Predictive Value of Oropharyngeal Cultures for Identifying Lower Airway Bacteria in Cystic Fibrosis Patients1- 3
BONNIE W. RAMSEY, KIM R. WENTZ, ARNOLD L. SMITH, MARK RICHARDSON, JUDY WILLIAMS-WARREN, DONNA L. HEDGES, RONALD GIBSON, GREGORY J. REDDING, KAREN LENT, and KIM HARRIS
Introduction Chronic pulmonary infections, principally with Staphylococcus aureus and Pseudomonas aeruginosa, are the major cause of morbidity and mortality in patients with cystic fibrosis (CF) (1). Treatment of pulmonary infection with antibiotics, although not curative, is associated with a decrease in sputum bacterial density (2, 3), improved pulmonary function (4), and improved survival (5). Appropriate antibiotic selection requires identification of the pathogen(s) in the lower airways. In patients with CF, expectorated sputum specimens contain the same bacterial species as simultaneously collected bronchial secretions (6), lung parenchyma (7), or transtracheal aspirates (8). However, isolating bacterial pathogens from the lower airway of young, nonexpectorating patients with CF is a common and difficult problem. To avoid exposing these patients to the risk and discomfort of repeated bronchoscopic examinations, clinicians have relied on oropharyngeal cultures or used empiric antibiotic administration. Antibiotic selection has been based on microbiologic data obtained from patients of differing age and severity of illness. For example, clinicians assume that the young child with mild respiratory symptoms is colonized with S. aureus or Haemophilus influenzae, and the older child with more severe respiratory symptoms is assumed to be colonized with R aeruginosa. Oropharyngeal cultures from children without cystic fibrosis do not accurately identify lower respiratory bacterial pathogens (9). Cystic fibrosis patients, however, are more likely than unaffected individuals to be colonized with R aeruginosa in oral secretions (10, 11). Oral cultures have therefore been suggested as a way to identify lower respiratory tract pathogens in this population. The central question posed in this study is whether oropharyngeal cultures
SUMMARY Identifying lower respiratory pathogens In young, nonexpectorating cystic fibrosis (CF) patients has been problematic. Bronchial secretions are difficult to obtain, and little Is known about lower airway flora In these patients. Wecollected simultaneous bronchial and oropharyngealspeclmens In 43 CF patients In optimal respiratory status, Including both expectorating (17) and nonexpectoratlng (26) patients, to determine the predictive value of oropharyngeal cunures for Identifying lower airway pathogens. An additional goal was to characterize the lower respiratory flora of the.. patients. Predictive values were defined as the proportion of oropharyngeal culture results that accurately reflected the results of bronchial cultures. Predictive values of positive oropharyngeal eulturesln nonexpectoratlng patients were 83% (95% confidence Interval 36 to 100%) for Pseudomo· nas aeruglnosa and 91% (59 to 100%) for Staphylococcus aureus. Predictive values of negative oropharyngeal cultures were lower: 70% (48 to 86%) for Po aeruglnosa and 80% (52 to 98%) for S.aureus. A relatively high proportion of nonexpectoratlng CF patients 1888than 10 yr old had Po aeru· glnosa (11 of 24, 46%) or Klebsiella species (5 of 24, 21%) In their lower airways. The Isolation of Klebsiella was assoclatad with younger age (p = 0.03)and recent administration of antlstaphylococ0.05). Our results suggest that oropharyngeal cultures yielding Po aeluglnosa cal antibiotics (p or S. aureus are highly predictive, but such cultures lacking these organisms do not rule out the presence of these pathogens In the lower airways of CF patients.
AM REV RESPIA DIS 1991; 144:331-337
accurately reflect the presence or absence of bacterial pathogens in the lower respiratory tract of CF patients. To answer this question, we compared culture results of simultaneously collected bronchial and oropharyngeal specimens from 43 CF patients. In addition, we describe the lower respiratory tract flora of these patients. Methods
Subjects Subjects wererecruited at the CF Center, Children's Hospital and Medical Center (CHMC), Seattle, Washington, between April 1986and June 1988from all patients coming in for routine care who were also undergoing an elective surgical procedure. All subjects were diagnosed as having cysticfibrosis based on two sweat chloride values greater than 60 mEq/L and at least two clinical features (1). The procedures included the following: diagnostic bronchoscopy (18subjects), nasal polypectomy and sinus drainage (10 subjects), gastrostomy tube placement (4 subjects), central venous line placement (4 subjects), lobectomy (2 subjects), and other (6 subjects). Informed consent, approved by the CHMC Human Subjects Review Board, was obtained from the patient or the family. To minimize the risk from administration of general anesthesia and the bronchoscopic
procedure, patients were recruited at a time of optimal respiratory status. Each subject met the following criteria in the prior 48 h: (1) no oral temperature> 37° C; (2) no symptoms of coryza or sore throat; (3) normal appetite; (4) usual activity level; (5) no increase in cough or sputum production; and (6) oxygen saturation ~ 92070 by pulse oximeter. The following criteria excluded subjects from study participation: (1) a prothrombin time ~ 1.5times the control value; (2) hematocrit " 30070; (3) history of expectorating ~ 30 ml of bright red blood within a 24-h period; and (4) forced vital capacity" 50% of that predicted for height. Over the 2-yr period, 46 patients met the
(Received in original form May 14, 1990 and in revised form November 26, 1990) 1 From the Divisions of Ambulatory Pediatrics, Neonatology, and Infectious Diseases, Children's Hospital and Medical Center, and the Departments of Pediatrics and Epidemiology, University of WashingtonSchool of Medicineand School of Public Health and Community Medicine, Seattle, Washington. 1 Supported by a grant from the Cystic Fibrosis Foundation. 3 Correspondence and requests for reprints should be addressed to Bonnie W. Ramsey, Children's Hospital and Medical Center, 4800 Sand Point Way NE, Seattle, WA 98105.
332 criteria and were invited to participate; of these, 44 (950/0) were enrolled. One 17-yr-old male was excluded because of hemoptysis and a 17-yr-old female declined to participate. One subject was enrolled but not included in the analysis because her oropharyngeal specimen was misplaced,leaving 43 subjects. The number of subjects studied represents 20% (43 of 210) of the number of patients seen for routine evaluation in our clinic during the study period. After enrollment, subjects were classified as expectorators or nonexpectorators based on history obtained from their family and medical records. Expectorators were defined as having produced a sputum specimen for culture on at least one occasion in the past. There were 17 expectorators and 26 nonexpectorators. A research nurse recorded antibiotics received by the subjects for the 2 days before anesthesia. Antibiotic administration was classified in one of four categories: (1) none; (2) antistaphylococcal antibiotics (firstgeneration cephalosporins, trimethoprim-sulfa, dicloxacillin, erythromycin, or ampicillin with clavulanic acid); (3) anti-pseudomonal antibiotics (aminoglycosides,third-generation cephalosporin, ureidopenicillins, or fluoroquinolones); or (4) both antistaphylococcal and antipseudomonal antibiotics.
Collection of Specimens Within 4 h before receiving sedation and anesthesia, the research nurse recorded the subject's resting respiratory rate, pulse, and temperature, as well as height and weight. A Shwachman-Kulczyki (12) clinical score was determined by one of the authors. A chest x-ray wasobtained (if none was availablewithin the previous 2 months) for determination of a Brasfield (13) score. The clinical and radiologic scores were determined by Ramsey without knowledge of the microbiologic results. Whole blood was collected for measurements of white blood count with differential and serum for measurement of immunoglobulin G anti-pseudomonas exotoxin A titers. The research nurse collected the throat cultures within 1 h before bronchoscopy and at least 1 h following chest physiotherapy and postural drainage. The procedure for collecting specimens was as follows. Twocotton-tipped swabs wereplaced simultaneously in the posterior oropharynx swabbing the tonsillar pillars and posterior wall. Each swab was placed immediately in 1 ml phosphate-buffered saline (PBS) and delivered to the infectious disease laboratory for culture (see Laboratory procedures). Immediately following collection of throat cultures, those subjects able to expectorate sputum were asked to provide a sputum specimen. A minimum of 0.5 g sputum was collected in a sterile cup for quantitative bacterial culture (14). All bronchoscopic examinations were performed under the supervision of Dr. Mark Richardson. A ventilating, rigid bronchoscope was used in all 36 subjects receiving general anesthesia so that the anesthesiologist had ac-
RAMSEY, WENTZ, SMITH, RICHARDSON, WILLlAMS·WARREN, HEDGES, GIBSON, REDDING, LENT, AND HARRIS
cess to the airway for administration of oxygen and positive pressure. Immediately following induction with general anesthesia, the scope was introduced through the oropharynx and vocal cords beforeintubation. Those eight subjects requiring only sedation with midazolam underwent bronchoscopic examination with a gas-sterilized, flexible Olympus BC 3FP bronchoscope (Olympus Corp., Sunnyvale,CA) introduced through the nasopharynx and vocal cords. In all subjects, 2% xylocaine was applied to the vocal cords. Flexible and rigid bronchoscopes that can be introduced into a pediatric airway are too narrow to allow passage of a double-lumen catheter. To minimize contamination with upper airway flora, no suctioning occurred until after the bronchoscope had been placed in the right lower lobe. The sample was collected from the right lower lobe in an area of purulent secretions. After instillation (via sterile tubing) of 5 to 10ml nonbacteriostatic normal saline, the sample was retrieved with a separate sterilesuction apparatus into a Luki trap. This specimen was sent to the CHMC infectious disease laboratory for quantitative bacterial culture (14).
was performed as follows. Each micro titer plate well was coated overnight with 10 ng exotoxin A in carbonate buffer, pH 9.6. Plates were washed six times with phosphatebuffered saline with 5% Tween'"(PBST; Sigma Chemical, St. Louis, MO) between each step. Nonfat dry milk (5%) solution was added to block nonspecific binding sites. Human sera, diluted serially in PBST from 1:100to 1:12,800, were incubated for 1 h at 37° C. Horseradish peroxidase-conjugated goat antihuman IgG was added and incubated for 30 min at 37° C. Peroxidase substrate, phenylene diamine, and hydrogenperoxide werethen added and the reaction was allowed to continue for 10 min at room temperature in the dark. The reaction was stopped by adding 50 III of 2.5 M sulfuric acid. Optical densities of wells were read at 492 nm with a Titertek Multiskans spectrophotometer (Eflab Oy Laboratories, Helsinki, Finland). All serum samples were simultaneously analyzed in triplicate and a mean value determined. The first serum dilution at which the absorbance was less than 0.500 OD (optical density) was recorded for each patient. The negative control was a pooled sample of sera from healthy children obtained from the clinLaboratory Procedures ical chemistry laboratories at Children's HosBacterial culture technique. Specimens col- pital and Medical Center. The positive conlected from the oropharynx (throat) werecul- trol samples were from two cystic fibrosis patured using two techniques: qualitative and tients with documented R aeruginosa .quantitative throat cultures. The qualitative colonization of greater than lO-yr duration. culture method is as follows: throat swabs were Analysis. Sensitivity,specificity, and predicstreaked directly onto three culture media: tivevaluesof positive and negativeresults were MacConkey's agar, sheep blood agar, and calculated for quantitative and qualitative chocolate agar. They were incubated at 37° throat cultures using the bronchoscopy specC and read at 24 and 48 h. Standard microbi- imen culture as the primary standard. Sensiologic techniques (15) were used to identify tivity is defined as the percentage of subjects R aeruginosa, S. aureus, H. influenzae, and with positive bronchial cultures whose oroKlebsiella species. pharyngeal cultures yield the same pathogen. The quantitative culture technique was as Specificity is the percentage of subjects with follows (14).The throat swabs, placed initial- negative culture results whose oropharyngeal ly in phosphate-buffered saline, were diluted cultures lack the same pathogen. Predictive 1:10, 1:1,000, and 1:10,000 with sterile PBS value is defined as the percentage of positive containing 0.1070 gelatin. From each dilution, or negative oropharyngeal culture results that 0.1 ml was spread on the following six solid correctly reflects the results of the bronchial media using a sterile glass rod: (1) mannitol culture. The 95% confidence intervals were salt; (2) Cetrimides (Difco, Detroit, MI); (3) calculated around the values for sensitivity, MacConkey's; (4) brain-heart infusion with specificity, and predictive values using Wil300 ug/ml of bacitracin; (5) sheep blood agar son's approximate method for n > 15 and with 10ug/ml each of neomycin and gentami- Fisher's exact method for n ~ 15 (18, 19). cin (NG); and (6) OFPBL (oxidative fermenPearson's correlation coefficient was caltative basal media containing 300,000 U/L culated to compare the bacterial density in of polymixin, 200 U/L of bacitracin, and 1% bronchial cultures with that in the throat cullactose). The plates were incubated at 37° C tures. Logistic regression (20) was used to de(NG and bacitracin incubated in an anaero- termine whether information about clinical bic chamber) and read at 24 and 48 h. All characteristics could be added to throat culorganisms wereidentified using standard tech- ture results to improve their ability to predict niques (15). the presence of lower airway pathogens. The Serology. An ELISA (16, 17) was used to Statistical Package for the Social Sciences determine the study subject's antibody titers (SPSS), the Epidemiologic Graphics Estimaagainst R aeruginosa exotoxin A. The sera tion and Testing Program (EGRET, Statiswere stored at - 20° C and thawed immedi- tics and Epidemiology ResearchCorporation, ately before the assay was performed. The R 909NE 43rd St., Suite 310, Seattle, WA98105), aeruginosa exotoxin A was obtained from and the Rothman/Boice epidemiologic analMark Strom, M.S., Department of Microbiysis calculation programs (19) were used for ology, University of Washington School of computer analysis of data. Medicine. Clinical characteristics of patients were The indirect microtiter ELISA method (17) compared between patients who could and
PREDICTIVE VAWE OF OROPHARYNGEAL CULTURES
333 TABLE 1
SELECTED CHARACTERISTICS OF EXPECTORATING AND NONEXPECTORATING PATIENTS
Age, yr Brasfield score" Schwachman-Kulczyki score
Expectorators (n = 17)
Nonexpectorators (n = 26)
Mean ± SEM
Mean ± SEM
13.8 16.5 75.8
6.6 4.2 15.7
4.5 18.7 71.3
< 0.001 < 0.01
4.2 3.7 18.2
Males Antibiotic therapy None Antistaphylococcal Antipseudomonal Both P. aeruginosa in bronchial secretions
5 2 8 2 12
29 12 47 12 70
15 9 1 1 11
58 35 4 4 42
NS NS < 0.01 NS NS
• Covaried for age. Not significant, p > 0.05.
TABLE 2 SENSITIVITY AND SPECIFICITY OF THROAT CULTURES BY ORGANISM·
Quantitative cultures (n = 43) Expectorators Nonexpectorators Qualitative cultures (n = 41)t Expectorators Nonexpectorators
9/12 75% (43-95)
80% 4/5 (28-100)
4/7 57% (18-90)
9/10 90% (56-100)
4/6 67% (22-96)
10/11 91% (59-100)
5/11 46% (17-77)
14/15 93% (68-100)
10/13 77% (46-95)
12/13 92% (77-100)
5/10 50% (19-81)
13/16 81% (57-93)
9/12 75% (43-95)
80% 4/5 (28-100)
217 29% (4-71)
10/10 100% (74-100)
0/6 0% (0-39)
11/11 100% (3-100)
5/9 56% (21-86)
14/15 93% (68-100)
7/13 54% (25-81)
11/11 100% (76-100)
1/9 11% (0-48)
13/15 87% (6Q-98)
• Values represented as n, 0lb; 95% confidence interval in parentheses. were not performed on two patients.
t Routine throat cultures
could not expectorate and between patients with and without specific pathogens isolated from the bronchial specimen. Because many patients were colonized by more than one pathogen and the ability to expectorate was related to age, copathogens and age were covaried when making the comparisons. Groups of patients with different pathogens could not becompared with each other because these groups were not mutually exclusive, since many patients had copathogens. Differences were tested for statistical significance using analysis of variance (ANOVA)for continuous data and a chi-square test with Yate's continuity correction for proportions (18).
Patient Characteristics Subjects ranging in age from 4 months to 25 yr, with a mean age of 8.2 ± 6.9 (SEM) yr. There were 20 males and 23 females. Illnessseverityscores wereas follows: Shwachman-Kulczyki (S-K) score range 34to 98, mean 73.0 ± 17.2(SEM); Brasfield chest roentgenogram range 8 to 23, mean 17.8 ± 4.0 (SEM). Within 48 h before bronchoscopy, 20 subjects had received no antibiotics, 11 received anti staphylococcal antibiotics, 9 received
antipseudomonal antibiotics, and 3 received both types of antibiotics. The clinical characteristics of nonexpectorating and expectorating patients are described in table 1.Patients who did not expectorate were younger and had higher Brasfield scores (comparison adjusted for age). They were also less likely to have received antipseudomonal antibiotics in the 48 h before bronchoscopy.
Sensitivity and Specificity of Oropharyngeal Cultures For each study subject, results of oropharyngeal cultures were compared with results from concurrently obtained bronchial secretions. Sensitivity and specificity (using the bronchial culture as the standard) of oropharyngeal cultures are presented in table 2 for each pathogen, in expectorators and nonexpectorators and by type of oropharyngeal culture (quantitative versus qualitative). Sensitivity, specificity, and predictive value were calculated only for organisms isolated in at least 7 patients, which excluded Pseudomonas and Klebsiella species. In general, oropharyngeal cultures had
higher specificity than sensitivity. In nonexpectorators, specificityof quantitative cultures for R aeruginosa was 93070 (14 of 15), for S. aureus 92070 (12 of 13), and for H. influenzae 82070 (13 of 16). In nine subjects, a bacterial pathogen was isolated from the upper airway, but not from bronchial secretions (i.e., a false positive culture). These isolates included two R aeruginosa, two S. aureus, four H. influenzae, and one Klebsiellaspecies. Sensitivity of quantitative cultures in nonexpectorators for R aeruginosa was only 46% (5 of 11), for S. aureus 77070 (10 of 13), and for H. influenzae 50070 (5 of 10). A total of 65 bacterial pathogens were isolated from bronchial secretions, of which 42 (65%) were correctly identified by quantitative oropharyngeal culture. The remaining 23 pathogens not identified by oropharyngeal cultures included nine R aeruginosa, six S. aureus, seven H. influenzae, and one Klebsiella species. Quantitative oropharyngeal cultures wereneither more sensitivenor more specific than qualitative oropharyngeal cultures. Sensitivity and specificity of
RAMSEY, WENTZ, SMITH, RICHARDSON, WILLIAMS-WARREN, HEDGES, GIBSON, REDDING, LENT, AND HARRIS
TABLE 3 PREDICTIVE VALUES OF POSITIVE AND NEGATIVE THROAT CULTURES BY ORGANISM·
+ Quantitative cultures (n = 43) Expeetorators
9/10 90% (56-100)
Nonexpectorators Qualitative cultures (n = 41)t Expectorators Nonexpectorators
4/5 80% (28-100)
9/12 75% (43-95)
4/5 80% (28-100)
10/12 83% (52-98)
83% 5/6 (36-100)
14/20 70% (48-86)
10/11 91% (59-100)
12/15 80% (52-96)
5/8 63% (25-92)
13/18 72% (49-88)
9/10 90% (56-100)
4/7 57% (18-90)
2/2 100% (22-100)
10/15 67% (38-88)
11/17 65% (41-83)
83% 5/6 (36-100)
14/18 78% (55-91)
11/17 65% (41-83)
1/3 33% (1-91)
13/21 62% (41-79)
• Values represented as n, %; 95% confidence interval in parentheses. Routine cultures were not performed on two patients.
TABLE 4 DENSITY· OF BACTERIA IN UPPER AND LOWER AIRWAY SPECIMENS
Bronchial specimen Oropharyngeal specimen Sputum specimen
101-109 103-108 108-107
10 4 103.8 1Q3·4
101-107 101-107 1~-107
r Correlationt between oropharyngeal and bronchial specimens Correlation between sputum and bronchial specimens
• Valuesexpressed in colony-forming units/gram sputum.
t Correlationexpressedas Pearson's correlationcoefficient (r).Significancelevelof all r valueswasp < 0.001,exceptfor H. influenzae isolates, which were p < 0.01.
oropharyngeal cultures did not vary between expectorating and nonexpectorating patients.
Predictive Value of Oropharyngeal Culture Results The predictive values of positive and negative oropharyngeal culture results are presented in table 3. For quantitative oropharyngeal cultures in nonexpectorating patients, the predictive values for positive results were 83070 for R aeruginosa, 91070 for S. aureus, and 63070 for H. influenzae. In other words, a report of R aeruginosa isolated from an oropharyngeal culture had an 83070 chance of being associated with the isolation of R aeruginosa from a simultaneous bronchial specimen. The predictive values for negative results in quantitative oropharyngeal cultures were 70% for R aeruginosa, 80% for S. aureus, and 72070 for H. influenzae. Predictive values of oropharyngeal cultures did not vary between the expec-
torators and nonexpectorators or by quantitative versus qualitative culture technique. To determine whether clinical correlates could be used to improve the predictive ability of oropharyngeal cultures to correctly identify R aeruginosa, the following parameters wereincluded individually and in combination, in a logistic regression model: respiratory rate, age, Shwachman-Kulczyki score, Brasfield score, white blood count, immature neutrophil count, and serum anti-pseudomonas exotoxin A antibody titer, in addition to oropharyngeal culture results. Of the eight variables listed, only two were significantly related to the presence of R aeruginosa in bronchial secretions: (1) oropharyngeal culture and (2) serum anti-pseudomonas exotoxin A antibody titer. The serum anti-exotoxin A titer was a good predictor of R aeruginosa colonization when included alone in the model. However, it did not improve the predictive value of oropharyngeal cultures when
both were included in the regression model.
Comparison of Bacterial Density between Lower and Upper A irway Secretions Bacterial colony densities in bronchial, oropharyngeal, and sputum specimens are presented in table 4. Median colony density tended to be lower in oropharyngeal isolates than in either expectorated sputum or bronchial isolates. There was a higher correlation between the bacterial densities of sputum and bronchial specimens than between oropharyngeal and bronchial specimens. Bacterial Isolates from Lower Respiratory Secretions A summary of the bacterial pathogens isolated from bronchial secretions of expectorating and nonexpectorating subjects is depicted in figure 1.Ofthe 43 subjects completing the study, 40 had pathogens isolated from the lower airway
PREDICTIVE VAWE OF OROPHARYNGEAL CULTURES
nontypable H. influenzae. 1\vo children had K. pneumoniae as the only pathogen, and a third child was also colonized with P. aeruginosa. Discussion no pathogens 1
Fig. 1. Pathogens presentin bronchial specimensof expectorating and nonexpeetorating cystic fibrosis patients:pa, P. aeruginosa, sa, S. aureus; hf, H. influenzae; kb, Klebsiella species. Top = expeetorators (n = 17); bottom = nonexpectorators (n = 26).
no pathogens 2
secretions. The three subjects not yielding jects colonized with S. aureus had highpositive cultures included two expectora- er S-K scores and weremore likely to have tors, a 21-yr-old female and a 25-yr-old received no antibiotics in the 48 h before male, and a nonexpectorating 5-month- bronchoscopy than patients without old male infant. All three individuals re- S. aureus in bronchial secretions. Subceived 7 to 10 days of intravenous anti- jects colonized with H. influenzae had pseudomonal antibiotics before the bron- higher Brasfield scores and were less likechoscopic procedure. Although their ly to have received antibiotic therapy in bronchial secretions appeared purulent, the 48 h before enrollment than patients no pathogens wereisolated. P. aerugino- without H. influenzae. sa was previouslyisolated in sputum samThe six subjects whose lower airway ples from the two expectorating subjects. secretions yielded Klebsiella species inOropharyngeal cultures from the 21-yr- . eluded three males and three females. old female grew P. aeruginosa; from the Their ages ranged from 5 to 60 months 25-yr-old male, mixed mouth flora; and (mean 2.3 ± 2.0 yr SEM). They were sigfrom the male infant, Klebsiella pneu- nificantly younger (p = 0.03) than submoniae. These 3 subjects were included jects not colonized with Klebsiella (mean in the analysis. age 9.2 ± 7.0 yr SEM). These six chilThe clinical characteristics of patients dren had S-K scores ranging from 35 to with and without bronchial isolates of 88 (mean 59.5 ± 21.8SEM). Their Braseach pathogen are described in table 5. field scores ranged from 10 to 22 (mean Because many subjects were colonized 18.0 ± 4.4 SEM). Four of the six chilwith multiple pathogens, the analyses of dren were receiving antistaphylococcal age, S-K score, and Brasfield score were antibiotics at the time of bronchoscopy, covaried for copathogens. a frequency higher (p = 0.05) than in the Subjects colonized with P. aeruginosa subjects not colonized with Klebsiella. were more likely to have high titers of se- Three patients were colonized with Klebrum anti-exotoxin A and had lower S-K siella oxytoca and three with K. pneuscores than subjects without P. aerugi- moniae. The three children colonized nosa in their bronchial secretions. Sub- with K. oxytocawerealso colonized with
The present study compared bacterial pathogens collected simultaneously from the lower airways and oropharynx of patients with cystic fibrosis. We are encouraged by the high positive predictive value of an oropharyngeal culture yielding P. aeruginosa (830,10) or S. aureus (910,10). These predictive values reflect a low frequency of false positives (i,e., high specificity) and the high prevalence of these two bacterial species in the lower airways of our subjects. However,the predictivevalues of negative oropharyngeal culture results were discouraging, as a result of high frequency of false negatives. Our results suggest that the clinician can be confident that a child with an oropharyngeal culture yielding P. aeruglnosa or S. aureus has the pathogen in the lower respiratory tract, whereas a negative throat culture does not rule out the presence of these pathogens. The high frequency of oropharyngeal cultures not yielding pathogens isolated from the lower airway was not related to prior antibiotic administration: sensitivity for P. aeruginosa was highest in patients treated with antipseudomonal antibiotics. Three previous studies (6,8,21) compared simultaneous cultures from the upper and lower airway in CF patients. Iacocca and coworkers in 1963 found that oropharyngeal culture results were concordant with sputum cultures for S. aureus in 28 of 29 expectorating CF patients and for P. aeruginosa in 16 of 18expectorating CF patients (6). Huang and coworkers in 1961 noted that throat and bronchial cultures yielded pathogens in over 900,10 of 28 CF patients aged 3 months to 16 yr; however, they did not examine concordance between oropharyngeal and bronchial cultures in individual patients (21). Baren and Cordiar in 1973compared 14 transtracheal aspirate cultures with oropharyngeal cultures in a group of young nonexpectorating CF patients (8). Only 3 of 14oropharyngeal cultures were concordant with aspirate cultures; the remaining 11 oropharyngeal cultures were sterile. The studies by Baren and Cordiar and Iacocca's group include up to 10cultures per patient, making their results difficult to compare with our study design, which includes only in-
RAMSEY, WENTZ, SMITH, RICHARDSON, WILLIAMS-WARREN, HEDGES, GIBSON, REDDING, LENT, AND HARRIS
TABLE 5 CLINICAL CHARACTERISTICS OF PATIENTS WITH AND WITHOUT SELECTED PATHOGENS IN BRONCHIAL SECRETIONS
With (n = 23)
Without (n = 20)
With (n = 20)
Without (n = 23)
Sehwaehman-Kulczyki Seore,n mean (SEM)
22 35 13
69 19 6 6
33 30 30 7
17 67 17 0
51 19 22 8
Brasfield seore,n mean (SEM)
Antibiotic therapy, % None Antistaphyloeoeeal Antipseudomonal Both
30 30 4
60 20 10 10
75 15 10 0
Exotoxin A titer' > 1:800, %
With = 16)
Without (n = 27)
With (n = 6)
Without (n = 37)
• p < 0.05. t p < 0.001.
Covaried for copathogens.
§ Standard error of the mean.
UCovaried for copathogens and age. • Based on 35 patients; 8 patients' specimens were missing.
dependent observations (i.e., one culture per patient). Certain limitations to our study should be recognized. The number of nonexpectorating patients studied was relatively small (n = 26), with only 11 having R aeruginosa colonization of their lower airways, so the confidence intervals around our estimates of sensitivity, specificity, and predictive value are wide. Second, for the safety of the patients, all subjects were recruited while at optimal respiratory status. The relationship between colonization of the lower and upper airways may not be the same as in acutely ill children, for whom oropharyngeal cultures would be used to identify lower airway pathogens. Acutely ill children may be more likely than well children to have both upper and lower airway colonization; hence, our estimates of positive predictive value may underestimate the true values for oropharyngeal cultures. The finding of many false negative oropharyngeal cultures raises the question of whether our technique for sampling secretions from the posterior oropharynx was optimal. We made every effort to maximize the yield of oropharyngeal cultures. First, wechose the posterior pharynx based upon Huang's (21) data that pharyngeal cultures were more likely to yield positive pseudomonas cultures than nasopharyngeal swabs. Second, weswabbed both the tonsils and posterior phar-
ynx. This technique was found to yield the highest number of positive cultures for identification of Group A streptococcal (GAS) pharyngitis (22). Third, we performed duplicate swabbing of the oropharynx to compare two culture techniques (quantitative versus qualitative). Several GAS pharyngitis studies have found the discordance rate between duplicate cultures to be as high as 24070 (23) and that a single culture may miss 8.70/0 of positive cultures (24). There was no discordance between the two types of oropharyngeal cultures in this study for any pathogen, suggesting that optimal collection techniques were performed. The size of pediatric fiberoptic and rigid bronchoscopes does not permit the introduction of a double-lumen catheter into the side channel, thereby introducing the possible risk of contamination of lower airway secretions during passage through the scope. This limitation of fiberoptic bronchoscopic collection technique has been recognized (25-27). Bartlett and colleagues (25) noted in their study of 16 healthy volunteers that contaminated fiberoptic aspirates averaged five bacterial species per aspirate, with a predominance of a-hemolytic streptococcus, Neisseria species, and peptostreptococcus. We did not find any lower airway specimens with these mouth flora. All samples had three or fewer bacterial species. Our results indicate that among the
nonexpectorating patients less than 10yr of age, 460/0 (11 of 24) have lower airway colonization with R aeruginosa. We could not find an appropriate agematched comparison group in the literature, so we compared our subjects with CF children (ages 0 to 10yr) included in the Western Consortium (Washington, California, Arizona, Michigan, New Mexico, and Utah) data registry.Of 1,842 children, 1,412 (77%) had respiratory culture data reported to the 1987 National Cystic Fibrosis Data Registry: 1,069 from sputum specimens and 343 from oropharynx specimens. Whether these cultures were collected during an acute illness or well visit was not reported. Of the patients who produced sputum, 720/0 (771 of 1,069) werecolonized with R aeruginosa; 32% (111 of 343) of patients with oropharyngeal specimens werecolonized with R aeruginosa. The frequency of R aeruginosa in our study appears to be within the same range as other Western CF centers in the United States. The additional time and expense of quantitative, as opposed to routine qualitative, throat cultures appear not to be warranted. We found no improvement in sensitivity, specificity,or predictivevalues for quantitative over qualitative cultures. Although Klebsiella species have been reported in the CF literature (28-32), this genus has received much less attention than S. aureus, R aeruginosa, and H. influenzae. In 1968, a British study (28)
PREDICTIVE VAWE OF OROPHARYNGEAL CULTURES
References found that 10% (3 of 30) of CF patients tested had precipitins directed against . 1. Wood RE, Boat TF, Doershuk C. Cystic Fibrosis Klebsiella species. These individuals had (state of the art). Am Rev Respir Dis 1976; 113: mucoid K. ozoonea isolated from spu- 833-78. 2. McLaughlin FJ, Matthews WJ, Strieder OJ, et tum. The Danish center in 1980 (31) and al. Clinical and bacteriological response to three Brussels in 1973 (8) reported Klebsiella antibiotic regimes for acute exacerbation of cystic in 5 to 10070 of sputum samples. The fibrosis: ticarcillin-tobramycin, azlocillin-tobraToronto center (32) reported a decline in mycin, azlocillin-placebo. J Infect Dis 1983; 147: the rate of infection with Klebsiella spe- 559-67. 3. Smith AL, Redding G, Doershuk C, et al. Spucies from 40070 per year in 1970 to less tum changes associated with therapy for endobronthan 10070 per year in 1981 in clinic pa- chial exacerbation in cystic fibrosis. J Pediatr 1988; tients of all ages, using data from spu- 112:547-54. tum and nasopharyngeal samples. The 4. Redding GJ, Restuccia R, Cotton EK, Brooks JG. Serial changes in pulmonary functions in chilreason for this decline is not known but dren hospitalized with cysticfibrosis. Am Rev Respir is associated with an increase in the Dis 1982; 126:31-6. rate of isolation of R aeruginosa and 5. Szaff M, Hoiby N, Flensborg EW. Frequent
R cepacia. The role of Klebsiella as a pathogen in CF has not been defined. Lloyd-Still and coworkers (30) noted that 11 of 17 infants with severe respiratory disease had K.pneumoniae in tracheal aspirates obtained during hospitalization for respiratory distress. Of these 11 patients, 6 were also colonized with R aeruginosa and 5 with S. aureus. The mortality rate among the K.pneumoniae-colonized infants was over 50070 (6 of 11). In the present study, four of these six patients with Klebsiella receivedantistaphylococcal antibiotics, an association also noted by Lloyd-Still and coworkers (30). In summary, oropharyngeal cultures yielding R aeruginosa or S. aureus are highly predictive of the presence of those pathogens in lower airways of patients with CF. Negative oropharyngeal cultures do not rule out the presence of such pathogens in the lower airways. These CF patients' lower airways were colonized with P. aeruginosa at a young age, frequently preceding the ability to expectorate. Young patients may also be colonized with Klebsiella species, at least transiently. Acknowledgment The authors thank Ivan Harwood, M.D. and Michael Light, M.D. for their contribution of Western Consortium data, and Steve Moseley, Ph.D. for his helpful criticism.
antibiotic therapy improves survival of cystic fibrosis patients with chronic Pseudomonas aeruginosa infection. Acta Paediatr Scand 1983; 72:651-7. 6. Iacocca VF, Sibinga M, Barbero G. Respiratory tract bacteriology in cystic fibrosis. Am J Dis Child 1963; 106:115-24. 7. Thomassen MJ, Klinger JD, Badger SJ, van Heeckeren OW, Stern RC. Cultures of thoracotomy specimens confirm usefulness of sputum cultures in cystic fibrosis. J Pediatr 1984; 104:352-6. 8. Baren 0, Cordiar N. Usefulness oftranstracheal puncture in the bacterial diagnosis of lung infections in children. Helv Paediatr Acta 1973;28:391-9. 9. Silverman M, Stratton 0, Diallo A, Egler LJ. Diagnosis of acute bacterial pneumonia in Nigerian children. Arch Dis Child 1977; 52:925-31. 10. Lindemann RA, Newman MG, Kaufman AK, Le TV. Oral colonization and susceptibility testing of Pseudomonas aeruginosa. Oral isolates from cystic fibrosis patients. J Dent Res 1985; 64:54-7. 11. Woods DE, Bass JA, Johanson WO Jr, Strauss DC. Role of adherence in the pathogenesis of Pseudomonasaeruginosa lung infection in cystic fibrosis patients. Infect Immun 1980; 30:694-9. 12. Shwachman H, Kulczyki LL. A report of one hundred and five patients with cystic fibrosis of the pancreas studied over a five- to fourteen-year period. Am J Dis Child 1958; 96:6-15. 13. Brasfield 0, Hicks G, Soong S, Tiller RE. The chest roentgenogram in cystic fibrosis: a new scoring system. Pediatrics 1979; 63:24-9. 14. Wong K, Roberts MC, Owens L, Fife M, Smith AL. Selective media for the quantification of bacteria in cystic fibrosis sputum. J Med Microbiol 1984; 17:113-9. 15. Treagan L, Pulliam L, eds. Medical microbiology laboratory procedures. Philadelphia: W. B. Saunders, 1982; 5-9, 23-42. 16. Jaffe KS, Robinson DL, Franz MN, Warren RL. Detection by enzyme-linked immunosorbent assays of antibody specific for Pseudomonas proteases and exotoxin A in sera from cystic fibrosis
patients. J Clin Microbiol 1982; 15:1054-8. 17. Vollmer A, Bidwell DE, Bartlett A. The enzyme linked immunosorbent assay (ELISA): a guide with abstracts of microplate applications. Alexandria, VA:Dynatech Laboratories, Inc, 1979;23-40. 18. Fleiss JL. Statistical methods for rates and proportions. 2nd ed. New York: John Wiley and Sons, 1981. 19. Rothman KJ, Boice JD. Epidemiologic analysis with a programmable calculator. Boston: Epidemiology Resources, Inc., 1982. 20. Breslow N, Day N. Statistical methods in cancer research. Vol. 1. The analysis of case-control studies. Lyon, France: International Agency for Research on Cancer, 1980. IARC Scientific Publications No. 32. 21. Huang NN, Van Loon EL, Sheng KT. The flora of the respiratory tract of patients with cysticfibrosis of the pancreas. J Pediatr 1961; 59:512-21. 22. Brien JH, Bass JW. Streptococcal pharyngitis: optimal site of throat culture. J Pediatr 1985; 106:781-3. 23. Gerber MA. Diagnosis of pharyngitis: methodology of throat culture. In: Shulman ST, ed. Pharyngitis: management in an era of declining rheumatic fever. New York: Praeger, 1984; 61-72. 24. Peter G. The child with group A streptococcal pharyngitis. In: S. Aronoff, ed. Advances in pediatric infectious diseases. Chicago, Year Book Medical Publisher, 1986; 1-18. 25. Bartlett JO, Alexander J, Mayhew J, SullivanSigler N, Sherwood O. Should fiberoptic bronchoscopy aspirates be cultured? Am Rev Respir Dis 1976; 114:73-8. 26. Teague RB, Wallace RJ, Awe RJ. The use of quantitative sterile brush culture and gram stain analysis in the diagnosis of lower respiratory tract infection. Chest 1981; 79:157-61. 27. Winterbauer RH, Hutchinson JF, Reinhardt ON, et al: The use of quantitative cultures and antibody coating of bacteria to diagnose bacterial pneumonia by fiberoptic bronchoscopy. Am Rev Respir Dis 1983; 128:98-103. 28. Burns MW. Precipitins to Klebsiella and other enterobacteria in the serum of patients with chronic respiratory disorders. Lancet 1968; 1:383-5. 29. Maj JR, Herrick NC, Thompson D. Bacterial infection in cystic fibrosis. Arch Dis Child 1972; 47:908-13. 30. Lloyd-Still JD, Khaw K-T, Shwachman H. Severe respiratory disease in infants with cystic fibrosis. Pediatrics 1974; 53:678-82. 31. Hoiby N. Microbiology of lung infections in cystic fibrosis patients. Acta Paediatr Scand (Suppl) 1982; 301:33-54. 32. Corey M, Allison L, Prober C, Levison H. Sputum bacteriology in patients with cystic fibrosis in a Toronto hospital during 1970-1981. J Infect Dis 1984; 149:283.