Airborne Infection with Bacillus anthracis -- from Mills to Mail

Kevin P. Fennelly; Amy L. Davidow; Shelly L. Miller; Nancy Connell; Jerrold J. Ellner


Emerging Infectious Diseases. 2004;10(6) 

In This Article


These analyses emphasize that the risks for airborne infection are determined not only by the virulence of the agent but also by environmental factors, e.g., room size and ventilation rates, and host factors, e.g., pulmonary ventilation rate. Little can be done to change one´s pulmonary ventilation rate, but these data suggest that risks for inhalation anthrax are likely to decrease considerably with fairly modest increases in room ventilation for low-concentration exposures. These data might be helpful in educating both the general public and policymakers regarding strategies to reduce risks from aerosolized bioweapons.

General dilution ventilation reduces the risk of transmitting airborne pathogens; however, as described for M. tuberculosis, room ventilation is theoretically limited in decreasing this risk, especially when the exposure concentration is high.[14] UV germicidal irradiation has been suggested as a potential bioterrorism countermeasure.[18] Although we are not aware of data regarding the use of irradiation to inactivate B. anthracis, Peng and colleagues found that UV germicidal irradiation of upper-room air reduced B. subtilis spore concentrations by 46% to 80%, depending on the ventilation rate.[19] Thus, further research on the use of UV germicidal irradiation and other interventions seems warranted.

These modeling data suggest that approximately 0.7–63.4 million total airborne spores and approximately 0.2–20 million spores <5 µm may have been produced by contaminated mail during work shifts at the two USPS processing and distribution centers where inhalation anthrax developed in postal workers. These data are plausible, given the contamination with B. anthracis documented by environmental sampling of both the New Jersey and the Washington, D.C., facilities.[6,20] Although the results of air sampling were negative, dust samples grew 3–9.7 million CFU/g.[20] A contaminated letter in Washington, D.C., purportedly contained 2 g powder with 100 billion to 1 trillion spores per gram,[21] which suggests that letters can be a source of infectious material.

These data also suggest that a few cases of inhalation anthrax can be expected if large numbers of people are exposed to a virulent strain at low concentrations within homes or offices. We will probably never know how many persons were exposed to mail contaminated with B. anthracis in the fall of 2001. If 10,000 persons were exposed in a home or office to secondarily contaminated mail, and two persons were infected, these models suggest that exposure was probably to 0.1 or 0.01 quantum, or 18–863 airborne spores. Above and below that range of exposures, the number estimated to be infected is very large or very small. Even if 1,000 persons were exposed, this range of exposures appears most reasonable, except for exposures to 1 quantum in larger or well-ventilated environments.

Glassman suggested that 100 spores may be sufficient to cause infection,[16] and a more recent analysis suggested that as few as 1–3 spores may be sufficient to cause infection.[22] Our data are consistent with these conclusions, as 0.001 quantum would be 2 to 9 bacilli in our model.

We assumed in our analyses that the strain used in the mailings was at least as virulent as the strain reported in the outbreak of inhalation anthrax at the Manchester mill.[15] Genomic analysis recently identified the isolate used in the mailings as the Ames strain, which is used in multiple research laboratories.[23] However, the strain was prepared for aerosolization in powder form, and we are not aware of published data on infectious doses for powdered preparations. The method of preparing bacilli for aerosolization may affect retention of aerosolized agents and their virulence of B. anthracis. For example, the surface-active compound Tergitol was found to increase virulence 10-fold in guinea pigs.[24]

The two women with no identified exposures in the recent bioterrorism-related outbreak may have been unusually susceptible to inhalation anthrax. Risk factors for increased susceptibility or resistance to inhalation anthrax are not known, although age has been suggested.[25] One of these women was 61 years old,[2] and the other was 94 years old.[1] Aging is associated with a decrease in mucociliary clearance as well as alterations in immune responses, but given the large numbers of elderly persons who were likely exposed to contaminated mail, age alone seems inadequate to explain the epidemiology of this outbreak. We are reluctant to accept that children are less susceptible than elderly adults without additional data. Previous reports of cases associated with industrial sources or with materials contaminated by B. anthracis suggest that some susceptible persons, including children, were infected during relatively brief exposures.[26–28] We speculate that the combination of thin body type, age, and female sex might be a risk factor for inhalation anthrax, as it is for pulmonary infections with environmental nontuberculous mycobacteria, e.g., with M. avium complex.[29] Women with asymptomatic bronchial hyperresponsiveness have increased deposition of inhaled particles of ≈ 1 µm,[30] but we are not aware of data that show this to be a risk factor for respiratory infections.

Modeling studies of transmission of tuberculosis or measles have usually assumed a single uniform pulmonary ventilation rate in exposed persons as well as a homogenous concentration of infectious aerosol over time and space. Riley´s choice of 10 L/min to model human ventilation has been used by other modelers,[14] but we have added a broader range of values to assess the relative importance of variability in minute ventilation. However, human ventilation is not uniform, being punctuated by sighs in which the tidal volume may be three times greater than the volume at rest.[12] If a piece of contaminated mail were opened, aerosol was likely concentrated immediately afterwards and in the immediate vicinity. Thus, if a person were to sigh when exposed to the maximum number of airborne spores, spores would likely deposit in greater concentration. However, we are not aware of data documenting deposition associated with short-term irregularities in ventilation. This is another area for further research.

Druett first applied the probit model in early studies of inhalation anthrax in animal models.[31] He acknowledged that this model best fit the data within the range of the LD25 –LD75 but cautioned about interpreting data outside that range. Our concern with estimating the number of bacilli required to kill roughly 0.1%–1.0% of exposed persons (10–100/10,000) was one of the reasons we chose not to apply the Druett model. We selected the Wells-Riley model because of a number of clear and compelling traits. First, data generated by the Wells-Riley model agree with observed data in airborne infection with measles and tuberculosis.[4,5] Neither the Druett model nor other models of airborne infection with more specified variables and more complex mathematical forms[32,33] have been validated against epidemiologic data of any airborne infection. Second, the Wells-Riley model includes variables for environmental conditions and aerosol production rates not accounted for in the Druett model. Alternative models might allow for a sigmoidal growth in the probability for infection, i.e., the probability of infection increases monotonically; however, the rate of change first increases to a maximum and then decreases steadily to zero. While such a model is biologically plausible, experimental data to justify such a model are lacking. No models have been validated for inhalation anthrax in humans, which differs markedly from pulmonary tuberculosis and measles in its pathogenesis.

One possible approach to extending the Wells-Riley model in the future would be adding a susceptibility factor, y. With the current model formulation, as q increases, airborne infection approaches 100% certainty, i.e., the probability approaches 1. Increases in the other parameters (I, t, p and 1/Q) have a similar effect on the probability for airborne infection. An alternative model might be given by C/S = y[1-exp(-Iqtp/Q)] where 0 < y < 1. With such a model, as any of the parameters I, q, t, p, or 1/Q increase, the probability of airborne infection approaches 1. The factor y may be conceived as a fixed or random effect, or one that is conditional on measurable, individual-level susceptibility factors such as age or coexisting conditions. Additional data on susceptibility are needed, however, before such a parameter can be included with confidence.

Modeling depends on the assumptions used; the most critical of our assumptions is that all affected persons were exposed to similar quanta of B. anthracis from contaminated mail. More likely, the number of quanta decreased with "postal distance" from the index letters. Webb and Blaser allowed for this possibility by modeling the cross-contamination of letters resulting from contact with index letters passing through a prototypical postal system.[25] Their model, however, did not allow for the environmental considerations that can mediate between aerosolization and infection. The Wells-Riley model allows for environmental considerations, and it can be adapted to variable exposure scenarios by stratifying by persons sharing a common quantum of exposure. To do that would require further modeling assumptions in the fashion of those made by Webb and Blaser. Epidemiologic data will likely never become available to test the validity of any of these models, but this modeling exercise demonstrates how the risk for infection is sensitive not only to the infectious dose but also to environmental parameters.