Analysis of Evidence

Studies located in the above search were classified according to agent, disease, species, and study method, and these data were then entered into an online database of animals as sentinels of human environmental health hazards (available from http://canarydatabase.org/) for further analysis. For the purposes of this review, we created a taxonomy for evidence regarding animals of sentinels based on existing evidence-based medicine taxonomies that provide guidelines for assigning levels of evidence based on quality and consistency of scientific studies (5). We considered level 1 evidence studies to include experimental studies, cohort studies, and systematic reviews of high-quality studies with consistent findings. We classified case-control studies and cross-sectional surveys of animals as level 2 evidence. Evidence from professional consensus statements, textbooks, and descriptive case reports was classified as level 3 evidence. To arrive at an overall strength of recommendation based on a body of evidence, we used these levels of evidence to determine the overall strength of the recommendation that animals could serve in the 3 sentinel capacities of early warning signal, ongoing exposure indicator, and potential propagator and spreader of a bioterrorism agent.

Evidence That Animals Provide Early Warning of an Acute Bioterrorism Attack

For a number of agents, this review found evidence that animals might be affected before human populations. For Bacillus anthracis, whether animals would have a shorter incubation period in the event of an aerosol release was not clear, since the incubation period in the 2001 mail attacks was <4 days, while during the 1979 release of B. anthracis from a Soviet military laboratory, human symptoms began in 2 days, with death in as few as 6 days (6). At the same time, while human cases in Sverdlovsk were concentrated along the path of the prevailing wind <4 km from the laboratory, livestock, including sheep and cows, began dying 3 days after the release in 6 villages located along the path of the aerosol at a distance <50 km downwind from the facility. No human cases were reported in these villages. Calculations of the airborne B. anthracis dosage at a town where several sheep and a cow died indicate that the animals were exposed to a dose more than an order of magnitude lower than humans received near the weapons facility. This finding suggests that sheep and cows are more susceptible than humans, although they could have also remained outside in the path of the aerosol for a longer period, which led to greater exposure (6).

For Yersinia pestis, evidence from experimental inhalation studies in cats indicates that the usual incubation period for symptoms of plague to develop after an inhalation exposure may be shorter (1–2 days) than the presumed incubation time for humans (1–6 days), which provides evidence that symptoms develop in cats at the same time as in humans, and thus may have some sentinel value. In contrast to the findings for anthrax and plague, however, we were unable to find evidence that animals could provide early warning of infection with airborne Francisella tularensis. During a prolonged outbreak of pneumonic tularemia in Scandinavia, for example, febrile illness developed in a number of horses, a cow, and a pig, but apparently not before the onset of disease in humans living nearby (13).

For foodborne illnesses, including botulism, animals would likely not manifest illness before humans if an attack were directed at humans, since in a typical attack scenario, food would be infected during the distribution pathway before consumption by humans and not necessarily allow for animal consumption before this. We did not locate reports of animals becoming symptomatic with foodborne illness before the onset of human cases. However, if the attack on the food supply were directed at the animals themselves, they could potentially manifest symptoms before humans who would consume the meat, eggs, or dairy products (24). Attacks on water supplies with agents such as Clostridium botulinum could put humans at risk as well, although dilution and water treatment would reduce the risk (37). An exception may be Cryptosporidium spp., which have caused widespread outbreaks through the water supply. Waterfowl die-offs from type C and type E botulism have been well documented, although these types are not well recognized as causes of clinical C. botulinum poisoning in humans, but the fact that primates are susceptible to type C makes C. botulinum poisoning a possibility. On the whole, however, an attack on human populations with C. botulinum would probably not be first detected in animals; the illness would have such a short incubation period in humans that they would become symptomatic at the same time as or before the animals.

For alphaviruses, natural outbreaks have often appeared in animal populations before they affected humans, for example, eastern equine encephalitis virus often appears in equines 2 weeks before humans become symptomatic (28). Whether the same pattern would hold true during an attack with an aerosol is not clear. For certain newer agents, such as filoviruses and Nipah virus, current evidence is insufficient to state whether after a generalized release of an aerosolized agent, animal infection would precede that in humans. Studies in Africa have demonstrated that Ebola virus outbreaks can be preceded by deaths in primates as well as in other animal species such as duikers (type of antelope) (17), but whether a generalized attack that used Ebola virus in the United States would affect certain animal species first is unknown.

For a number of agents, including Brucella spp., Coxiella burnetii, and hantavirus, infection in animals is either asymptomatic or develops so slowly that recognizable human cases seem certain to precede animal cases if the agents are released as an aerosol. Finally, the illnesses caused by some agents appear to have shorter incubation times in animals, for example, the 12-hour incubation period for Rift Valley fever in calves and lambs (23) compared to the incubation period of several days in humans.

Evidence That Animals Could Be Markers for Ongoing Exposure Risk

After the acute release of a bioterrorist agent, public health officials could be faced with the possibility of an agent persisting in the environment. Anthrax spores can survive for years in soil. Therefore, monitoring for sporadic cases in animal populations such as livestock could indicate so-called exposure hot spots. Although dogs and cats are less susceptible to B. anthracis than ruminants, their proximity to humans and their contact with soil could make them sentinels; for example, anthrax developed in a Labrador retriever after the dog hunted in a freshly plowed field (8). In the case of ongoing exposures to an agent that has become established in an animal population, case detection could be useful; this situation has been seen for plague in cats (9). Similarly, agents like Brucella spp. and C. burnetii, although they do not cause severe acute illness in animals, could be detected by recognizing increased rates of abortion among a variety of species.

Aside from case detection, active surveillance with surveys of animals may be useful; this surveillance may require testing wildlife as well. Such testing could involve antibody seroprevalence or use of polymerase chain reaction techniques to detect antigen. Evidence about the usefulness of such an approach was inconsistent. During an epidemic of pneumonic tularemia, attributable to contaminated hay, the etiologic agent, F. tularensis, can persist in the environment; a serosurvey of asymptomatic livestock (horses and cows) did not show evidence of exposure (13). By contrast, in a more recent tularemia outbreak, serosurveys in the wildlife population did show antibodies in a skunk and a rat that lived near persons who had become infected after mowing fields (14).

Evidence That Animals Could Propagate an Epidemic Caused by a Bioterrorism Agent

A number of biological terrorism agents have little potential for secondary spread in either animal or human populations, including B. anthracis and C. botulinum. For other agents, however, we found evidence that their introduction into an animal population could cause an epizootic that would then place additional human populations at risk. For example, studies of mosquitoes native to the United States have demonstrated their potential to spread a disease such as Rift Valley fever through livestock and other animal populations (29), even though person-to-person transmission does not occur. The results of animal surveillance for Ebola virus in Africa found that ongoing outbreaks in both primates and duikers suggest that the virus may be able to propagate in a wildlife population (17), however, this characteristic has not been demonstrated in US wildlife species.

Agents such as C. burnetii and Brucella spp. spread easily in animal populations through direct contact and can then pose a wider risk to humans, even though human-to-human transmission does not occur. Agents such as alphaviruses that are prevalent in wild bird populations can spread over a wide area in a short time (27). Experimental studies have documented that viruses such as West Nile virus can easily spread from animal to animal in bird populations (36).