative status 2 (6.6) Trauma 2 (6.6) Miscellaneous 4 (13.3) Multiorgan dysfunction 15 (50) Use of steroids 4 (13.3) Immunosuppressed state 2 (20) Use of H2 blockers 9 (30) Central lines 30 (100) Arterial lines 30 (100) Chest x-ray At admission Normal 21 (70) ARDS 3 (10) Others 6 (20) At initiation of MV Normal 14 (44.6) ARDS 9 (30) Others 7 (23.3) CPIS mean ±SD (range) 7.8 ± 1.5; 6 - 10

Figures in parentheses indicate percentage unless mentioned

Table 2 - Clinical characteristics of study cases on the day of lower respiratory tract aspirate collection (n= 30)

No. of VAP episodes 40 Temperature °C a 38.9 ± 0.81 (36 - 40.2) TLC a 17995± 6840 (3200 - 32700) PaO2 / FiO2a 185 ± 65.3 (112 - 320) Purulent tracheo- bronchial secretions Scanty 1 Moderate 14 Profuse 25 Abnormal chest X-ray Collapse 21 Consolidation 8 B/L pulmonary haziness 6 ARDS 4 Cavitatory lesions 1 CPIS a (range) 7.8 ± 1.5 (6-10)

a mean ± SD . Figures in parentheses indicate range

Table 3 - Operative indices of simplified clinical pulmonary infection score for the definite diagnosis of ventilator associated pneumonia (Reference standard BAL colony counts of ≥ 104 cfu / ml)

CPIS Sensitivity Specificity PPV NPV Accuracy

6 100 0 62.5 0 62.5

7 88 60 78.5 75 77.5

8 80 80 87 70.5 80

9 61 87 87.5 54 67.5

10 24 93 85.7 42 50

Table 4 -. Relation between polymorhonuclear cells on Gram stain samples of lower respiratory tract aspirate obtained by different techniques with positive culture of BAL (≥104 cfu /ml) (n= 40 in each category)

Technique Gram stain PMN a No. of BAL culture Contingency p value

• 104 <104 coefficient

Endotracheal Few 0 3 Aspirate Moderate 0 1 0.39 0.02 Many 25 11

BBS Few 0 3 Moderate 1 2 0.38 0.02 Many 24 10

Blind BAL Few 1 6 Moderate 3 2 0.42 0.03 Many 21 7

BAL Few 0 7 Moderate 2 1 0.51 0.003 Many 23 7

a Reference 25

Table 5. Microorganisms cultured from different lower respiratory tract aspiration techniques ( n= 40 in each category)

Organisms ETA BBS Blind BAL BAL

Pseudomonas aeruginosa 17 15 13 13 Acinetobacter baumannii 9 7 7 6 Klebsiella pneumoniae 6 5 2 5 Enterobacter species 5 4 4 4 MRSA 3 2 4 2 Proteus mirabilis 1 1 1 1 Escherichia coli 2 0 0 0 Candida species 3 3 3 3

Number of isolates exceeds no. of aspirates due to polymicrobial growth

Table 6. Operative indices of three sampling methods for the diagnosis of definite ventilator associated pneumonia compared with reference standard (BAL colony counts ≥104 cfu / mL)

Technique cfu / mL Sensitivity Specificity PPV NPV Accuracy

Endotracheal >103 100 27 69 100 72.5 Aspirate >104 96 60 80 90 81 >105 84 77 87.5 73 80 >106 8 100 100 39.5 42.5

BBS >102 100 60 81 100 85 >103 100 73 86 100 85 >104 88 82 88 83 87 >105 8 100 100 39.5 42.5

Blind BAL >102 96 73 86 91.6 87.5 >103 96 80 88 92.3 90 >104 32 100 100 47 57.5

A B

C D Figure 1. Receiver operative curves depicting the relation between reference standard BAL culture colony count ≥ 104 cfu/ ml and (A) CPIS (AUC 0.81± 0.06, p = 0.001), (B) ETA ( AUC 0.87± 0.06,p = 0.00), (C) BBS (AUC 0.89± 0.06, p= 0.01), (D) blind BAL ( AUC 0.89± 0.05, p= 0.00).

AUC- Area under curve

Inclusion Criteria:

• Children on mechanical ventilation for more than 72 hours and having simplified clinical pulmonary infection score more than 6 were included in study.

Exclusion Criteria:

• Age less than 1 month.