Environ. Sci. Technol. 2009, 43, 1783–1787
Framework for Evaluating Anthrax Risk in Buildings PHILLIP N. PRICE,* MICHAEL D. SOHN, KRISTINA S.H LACOMMARE, AND JENNIFER A. MCWILLIAMS Indoor Environment Department Environmental Energy Technologies Division Lawrence Berkeley National Laboratory Berkeley California 94720
Received May 30, 2008. Revised manuscript received January 5, 2009. Accepted January 22, 2009.
If Bacillus anthracis (BA), the organism that causes anthrax, is known or suspected to have contaminated a building, a critical decision is what level of contamination is unacceptable. This decision has two components: (1) what is the relationship between the degree of contamination and the risk to occupants, (2) and what is an acceptable risk to occupants? These lead to a further decision: (3) how many samples must be taken to determine whether a building is unacceptably contaminated? We discuss existing data that bear on these questions, and introduce a nomogram that can be used to investigate the relationship between risk of contracting anthrax, the surface concentration of BA, the probability of detection, and the number of samples needed to ensure detection with a given degree of certainty. The same approach could be used for other agents that are dangerous due to resuspension of deposited particles.
Introduction If Bacillus anthracis (BA), the organism that causes anthrax, is known or suspected to have contaminated a building, a critical decision is what level of contamination is unacceptable: “how clean is safe,” in the words of Vogt and Sorenson (1). This decision has two components (1): what is the relationship between the degree of contamination and the risk to occupants (2), and what is an acceptable risk to occupants? Once these decisions have been made, one must also decide how many samples to take to decide whether the building is unacceptably contaminated. With the sampling techniques and technologies applied in the response to anthrax releases in the United States in 2001, low spatial densities of BA spores (under about 1000 “colony-forming units” (CFU) per square meter) were unlikely to be detected. In essence, the “how clean is safe” decision did not have to be made explicitly, because the decision was implicitly determined by limitations on detectability. Even buildings that tested negative for BA in many samples may have had a moderate amount of BA contamination. As technologies and techniques improve, previously undetectable concentrations of BA will become detectable. At some point, it might reasonably be considered too expensive or disruptive to decontaminate an area with a very low, but detectable, level of contamination. At that point the “how clean is safe” decision must be made explicit. It is not reasonable to set “zero risk to occupants” as the threshold of acceptability when determining whether a * Corresponding author phone: 510-486-7875;e-mail: pnprice@ lbl.gov. 10.1021/es802506p CCC: $40.75
Published on Web 02/19/2009
2009 American Chemical Society
building needs to be decontaminated: there is a (very small) possibility that a single inhaled spore can cause anthrax. No detection technology can detect a single spore, so it is impossible to sample sufficiently to guarantee that there is no anthrax risk in a building. In this paper, we discuss the subdecisions that should be made to determine what level of contamination is acceptable, and how many samples must be taken to ensure that a building is not unacceptably contaminated. We also present a nomogramsa graphical method for finding the relationship between multiple variablessthat illustrates the relationship between risk of contracting anthrax, the surface concentration of BA, the probability of detection, and the number of samples needed in order to ensure detection with a given degree of certainty. This tool does not make the decision of what level of contamination to accept, but it does help explore the implications of a given decision. For example, the assumption that a given number of samples is sufficient to determine whether a building is unacceptably contaminated will be reasonable only if the BA dose-response relationship meets certain criteria, which can be determined using our approach.
Dose-Response Relationship for Anthrax BA can lead to two types of infection: “cutaneous” anthrax causes skin lesions and “inhalation” anthrax attacks the lungs. Inhalation anthrax is far more dangerous than cutaneous anthrax, and is responsible for essentially all anthrax deaths. Untreated inhalation anthrax is very likely to be fatal, so many authorities do not distinguish between LD50 (a degree of exposure that is lethal to half the people exposed) and ID50 (a degree of exposure that infects half the people exposed). In the present paper, we consider only inhalation anthrax and are thus concerned only with airborne spores. The dose-response relationship for inhaled BA is extremely uncertain, especially at low dose: there are few circumstances in which a large number of people are exposed to known numbers of BA spores. Extrapolation from animals is not very reliable because of differences in immune response and lung morphology between people and animals. LD50 and ID50 are uncertain by at least a factor of 5, and the dose-response relationship at low doses is much more uncertain because it relies on extrapolation by a factor of 100 or 1000 downward in dose. A complication in determining the dose-response relationship is that different populations (such as healthy people, immunocompromised people, children, etc.) have different dose-response relationships, which will be virtually impossible to determine in detail, so no single dose-response curve will apply to every situation. Several researchers (see refs 2-5 for example) suggest that 100 spores or even fewer might be enough to cause infection in rare cases. Wilkening (4) discusses five doseresponse models that predict the probability of infection from 100 spores to be from 10-1 to about 10-5, and applies these models to an analysis of an accidental outdoor BA release in the Russian city of Sverdlosk. Wilkening suggests that risk estimates below 10-3 from 100 spores appear to be inconsistent with Sverdlosk data: some people apparently contracted anthrax from very low exposures. Wilkening’s analysis suggests that the chance of contracting anthrax from 100 spores may be somewhere in the range of 0.1-10%. Although there is evidence that even 100 spores can cause anthrax at least in particularly susceptible people, data also VOL. 43, NO. 6, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1783
suggest very low infection risksprobably much less than 10-4sfrom inhaling 10 or fewer spores. The discussion above refers to the numbers of viable spores, i.e., spores that can cause infection. BA spore viability depends on environmental conditions, and viability will decrease with time, but there is very little quantitative information about this issue. It is known that BA spores are extremely hardy and can survive for decades in the right conditions.
Relationship between Surface Concentration and Dose BA spores deposited on surfaces do not present a direct threat of inhalation anthrax: they must first be resuspended into the air so that they can be inhaled. The relationship between surface and airborne concentrations depends on the type of surface and on the activity that causes resuspension. In this paper, we will consider resuspension from floors, and discuss sampling probabilities for samples taken from floors, for two reasons. First, floors present large, easily sampled flat surfaces that have high particle concentrations due to gravitational settling. Second, most published sampling protocols have been tested either on nonporous surfaces (such as tile) or on surfaces such as carpets. Considering floors alone may not give a complete picture: an anonymous reviewer notes that resuspension from furniture may be a significant source of airborne contamination. Information about resuspension rates and detection probabilities for furniture (or other objects such as contaminated papers) could be used to create an alternative to the surface-airborne concentration relationship that we discuss below. Several experiments have quantified the relationship between airborne and surface concentrations of biological agents (refs 6, 7 and discussion in the Supporting Information). Rather than measuring the relationship between surface concentration and airborne concentration directly, many experiments measure resuspension rates: the fraction of surface contamination that becomes airborne within a given time period. We calculate airborne indoor concentrations from resuspension rates as follows. Let Ca be the airborne concentration of particles, r the resuspension rate (the fraction of surface particles that become airborne, per unit time), A the floor area of a room, V the room volume, f the removal rate (the sum of deposition and ventilation, in room volumes per unit time), and Sd the number of particles per square meter on the floor. Then Ca )
rA S fV d
9
The high degree of variability in resuspension rates among surfaces and levels of disturbance is a significant challenge to determining an acceptable level of surface concentration. Perhaps experiments or theory can lead to a better ability to predict the resuspension rate for a given type of surface. Or, in situ experiments can be performed—along the lines of Weis et al. (6), using an air sampler to detect airborne concentrations while contaminated surfaces are disturbedsto empirically determine the relationship between surface and airborne concentration at the contaminated site.
(1)
For small particles (diameter 0.9) of resulting in a positive sample, using wet wipes and current culture techniques, for a nonporous surface. Thus, for a wipe that samples 100 cm2, detection is likely only if the surface is contaminated at 104 CFU per m2. If an entire meter can be sampled then detection is likely if the surface is contaminated at 100 CFU per m2, in good agreement with the findings of Buttner et al. (9). The results discussed above are from laboratory experiments on steel (which is nonporous) and wood (which is moderately porous). Detection limits for other common materials, such as carpet or paper, are not known. For such materialssespecially carpetsit is possible that a vacuum system would be more effective than wipe samples. Buttner et al. (8) found wipe samples to be about as effective on carpet as on tile, but they did not investigate vacuum samples. Using a standard logistic model (see ref 11 for example) for the relationship between concentration and detection probability, Figure 2 shows the approximate current situation as described in the literature, as of late 2008.
Number of Samples Required to Detect Contamination We can now attempt to determine the number of samples necessary in order to confirm that the level of contamination in a building or a region of a building is low enough that the risk of occupancy falls below a specified level. We do this as follows: (1) Decide on a level of risk that is acceptable. (2) Convert the risk to an airborne spore concentration, via an assumed dose-response curve. (3)Convert the airborne spore concentration to a surface concentration, using an assumed resuspension rate. (4) Convert the surface concentration to a probability that any single sample is positive for BA, using a sampling effectiveness curve. (5) Find the number of samples needed, for the single-sample probability determined in the previous step, such that if all the samples are negative then the building or region is “safe” with a specified certainty. The decision-maker can choose how certain they VOL. 43, NO. 6, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1785
FIGURE 3. Determine the number of samples needed to ensure occupant safety: begin in the upper left portion of the figure by selecting an acceptable risk of infection for a person who will occupy the region for one hour. Read downward to a dose-response curve, then horizontally to the right to a “resuspension” curve. Continue downward into the plot below to a sampling effectiveness curve. Finally, read over to the plot at the bottom right, to a curve corresponding to probability of 0.99 or 0.999 of obtaining at least one positive sample. Read downward to the resulting x-value to determine the number of samples needed (all of them negative) to attain the selected risk level. The dotted line on the figure shows an example. want to be. A single positive sample is assumed to be enough to detect that the building or region is contaminated to an unacceptable level, although in practice confirmatory sampling should be performed, of course. Figure 3 allows the calculation of the necessary number of samples: it connects the risk from a 1 h exposure to the number of samples that must be taken in order to confirm that a building or region is safe, with a specified level of certainty. It is an extension of Figure 1 and works the same way; in fact, the top two panels of Figure 3 are the same as Figure 1. The detection probability curves on the lower left panel are approximations to the results in refs 10 and 12. For the lower right panel, we assume samples are statistically independent. The probability Ppos of obtaining at least one positive sample out of n samples, given a probability p that a given sample is positive, is Ppos ) 1 (1 - p)n. We illustrate with Ppos ) 0.999 and 0.99 (a 99.9% and 99% chance that unacceptable contamination would be detected if it is present); other curves may be generated as desired. Figure 3 assumes the worst case, from the standpoint of detecting a very low level of contamination: it assumes that the building or region is uniformly contaminated (and that the resuspension rate is also uniform), so that the risk is equal everywhere. In the more realistic case that the level of contamination is high in some regions and low in others, the average risk can be the same as in the uniform case but the number of samples required may be smaller. If some area is much more heavily contaminated than other areas, a sampling scheme that causes samples to be taken in the 1786
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 6, 2009
heavily contaminated area will reduce the number of samples needed for detection. Such nonuniformity is expected in the case of an indoor release. For example, Sajo et al. (13) performed aerosol release experiments in a two-story, 21room office building served by a single HVAC sytem, using particles with a geometric mean diameter of 1.1 µm and geometric standard deviation of 2. They found that the deposited mass density was at least 10 times higher near the release location than anywhere else in the building. Away from the release location, spatial variation was modest: the highest concentrations were about 20 times higher than the lowest. If most of a building is contaminated at a level that would lead to a large fraction of negative tests, a “hot spot” with 50 or 100 times the surface concentration of the rest of the building should still be easily detectable. Given the large potential advantage to sampling at spatially dispersed sites, a dispersed sampling scheme should certainly be used. In many or most scenarios where contamination with BA is suspected, it will be possible to concentrate measurements in the area where the concentration of contamination is highestsin the mailroom, in the office of a victim, and so on. It is likely that the level of contamination in these areas will be high enough that the probability that occupants will contract anthrax is unacceptably high and decontamination decisions will be straightforward. However, even in these scenarios there may still be other wings or floors of the building in which the spore concentration is very low and fairly uniform. Figure 3 will apply to such areas. The figure will also apply if BA is introduced to a building in a way that
produces a rather uniform distribution of surface contamination, such as through a ventilation system.
Discussion The “how clean is safe” decision is hard. The BA dose-response is very uncertain even at low doses, but it is apparent that exposure to quite low numbers of sporessprobably on the order of dozens or hundredsscan cause fatal anthrax infection in rare cases. Decisions concerning the amount of sampling to perform, and how to respond to very low levels of contamination detected by that sampling, may involve a level of contamination so low as to be barely detectable, but that may still be judged unacceptable. Once a risk level is chosen, Figure 3 can be used to determine the required number of samples to be sure the building is safe at that level. Attempting to be very “conservative” (i.e., highly protective) in all of the subdecisionssthe choice of an acceptable risk, the dose-response relationship, the resuspension rate, the assumed sampling effectiveness, and the level of certainty that a building is “safe enough”swill lead to infeasible or even ludicrous decisions. For example, choosing 10-8 as an acceptable risk per person-hour, and assuming high risk per dose and a high resuspension rate, will lead to the conclusion that even 0.1 spores per square meter is an unacceptable level of contamination. Thousands of measurements could be made in a building uniformly contaminated at this level, without necessarily finding even a single positive result. Decision-makers faced with such a calculation will be forced to revisit their assumptions: could the risk per dose really be that high? Is there suspension rate really that high? If the answers are “yes,” they may either have to decontaminate or at least shut down buildings for which there are no positive results but for which contamination is strongly suspected, in the absence of sampling methods that are far more effective than those presently available. In addition to determining the number of samples to take as a function of the acceptable risk, Figure 3 can also be used in the reverse direction. For instance, some of the buildings that were tested for BA as part of the 2001 anthrax incident had a total of five samples taken. What level of safety is implied by five negative samples in a potentially contaminated building? Begin in the lower right panel of Figure 3: this figure tells us that if we take five independent samples in a contaminated building, then there is a 99.9% probability of obtaining at least one positive sample as long as each sample has a 75% chance of testing postitive. Moving to the left panel, we see from the “field experience” curve that the probability of a positive detection probably exceeded 75% only if the surface concentration was above about 10 000 spores per square meter. Thus, classifying a building as safe on the basis of five negative samples was, implicitly at least, expressing a willingness to accept a concentration of about 10 000 spores per square meter. Perhaps decision-makers would have called for more samples in the 2001 response if this implication had been understood. In this paper we have not proposed an answer to the question “how clean is safe”; instead, we have explored the key relationships that, if quantitatively known, would de-
termine the answer to that question. We see the logical relationships exemplified in Figure 3 as our major contribution to the discussion. The same approach could be applied to other chemical or biological agents for which resuspension of deposited particles is hazardous.
Acknowledgments This work was supported in part by the Office of Chemical and Biological Countermeasures of the Science and Technology Directorate of the Department of Homeland Security and performed under U.S. Department of Energy Contract No. DE-AC02-705CH11231.
Supporting Information Available Selected references and discussion of the anthrax doseresponse relationship, resuspension data, etc. This material is available free of charge via the Internet at http:// pubs.acs.org.
Literature Cited (1) Vogt BM and Sorensen JH, “How Clean is Safe? Improving the Effectiveness of Decontamination of Structures and People Following Chemical and Biological Incidents”, ORNL/TM-2002/ 178, Oak Ridge National Laboratory, 2002. (2) Fennelly, K. P.; Davidow, A. L.; Miller, S. L.; Connell, N.; Ellner, J. J. Airborne infection with Bacillus anthracissfrom Mills to Mail. Emerging Infect. Dis. 2004, 10, 996–1001. (3) Glassman, N. H. Discussion of industrial inhalation anthrax. Bacteriol. Rev. 1966, 30, 657–659. (4) Wilkening, D. A. Sverdlosk Revisited: Modeling Human Inhalation Anthrax. Proc. Nat. Acad. Sci. U. S. A. 2006, 103, 7589–7594. (5) Peters, C. J.; Hartley, D. M. Anthrax inhalation and lethal human infection. Lancet 2002, 359, 710–711. (6) Weis, C. P.; Intrepido, A. J.; Miller, A. K.; Cowin, P. G.; Durno, M. A.; Gebhardt, J. S.; Bull, R. Secondary aerosolization of viable Bacillus anthracis spores in a contaminated U.S. Senate office. J. Am. Med. Assoc. 2007, 288, 2853–2858. (7) Buttner, M. P.; Cruz-Perez, P.; Stetzenbach, L. D.; Garrett, P. J.; Lutke, A. E. Measurement of airborne fungal spore dispersal from three types of flooring materials. Aerobiologia 2002, 18 (1), 1–11. (8) Buttner, MP,; Cruz-Perez, P, and; Stetzenbach, L. D. Enhanced detection of surface-associated bacteria in indoor environments by PCR. Appl. Environ. Microbiol. 2001, 67, 2564–2570. (9) Buttner, M. P.; Cruz, P.; Stetzenbach, L. D.; Klima-Comba, A. K.; Stevens, V. L.; Emanuel, P. A. Evaluation of the biological sampling kit (BiSKit) for large-area surface sampling. Appl. Environ. Microbiol. 2004, 70, 7040–7045. (10) Brown, G. S.; Betty, R. G.; Brockmann, J. E.; Lucero, D. A.; Souza, C. A.; Walsh, K. S.; Bougher, R. M.; Tezak, M.; Wilson, M. C. Evaluation of wipe surface sample collection method for bacillus spores from non-porous surfaces. Appl. Environ. Microbiol. 2007, 73, 706–710. (11) Ellison, S. L. R.; Fearn, T. Characterising the performance of qualitative analytical methods: Statistics and terminology. Trends Anal. Chem. 2005, 6, 468–476. (12) Buttner, M. P.; Cruz, P.; Stetzenbach, L. D.; Klima-Comba, A. K.; Stevens, V. L.; Cronin, T. D. Determination of the efficacy of two building decontamination strategies by surface sampling with culture and quantitative PCR analysis. Appl. Environ. Microbiol. 2004, 70, 4740–4747. (13) Sajo, E.; Zhu, H.; Courtney, J. C. Spatial distribution of indoor aerosol deposition under accidental release conditions. Health Phys. 2002, 83, 871–883.
ES802506P
VOL. 43, NO. 6, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1787
