Bioterrorism Detection The Smoke by wulinqing


									                                             Bioterrorism Detection: The Smoke Alarm and the Canary

Bioterrorism Detection:
The Smoke Alarm and the Canary

Brian M. Sullivan
Northrop Grumman Mission Systems, Tactical Systems Division

    The bioterrorism incidents of October 2001 brought our vulnerability to
    biological agents into sharp focus and elevated our level of concern.
    Biological agents are easy to manufacture, conceal, and release, making
    bioterrorism difficult to prevent. As a result, the practical front line of
    biodefense is to determine the presence of an attack as quickly as possible,
    enabling optimally effective decontamination, containment, and treatment.
    Agents can be detected either directly, using sensors, or indirectly by diagnos-
    ing their effects on a target population. Both approaches currently have
    significant drawbacks, but they are being combined to allow an effective,
    synergistic, affordable defense. This article assesses the combination of
    biosensors and epidemiology, which appears to provide a more cost-
    effective capability to detect, respond to, and recover from a bioterrorist
    attack than does either approach used alone. The combination will be less
    expensive than systems that use only sensors and more effective than
    systems that use only epidemiology. The next steps are to quantitatively test
    the asserted benefits of the combination, using modeling, and to implement
    the system in a real-world application, such as defense of fixed military

In planning a defense against biological terrorism (bioterrorism)—that is, the deliberate
introduction of a biological agent (toxin, virus, or bacterium) to sicken or frighten a
population—prevention is a laudable but idealistic goal. Unfortunately, toxic agents are
available from naturally occurring sources and relatively easy to produce or grow.
Weaponization of biological agents to deliver them efficiently to large numbers of
people is more challenging. Nevertheless, it is significantly easier than constructing a
nuclear weapon. As a result, preventing bioterrorism is virtually impossible. An aggres-
sor with either very good connections or a combination of expertise, determination, and
funds can mount a successful debilitating attack. Given the animosity in the geopolitical
climate and the apparent willingness of certain enemies to use bioterrorism, it is not
unreasonable to think that a repeat or variant of the anthrax letter attacks of 2001 is at
least a realistic concern—and probably inevitable.
As a first step in protecting against bioterrorism, we have developed an architectural
design for a system to protect a fixed military installation from bioterrorism. Here we
describe the advantages and disadvantages of the two main building blocks of that
system: networked biosensors and an epidemiological monitoring system. We detail
how the two building blocks can work together to compensate for the failings of each
working alone. Finally, we summarize our ongoing efforts in this area and provide a view

                     Technology Review Journal • Spring/Summer 2003                            135
Bioterrorism Detection: The Smoke Alarm and the Canary

 for the future, given the anticipated near-term need to protect fixed military installa-
 tions—and the communities of which they are a part—against bioterrorism.

 Detection of Bioterrorism
 Many terrorist attacks are obvious. When airplanes flew into the World Trade Center
 towers, people quickly realized what had happened and acted to minimize the damage. As
 a result, even though the buildings were dramatically destroyed, the number of people
 who died was much smaller than the total number of occupants at the time of attack.
 Bioterrorism, on the other hand, is insidious. A deadly biological agent can easily be
 released and spread—using a leaf blower, for example—long before an attack is even
 suspected. Without sensor-based detection, the first indication that something is wrong
 would be an unusually large number of people seeking treatment for the same unusual
 Sensor-based Detection: The Smoke Alarm. Sensor-based approaches would be
 preferable if biosensors were as functional, small, and inexpensive as smoke alarms.
 Leap-ahead sensor technologies that are small, capable, and inexpensive are expected to
 be available between 2007 and 2012. In the meantime, customers are forced to make
 tradeoffs between what they want and what they can afford. As a result, there is a
 mapping between applications and the optimal technology for each, based on perfor-
 mance, cost, and size.
 Epidemiological Monitoring: The Canary. For the general civilian population, a less
 expensive near- and medium-term approach is needed. People would naturally seek
 medical care for early symptoms after an attack. Therefore, monitoring of large-scale
 health care patterns can provide a valuable alternative or complementary method for
 discerning a bioterrorist attack.
 Sufficiently large shifts from regular epidemiological patterns would focus more
 detailed sensor-based analysis, possibly followed by coordinated treatment, contain-
 ment, and decontamination. The clear disadvantage of this approach is that people must
 become sick or even die, like canaries in coal mines, before a response is possible.
 Nonetheless, their unusual symptoms would be noticed and combined to contribute to
 the greater good. Diverse types of data and analytical approaches can contribute to such
 an epidemiological system, with the ancillary benefit of augmenting the civilian health
 Advantages of Timely Detection and Consequences of No Detection. For every type
 of biological agent, an absence of immediate detection severely limits the effectiveness
 of decontamination, intervention, and treatment. Decontamination is impossible until an
 attack is detected, so the primary attack lingers, continuing to infect new victims as long
 as it is undetected.
 For contagious agents with a long incubation period—such as smallpox—undetected
 infection is allowed to progress, not only harming the infected, but also making them
 infectious. Each person who was infected during the initial attack then becomes part of
 the biological weapon, creating a “secondary attack” by transmitting the disease until
 diagnosis and quarantine occur. Treatment is much more effective when the infection
 can be anticipated, instead of when it is treated retroactively after symptoms emerge.

136                        Technology Review Journal • Spring/Summer 2003
                                               Bioterrorism Detection: The Smoke Alarm and the Canary

For noncontagious agents with a long incubation period—such as anthrax—there is no
secondary attack, but treatment is much more effective when it is prompt and even
In the case of toxins, the rapid and specific effects of the toxin cannot be treated until
the toxin has been identified.

Expenses and Benefits of Biosensors (The Smoke Alarm)
Effective detection offers dramatic advantages in defending against and responding to a
bioterrorist attack. However, large-scale deployment of biosensor systems will incur
two main costs, financial and organizational.
Financial Costs. Biosensors are expensive. One class of devices—those that use
fluidics—includes air sampling, fluid handling, and analyses based on either antibodies
or deoxyribonucleic acid (DNA). Such systems cost about $200,000. Operation and
maintenance (O&M) costs for fluidics-based systems range from about $2 per test for
antibodies to tens of dollars per test for polymerase chain reaction (PCR), the most
common DNA-based test.
Such detectors also frequently include an optional generic detector, called a trigger,
able to determine that a threat might exist and initiate a more detailed analysis. Suitable
for use as simple stand-alone detectors, triggers currently cost between $20,000 and
$50,000. The O&M costs for triggers are much more reasonable, because they usually
require only electricity and occasional replacement parts.
A third class of devices is based on mass spectrometry. Nearly ready for widespread use,
such devices are expected to be similar in size and cost to those of the fluidics-based
detectors, but with much smaller O&M costs—pennies per test.
Organizational Costs. Biosensors can be disruptive. If biosensors are biological
alarms analogous to smoke alarms, then activation of biosensors will initiate a
“biodrill,” similar to a fire drill. Unfortunately, a biodrill is much more disruptive than a
fire drill. Instead of employees simply going outside for 20 minutes until firemen
ascertain that all is safe, operations are halted for at least a day—and possibly as long as
a week—while scientists determine whether, for example, the area is contaminated by
anthrax. Deployment of biosensor equipment for many real-world applications is
feasible only if false alarms are extremely infrequent. Plus, false alarms for anthrax are
exacerbated by the bacteria’s natural occurrence in much of the United States, particu-
larly near large concentrations of cattle.
Downstream Benefits of Biosensors. Deploying biosensors in various locations
will provide significant future benefits to human health. When the threat of biological
terrorism subsides, the sensors can be used to perform tasks such as detecting new
strains of influenza- or drug-resistant bacteria and monitoring the spread of naturally
occurring diseases.

Epidemiology-based Detection: Collecting and Synthesizing
Data (The Canary)
Epidemiology—or the study of the causes, distribution, and control of disease in
populations—can synthesize huge amounts of data to paint a picture of the current state

                      Technology Review Journal • Spring/Summer 2003                             137
Bioterrorism Detection: The Smoke Alarm and the Canary

 of health of a target population. The clarity of the resulting picture is highly dependent
 on the amount of natural variation in the data (e.g., from the annual flu season) and the
 quality of both the data and the data analysis. From such a picture, though perhaps
 unclear, we hope to be able to infer the existence of a recent bioterrorist attack.
 Typical data streams that indicate the health of a population are absenteeism, over-the-
 counter drug sales, and medical diagnoses, discussed below. However, large-scale
 epidemiological monitoring is a rapidly evolving field. Certainly, more data streams will
 be found useful in the future and data analysis techniques will improve.
 Absenteeism. When the symptoms of a bioterrorist attack become apparent in the
 target population, one of the first things that infected people will do is to decide to take
 the day off from school or work. Entering absenteeism rates into an epidemiological
 monitoring system can thus provide useful information.
 Over-the-Counter Drug Sales. After, or even instead of, deciding to stay home for
 the day, infected people will frequently seek over-the-counter medications. The target
 symptoms of these medications can be correlated to particular diseases.
 Medical Diagnosis: Horses and Zebras. When symptoms persist, people will visit a
 physician. Interestingly, physicians are explicitly told to disregard the possibility of
 low-probability diseases. In particular, they are cautioned, “When you hear hoofbeats,
 think horses, not zebras.” The challenge is to encourage physicians to report symp-
 toms—including those they consider routine. It is important to consider the context of
 that goal. Physicians are currently required to report cases of a number of diseases, but
 rarely do so.

 Our Roles in Biosensors and Epidemiology-based Detection
 The fields of biosensing and epidemiology are rapidly evolving. Northrop Grumman is
 playing an active role in “defining the future” of both areas.
 Biosensors. Northrop Grumman is significantly involved in developing the full range of
 technological approaches for biosensors, with examples for all phases in the technology
 development life cycle, starting with concept development and research, continuing with
 system engineering, then integration and test, and sustaining with O&M.
 Attractive technical options, summarized in Table 1, are currently available to detect
 biological threat agents. Each has its strengths and weaknesses. The technologies are
 ordered in Table 1 according to the time required for detection. (For a more detailed
 description of detection technologies, see Sullivan et al. [1].)
 • Standoff sensors detect in advance of an agent cloud passing over the sensor. They
      usually detect bioagents, based on passive or active optical fingerprints of biologi-
      cal materials, such as infrared emission or ultraviolet fluorescence, respectively.
      Standoff sensors can broadly discriminate large amounts of biological and
      nonbiological materials.
 • Aerosol trigger sensors discriminate biological and nonbiological materials within
      a few seconds after an agent cloud passes over the sensor. Aerosol trigger sensors
      can operate using some of the same physical principles as standoff sensors, with
      additional discriminators such as particle size or shape, based on low-angle scatter-
      ing of laser light.

138                        Technology Review Journal • Spring/Summer 2003
                                                      Bioterrorism Detection: The Smoke Alarm and the Canary

Table 1. Options for sensor-based detection

                                 Anticipated                                                   Northrop
      Class of        Speed of Accuracy of      Example     Example             Primary       Grumman
      Sensor          Detection Identification Application Technology          Drawback      Involvement

    Standoff:         Detects in    Crude      Detects to       High-power     Not yet       Provide ultra-
    Biological/       advance                  warn from        laser          proven        violet laser, air
    nonbiological                              distance                                      traffic control
                                                                                             radar R&D

    Aerosol           <30 s         Accurate   Detects to       Miniaturized   Not sensitive/ R&D
    trigger:                                   warn on          lasers for     discriminating prototype
    Biological/                                location,        scattering,    enough
    nonbiological                              e.g., building   fluorescence
                                               and air

    Aerosol           3 to 5 min    Crude      Detects to       Mass spec-     Not            R&D
    identification:                            warn/treat       trometry       discriminating
    Mass                                       on location                     enough

    Fluidics          15 to 60      Accurate   Detects to       Anitbodies,    O&M too       Biohazard
    identification:   min                      treat on         aptamers, or   expensive     detection
    Wet                                        location         PCR                          system
    chemistry                                                                                and R&D
    a R&D = Research and development

•      Aerosol identification sensors are based on mass spectrometry. They measure the
       amount of material in a sample at each molecular mass, producing data in the form
       of a mass spectrum. Mass-spectrometry sensors can discern the presence of
       specific biological agents in a sample in a few minutes (five or so).
•      Fluidics identification sensors are based on wet chemistry, detecting agents based
       on molecular interactions between the sensor and the target. Such interactions
       require 15 minutes to one hour.
Epidemiology. Historically, we have made significant contributions to the development
and operation of medical information databases. We are now extending our historical
stronghold into the areas of analysis, security, and integration of that information.
Analysis is necessary to transform information about the symptoms of individuals into
patterns of symptoms across a population. Security is required to safeguard precious
personal medical information from unwarranted access. That desire has taken on a new
urgency since the passage and enforcement of the Health Insurance Portability and
Accountability Act of 1996. Integration of epidemiological systems with related sensor
systems provides the benefits of a broader range of information.

Complementary Detection Approaches
At the system level, it is useful to consider the advantages and disadvantages of systems
that contain either sensors or epidemiology only. Sensor-only systems are desirable
because they continuously monitor for the presence of threat agents. Such systems
offer the best protection but also the highest cost. Because of their expense, sensor-

                              Technology Review Journal • Spring/Summer 2003                               139
Bioterrorism Detection: The Smoke Alarm and the Canary

 only systems are appropriate for protecting assets whose extreme value justifies the
 cost of biosensor hardware, operation, maintenance, and organizational disruption due to
 false alarms.
 Because of their relative lack of expense, epidemiology-only systems are appropriate
 for protecting assets that are valuable enough to justify defense and perhaps some level
 of associated up-front expense, but not significant recurring costs. In between are a
 number of applications that can afford more than epidemiology alone but also expect
 more in return.
 At a conceptual level, the potential benefits of a system that contains both biosensors
 and epidemiological monitoring are most obvious when compared with a “bare bones”
 system that contains only epidemiology. The main disadvantage of epidemiology is its
 slow speed. Addition of sensors can increase speed, while keeping biosensor expenses
 within a given budget. From the perspective of starting with a sensor-only system,
 addition of epidemiology can reduce costs. Those benefits can be tested in modeling
 environments that reflect the properties of the sensors, the traits of an epidemiological
 monitoring system, and realistic representations of both the threat and human behavior.
 Useful feedback can occur between the smoke alarms and the canaries in combined
 systems. For example, an observation or inference of a large number of certain symp-
 toms in a target population can influence the threshold of networked detectors that are
 monitoring for the presence of agents able to produce those symptoms. Similarly,
 sensors can trigger alerts to medical professionals in specific geographic areas to
 increase vigilance and accelerate symptom reporting. Sensor results can also influence
 the tendency to test for diseases that usually occur infrequently.

 The threat of biological terrorism is real. To minimize the consequences of an attack,
 we can significantly augment our defenses, particularly with respect to detection.
 Augmentation of sensor-based detection is complementary to epidemiology-based
 detection. Epidemiology can set thresholds for networked sensors that can be used to
 encourage doctors to test for bioterrorism-associated illness. Such improvements can
 work together to prevent the spread of naturally occurring diseases and more effectively
 treat the infected. Despite the costs, both monetary and operational, the recommended
 augmentations are justified by both the continued terrorism threat and the potential for
 downstream improvements to public health. We look forward to the opportunity to
 validate the benefits of systems that contain both canaries and smoke alarms through
 modeling and implementation in a real-world application, such as defense of fixed
 military installations.

 1.   B.M. Sullivan, B.W. Evans, and P.W. Allen, “Biological and Chemical Warfare
      Defense Sensors and Systems,” Technology Review Journal, Vol. 8, No. 1, Spring/
      Summer 2000, pp. 103–110.

140                        Technology Review Journal • Spring/Summer 2003
                                        Bioterrorism Detection: The Smoke Alarm and the Canary

Author Profile
                      Brian M. Sullivan joined Northrop Grumman Mission
                      Systems in 1998 as a software engineer and has contributed
                      to projects focusing on biodefense and biometrics. He is
                      currently the Biodetection Integrated Product Team lead for
                      the U.S. Postal Service’s Biohazard Detection System and is
                      working, within Integrated Physical Security Systems, to
                      protect fixed military installations from weapons of mass
                      destruction. He holds a BS from Allegheny College and a
                      PhD from the California Institute of Technology, both in
                      biology. Dr. Sullivan has one patent issued and five pending.
                      He has published a number of articles in professional
                      technical journals.


                 Technology Review Journal • Spring/Summer 2003                           141

To top