Reproducibility of protected brush catheter specimen cultures in critically ill patients with suspected nosocomial pneumonia

OBJECTIVE: To determine the reproducibility of two protected brush catheter (PBC) specimens obtained during the same bronchoscopy in critically ill patients with suspected ventilator associated pneumonia.

growth (positive/positive) in :n cases, nonsi gnilicant growth in 45 cases (negative/negative) and discordant results in six (positive/negative, n=3 or negative/positive, n=3 ).There was a significant relationship (P<0.05)betwee n the conrnrrenl use of antibiotics and a negative PBC result.Howeve r, after exclusion of patients on antibiotics, the overall proportion of agreement between the two PBCs was 0 .94with K=0.875 (P<O.O I versus K=0.4,95 % CL 0.721.1.029), which also indicates excellent agreement betwern the two tests .CONCLUSIONS : Although discordant results were observed in 7.2% cases.the overall reproducibility of the PBC results appears to be high.The signific ant relationship between concurrent antibiotic use and a negative PBC result is of concern clinically since many patients arc being treated with antibiotics al the time of bronchoscopy.Therefore.whe n the d iagnosis of nosocnmial pne umonia in critically ill patients is established.the PBC result should be considered unly in association with all the other clinical data, particularly in patients receiving concum.•ntantibiotics.( Pour le res11111,;.l'oir pag<' I 7-:f.J
E STAB LISIII NG THE C LI N ICAi.lll.-\G NOSIS OF PNF.U MON lA in mechanically ventilated patients is both inaccurate and often difficult.Oropharyngeal and cndotrachea l tu be colonization with pathogenic bacteria ocrnrs withi n 24 h of intubation and decreases the reliability of expectorated or aspirated tracheal secretions ( I).Furthermore, critica lly ill patients often have pulmonary infiltrates, fe ver and leu kocytosis clue to other conditions (pulmonary embolism, atclectasis, pulmonary edema, etc).which complicates the diagnosis of pneumonia.
Fihreoptic brnnchoscopy has been used extensively in intubated patients with suspccrccl pneumonia to obtai n samples of lower respiratory trac t secretions for bacterial cu lt ure (2-4).The development of the protec ted brush catheter (PBC) has further enhanced the usefulness of bronchoscopy in the diagnosis of nosocomial pneumonia (5-9).In addition , many investigators ( I 0-12) believe that the PB C used in conjunction with quantitative bacterial cu lture techniq ues improves the accuracy of the diagnosis or ventilator associated pneumonia.
The rcliahility or precision of a diagnostic test depends on how close repeated measurements of a stabl e phenomenon arc to each other ( I 3 ).To date, the reliability of the PBC technique re mains uncertain in the clinical setting.There has hccn limited expe rience with collecting mult iple PB C specimens from the same individual to dc te nnine the repro-PRINC IPA UX RrA .-.u LTATS : La proportion globale de concordance des resul tats des deux specimens obtenus par BP etait de 0.928 avec une stati stique kappa calculee (K)=0,853 (P<O,O I par rapport i1 K=0.4,interval le de confiam:e a 95 % 0,692-1.014)dcmontrant une concordan e excellcnte entre les dcux spec imens.Les deux specimens obtenus par BP 0111 revelc une croissancc significative (plus de I 0 3 unites formant colonies/ml) dans 33 cas (positif-positif), unc croissance non significative (negative-negati ve) dans 45 cas et des resultats discordants dans six cas (positif-ncgatiL n = 3 ou ncgatif-pos itif, n = 3).On a observe une relation significativc (P<0,05) entre l"utili sation concomitante d'anti biot iq ues ct un resultat ncgatif du BP.Ccpcndant.apn:s avoir excl ude l'e tude les patients sous antibiotiques, la proportion globale de concordance entre lcs deux spec imens obtenus par BP eta it de 0.94 avec K=0,87.'i(P<O,O I par rapport a K=0,4, intervalle de confiance a 95 % 0,721 -1,029), ce qui indique egalemcnt une concordance cxccllente entre les deux tests.CONCLUSIONS : Bien que des res ultats discordants aient ere observes dans 7.2 % des cas.la reprod uctibilitc globalc des resultats du BP sem blc elevee.La relation significativc cn tre (' utili sation concomitante d'antibiotiques et un resultat du BP negatif semble importantc cl iniqueme.ntpuisque de nombrcux patients son! sous antib iotiqucs au moment de la bronchoscopie.Done, lorsque le diagnostic de pneumonie nosocomialc a ete pose chCI.des patients des soi ns intensifs, le resultat du BP devrait sculcmcnt ftre interprctc en fon ction de toutes les autres donnces cliniqucs, surtout chez lcs patients qui son! dei1 sous traitcmcnt antibiotiq ue.ducibility of thi s technique (14)(15)(16).Marquette cl al (l:'i) demonstrated excellent qualitative reproducibility of the PBC techn ique ; however.quantitative cultures varied by more than I log 10 in over half the patients studied.and the presence or absence of pneumonia (based on the critical 'c utoff' of greate r than I 0 1 colony-fa nn ing uni ts I CFU 1/mL) differed in about 14% of patients.Similarl y, Timsit et al ( 16) observed discordant results between two PB C specimens in abou t I 7% of patients with suspected ventilator acquired pneumonia.We previously collected two PBC specimens during bronchoscopy in 48 critically ill pat ients with 53 epi sodes of suspected ventilato r acqu ired pneumonia ( 14) and noted a di screpancy between the two brush specimens in 7 .5% of cases.In the present paper.we describe the compldcd data of th is study (Ht 75 paticnts with 84 episodes or su, peeled ventilator acqu ired pneumonia in which we col • lected two sequential PBC spec imens at bronclwscopy.

PATIENTS AN D METHODS
The study protocol was approved by the human investi gation committee at the University of Western Onrnrio.Patient selection: Data were collected during 84 epi sodes of suspected nosocomial pneumonia occ urring in 75 criti ca lly ill pat ients between January I, 1991 and June 30, 1992.Patients were enrolled from two multidisciplinary intcnsi vc care units at the Un iversity of Western Ontario with a com -bi ned total or 50 beds and ap proximately 2200 annual admissions.The clinical diagnosis of nosocomial pneumonia was suspected in patients receiv ing mechanical ventil ation for more than 48 h if they developed a new pulmonary infiltrate that per isted longe r than 24 h.developed mac roscopica ll y purulent tracheal secretions.and either leukocytosis (white blood cell count greater than I I ,000/mm 3 ) or elevated tempe rat ure (greater than 38.3°C).Pat ients were excluded if they met any of the followin g criteria: clinical condition that d id not pem1it flex ible bronchoscopy ; diagnos is of pneumoni a upon admission to the intensive care unit; immunocompromised cond itio n (organ transplant.hematological mal ignancy, humora l or cell ular immune defic iency and patients rece iving more than 0.5 mg/kg/day of prednisonc ur its equi valent); or age less than 18 years.
All pat ients were ventilated with a Bird (Bird Products Corporation , Cali fumia) , Veolar (Hamilton Med ical, Nevada) or a Puritan-Bennett 7200 (Puritan-Bennett Corporation.Ca li forn ia) ventilator through an orotrachea l. nasotrac heal or tracheostomy tu be.The vent ilator tu bing (Marques!Med ica l Produc ts, Colorado) was changed e very 48 h according to routine protocol.Traeheobronchi al suctioning, chest phys iothe rapy and drainage of condensate from the ve nt ilator tubing were pe rfonned every 3 h or more frequentl y if it was deemed necessary by the attend ing staff.Experimental protocol: On entry into the study , the patient ' s hosp ital chart was re viewed and the fo llow ing variables recorded: age.sex.unde rlying med ical illnesses, da te of hospital and intens ive care unit admi ssion, elate of intubat ion, APAC HE II score (17) and an tibiotics used within 4 8 h be fore bronc hoscopy.Speci fic physio log ical vari ables (ie.Acute Physio logic Score) (17 .18) and a radiograph ic assessment (radi ographic score) ( 19,20) were recorded on the day nf bronc hoscopy .
All potentially eli gible patie nts were rev iewed on a daily basis for the signs and symptoms of pneumonia descri bed above.Once these signs developed and persisted more than 24 h. the patient (if awake) or the next of kin were approached fo r possible enrolment into the study.After in formed consent was given.lowe r respiratory tract secretions we re obtai ned for culture from all patie nts within 12 h of inc lus ion into the study and within 36 h after the diagnosis of pneumonia was s uspected.Bac teriological samples we re collected by both aspi ration through the endotracheal tube (tracheal aspirates) and fl exible bronchoscopy.with separate samples obtained by two PBCs (PBC-A and PBC-B).Tracheal aspirates were obtained immediately before bronehoscopy from all cases by an aseptic technique us ing dee p suc tioning via the endotracheal tube and collected in a sterile suction trap.
In preparati on for bronchoscopy.all patients were premedicated wi th intravenous cliazepam (5 to IO mg) or midazelam (2.5 to .5 mg ) with or without fenta nyl (.50 to I 00 µ g).Topical xy locainc ( I to 2%) was used minimally for local anesthesia only when patients continued to cough despi te int rave nous sedation.Arterial oxygen saturat ion and card iac rhythm were cont inuously monitored throughout the procedure with a pul se oximeter (SpaceLabs.Washington; Hewlett  He wlett Packard).W ith the fraction of inspired oxyge n adjusted to 1.0, the fl exible fib reoptic bronchoscope (Pentax model FB l 9H) was introduced into the endotrachea l tube th rough a special adapter (Rusch-Kernen , Germany ) allowing sim ul taneous mechanical ventilation and the ma intenance of positive airway press ure.The bronchoscope was advanced into the bronchial ori fice of the involved pulmonary segment that had been identified radiograph ically.Te le scoping PBCs (M ill-Rose; lnstrumecl) wit h distal plugs were inserted th roug h the inner suction channel of the bronchoscope .O nce in the involved bronchopulmonary segment, the di ssolvabl e cellulose plug was dislodged and the brush advanced under vis ual gu idance into the area of suspected pneumonia .After the involved segment was brushed, the brush (PBC-A) was withdra wn into the protective sheath and the entire catheter was removed fro m the bronchoscope ( 19).
In eac h patient, a second protected brush spec imen (PBC-B ) was immediate ly obtai ned from the same bronchopulrnonary segme nt by an ide ntical techn ique.
To m inimize potent ia l contami nation of the bronchoscope, suctioning was ke pt lo a minimum throughout the procedure.After withdrawal, the outer catheter was immediate ly cleansed with an a lcohol swab and severed with sterile sc issors.T he distal portion of the inner telescoping catheter was advanced, cleansed and cut in a similar manner.The brush wire was then advanced and cut with .,terilescissors into a sterile glass vial containing 1.0 mL of lactated sterile s::iline solution .
Between cases the bronchoscope was cleaned according to rou tine protocol.Briefly, the bronchoscope is di smantled and cl eaned with a brush and soap and water, then soaked in C iclex (Johnson and Johnson) for 30 mins before being rinsed with ste rile water.Bacteriological assessment: After colkction, specimens were immediate ly transported to the microbiology department.where tracheal aspirates were Gram stainL~d and cul- Specimen vials containing the protec ted brush tips were placed in a vortex m ixer fo r 60 s to suspend all materi al from the bru sh .A ft er Gram staini ng, two serial I 00-fo ld d il utio ns were pe ,fo rmed and 0. I mL of each dilution was plated on MacConkey agar, 5% horse blood agar and chocol ate agar for q uantitat ive c ultu re.Cultu res we re cons idered positi ve if mo re than 10 3 FU/mL of one or more po te ntially pathogenic bacteria were isolated.
Statistical analysis: Al l data are presented as mean ± SEM .
Qualitative comparisons were performed with x2=0 .05statistical analysis.Ag reement bet wee n the fi rs t (PBC-A) and second (PBC-8 ) c ult ure resu lt s was assessed by calcul ati ng the kappa statistic (K) (21 ).

RESULTS
Eighty-four episodes of cli nically sus pected pneumon ia were iden tified in 75 pat ie nts .T he demographic characteristics o f these patients a re presented in Table I.There \-Vere 50 males and 25 females with a mean age o f 60.7± 1.9 years.Twenty-four pat ients were admitted from various medical services, 19 fro m surgical services includ ing the ope rating room.16 w ith central nervous syste m d isease and 16 following multiple trauma.Twenty pat ie nts we re transfe rred direct ly to the intens ive care unit from periphe ral hospitals.T he mean APACHE II score at the time or inte ns ive care unit admiss ion was 22.4±0.8.Pati e nts were ve ntilated fo r a n average of 14.5±2.0days before the deve lopmen t o f cl inical pne umon ia .
No speci he clini ca l fe at ures could be identified that wou ld diffe ren tiate pat ients with pne umonia (based o n s ign ificant PBC speci men cu ltu re) from those without pneumon ia (no nsig nifica nt PBC spec ime n cu ltu re) .Speci fica lly, the re we re no differe nces in tempe rat ure.wh ite blood ce ll count, sput um purule nce or radiographic infi lt rates among the three groups (Table 2).
Bacteriological results: In all patients, sem i4uantit ati vecultures o f trachea l aspirates were pos itive (T able 3).ln ni ne of the 84 cases, the trac heal asp irate specime ns g re w multiple organi sms.The bacte ri olog ical crite ri on of more than 10 3 CFU/mL of potent iall y pathogen ic bacteria was used to idcn-176 while 96 (57.I% ) were considered non sig ni ficant.In 33 cases, both PBC specimens had s ign ificant growth (positi ve/posi tive}. in 45 cases , both had non significant growth (negati ve/negative), and culture results were di scordan t in six (posit ive/negative , n=3 o r negat ive/pos iti ve, n= 3).
Q uan titative cult ure of the PBC specimen grew Grampos itive cocc i in 13 , Gram-negative bacilli in 25 and Gramnegati ve dipl ococci in one (Table 3).
O f six patie nts w ith di scordant res ults , both PBC-A and PBC-8 grew the sa me o rganism in fou r.T he quantitative cul tures in these four cases d iffe red by I log 10 ;:ibove and below the critical cutoff of greater than I 0 1 CFU/mL.In the; re maining two cases, the fi rst PBC spec imen was s ignificant ( i(•.greater than I 0 3 CFU/m L), while the.second specimen was ste rile (T able 4).T he overall proportio n or agreement be tween the two PBC specime ns was 0.928 (Ta ble 5) with a calcul ated kappa statistic of K=0.853 (P<O.O I versus K=O.-L 95 % CL 0.692, 1.014) indicating excelle nt agreement between the tests.
In 46 (54.7% ) cases of pneumonia.antibiotics had been adm ini ste red wit hin 48 h before bronchoscopy.The use of ;intibiotics was s ign ificantly (P<0.05)associated with a nonsi g nificant PBC result (Table I) .O f the 33 patients wit h pos itive/ pos itive PBC resul ts, only I I had rece ived anti biotics, whereas 35 of the 51 patients with eithe r negative/negative, positive/negative or negative/positi ve PBC results had been on anti biotics before bronchoscopy.O f these 46 cases rece iving antibiotics.11 were pos itive/posi tive, .,Owere negat ive/negative and four were pos itive/negative.The o vera ll proportion of agreement (0.94) and the kap pa statistic were recalcu lated usi ng the 38 cases that did not receive antibiotics (T able 5).T hus, e xcluding prior antibiotics.K=0.875 (P<O.O I ve rsus K=0.4,95% CL 0.721.1.029) aho indicates exce llent agreement bet ween the two tests.

DISCUSSION
Nosocom ial pneumon ia continue.\to he a s ignificant problem in the cri tically ill patient.As a resul t or inaccurate clinic al fe atures and limi ted acceptance of invasive diagnos-1ic procedu res (22) the diagnosis o f pneumonia in mec hani-La lly ventilated pat ients is a common prohlem ( 11,23.24).O ur data concur with existing li terature ( l l .I 9,24-27) and demo nstrate that, ba •ed on clinical fea tures alone, the d iagnosis of ventilator as •oc iated pneumonia is unreli able compared with the PBC as the ' gold standard'.We have al so shown that the PBC used to obtain lower respiratory tract secre tio ns at 1he time of bronchoscopy is associated wi th a high degree o f reprod ucibility and interobserver ag reement.However. the concum:nt use of antibiotics be fore bronchoscopy may decrease the reliability of the PB C in the clinical setting .
If PBC is assumed to be th e go ld standard for diag nosing nosocomial pneumonia in critically ill ventilated patients.we would have o ve rd iagnosed ventilator associ ated pneumonia in the majority of our study patie nts on the basis of trac hea l aspirates and clinical features.Similarly, Fagon ct al ( 19) obtained PBC specimens in 162 cases of suspected venti lator associated pne umonia and obtained significant bacteri ological results in only 35%.In our study, significant PBC c ultures were obta ined in on ly 72 of 168 ( 42.9% ) spec imens.Whe n wc namincd the two groups separately.we could not identify ,my distingu ishi ng clinical features hetwecn them.T hese results n:inforce the diagnostic dilemma raced by c li nicians in the critical care setti ng when they try to es tablish the diagnosis or ventilator assoc iated pneumonia.T his is pa11icularly important where inappropriate antibiotic• therapy in pati en ts without pm'umonia may lead to added morbidity and health care costs.
Since the ro le of the PBC in clinical practice is to guide ant ibiotic therapy in patients with nosocomial pneumonia, irreproducible or unre liabl e results may h,1 vt' a de tri menta l impact on pati e nt outcome.This is rhe largest series to date examining the reproducibility of PBC result s in patien ts wit h su spec ted ventilator acquired pneumonia.We observed an overall proportion of agreement hetween the two PBC spcci -  me ns of 0.928 for all specime ns and 0.94 for spec imens in patients not treated with ant ibiotics (Table 5 ).The kappa statistic for inte robse rvcr agreement ind icates cxcclknt agreement bet ween the two PBC spec imens.Our res ults arc in cont rast to those o f o thers ( 16,2 8) in that we only observed disrnrdant results in six ( 7. 2'/o) cases.We be lieve, howe ver, that the kap pa statisti c fo r inte robserver agreement is a much mor ri go rous statist ical test (2 1) than s imply measuring the rates of disco rd ance ( 15,16), because the kappa stati sti c determines the amo unt or ag reement observed in re lation to that expected on the basis o f chance a lone.The observed d iscrepancy between the two brushes in our study may be the result of e ither sampling e rror or biological variat ion in the bacte rial popu lation ( 13).Potential so urces of sam pling error may be secondary to tec hnical facto rs (ie, cathe ter placeme nt, suction ing of tracheal sec re tions through the bronchoscopc .the use of topic al ane.thct ics) or observe r variation (ie, laboratory person ne l).In an effort to circ umvent th ese poten ti al e rro rs all bro nchoscopic proced ures were performed hy three ind iv iduals .T he PBC samp le was obtai ned only whe n endobronL•hia l pus was seen and the same technique was used for both specimens.Fu11hermorc .topica l anesthe tics we re used sparing ly and suc tioning was minimi zed throughout the proced ure .Biological vari ati on may be due to the concurrent use of an tibioti cs.presence or absence of comorbid disease and the presence o f bacteri al coloni zation.Furthe rmore.bacteriological counts may vary within till' local mclicu of the lowe r respirat o ry tract (29,30).
We o bserved a significant (P<(U)5) and cl isprnportionatcly hig h freque ncy of an ti biot ic use in patients with negati ve/negative.posi ti ve/negative or negative/pos itive PBC results (Table I).Forty-s ix of 84 pati e nts had rece ived antibiotics with in 48 h of bronc hoscopy , o f whom 35 had a nons ig nificant PBC result and I I had sig nifica nt PBC res ults (posi tive/pos iti ve ).Al though not conclus ive, these data raise the possibility that prior or co nc urre nt antibiotic therapy may be associated wit h nons ign ificant PBC resu lts.Unfortunately, in th is study we were unable to cle tem1ine whether this represents a true l'al sc-negative res ult s ince ti ssue sam ples (the ultimate gold standa rd) were not col lected .

CONCLUSIONS
Even though we observed discordant results in 7.l clo of our patients the overall reprod ucibility of the PBC results appears to be high with excellent agreement between the two tests.More important, however, we noted a significant relationshi p between the concurrent use of antibioties and nonsi gnificant PBC results.Therefore , although we be lieve that the PBC is a useful test.we emphasize that when ventil ator assoc iated pneumonia is d iagnosed.the PB C results should be considered along with other cli nical informat ion .particularly in patients receiving antibiotics.In addition, these data add to the recent debate over the use of invas ive versus noninvasive diagnostic procedures in thi s setting, and should sti mulate future research on the effects of concurre nt antibiotic use and the reliability of PBC specimens.

TABLE 1
Demographic data of patients with suspected ventilator associated pneumonia SEMPackard.Massachusetts) and cardiac nrnnitnr (SpaceLabs;

TABLE 2
Clinical features of patients with suspected ventilator associated pneumonia

TABLE 3
Bacteriology of tracheal aspirate (TA) and significant protected brush catheter (PBC) cultures a si gn ifi cant bacterial growth in lower respiratory tract secret ions obtai ned with the PBC.O f the 168 PBC specime ns.72 (42.9%} demonstrated sign ificant bacterial growth. tify