APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Aug. 1992, p. 2543-25510099-2240/92/082543-09$02.00/0Copyright © 1992, American Society for Microbiology

Comparison of Bioaerosol Sampling Methods in BarnsHousing Swine

PETER S. THORNE,* MARGARET S. KIEKHAEFER, PAUL WHITTEN,AND KELLEY J. DONHAM

Department ofPreventive Medicine and Environmental Health, The Institute for AgriculturalMedicine and Occupational Health, University of Iowa, Iowa City, Iowa 52242

Received 15 November 1991/Accepted 15 May 1992

The air in livestock buildings contains bioaerosol levels that are sufficiently high to cause adverse healtheffects in animals and workers. These bioaerosols are complex mixtures of live and dead microorganisms andtheir products as well as other aeroallergens. The effectiveness of sampling methods used for quantifying thevery high concentrations of microorganisms in these environments has not been well studied. To facilitate anaccurate assessment of respiratory hazards from viable organisms in agricultural environments, threebioaerosol sampling methods were investigated: the Andersen microbial sampler method (AMS), the all-glassimpinger method (AGI), and the Nuclepore filtration-elution method (NFE). These methods were studied in aparallel fashion in 24 swine confinement buildings. Measurements were taken in two seasons with three typesof culture media in duplicate to assess total bacteria, gram-negative enteric bacteria, and total fungi. Methodswere analyzed for the proportion of samples yielding data within the limits of detection, intraclass reliability,and correlation between methods. For sampling viable bacteria, the AMS had a poor data yield because ofoverloading and demonstrated weak correlation with the AGI. Conversely, the AGI and NFE gave sufficientnumbers of valid data points (90%o), yielded high intraclass reliabilities (at 0.92), and were highly correlatedwith each other (r = 0.86). The AGI and the NFE were suitable methods for assessing bacteria in thisenvironment, but the AMS was not. The AMS was the only method that consistently recovered enteric bacteria(73% data yield). For sampling fungi, the AGI and AMS both yielded sufficient data and all three methodsdemonstrated high intraclass reliability. The AGI and AMS correlated moderately with each other, but eachcorrelated well with the NFE. However, the AGI measured significantly higher airborne fungal concentrationsthan did the AMS. Thus, the AGI was the preferred sampling method for viable fungi. Selection of an

appropriate method depends on the purpose of sampling, expected bioaerosol concentrations, and environ-mental conditions. In this study, the AGI was best for total bacteria and fungi and the AMS was preferred forsampling enteric organisms.

Agricultural work often involves life-long exposure todusts at concentrations that frequently exceed 5 mg/mr3 (8).These high dust exposures occur especially during harvest-ing, grain and silage handling, and work in livestock andpoultry buildings. Since agricultural dusts contain animalproteins or waste products, pollens and plant fragments,arthropods, and environmental and fecal microorganisms,they have high biological activity and are therefore morehazardous than common nuisance dust (8). The microbialcomponent contributes significantly to the pulmonary dis-eases associated with inhalation of agricultural dusts (23,36). In the past 8 years, a number of investigators haveattempted to quantify levels of airborne microorganisms inagricultural environments (5, 7, 9, 10, 12, 17, 21, 31, 33, 46,48). One result has been the realization that methods devel-oped for bioaerosol sampling in other settings do not neces-sarily work well in the agricultural environment. Thus, thisstudy was undertaken in conjunction with an exposureassessment of swine confinement workers to systematicallyevaluate the three commonly used bioaerosol samplingmethods described below.AGI. Various devices for assessing airborne concentra-

tions of viable microorganisms have been developed andtested over the past 70 years (2, 3, 16, 24, 35, 37, 41, 43). Theall-glass impinger method (AGI) was introduced in the 1920s

* Corresponding author.

(16) and modified in the 1950s (35) and continues to be widelyused. This sampling device is inexpensive and easy to useand allows for a considerable range of ambient concentra-tions when serial plating of the solution is performed. Inaddition, a single sampling solution can be plated onto avariety of media plus subjected to chemical tests or toxinanalyses. Limitations of the AGI were noted as loss ofviability due to impingement of the microorganisms (45, 49,50) and loss by re-entrainment in the exhaust flow caused byhydrophobicity, agitation, and misting within the impinger(38). In addition, some organisms may suffer from the effectsof sudden hydration upon impingement or osmotic shock (6).Viable sampling must be done for short periods of time, lessthan about 30 min, and preferably on ice to impede replica-tion of microorganisms.AMS. The six-stage Andersen microbial sampler method

(AMS) was developed to allow simultaneous sizing andcounting of viable microorganisms (2). The AMS was stud-ied in detail by a number of researchers, and this work led toimprovements in the design (34) and in colony enumerationtechniques (25, 29, 42). In addition, Solomon (47) and Joneset al. (20) tested the use of the six-stage AMS in alternateconfigurations with just the last stage. Today six-stage,two-stage, and single-stage AMSs are sold commercially andwidely used. The AMS has the advantage of collectiondirectly onto culture media for incubation and analysis withno further dilution or plating. The main problem associatedwith the AMS is that one can only sample for viable

2543

Vol. 58, No. 8

on May 20, 2020 by guest

http://aem.asm

.org/D

ownloaded from

APPL. ENVIRON. MICROBIOL.

microbes, and many of the organisms may be nonviable butstill have harmful effects when inhaled. Another majorproblem is the rapid overloading of the plates that occurswith high viable microbial levels. For instance, the two-stageAMS, with 200 holes per stage, reaches overload at aconcentration of 2.8 x 105 CFU/m3 in just 10 s (corrected forcoincidence and not exceeding a 10% coefficient of varia-tion). One approach developed for expanding the samplingrange of the AMS was to homogenize the agar after samplingand to plate serial dilutions to establish total viable counts(27). This technique, however, can decrease the viability ofmany organisms and suffers from the drawbacks of otherserial plating methods. Other potential problems with theAMS are agglomeration of microorganisms (24), the stress ofimpaction, and electrostatic attraction of particles to theplastic agar plates (2).NFE. Collection of airborne microorganisms onto filter

media followed by elution and plating was studied by Wolo-chow in 1958 (52) and found to be suitable for some organ-isms within certain environmental limits. Major problemswith this membrane filter method included loss of viabilityand poor recovery of the organisms from the filters. Laterstudies for the National Aeronautics and Space Administra-tion (13) indicated that this method was a reasonable way tosample for microorganisms, given the minimal equipmentneeded and its suitability for a variety of environments.Johnston et al. (19) compared sampling methods and notedthe simplicity and versatility of the Nuclepore filtration andelution method (NFE) as its major advantage.

Initial studies with membrane filters indicated that micro-bial recoveries were lowered as a result of trapping in thefilter matrix. In comparative studies, Lundholm found thatmembrane filter methods gave consistently lower recoverythat varied with the test environment (28). The availability ofthe Nuclepore polycarbonate filter beginning about 1975offered standardized pore size combined with a smoothsurface to allow greater retrieval of the microorganisms.Hobbie and coworkers (18) developed a Nuclepore filtermethod for microbial analysis that has been used primarilyfor aquatic microorganisms. Palmgren et al. (40, 41) adaptedthis method for evaluation of aerosolized microorganismsand used it extensively to study agricultural dusts in Swe-den. Use of the NFE for viable organisms has shown thatthere is as much as a fourfold loss of viability, especially forspores (39, 40). Despite variable results with this method incomparative studies (27), filter methods continue to be usedbecause of their simplicity and their ability to provideinformation on both viable and nonviable organisms (40).To quantify exposures to microorganisms in swine con-

finement buildings for human health studies, we needed tovalidate potential sampling procedures. The literature didnot contain reports of systematic testing of bioaerosol sam-pling methods in agricultural buildings. Thus, this study wasundertaken to test the AGI, AMS, and NFE in each of 24buildings during two different seasons. Three types of mi-crobes were measured in each building: total viable bacteria,enteric bacteria, and viable fungi. For comparison purposes,viable fungi were measured outdoors as well and nonviableplus viable organisms were assessed with a fluorescencemicroscopy NFE method similar to the CAMNEA method(40). The overall aim of our study was to determine the bestsampling method with respect to each class of viable micro-organism in swine confinement buildings.

(Portions of this work were presented at the 1990 Ameri-can Industrial Hygiene Conference and at the 1991 PanAmerican Aerobiology Association Conference.)

TABLE 1. Synopsis of the study design

Parameter studied n Specification

Swine barns (3 24 Each sampled twicetypes of barn) 9 Farrowing barn

9 Nursery grower barn6 Finishing barn

Microbial sampling 4 AGImethods AMS

NFEFluorescence microscopy NFE

Microbial media 3 TSA(for AGI, AMS, MACNFE) MEA

MATERUILS AND METHODS

Environmental sampling was performed in each of 24swine confinement buildings during the fall (late August toearly December) and then repeated during the winter (earlyJanuary to March). Table 1 contains a synopsis of the studydesign. The swine barns were divided among three barntypes: farrowing buildings, nursery grower buildings, andfinishing buildings. Sampling of airborne microorganismswas performed concurrently with the three different methodsand duplicate samples. Thus, at each location two AMSs,two AGIs and two NFEs were operated simultaneously. Inaddition, air samples were collected in duplicate for analysisof total organisms with the fluorescence microscopy NFEand for outdoor fungi with the AMS. The methodology foreach of these sampling procedures follows.AMS. Paired sampling was performed with the six-stage

AMS in the one- or two-stage configuration without the inletcone at a calibrated flow rate of 28.3 liters/min. Stages 2 and6 (-30-,um entrance limit, 4.7- and 0.65-pum cutoff diameters)were used for total bacterial counts, and stage 6 was usedalone for quantifying fungi and enteric bacteria. Severalstudies have shown that these methods are comparable tototal counts determined with the six-stage device (15, 20).The sampling media consisted of 20 ml of Trypticase soyagar (TSA), MacConkey's medium (MAC), or malt extractagar (MEA) in 100-mm disposable plastic petri plates. TSAwas selected because it had been used extensively andsuccessfully for the isolation and cultivation of a variety offastidious environmental microorganisms. MAC is a nonse-lective differential medium for isolation of gram-negativeenteric bacilli and will isolate members of the family Entero-bacteriaceae, non-lactose-fermenting gram-negative bacilli,and some enterococci. MAC is an appropriate choice for theswine barn environment, since the majority of gram-negativebacteria are of enteric origin (9). MEA has been widely usedfor sampling airborne fungi and is recommended by theAmerican Conference of Governmental Industrial Hygien-ists bioaerosols committee because it exceeds other myco-logic media in recovery of common saprophytes and is oneof the diagnostic media for Aspergillus spp.Andersen samplers were autoclaved and unwrapped just

before use. Sampling times ranged from 15 to 90 s. Controlpetri plates were handled similarly, except that the pumpwas not turned on. Bacterial plates (TSA, MAC) wereincubated at 30°C, and fungal plates (MEA) were incubatedat room temperature. Colonies were enumerated with aQuebec colony counter at 24-h intervals until growth hadstabilized, usually by 5 to 7 days. Corrections for coinci-

2544 THORNE ET AL.

on May 20, 2020 by guest

http://aem.asm

.org/D

ownloaded from

BIOAEROSOL SAMPLING METHODS IN BARNS HOUSING SWINE 2545

dence were made by the positive-hole method (2, 42), whichcorrects for the decreasing probability of a particle passingthrough a "naive" hole with increasing bioaerosol concen-tration.AGI. Paired, autoclaved all-glass impingers with 30-mm

jet-to-bottom spacing (AGI-30) and 12.5-liter/min criticalorifices were used with 20.0 ml of 1% peptone-distilled waterwith 0.01% Tween 80 and 0.005% antifoam A (Sigma Chem-ical Co., St. Louis, Mo.) by the widely used method in whichthe bioaerosol-laden air stream impacts on the liquid surface.Sampling was performed for 30 min or less with the impingeron ice, and then the impingers were transported to thelaboratory under refrigeration. The neck of each impingerwas flushed with the impinger solution to wash the nonre-spirable fraction of the dust into the impinger containing therespirable material. The final volume was measured andcorrected for evaporation. Suspensions were shaken, and0.1-ml samples of serial 10-fold dilutions in 0.1% peptone-distilled water were plated onto TSA, MAC, and MEA andthen incubated and counted. Optimally, colony counting wasdone on plates containing between 30 and 300 colonies.Concentrations of airborne organisms were then calculatedfrom the number of colonies, the volume of air sampled, andthe dilution factor.NFE. The NFE was based on a sampling procedure

described by Palmgren et al. (41). Nuclepore filters (0.4-,umpore size, 37-mm diameter) supported by cellulose padswere loaded into closed-face, collared cassettes. Cassetteswere sterilized with ethylene oxide and off gassed for 1month before use. Paired samples were collected at 2 liters/min with calibrated personal sampling pumps. The samplingtimes ranged from 15 to 30 min. By sampling for 30 min orless at 2 liters/min, both viable and nonviable organismswere recovered while the loss of viability of organisms wasminimized. Samples were transported within 4 h at ambienttemperatures to the laboratory, where they were eluted fromthe filters by injecting 1 ml of filtered, sterile 0.1% peptone-distilled water with 0.01% Tween 80 into the support pad and5 ml of the same medium onto the filter surface. Thecassettes were recapped and shaken for 15 min, and thewash solution was drawn off, measured, and then plated inserial 10-fold dilutions onto the culture media. Airborneconcentrations were determined from the plate counts bymultiplying the CFU by the dilution factor and by the eluatevolume (5 ml) and dividing by the volume of serial dilutionmaterial plated and the volume of air sampled.

Fluorescence microscopy NFE. The viable plus nonviablemicroorganisms were quantified by a previously describedmethod (1) modified in a manner similar to that described byDonham et al. (9) and Palmgren et al. (40, 41). Bioaerosolswere sampled as described above for the NFE, except thatthey were fixed with 1% formaldehyde and stained in situ for2 min with filtered acridine orange (0.1 mg/ml) in pH 7.2phosphate buffer. Filters were then dried and mounted onglass slides and cleared with 1 drop of Cargille A immersionoil. Between 20 and 300 fields were then counted, such that400 to 1,000 microorganisms were tallied. Fluorescence andproper morphology were used to judge whether an objectwas a microorganism. The total organisms counted in thesefields were adjusted to microorganisms on the samplingsurface of the filter by multiplying by the ratio of the areacounted to the total sampling area. Airborne concentrationswere calculated by dividing total organisms on the filter bythe volume of air sampled.

Statistical analyses. All statistical analysis were performedusing SAS/STAT V.6 or SPSS/PC+. SAS analyses included

10

6T' 6E~ 10

05

w 10U.z

104

13

10~

1

w----

-Ar Viable Bacteria: TSA

U0 Gm- Bacteria: MAC

1 10 100Filter Storage Time, hrs

FIG. 1. Effects of filter storage time after sampling on the con-centrations of bacteria for 14 samples collected in parallel. For totalbacteria (on TSA) and gram-negative enteric bacteria (on MAC), thecounts were reasonably stable up to 38 h after sampling. Theseresults indicated that storage of filters at room temperature for 4 hbefore analysis did not influence the concentrations represented.

the following statistical programs: MEANS, FREQ,UNIVARIATE, ITEST, NPARlWAY, REG, and GLM.Tests of normality of data were performed by using the Wstatistic developed by Shapiro and Wilk (44) as computedwithin SAS UNIVARIATE. The Wstatistic tests for the nullhypothesis that the input data are drawn as a random samplefrom a normal distribution. This statistic ranges from zero toone with low Wvalues (values yielding probability below the5% point), leading to rejection of the null hypothesis andsuggesting the use of nonparametric methods for furtheranalysis of the data. SPSS/PC+ RELIABILITY was used tocompute Cronbach's coefficient ot.

RESULTSBefore the comparative studies between the three sam-

pling methods were begun, pilot studies were performed tofine tune each of the sampling protocols. One such studyevaluated the influence of filter storage time in the NFE onairborne concentration determinations for total bacteria andgram-negative enteric bacteria after sampling in a swinebarn. Fourteen pumps with sampling filters and cassetteswere operated simultaneously for 30 min; they were thenstored in pairs at room temperature for 0.5, 2, 8, 20, 38, 75,or 168 h before analysis. For total bacteria (TSA) andgram-negative enteric bacteria (MAC), the counts werereasonably stable up to 38 h after sampling (Fig. 1). Geomet-ric mean values for the first five samples were 2.1 x 106CFU/m3 for bacteria and 0.5 x 106 CFU/m3 for entericbacteria and standard deviations were between 0.17 and 0.13log unit. At the 75-h time point, the drops in the numbers ofviable organisms recovered exceeded 1 order of magnitudefor bacteria and 2 orders of magnitude for enteric bacteria.These results indicated that storage of filters at room tem-perature for the planned 4 h before analysis did not influencethe concentrations represented.

Table 1 illustrates the study design for the comparison ofsampling methods in the three types of swine barns. Farrow-ing buildings house sows and piglets from birth to weaning.

VOL. 58, 1992

on May 20, 2020 by guest

http://aem.asm

.org/D

ownloaded from

2546 THORNE ET AL.

TABLE 2. Testing for differences between simultaneousduplicate samples

No. ofOrganismclassMethod valid Reliability t test

datum (ax) P valuepointsa

Bacteria AGI 43 0.98 0.54AMS 21 0.92 0.63NFE 42 0.92 0.09

Enteric bacteria AGI 11 0.17 0.20AMS 35 0.95 0.68NFE 2

Fungi AGI 34 0.86 0.04AMS 45 0.94 0.22NFE 26 0.86 0.18

Total organisms Fluorescence 41 0.88 0.27microscopy NFE

Outside fungi AMS 43 0.94 0.92

a For a paired sample to be considered valid, each of the two values had tobe nonzero and within the limit of detection. Total possible, 48.

The young swine are then moved to the nursery grower,where they remain until 12 weeks. Finishing buildings housepigs until they are sent for slaughter at about 6 months of ageand 100 kg. Environmental conditions and bioaerosols differamong these barn types. To guide the analysis of theenvironmental data and their subsequent use in testing fordose-response relationships for adverse health effects in thefarmers, we needed to determine which sampling methodsworked best in this setting. Thus, we developed the follow-ing selection criteria, against which the acceptability ofmicrobial sampling methods was judged.

Criterion 1: at least two-thirds of the measurements mustyield valid datum points for both duplicate samples. Individ-ual measurement values were deemed acceptable if they fellwithin the limits of detection for the method (i.e., were nottoo numerous to count) and were nonzero for microbialgrowth. The requirement of nonzero values is reasonablesince, in the swine barn environment, it is nearly impossibleto obtain a value of zero bacterial or fungal growth onnonselective media if the method is appropriate and properlyexecuted. To ensure highly reliable data, we further requiredthat both duplicates of the paired measures be determinateand nonzero for the mean datum point (geometric mean ofthe duplicates) to be valid. Table 2 lists the number of validdatum points obtained (out of 48 possible) for each of themicrobial sampling protocols. For sampling bacteria, morethan 87% of the concentration values were valid for the AGIand NFE, whereas less than half of the AMS values werevalid. Most of the invalid AMS data were too numerous tocount because of overloading, even at sampling times asshort as 15 s. For enteric bacteria, only the AMS yieldedadequate valid data to satisfy criterion 1 (35 of 48 samples).For assessment of airborne fungi, both the AGI (71%) andthe AMS (94%) gave sufficient percentages of valid data tomeet criterion 1 and the NFE fell short (54%).

Criterion 2: intraclass reliability coefficients must exceed0.85. The intraclass reliability (26), calculated as Cronbach'scoefficient ao, was determined for each sampling protocol.Each measure of bioaerosol concentration contains somedegree of measurement error. Two simultaneous measures

TABLE 3. Tests of correlation between sampling methods foreach organism class

Pearson's r (n)Organism class AGI vs

AMS AGI vs NFE AMS vs NFE

Bacteria 0.33 (20) 0.86 (38) 0.53 (18)Enteric bacteria 0.70a (11) (2) (2)Fungi 0.43 (31) 0.78 (21) 0.69 (25)

a Spearman's r is given because the AMS data were not normally distrib-uted after logarithmic transformation.

on the same variable will usually not have the same value.Repeated simultaneous measurements in a series of barnswill show some consistency that can be assessed by usingCronbach's coefficient a. This reliability coefficient indicatesthe degree to which the two paired samples are measuringthe same value. A high at value (20.85) indicates a greatdegree of precision for the measurement protocol. All of thesampling protocols yielded values for a in excess of therequired 0.85, except for the AGI with enteric bacteria(Table 2). The 11 valid pairs in this protocol demonstratedvery poor agreement. Total organisms determined by thefluorescence microscopy NFE and outside fungi sampled bythe AMS both had high a values (0.88 and 0.94). Table 2 alsolists the results of a paired t test for the duplicate measures.Only the P value for the measurement of fungi by the AGIwas less than 0.05. Thus, for this AGI, the duplicates werewell correlated (ao = 0.86) but the difference between themshowed a poor central tendency (P = 0.04).

Criterion 3: the data obtained with an acceptable methodshould correlate with some other acceptable method with atleast a moderate correlation coefficient. Univariate analysesand computation of the W statistic were performed on thegeometric means of the logarithmically transformed datumpairs and verified normal distribution in all but one of theparameters tested: the AMS for enteric bacteria (P < W =0.045). The distribution of this variable was somewhattruncated on the lower end. The other 21 parameters hadhigh values for the W statistic, providing strong evidencethat they were sampled from a normal distribution. Thus, thedegree of correlation was assessed with Pearson's r for allbut one comparison in Table 3. For this correlation coeffi-cient, values above 0.75 indicate excellent agreement andvalues from 0.40 to 0.75 are generally considered to indicatefair to good agreement for environmental sampling data (14).The results of tests of correlation between sampling meth-

ods within organism class are listed in Table 3 and plotted inFig. 2. For sampling bacteria, the correlation between theAGI and the AMS was poor (r = 0.33), although the valueswere centered about the identity line (Fig. 2a). The AGIresults plotted against the NFE results illustrate excellentcorrelation (r = 0.86), with most values falling close to theidentity line. In the five cases where there was the greatestdeviation from the identity line, the values yielded by theAGI were higher than those of the NFE. Calculation ofCronbach's ao for the three sampling methods analyzed as agroup showed that dropping the AMS-sampled measuresfrom the data set improved the reliability from 0.88 to 0.91.Dropping either of the other measures resulted in a loss ofreliability.For sampling airborne fungi, the AGI correlated poorly

with the AMS (r = 0.43) but very well with the NFE (r =0.78). The AMS substantially (but not systematically) under-

APPL. ENVIRON. MICROBIOL.

on May 20, 2020 by guest

http://aem.asm

.org/D

ownloaded from

BIOAEROSOL SAMPLING METHODS IN BARNS HOUSING SWINE 2547

3AMS (CFU/m )

UL.0

0--

cvE

IL

0

NFE (CFU/m3)

10io2

AMS (CFU/m3) NFE (CFU/m3)FIG. 2. Results of tests of correlation between sampling methods within organism class. a and b, bacteria; c and d, fungi. The lines indicate

identity between the methods, and the data from the fall (-) and winter (K) rounds of sampling are differentiated. (a) Poor correlation betweenthe AGI and the AMS for sampling bacteria (Pearson's r = 0.33). (b) AGI results plotted against NFE results, illustrating excellent correlation(Pearson's r = 0.86) with most values falling close to the identity line. Sampling results for airborne fungi are shown in panels c and d. TheAGI correlated only moderately with the AMS (r = 0.43) but very well with the NFE (r = 0.78). The AMS underestimated the concentrations(c). Conversely, the data for the AGI versus the NFE fell close to the identity line (d). No systematic seasonal difference was noted.

estimated the concentrations relative to the AGI (Fig. 2c andd). Conversely, the plot for the AGI versus NFE shows thatthe data fell close to the identity line. For both the bacteriaand the fungi, the correlation of the AMS with the NFE was

stronger than that of the AGI with the AMS (Table 3). Forenteric bacteria, the AGI and the AMS correlated well (r =0.70) but the sample size was small (n = 11). Table 3indicates that, overall, the AGI and NFE demonstrated the

strongest correlations and the AMS demonstrated the poor-est correlation.To compare the microbial yields of the sampling methods,

mean concentrations were calculated with only those datafor which all three methods provided valid pairs of mea-

sures. These values demonstrate that the bacterial concen-

trations were not significantly different for the three mea-

sures but that for fungi the AMS was significantly lower than

cE

C.

2

C)

105

VOL. 58, 1992

on May 20, 2020 by guest

http://aem.asm

.org/D

ownloaded from

APPL. ENvIRON. MICROBIOL.

TABLE 4. Mean concentrations of bacteria and fungi determined by the three test methodsa

Mean concn (CFU/m3) POrganism class

AMS AGI NFE AMS vs AGI AGI vs NFE

Bacteria (n = 18) 73,200 96,400 77,800 0.33 0.55Fungi (n = 20) 1,970 5,380 5,850 0.002 0.77

a For this comparison, concentrations were calculated with only data from barns for which all three methods yielded valid datum pairs.

the AGI (P = 0.002) (Table 4). The fact that the AMS yieldedapproximately one-third the number of CFU per cubic meterobtained with the AGI or the NFE was particularly disturb-ing. In our experience, the AMS generally performs well forsampling fungi in other settings.

This analysis, guided by the three evaluation criteria,illustrates that for sampling viable bacteria in swine barnsthe AGI and NFE gave sufficient numbers of valid datumpoints, all three methods yielded high intraclass reliabilities,and the AGI and NFE correlated well. For enteric bacteria,only the AMS generated sufficient data with a high enough a

value. For sampling fungi, the AGI and AMS both yieldedsufficient data, all three methods had high Cronbach's aLvalues, and the AGI and NFE correlated well. Since therecovery of airborne fungi measured with the AGI wassuperior to that of the AMS (Table 4), the AGI was thesampling method of choice for viable fungi.

DISCUSSION

Three bioaerosol sampling methods, each run in duplicate,were performed simultaneously in 48 field sampling sessionsto determine the most reliable and predictive method for airmonitoring in livestock confinement facilities. Analysescompared the proportion of the data falling within the limitsof detection, the intraclass reliabilities, and the correlationsbetween methods. Selection criteria based on these analyseswere used to judge which method was the most suitable.Subsequent evaluations of possible relationships betweenthese bioaerosol exposures and the health status of the swineconfinement workers studied at these sites will use only thedata from sampling methods judged acceptable by thesecriteria.When bacteria were sampled, the AMS overloaded even

with very short sampling times (15 s) and demonstrated poorcorrelation with the AGI (Table 3, Pearson's r = 0.33). TheAMS data in Fig. 2a appear to be truncated at about 2 x 105CFU/m3. Using stages 2 and 6 of the 400-hole 6-stage AMS,the maximum concentration that can be quantified in a 15-ssample is 5.2 x 105 CFU/m3 (coincidence corrected; coeffi-cient of variation, <10%). Since the total bacterial count isconsidered over the upper limit for the instrument if eitherstage overloads, one can exceed the limit of detection at a

bacterial concentration as low as 2.6 x 105 CFU/m3. Thus,the strange appearance of Fig. 2a is attributable to exclusionof data exceeding the upper limit of detection of the AMSthat occurred between 2.6 x 105 and 5.2 x 105 CFU/m3. Themean level of viable bacteria in the 24 buildings sampled inthis study determined by the AGI was 2.1 x 105 CFU/m3 (n= 43), and levels ranged to 8.7 x 106 CFU/m3. Only 26 ofthese 43 valid datum points for the AGI fell within the 2.6 x105-CFU/m3 limit of detection for the AMS. In grain han-dling facilities where bacterial concentrations ranged from200 to 200,000 CFU/m3, the results of parallel samplingdemonstrated good agreement between the AMS and theAGI (Spearman's r = 0.73, n = 19) (48).

Prior testing of bioaerosol sampling protocols has gener-ally been performed in environments with low microbialconcentrations, under laboratory test conditions with onlyone or two species, or without comparisons among the AGI,AMS, and NFE. Testing that fails to reflect the harsh natureof environments such as barn atmospheres does not shedlight on how the methods will operate under these fieldconditions. Lembke et al. (24) performed air sampling withthe six-stage AMS (with TSA) and later with the AGI (withblood agar) in a municipal solid waste recovery system.They compared the duplicate sampling data within themethod and gave 95% confidence interval estimates of thevariance (which they called coefficients of variation), whichwere 0.38 for the AGI and 0.23 for the AMS. We analyzedtheir data and obtained a Cronbach's ao of 0.99 for both theAGI (neck rinsed) and the AMS. Unfortunately, samplingswith the AMS and the AGI were not simultaneous, the mediawere different, and no between-methods comparisons werepossible.

Several investigators have published work in which side-by-side sampling in some form was performed (Table 5).Lundholm (27, 28) compared sampling with the six-stageAMS, the AGI, and a method in which filters were placeddirectly on the agar plates. Suspensions of Staphylococcusepidernidis (gram-positive bacteria) or Serratia marcescens(gram-negative bacteria) were aerosolized into a 37-literexposure chamber in which the samplers were placed. Thus,the chamber held an AMS drawing air at 28.3 liters/min, a slitsampler at 30 liters/min, an impinger at 11.5 liters/min, and afilter sampling at 15 liters/min, totaling 85 liters/min. This

TABLE 5. Bioaerosol sampling studies that allow comparison among methods

Sampling methods Sampling conditions Organism(s) Reference or source

AMS, AGI, filter Chamber Staphylococcus epidermidis, Serratia marcescens Lundholm (28)AMS, AGI, filter Chamber Bacillus subtilis, Escherichia coli Macher and First (30)AMS, AGI Chamber Faeni rectivirgula Kaliszewski et al. (22)AMS, MAY Staged field test E. coli Zimmerman et al. (53)AMS, AGI Staged laboratory test Pseudomonas syringae Buttner and Stetzenbach (4)MAY, filter 6 swine barns Bacteria, fungi, thermophilic actinomycetes Crook et al. (7)AMS, AGI, NFE 24 swine barns, each Bacteria, fungi, enteric bacteria This study

sampled twice

2548 THORNE ET AL.

on May 20, 2020 by guest

http://aem.asm

.org/D

ownloaded from

BIOAEROSOL SAMPLING METHODS IN BARNS HOUSING SWINE 2549

situation likely led to the samplers having a strong influenceover the concentrations. With S. epidermidis, the AMSyielded higher values than the AGI or filter method (AMS,15,700 CFU/m3; AGI, 6,300 CFU/m3; filter method, 6,100CFU/m3), whereas the values for the AMS and AGI were

much higher than those for the filter method for S. marces-

cens (AMS, 4,200 CFU/m3; AGI, 3,200 CFU/m3; filtermethod, 70 CFU/m3). These results do not agree with our

findings (Table 4), where the mean values for the threemethods were not significantly different for bacteria. Lund-holm also sampled for gram-negative bacteria in 20 sewageplants with the AMS and the filter method; she reported9,900 + 4,100 CFU/m3 and 300 + 200 CFU/m3, respectively,indicating poor recovery with this filter method. We alsoexperienced poor recovery of gram-negative bacteria with a

filter method, as indicated by the numbers of valid measures

for the NFE indicated in Table 2.Three studies were performed in laboratory test chambers

by Macher and First (30). In the first, Bacillus subtilis spores

and Escherichia coli were generated from 107-CFU/ml solu-tions into a test chamber. Using a gelatin filtration method,Macher and First reported recoveries of 0 CFU/m3 for E.coli versus 174,000 to 202,000 CFU/m3 with the AGI. For B.subtilis, the filtration method yielded 314,000 and 105,000CFU/m3, compared with 60,000 and 88,000 CFU/m3, respec-

tively, for the AGI. Thus, recoveries were poor for E. coliand variable for B. subtilis. The second study compared thesix-stage AMS or the AGI with a modified personal impingermethod. The AMS yielded data that were 10 and 50% higherthan those of the modified personal impinger method (in twodatum points), while the data for the AGI were 1.5- to 2-foldlower than those for the modified personal impinger methodin seven trials. A third experiment performed with B. subtilisallowed direct comparison of the results from the AGI andthe sum of the results from stages 4 to 6 of the AMS. Theratio of concentrations determined with the AGI to thosedetermined with the AMS were 1.2, 0.4, and 0.9 for AMSlevels of 2,600, 27,000, and 42,000 CFU/m3, respectively.Since spores of B. subtilis are 1-,um ovoids, it is notsurprising that the AMS generally recovered more organ-

isms. Our comparison of these two samplers with spores ofMicropolyspora faeni (Faeni rectivirgula) (1.2 ,um) yieldedthe same result and indicated that small spores pass throughthe AGI (22).

Several investigators have compared the AMS or the AGIwith the three-stage May liquid impinger method (MAY)with favorable results. Zimmerman and coworkers (53)sampled a high-pressure spray of E. coli solution (6 x 106CFU/ml) with the two-stage AMS and the MAY at very lowbioaerosol concentrations. Our analysis of their tabulateddata indicated that the AMS correlated with the MAY with a

Spearman coefficient of 0.94, which was very high. Inaddition, the geometric means were not significantly differ-ent for the AMS and the MAY (125 and 117 CFU/m3,respectively). This close agreement may reflect the idealizedsampling conditions, the single organism approach, and thelow concentrations used. A study similar to that of Zimmer-man et al. was carried out by Buttner and Stetzenbach (4),who evaluated the AGI and the six-stage AMS by samplingaerosolized Pseudomonas syningae. They found that theconcentrations detected decreased 5-fold with increasingsampling times and that the AMS recoveries exceeded thoseof the AGI by 3.2-fold at the 5-min sampling times (AGI,2,512 CFU/m3; AMS, 7,943 CFU/m3). These authors con-cluded that the upper quantitation limit for the six-stageAMS was 105 CFU/m3, allowing 100% positive holes, but

they based this on a sampling time of 5 min. Although AMSsampling times well below 5 min are not optimal, manyresearchers use them to extend the range of applicability ofthe AMS.

Recent work by Crook and colleagues (7) included simul-taneous bioaerosol sampling in six swine confinement build-ings with the MAY and a filtration method similar to theNFE. Although direct comparisons of the bioaerosol sam-pling methods were not performed, their data did allow fortests of correlation for fungi isolated at 25°C. The NFE andMAY impinger data (three stages were summed) for thesesix sites demonstrated good correlation between the meth-ods (Spearman's r = 0.71). This is consistent with our data inwhich the NFE and a liquid impinger method (AGI) werewell correlated (r = 0.78). Geometric means for fungi incu-bated at 25°C were somewhat higher for the NFE (8,700CFU/m3) than for the MAY (5,100 CFU/m3). For bacteria at37°C the MAY and NFE did not differ significantly (1.6 x 106and 1.2 x 106 CFU/m3, respectively). These results forairborne fungi are close to the values shown in Table 4 forthe AGI and the NFE. Table 4 also illustrates that thebacterial counts for the AGI and the NFE were not signifi-cantly different. Crook et al. obtained the same result butobserved 16-fold-higher concentrations of bacteria.Although comparison studies of bioaerosol sampling

methods performed under controlled environmental condi-tions give useful information, they do not indicate thesuitability of the methods for the particular field application.By performing parallel comparison studies in the atmo-spheres we wished to characterize, we gained directly appli-cable information about the utility of these methods. Whensampling for microorganisms in highly contaminated envi-ronments, many sampling problems that are not found in lesshostile environments may arise. The airborne dusts them-selves are a significant factor affecting the tenacity of theorganisms. Components of the dust lead to larger aggrega-tions that can serve as protectants and absorb gases andvapors, thus complicating the quantitation. Work by Dossowand Muller (11) showed that agricultural dusts, calciumcarbonate, ammonia, and humidity all influence the survivalof microorganisms. Environmental factors have various ef-fects on microorganisms, and these effects must be consid-ered to optimize collection. Marthi et al. (32) and Walter etal. (51) showed that aerosol droplet size, relative humidity,and temperature affect viability of organisms and that sur-vival rates depend on the type of organism. Temperaturesand relative humidities comparable to those found in live-stock confinement facilities prolong the viability of airbornemicroorganisms. On the other hand, in the swine barnsstudied in this work, concentrations of ammonia reached 40ppm, concentrations of CO2 reached 5,500 ppm, insidetemperatures were as low as 10°C, and relative humiditieswere as low as 22%. These extremes clearly have an impacton bioaerosol sampling results when, for instance, the sam-pling method concentrates ammonia from the air in the liquidsampling media (as with the AGI) or draws contaminated,cold, or dry air across microorganisms that have alreadybeen trapped within the filter matrix (as with the NFE) or onthe surface of media (as with the AMS). The effects ofspecific environmental conditions on the ability of thesesampling devices to yield accurate information remain un-known.

It has become imperative that we determine the specificroles of microorganisms and their components in occupa-tional pulmonary disease. Bioaerosols need to be sampled tocharacterize the specific hazardous components and their

VOL. 58, 1992

on May 20, 2020 by guest

http://aem.asm

.org/D

ownloaded from

APPL. ENVIRON. MICROBIOL.

sources, quantify the organisms, and establish dose-re-sponse relationships between these components and diseaseconditions. By quantifying problem agents, control mea-sures and remediation can be more effectively introducedinto the agricultural environment. From the analyses de-scribed herein, it was found that the AGI was the samplingmethod of choice for viable bacteria in livestock confinementbuildings but that the NFE was also satisfactory. For entericbacteria, the AGI and NFE methods rarely gave acceptablevalues but the AMS demonstrated a good yield of valid dataand high reliability. For sampling fungi, the AMS and theAGI both met the acceptance criteria, but the AGI demon-strated 2.7-fold higher recovery. Thus, for bioaerosol sam-pling in swine barns we found the AGI most effective fortotal bacteria and fungi and the AMS best for entericbacteria.

ACKNOWLEDGMENTS

We are grateful to the many Iowa farmers who have given sogenerously of their time and shared their knowledge with us to makethis and our other studies possible.

This research was funded in part by the National Pork Producer'sCouncil and by Public Health Service grant P30 ES05605-01 fromthe National Institute of Environmental Health Sciences.

REFERENCES1. American Society for Testing and Materials. 1987. Test method

for enumeration of aquatic bacteria by epifluorescence micros-copy counting procedure, p. 56. ASTM standards on materialsand environmental microbiology. Publication no. D4455-85.American Society for Testing and Materials, Villanova, Pa.

2. Andersen, A. A. 1958. New sampler for the collection, sizing,and enumeration of viable airborne particles. J. Bacteriol.76:471-484.

3. Brachman, P. S., R. Erlich, H. F. Eichenwald, V. J. Cabelli,T. W. Kethley, S. H. Madein, J. R. Maltman, G. Middlebrook,J. D. Morton, I. H. Silver, and E. K. Wolfe. 1964. Standardsampler for assay of airborne microorganisms. Science 144:1295.

4. Buttner, M. P., and L. S. Stetzenbach. 1991. Evaluation of fouraerobiological sampling methods for the retrieval of aerosolizedPseudomonas syringae. Appl. Environ. Microbiol. 57:1268-1270.

5. Clark, S., R. Rylander, and L. Larsson. 1983. Airborne bacteria,endotoxin and fungi in poultry and swine confinement buildings.Am. Ind. Hyg. Assoc. J. 44:537-541.

6. Cox, C. S. 1987. The aerobiological pathway of microorganisms,p. 24-77. John Wiley and Sons, Inc., New York.

7. Crook, B., J. F. Robertson, S. A. Travers Glass, E. M. Both-eroyd, J. Lacey, and M. D. Topping. 1991. Airborne dust,ammonia, microorganisms, and antigens in pig confinementhouses and the respiratory health of exposed farm workers. Am.Ind. Hyg. Assoc. J. 52:271-279.

8. Donham, K. J. 1986. Hazardous agents in agricultural dusts andmethods of evaluation. Am. J. Ind. Med. 10:205-220.

9. Donham, K. J., W. Popendorf, U. Palmgren, and L. Larsson.1986. Characterization of dusts collected from swine confine-ment buildings. Am. J. Ind. Med. 10:294-297.

10. Donham, K. J., L. J. Scallon, W. Popendorf, M. W. Treuhaft,and R. C. Roberts. 1986. Characterization of dusts collectedfrom swine confinement buildings. Am. Ind. Hyg. Assoc. J.47:404-410.

11. Dossow, A. V., and W. Muller. 1988. The influence of carryingparticles on the tenacity of microorganisms, p. 416-419. InEnvironment and animal health: proceedings of the 6th Interna-tional Congress on Animal Hygiene, Skara, Sweden. SwedishUniversity of Agricultural Sciences, Skara, Sweden.

12. Dutkiewicz, J., S. A. Olenchock, W. G. Sorenson, V. F. Gerenc-ser, J. J. May, D. S. Pratt, and V. A. Robinson. 1989. Levels ofbacteria, fungi, and endotoxin in bulk and aerosolized corn

silage. Appl. Environ. Microbiol. 55:1093-1099.13. Fields, N. D., G. S. Oxborrow, J. R. Puleo, and C. M. Herring.

1974. Evaluation of membrane filter field monitors for microbi-ological air sampling. Appl. Microbiol. 27:517-520.

14. Fleiss, J. L. 1981. Statistical methods for rates and proportions.John Wiley and Sons, Inc., New York.

15. Gillespie, V. L., C. S. Clark, H. S. Bjornson, S. J. Samuels, andJ. W. Holland. 1981. A comparison of two-stage and six-stageAndersen impactors for viable aerosols. Am. Ind. Hyg. Assoc.J. 42:858-864.

16. Greenburg, L., and G. W. Smith. 1922. A new instrument forsampling aerial dust. Publication no. RI-2392. U.S. Bureau ofMines, Pittsburgh.

17. Heederik, D., R. Brouwer, K. Biersteker, and J. S. M. Boleij.1991. Relationship of airborne endotoxin and bacteria levels inpig farms with the lung function and respiratory symptoms offarmers. Int. Arch. Occup. Environ. Health 62:595-601.

18. Hobbie, J. E., R. J. Daley, and S. Jasper. 1977. Use ofNuclepore filters for counting bacteria by fluorescent micros-copy. Appl. Environ. Microbiol. 33:1225-1228.

19. Johnston, J. R., A. M. Butchart, and S. J. Kgamphe. 1978. Acomparison of sampling methods for airborne bacteria. Environ.Res. 16:279-284.

20. Jones, W. G., K. Morring, P. Morey, and W. Sorenson. 1985.Evaluation of the Andersen viable impactor for single stagesampling. Am. Ind. Hyg. Assoc. J. 46:294-298.

21. Jones, W. G., K. Morring, S. Olenchock, T. Williams, and J.Hickey. 1984. Environmental study of poultry confinementbuildings. Am. Ind. Hyg. Assoc. J. 45:760-766.

22. Kaliszewski, S. D., P. S. Thorne, and S. A. Bleuer. 1992.Production of hypersensitivity pneumonitis in mice by inhala-tion of spores of Micropolyspora faeni. Toxicologist 12:223.

23. Lacey, J., and B. Crook. 1988. Review: fungal and actinomycetespores as pollutants of the workplace and occupational aller-gens. Ann. Occup. Hyg. 32:515-533.

24. Lembke, L. L., R. N. Kniseley, R. C. Van Nostrand, and M. D.Hale. 1981. Precision of the all-glass impinger and the Andersenmicrobial impactor for air sampling in solid waste handlingfacilities. Appl. Environ. Microbiol. 42:222-225.

25. Leopold, S. S. 1988. "Positive hole" statistical adjustment for atwo stage, 200-hole-per-stage Andersen air sampler. Am. Ind.Hyg. Assoc. J. 49:A88-A90.

26. Lindquist, E. F. 1953. Design and analysis of experiments inpsychology and education, p. 357-382. Houghton Mifflin Co.,Boston, Mass.

27. Lundholm, M. 1981. Airborne gram-negative bacteria in cottonmills and sewage plants: environmental sampling and exposurerelated effects, p. 13-21. Ph.D. thesis, Department of Environ-mental Hygiene and Department of Clinical Bacteriology, Uni-versity of Goteborg, Goteborg, Sweden.

28. Lundholm, M. 1982. Comparison of methods for quantitativedeterminations of airborne bacteria and evaluation of totalviable counts. Appl. Environ. Microbiol. 44:179-183.

29. Macher, J. M. 1989. Positive-hole correction of multiple-jetimpactors for collecting viable microorganisms. Am. Ind. Hyg.Assoc. J. 50:561-568.

30. Macher, J. M., and M. W. First. 1984. Personal air samplers formeasuring occupational exposures to biological hazards. Am.Ind. Hyg. Assoc. J. 45:76-83.

31. Malmberg, P., A. Rask-Andersen, U. Palmgren, S. Hoglund, B.Kolmodin-Hedman, and G. Stalenheim. 1985. Exposure to mi-croorganisms, febrile and airway-obstructive symptoms, im-mune status, and lung function of Swedish farmers. Scand. J.Work Environ. Health 11:287-293.

32. Marthi, B., V. P. Fieland, M. Walter, and R. J. Seidler. 1990.Survival of bacteria during aerosolization. Appl. Environ. Mi-crobiol. 56:3463-3467.

33. May, J. J., D. S. Pratt, L. Stallones, P. R. Morey, S. A.Olenchock, I. W. Deep, and G. A. Bennett. 1986. A study of silounloading: the work environment and its physiologic effects.Am. J. Ind. Med. 10:318.

34. May, K. R. 1964. Calibration of a modified Andersen bacterialaerosol sampler. Appl. Microbiol. 12:37-43.

2550 THORNE ET AL.

on May 20, 2020 by guest

http://aem.asm

.org/D

ownloaded from

BIOAEROSOL SAMPLING METHODS IN BARNS HOUSING SWINE 2551

35. May, K. R., and G. J. Harper. 1957. The efficiency of variousliquid impinger samplers in bacterial aerosols. Br. J. Ind. Med.14:287-297.

36. Merchant, J. A. 1987. Agricultural exposures to organic dusts.Occup. Med. State of the Art Rev. 2:409-426.

37. Miles, A. A., S. S. Misra, and J. 0. Irwin. 1938. Estimation ofbactericidal power of blood. J. Hyg. Camb. 38:732.

38. Muilenberg, M. L. 1989. Aeroallergen assessment by micros-copy and culture. Immunol. Allergy Clinics North Am. 9:245-268.

39. Nordic Council of Ministries. 1988. Harmonization of samplingand analysis of mould spores. Nord 1988:88.

40. Palmgren, U., G. Strom, G. Blomquist, and P. Malmberg. 1986.Collection of airborne microorganisms on Nuclepore filters,estimation and analysis-CAMNEA method. J. Appl. Bacteriol.61:401-406.

41. Palmgren, U., G. Strom, P. Malmberg, and G. Blomquist. 1986.The Nuclepore filter method: a technique for enumeration ofviable and nonviable airborne microorganisms. Am. J. Ind.Med. 10:325-327.

42. Peto, S., and E. 0. Powell. 1970. The assessment of aerosolcontamination by means of the Andersen sampler. J. Appl.Bacteriol. 33:582-598.

43. Rosebury, T. 1947. Experimental airborne infection. TheWilliams & Wilkins Co., Baltimore.

44. Shapiro, S. S., and M. B. Wilk. 1965. An analysis of variancetest for normality (complete samples). Biometrika 52:591-611.

45. Shipe, E. L., M. E. Tyler, and D. N. Chapman. 1959. Bacterial

aerosol samplers. II. Development and evaluation of the Shipesampler. Appl. Microbiol. 7:349-354.

46. Smid, T., E. Schokkin, J. S. M. Boleij, and D. Heederik. 1989.Enumeration of viable fungi in occupational environments: acomparison of samplers and media. Am. Ind. Hyg. Assoc. J.50:235-239.

47. Solomon, W. R. 1975. Assessing fungus prevalence in domesticinterior. J. Allergy Clin. Immunol. 56:235-242.

48. Thorne, P. S., S. E. Hosch, J. L. Watt, and D. A. Schwartz. 1991.Assessment of bioaerosol exposures in grain and dry vegetablehandlers. U.S. Surgeon General's Conference, Des Moines,Iowa. Centers for Disease Control, Atlanta.

49. Tyler, M. E., and E. L. Shipe. 1959. Bacterial aerosol samplers.I. Development and evaluation of the all-glass impinger. Appl.Microbiol. 7:337-349.

50. Tyler, M. E., E. L. Shipe, and R. B. Painter. 1959. Bacterialaerosol samplers. III. Comparison of biological and physicaleffects in liquid impinger samplers. Appl. Microbiol. 7:355-362.

51. Walter, M. V., B. Marthi, V. P. Fieland, and L. M. Ganio. 1990.Effect of aerosolization on subsequent bacterial survival. Appl.Environ. Microbiol. 56:3468-3472.

52. Wolochow, H. 1958. The membrane filter technique for estimat-ing numbers of viable bacteria: some observed limitations withcertain species. Appl. Microbiol. 6:201-206.

53. Zimmerman, N. J., P. C. Reist, and A. G. Turner. 1987.Comparison of two biological aerosol sampling methods. Appl.Environ. Microbiol. 53:99-104.

VOL. 58, 1992

on May 20, 2020 by guest

http://aem.asm

.org/D

ownloaded from