Participants and Methods

ParticipantsA total of 593 medical (first and second year) and undergraduate students at Wake Forest University were screened for nasal S. aureus carriage by nasal cultures from October 2000 to April 2004 [8]. Some 120 (20%) carried S. aureus for at least 4 weeks, as confirmed by follow-up nose swabs every 2 weeks. Eleven healthy nasal S. aureus carriers were subsequently recruited for participation in the study on the basis of susceptibility (serum-neutralizing antibody titer ⩽1:4) to 1 of 3 available challenge rhinoviruses

A self-administered questionnaire regarding the medical and medication history of the participants was obtained from all individuals. Each participant provided informed consent, and the human-experimentation guidelines of the US Department of Health and Human Services and those of the Wake Forest University School of Medicine (Winston-Salem, NC) were followed in the conduct of the study

Microbiologic techniquesThe methods used for the collection of mucosal and air cultures, virus challenge, and the assessment of rhinovirus infection and illness have been described elsewhere in detail [7]. In short, samples for quantitative nasal, pharynx, and skin (both axillae, palms of the hands, and groin) culture were obtained from each participant by swabbing. After surface plating and incubation for 24–48 h at 35°C, the total number of colonies and those that had morphological characteristics consistent with S. aureus and CoNS were counted as colony-forming units and evaluated by Gram’s stain, catalase test, and coagulase test (Staphaurex; Murex Biotech)

Air-sample cultures were performed in an airtight chamber ∼110 ft³ (3.1 m³) in volume that was built around the front of a class II biological safety hood (Purifier; Labconco) with sufficient room for a volunteer to sit in front of the workbench. Three 2-stage and one 6-stage Andersen air samplers were used to measure airborne bacterial particles by impact on blood agar plates (sample volume, 28.3 L [1 ft³]/min; Andersen Instruments) [9]. The agar plates were incubated for 24 h at 35°C, then held for an additional 24 h at room temperature [10]. Colonies were counted and evaluated as described above. One colony of S. aureus and CoNS from each site cultured, if available, was stored in 95% tryptic soy broth with 5% glycerol at −70°C for molecular typing

To test the recovery of S. aureus isolates in the test chamber, a saline solution that contained a defined concentration of S. aureus bacteria was dispersed into the air using a Collison nebulizer [11]. Air samples were collected using the Andersen air samplers for 20 min, and the recovery rate was calculated as the mean of 3 independent chamber runs. On average, 8 cfu/m³/min of S. aureus had to be dispersed by the nebulizer to recover 1 cfu of S. aureus

Molecular typing of selected S. aureus isolates was performed using the restriction endonuclease SmaI followed by pulsed-field gel electrophoresis (PFGE; BioRad) [12]. Cluster analysis of the banding patterns was used to determine clonal similarity (BioNumerics software version 3.0; Applied Math). Strains were considered to be different if they showed >3 fragment differences [13]. Isolates from the dominant strains were tested for methicillin susceptibility using the disk-diffusion test (1 μg of methicillin; control S. aureus strain ATCC 25923) [14]

Study designDaily bacterial air-sample cultures were performed for 16 consecutive days, with participants sitting in the chamber for two 20-min sessions, preceded by a control run without participants. In all, there were 528 chamber sessions. During the first or “nonsneezing” session, participants were wearing sterile garb (a fluid-resistant surgical gown [Medline Industries] over street clothes, surgical gloves [Ansell Healthcare Products], shoe covers [Spunbond Gripper; Medline Industries], and a bouffant cap [Spunbond Sheer-Fit; Medline Industries]). This was followed by a histamine-induced sneezing session with new sterile garb, gloves, shoe covers, bouffant cap, and the administration of intranasal histamine (25 mg/mL). The latter was applied to the anterior nares with a sterile cotton swab and distributed by gently pressing the exterior walls of the nose together [15]. This sequence of sessions was repeated on all 16 days. During chamber sessions, the number of sneezes, coughs, nose blows, talking, and unusual activities of the participant were recorded by directly observing study personnel. The participants were asked to avoid brisk body movements and to sneeze or cough in the direction of the safety hood without covering the facial area

Participants were exposed after day 2 to 1 of 3 rhinovirus serotypes (Hanks, 39, or 16) at an intranasal dose of ∼100 TCID₅₀, on the basis of their serosusceptibility (serum neutralizing antibody titers ⩽1:4 to the challenge virus) [7, 16]. The measurements of bacterial dispersal were repeated daily. Infection was determined by virus isolation from nasal lavage samples and by measurement of homotypic serum-neutralizing antibody titers on paired acute and convalescent (4-week) specimens. To assess illness, 8 symptoms (sneezing, rhinorrhea, nasal obstruction, sore or scratchy throat, cough, malaise, chilliness, and headache) were each evaluated daily on a scale of 0 to 4 [17]. The total symptom score is the sum of the individual symptom ratings. A cold was defined as being present if a subject had a total symptom score of ⩾6 and rhinorrhea on ⩾3 days and/or the subjective impression of a cold [17]

Data analysisData analyses were performed using STATA software (version 9.1; StataCorp). The outcome was defined as the daily dispersal of airborne bacteria—such as S. aureus CoNS, and other bacteria—at the individual level over the course of 16 consecutive days, in colony-forming units per cubic meter per minute. The 2 study sessions with or without exposure to histamine served as a categorical exposure variable. Model building and testing were based on random-effects negative binomial distribution for overdispersed count data [18, 19]. Effect measures are expressed as incidence rate ratios (IRRs). This distribution accounts for between-individual heterogeneity in rates of airborne bacteria, which can lead to overdispersion and an excess of zeros found in the outcome distributions. The random-effects negative binomial distribution also accounts for clustering due to repeated observations occurring within an individual over time and across the sessions

The actual model building was based on a hierarchical backward-elimination strategy that pared down from the most complex model to a final model. Potential risk factors influencing the exposure outcome association included the modifiable factors exposure to rhinovirus and allergies and the fixed factors sex and age. Effect-measure modification was assessed using the composite Wald test with an α-level set at .05 and Akaike’s [20] and Schwartz’s [21] information criteria [22]. The change-in-estimate technique was chosen to test for confounding, with a preset threshold of 10%. After calculation of the final models, model diagnostics were applied to detect potential influential cases. Because of the clustered nature of the data set, each individual participant was removed 1 at a time from the final models, and the change-in-estimate effects were determined. The threshold was preset at a 20% change-in-estimate effect

Results

Participants and measures of illness/infectionTwo women and 9 men (mean age, 24 years; range, 19–29 years) were studied. One volunteer smoked. All participants had ⩾1 symptom and met the criteria for having a cold. Figure 1 depicts the daily total cold symptom scores after exposure to rhinovirus. Table 1 lists the number of participants affected by specific cold symptoms and the maximum symptom scores during the 5 days after rhinovirus exposure

Volunteers most often had high symptom scores for nasal congestion, sore throat, and runny nose, followed by coughing, malaise, and headache. Sneezing and chills were reported to be less frequent and less severe

Seroconversion with an antibody increase of ⩾4 dilutions occurred in 6 volunteers. All participants were infected and had at least 1 nasal-wash sample positive for the challenge rhinovirus after exposure

Sneezing and other eventsDuring the histamine-induced sneezing sessions, volunteers sneezed 5.4±4.4 times before rhinovirus exposure and 4.9±2.8 times after exposure, respectively (P=.2622). Sneezing events were rare during the nonhistamine sessions (4 sneezes in 176 sessions). Nose blowing occurred more frequently during the sneezing sessions than during the nonsneezing sessions (1.61±1.38 vs. 0.01±0.15; P<.0001), but this did not change significantly after rhinovirus exposure (P=.1736). Coughing events were uncommon, with a total of 183 coughs during 352 chamber sessions and with no apparent difference between sessions or rhinovirus exposure (P=.7722). Reading was the only additional activity noted

Mucosal/skin culturesAll participants carried S. aureus in their nares before and after rhinovirus exposure (table 2). The mean number of S. aureus in the nose or pharynx did not change significantly after rhinovirus exposure (IRR, 1.02; P = .888). However, skin carriage with S. aureus doubled from before to after rhinovirus exposure (P<.001), whereas nose carriage with “other bacteria” decreased by 64% (P=.002). None of the other bacteria groups changed because of rhinovirus exposure

Airborne dispersal of S. aureus and other upper-respiratory-tract bacteriaA total of 788,179 cfu were detected in air cultures, including 2130 cfu of S. aureus (0.3%) and 12,205 cfu of CoNS (1.6%). Table 3 lists the mean counts of S. aureus, CoNS, and other bacteria for the sneezing and nonsneezing sessions before and after rhinovirus exposure

Tables 4 and 5 list the final models after testing for effect-measure modifiers and confounders. The comparison between sneezing and nonsneezing sessions showed a 4.71-fold increase in S. aureus a 1.43-fold increase in CoNS, and a 3.92-fold increase in other bacteria associated with sneezing (P<.0001). Individual S. aureus dispersal during the sneezing sessions varied from 0 to 1.77 to 278.96 cfu/m³/min. The presence of respiratory allergies multiplied the effect measure for S. aureus from a 4.71-fold to a 17.45-fold increase in airborne S. aureus dispersal in allergic participants (table 5). The histamine-induced frequency of sneezing did not differ between allergic and nonallergic participants (average number of sneezes per session, 5.15 for allergic participants vs. 4.70 for nonallergic participants). However, the airborne release of S. aureus by a single sneeze was significantly higher in allergic than in nonallergic participants (IRR_{nonallergic participants}, 1.13 [95% confidence interval {CI}, 1.095–1.18] [P<.001] vs. IRR_{allergic participants}, 1.34 [95% CI, 1.27–1.41] [P<.05]). Effect-measure modification was also noted for CoNS by age, with a decrease in airborne dispersal of CoNS with older age. Male participants shed larger amounts of other bacteria into the air than female volunteers. No influential cases were detected when model diagnostics were used

On average, 1 sneeze expelled 2.83 cfu of S. aureus/m³/min, 3.24 cfu of CoNS/m³/min, and 474.61 cfu of other bacteria/m³/min over and above the existing dispersal. The percentage of particles <5 μm carrying S. aureus increased from 16.2% during nonsneezing sessions to 37.6% during sneezing sessions (P=.0593). A modest increase was observed in other bacteria (42.9%–49.4%; P=.0827), whereas the percentage of CoNS did not change (30.5%–29.5%; P=.7740). Rhinovirus exposure led to a similar increase in small particles carrying S. aureus (0%–16.7% for nonsneezing sessions, P=.0161; 23.9%–38.6% for sneezing sessions, P=.0773). CoNS and other bacteria did not change significantly in the nonsneezing sessions. During the sneezing sessions, there was a statistically significant decrease in small particles for other bacteria (64.5%–48.0%; P < .001) but not for CoNS (P=.1730)

Sneezing, the common cold, and airborne bacterial dispersal Inoculation with rhinovirus at day 2 led to a 1.4–8.3-fold increase in the airborne spread of S. aureus CoNS, or other bacteria (table 3). However, the changes were not statistically significant in random-effects negative binomial models (P>.05). Furthermore, the cold-symptom score was also not associated with a change in the dispersal of bacteria into the air

S. aureus characterizationA total of 813 of 2130 cfu of S. aureus (38%) were studied by PFGE. The clonal similarity of airborne and skin/mucosal S. aureus isolates was 84%–100% in 9 volunteers. The remaining 2 volunteers carried 2 distinct S. aureus strains, each with a dominating clone accounting for 60% and 78%, respectively, of S. aureus isolates. All S. aureus strains were sensitive to methicillin

Discussion

Sneezing is a common physiological reflex that involves the sudden, uncontrolled, forcible expulsion of air through the nose and mouth. The causes are numerous and include allergic rhinitis, vasomotor rhinitis, nasal irritation, drug withdrawal, exposure to bright light, gastric distension, epilepsy, psychogenic factors, sexual excitement, menstruation, chilling, cervical tuberculous lymphadenitis, and the common cold [23]. In particular, the impact of the latter on sneezing and the airborne dispersal of bacteria provoked our interest, given that each adult has an average of 2 colds each year [24]. We focused primarily on S. aureus because it is among the most common nosocomial pathogens in the United States [25, 26]

Early studies demonstrated the large quantities of bacteria spread by sneezing into the environment, compared with other activities, such as quietly breathing, talking, or coughing. Hare and Mackenzie [27] found a 95-fold increase in all airborne bacteria due to sneezing: a 50-fold increase from talking to sneezing and a 34-fold increase from coughing to sneezing. However, sneezing did not yield the airborne dispersal of S. aureus. By contrast, Duguid and Wallace [28] found S. aureus in the air after induced sneezing in 2 volunteers. In the present study, we used more-sophisticated measuring techniques to show that sneezing causes a highly significant increase in S. aureus CoNS, and other bacteria in environmental samples. It is noteworthy that the range of responses varied widely among the individual participants, from <1 to 279 cfu of S. aureus/m³/min during sneezing sessions. Five volunteers dispersed >60 cfu of S. aureus/m³/min, whereas the remaining 6 volunteers did not expel >5 cfu of S. aureus/m³/min. This is likely explained by the higher frequency of sneezing in the first volunteer group, in which an average of 6.5 sneezes/session was seen, compared with 3.8 sneezes/session in the latter group (P<.05). It must also be noted that no other cold symptom or activity was statistically related to this increase

The actual number of bacteria set free by a single sneezing event may be a key parameter in risk assessment for cross-infection in nosocomial settings. Our results demonstrate that 1 sneeze expels, on average, mostly other bacteria, such as α-hemolytic streptococci (474.61 cfu/m³/min), but also 3.24 cfu of CoNS/m³/min and 2.83 cfu of S. aureus/m³/min. No clear threshold exists for the airborne infective dose of S. aureus. However, Forster and Hutt [29] conducted several experiments in which they inoculated skin lesions with broth dilutions that contained S. aureus. The lowest amount that produced infection in a standard skin lesion was 15 cfu of S. aureus/mL. Previous tests in our airtight chamber revealed that, for every colony-forming unit recovered by Andersen samplers, 8 cfu have to be dispersed into the air by the participant. Given that 2.83 cfu of S. aureus/m³/min were recovered per sneeze, ∼23 cfu/m³/min were expelled into the air by 1 sneeze. Thus, a sneeze could deposit a significant number of S. aureus on a patient. Furthermore, it is also possible that airborne S. aureus can be transferred to and colonize the mucus membranes of persons in close proximity. However, to our knowledge, no data are available to date supporting this route of cross-colonization

Multivariate analyses revealed several effect-measure modifiers for the different species of bacteria and groups. The presence of respiratory allergies led to an increase in airborne S. aureus. Shiomori et al. [30] found an association between perennial allergies and higher rates of S. aureus carriage. These higher rates may lead to increased airborne dispersal. This is supported by the results of Bassetti et al. [5], which showed a significant increase in airborne spreading patterns in a person with symptomatic allergic rhinitis. Independent predictors were sneezing and the number of tissues used. In our study, the increased spread was associated not with sneezing frequencies or nasal or skin carriage rates but with the amount of organisms expelled by a sneeze. One can speculate about differences in the intensities of the sneezes in allergic individuals, the mucous membranes, or mucus itself promoting the release of S. aureus. Further studies are necessary to clarify the underlying mechanisms of allergies of the respiratory tract and S. aureus airborne dispersal. This applies also to the effects of age and sex on the airborne spread of CoNS and other bacteria

When investigating the transmission of pathogens via air, the distance between source and potential target is crucial. Because of the closed environment of our test chamber, it was not possible to investigate how far S. aureus or other bacteria could be spread from participants. However, measurement of the particle sizes revealed that sneezing caused a significant increase in particles <5 μm in size carrying S. aureus from 16.2% to 37.6%. It is well known that organisms attached to these particle sizes are transmissible via the airborne route because of their capability to float through the air [31–33]. Therefore, prolonged and wide dispersal of S. aureus by air after sneezing can be assumed. Once in the environment, S. aureus isolates can survive for >60 days on common hospital materials, which presents an immanent risk for environment-to-person transmission [34, 35]

A further objective of our study was to determine the impact of a common cold on the airborne spread of upper-respiratory-tract bacteria during sneezing. We compared the number of infection-related and histamine-induced sneezing events during the chamber sessions before and after exposure to rhinovirus, and we found no significant change in frequency. This was also confirmed by the overall low score of sneezing symptoms over the course of a 24-h period that was reported daily by the participants. Induction of a common cold was not an independent risk factor for the airborne dispersal of S. aureus CoNS, and other bacteria in the comparison of nonsneezing with sneezing sessions. One has to mention that previous findings indicated a significant increase in the airborne dispersal of S. aureus due to a rhinovirus challenge [4, 6]. This was not confirmed in the present study, where the main focus was to study the effect of sneezing. The implementation of preventive measures for persons with mild to moderate common colds because of this higher bacterial air load appears to be unnecessary, unless sneezing is part of their symptom complex

The main source of airborne S. aureus could not be conclusively determined. High clonal similarity with >84% for single strain carriers did not allow us to distinguish between organisms generated from the nose, pharynx, or skin. However, when we compared the airborne spreading patterns of bacteria with their natural colonization areas, S. aureus and other bacteria usually found on mucus membranes were more frequently expelled during sneezing sessions than were skin colonizers, such as CoNS

It remains unknown whether other viruses, such as coronaviruses or influenza virus, might cause similar results. In addition, the significant increase detected in the group of other bacteria with ∼475 cfu/m³/min per sneeze warrants further study, given that pathogens such as penicillin-resistant pneumococci, Neisseria meningitidis and Streptococcus pyogenes can inhabit the human respiratory tract and might be similarly spread by airborne route [36–38]

In conclusion, our findings suggest that sneezing contributes to the risk of cross-infection by airborne transmission of S. aureus CoNS, and probably other upper-respiratory-tract bacteria, and they should be taken into consideration in future investigations of outbreaks. Mild to moderate common colds do not appear to increase the risk of bacterial spread unless sneezing is a part of the syndrome. However, individuals with respiratory allergies pose a potential source for airborne S. aureus spread, particularly when they sneeze