Summary of Conclusions and a Path Forward
Each previous chapter in this report concludes with findings and recommendations relating to the individual technologies and detection strategies discussed in that chapter. This final chapter begins by summarizing the main conclusions reached by the committee about the individual technologies. It then attempts to synthesize the committee's best judgment into a plausible series of steps toward the goal of practical, effective detect-to-warn systems for the scenarios considered in this report. The chapter concludes with the committee's overall findings and recommendations regarding this path forward.
DETECTION AND IDENTIFICATION SYSTEMS
Advances in detect-to-warn systems have the potential to reduce significantly the casualties associated with bioterrorism attacks on both indoor and outdoor targets, ranging from individual high-value buildings to extended military bases. Such warning, if provided rapidly enough (in less than 3 to 5 minutes, depending on the scenario), will enable a range of protective options, from turning off HVAC systems to sheltering in place to facility evacuation in the case of an interior release.
The committee considered two types of biosensors for detect-to-warn systems: (1) nonspecific detectors with a rapid response time and (2) specific identifiers with a somewhat slower response time. Both types of biosensors are viewed as having important roles to play in an integrated detect-to-warn sensor architecture.
Nonspecific detectors respond to bioaerosol particles present in the air at concentrations higher than their detection threshold. They offer a rapid response but provide little information about the nature of the particles detected (hazardous or nonhazardous; alive or dead; type of biothreat agent). Two kinds of nonspecific detectors are discussed in this report: standoff detectors and spectroscopic point detectors. The committee's major conclusions about these detectors are summarized below.
Standoff detection uses electromagnetic radiation to detect the threat agents at a distance. Most of the work to date has focused on the use of lasers either to detect the presence of an aerosol cloud at distances of tens of kilometers or to ascertain whether the cloud has a biological or nonbiological content, at distances of several kilometers. This work has the potential to provide the earliest possible warning for
extended targets such as military bases, but more work is needed to better understand just how such systems would be used (i.e., the concept of operations). Provided that robust concepts of operations can be formulated, the development of a hybrid infrared-ultraviolet (IR/UV) standoff biodetection capability should be expedited.
In addition to these nonspecific detection techniques, the Department of Defense has also recently begun to investigate advanced standoff techniques—ranging from ultraviolet resonant Raman scattering to passive infrared detection—in the hope of providing either longer ranges or more specific identification. These are still in the early research stage. More convincing laboratory data to support modeling projection of detection ranges and ability to discriminate against expected backgrounds is needed before considering any acceleration of these efforts.
Spectroscopic Point Detectors
Spectroscopic point detectors typically measure some properties of the suspended aerosol at the detector itself rather than at some standoff distance. Some of the simplest spectroscopic sensors use particle counting with size discrimination to detect a sudden increase in aerosol concentration in the sizes of interest. Although this information is accurate and rapid, normal fluctuations in particles within the 1- to 30-micrometer diameter range can result in an unacceptable level of false alarms. Also, because an aerosolized biological agent can produce morbidity and mortality in exposed personnel even when its concentration is very low (even lower than that of nonpathogenic microbes), nonspecific spectroscopic point detectors cannot protect personnel against such low-level attacks.
A more capable spectroscopic point detector uses an ultraviolet laser to excite the tryptophan, reduced nicotinamide adenine dinucleotide (NADH), and flavin fluorescences that are characteristic of biological materials and uses the time rate of change of the signal to differentiate a rapid biowarfare agent release from the more gradual fluctuations of the natural background. The false alarm rates of these bioaerosol detectors are a function of the detection threshold and the ambient bioaerosol background. When operated outdoors and at modest sensitivities (tens of ACPLA) they currently exhibit false alarms rates of between one and tens per day. Their false alarm rates in filtered indoor environments and for the 102 to 104 ACPLA levels discussed above may be dramatically lower. This would depend on the size and nature of fluctuations in the level of nonpathogenic microbes as a result of human or other activity in the indoor environment. Research on the inclusion and exploitation of additional spectroscopic signature information should reduce this false alarm rate even further. Therefore, the committee considers bioaerosol detectors to be the most promising near-term candidates for the 1-minute, nonspecific detection portion of the system.
Specific Detection and Identification Technologies
Four potential methods of identification are discussed that can provide specific information on what type of biological organism or toxin may be present in a sampled bioaerosol cloud: nucleic acid sequence-based detection; molecular recognition of identifying structures on the surface of the organism or toxin; unique chemical attributes of the organism or toxin; and biological responses to the organism or toxin. The committee's conclusions regarding these identification approaches are summarized below.
Sequence-based detection uses the genetic information contained in a pathogen's DNA or RNA to detect and identify the pathogen. In laboratory studies, such techniques have shown the best sensitivities (detecting as few as 5 cells) and very low false alarm rates (about 10-5 for a single signature and lower for multiple signatures). Typical analysis times range from 15 to 60 minutes. The key issue for detect-to-warn applications is the extent to which this sensitivity and specificity must be sacrificed in moving to the demanding detect-to-warn time lines of less than 5 minutes and preferably of 1 minute or so. The
committee noted that a 1-minute detection time means 1 minute for the total process: sample collection, sample preparation, detection, and subsequent analyses.
For techniques that amplify the genetic material (e.g., PCR) and hence require sufficient time for multiple amplification cycles, the committee did not see any way of meeting the 1 minute time requirement. However, research groups around the world are pursuing integrated PCR systems for identification in under 5 minutes. The committee believes that the probability of such prototype systems being realized within the next few years is relatively high. The committee is also relatively confident that by developing rapid, automated sample preparation and handling techniques and by reducing the number of amplification cycles, it should be possible to perform a single-target, real-time PCR assay within 3 minutes, starting with the collection of an aerosol sample. There are no data that would permit an extrapolation of such a system to estimate the time requirements of a deeply multiplexed (e.g., 15 target sequences) PCR assay with any confidence.
Because repeated amplification cycles take time, one is led to consider unamplified detection. The most promising of these techniques is the detection of ribosomal RNA (rRNA), which exists in multiple copies in each cell (about 10,000 copies per cell is typical for bacteria). Achieving 1-minute detection will require significant research and development of rapid, automated sampling techniques that remove assay inhibitors, lyse the target cells, and transfer their content to the detector. Even if the research and development effort is successful, such an RNA assay would not carry the same information content as a full identification using DNA.
Structure-based detection uses molecular recognition processes to detect the characteristic shapes and functional group distribution of biomolecules on the surface of pathogens or of toxins. Typically this molecular recognition is achieved using antibodies in a multistep immunoassay. Today, such assays take on the order of 15 minutes, have detection thresholds of 103 to 105 particles per milliliter of solution, and exhibit false positive alarm rates on the order of 10-3. However, significant improvements in response times, detection thresholds, and false positive rates should be possible. Current response times are limited by the time it takes to transport the antigen to the molecular recognition site and not by the antibody-antigen binding time, which takes seconds or less. Improved detection techniques that get around some of the noise issues in current approaches promise lower detection thresholds.
Because structure-based detection techniques detect molecules on the surface of pathogens, they have simpler sample preparation requirements than do sequence-based techniques. The committee believes that with sufficient research and development to overcome the mass transport issues, 1-minute structure-based identification should be achievable.
The other significant issue is that of false positives. Two factors contribute to the false positive rate. One is insufficient specificity and the other is nonspecific binding. There are promising paths for addressing both factors, including the use of multiple signatures instead of one. Given that 2-minute detection has already been demonstrated and that there are reasonable approaches for attaining very low false alarm rates, the structure-based detector is the committee's leading candidate for a 1- to 2-minute identifier. The trade-offs between fixed-surface and solution-based assays need to be examined.
Chemical-based detection techniques use molecular characteristics (e.g., size, mass) or chemical composition rather than biological activity to detect biomolecules. Biomolecules of interest include proteins, their peptide subunits, lipids, carbohydrates, and the small molecules involved in the everyday functioning of biological agents. The best known of these techniques uses mass spectrometry to fingerprint bioagents—that is, to match the mass patterns from an unknown sample to the mass patterns in a library of known samples. There has been considerable progress in this area, especially for time-of-flight mass spectrometry (TOF MS). However, the critical issue for these biomarker approaches is how
well they will work in complex mixtures of naturally occurring microorganisms and other background components.
Another possible approach for rapid identification of pathogens would involve the use of small, low-cost, semiselective sensors, but very little work has been done in this area and off-the-shelf solutions are not available. It is not known if the performance of these semiselective sensors will meet detection requirements (signal-to-noise ratio, sensitivity, minimal false positives). An unacceptable number of false positives is expected if these semiselective sensors are operated independently; however, the use of networked sensor arrays would likely improve accuracy.
Functional-based detectors use organisms, whole cells, or parts of cells to detect the biological activity of agents. As such they are less specific than sequence- or structure-based techniques but have the potential for detecting unknown chemical and biological agents. Initial research has focused mainly on chemical intoxicants and has only recently begun to explore applicability to biological warfare agents. At present, detection times for truly functional systems are tens of minutes to hours, much too slow for detect-to-warn applications. While some of these times might be reduced with additional research, other times are limited by the physiological response time of the organism or cell to that particular family of agents and so will not likely be improved. Therefore, for at least the next 5 years, these systems seem best suited to be sentinels for exposure to a wide range of unknown toxic materials rather than detect-to-warn sensors with response times of 1 minute. It can be expected that cell-based detectors will be attacked by naturally occurring, nonpathogenic microbes, so removal and replacement will be a critical issue, as will the rate of false positives.
Effectiveness of Detection and Identification Systems
Spectroscopic bioaerosol detectors will be able to detect, but not identify, all biological agents (known or unknown; natural or engineered) as long as they contain proteins. Techniques based on nucleic acid sequences can, in principle, detect and identify all the bacteria and viruses listed in Table 1.1 except the toxins, because these do not contain genetic material.1 In practice, any fielded sequence-based instrument will have signatures for a few to a few tens of agents and will only be able to detect these agents.
Structure-based recognition systems can, in principle, detect and identify not only all the agents listed in Table 1.1 but other agents as well, provided the agent is known and a recognition element (e.g., an immunoassay) has been developed for it. In practice, any fielded instrument will have signatures for a few tens of agents and will only be able to detect these agents.
Chemistry-based techniques have the greatest potential to serve as rapid, inexpensive sensors required for the distributed biological smoke alarm concept. However, individually these detectors are generally semiselective—that is, they would respond not only to biological agents but also to other molecules and organisms with common characteristics, and this would lead to high false alarm rates. Achieving greater selectivity may require using a collection of such sensors with different chemically interactive surfaces.
The function-based techniques are the only ones capable of detecting unknown or unanticipated agents (natural or engineered, biological or chemical), though they are unlikely to be able to identify the specific agents. This class of techniques detects chemical and toxin agents by their activity. For the foreseeable future, these function-based techniques are likely to be too slow for warning purposes and will be most useful in detect-to-treat applications.
DETECT-TO-WARN SYSTEMS FOR BUILDINGS AND EXTENDED MILITARY INSTALLATIONS
While the detect-to-warn architectures for both indoor and outdoor targets draw on a similar technology base, differences in characteristic target and attack parameters require that the technologies be used differently in these two scenarios. Below, the committee summarizes its conclusions about the indoor release scenario, from detection of bioaerosols inside building airspaces to a phased defense implementation strategy. The committee then explores the protection of extended military installations from outdoor releases of bioagents and examines how differences in these scenarios lead to modifications in the recommended implementation strategy (Figures 11.1 and 11.2).
The focus of the scenarios considered in this report has been covert attacks on facilities or installations. Various environmental and operational features of conflict or restoration operations will impact the requirements and performance of the detection systems considered here. However, the effects can be in opposing directions. For example, a conflict environment may complicate the environmental background against which the signal must be detected but, on the other hand, it may increase the readiness of response systems, increase the number of surveillance assets available, or perhaps allow greater tolerance of false alarms. Many of these operational and response issues are beyond the scope of the current study. The bioaerosol and particulate background associated with various operational states is of interest. As noted, the committee found only limited data on even normal environment backgrounds. No data that represented measured or analyzed wartime environments were discovered.
These data would be of interest and worthy of inclusion in the report, but their impact on the recommendations would probably not be very large. Various proven techniques exist for filtering or
otherwise separating out much of the data pertaining to particulates that are not threatening. As a result, robustness to variability in nonbiological particulate backgrounds is not expected to be a showstopper, although different levels of concern about these unknown environments have been expressed by members of the committee. Of potentially greater impact are possible enemy countermeasures to purposefully bypass the detection process. Although also beyond the charter of this study, an in-depth red team review of detection-based architectures is essential to address this problem.
Protection of Buildings
For building protection, the most commonly considered threat scenarios are the release of agent either inside a room or inside an open area within the facility, or release of agent directly into the exterior HVAC intake on the outside of a facility. Because of the confined spaces, even small releases can result in very high local concentrations—that is, greater than 105 particles per liter for a large room, or more than 103 particles per liter for a typical air-handling zone. These concentrations are usually well above typical ambient interior biobackgrounds (typically 1-100 particles per liter, but disturbances or movement of large numbers of people can temporarily increase levels to around 1,000 particles per liter). Hence, for these scenarios, relatively simple and rapid (about 1 minute) detection systems, e.g., bioaerosol detectors, may give a baseline facility detect-to-warn capability in the next year or two. Such a system would have the advantage of being independent of the detailed nature of the agent and hence would provide broad-spectrum coverage. Importantly, even though bioagent concentrations are high in the vicinity of the detector—making detection feasible—subsequent transport losses and filtration will reduce these concentrations by several orders of magnitude before the HVAC system circulates the contaminated air to adjacent rooms or air-handling zones.
A still more capable system would also make provisions for detecting lower-level attacks that might be used with more infectious agents, or slow-release attacks in which the perpetrator attempts to keep the bioagent concentration below the detection threshold of a simple bioaerosol detector. At these lower agent concentration levels, a bioaerosol detector would be increasingly prone to false positive or negative responses. In this case, the addition of another detector that can identify specific agents and hence
discriminate from ambient backgrounds becomes important. For example, structure-based assays currently have response times on the order of 15 minutes, but with considerable effort these times may be reduced to 1 to 2 minutes. This leads naturally to the concept of a system of detectors: a bioaerosol detector that can detect all bioagents (known and unknown) with low false alarm levels for modest to large attacks and a more sophisticated identifier that can detect the most dangerous known agents even for very small attacks—all in about 1 minute. Both the bioaerosol detector and the rapid identifier will be operating continuously, making measurements every 1 to 2 minutes. When either or both alarm at a high signal-to-noise level, high-regret responses such as sheltering in place or building evacuation will be initiated. If the detection or identification signal has a lower signal-to-noise level, then low-regret options such as HVAC shutoff or air sterilization will be initiated. In all alarm cases, an air sample will be collected and passed to a sequence-based analyzer for confirmatory analyses on the 5 to 15 minute time scale.
Because of the high cost of the associated detectors and the operational and maintenance costs associated with continuous operations, the above concept leads naturally to a centralized detection architecture in which a detection system is placed in—or samples—the HVAC system of each of the air-handling zones in a facility. The fact that the agent concentration level can be 10-100 times higher in a given room or region than in the air-handling unit also raises the intriguing possibility of a distributed detection system made up of less capable but inexpensive detectors (the biological smoke alarm concept). Additional systems analysis and research and development on such low-cost sensors are needed to better evaluate the potential of this option.
Protection of Military Installations
The most commonly postulated outdoor attack on a military base is a line release of an aerosolized agent generated by a ground vehicle or a low-flying aircraft passing upwind of the target. Placing the release path within a few kilometers of the target maximizes the concentration of the agent at the target and reduces the potential warning time for defensive measures such as donning protective masks and sheltering in place. The most obvious defensive architecture for such an attack is a perimeter detection system, i.e., a sparsely populated line of point detectors placed as far forward of the target area as possible.
As with facility protection, it seems useful to think of nonspecific detectors for warning of high-level attacks and identifiers for warning of low-level attacks. However, the agent concentration at the detector is likely to be significantly lower outdoors than in a confined building, with the actual agent concentrations dependent on whether the attack is aimed primarily at personnel who may be outdoors at the time or whether it also targets personnel inside. In addition, the variability in the concentration of background aerosols is likely to be higher outdoors than in the filtered air of a building. The net result is that the nonspecific detector will address a smaller portion of the threat space for extended installations than for buildings—for example, if its false alarm rate is low enough to support confident initiation of high-regret responses, the nonspecific detector may be able to detect only large attacks. More of the warning burden will fall on the rapid identifiers, which are about 5 years away. Fortunately, the rapid identifier need not be quite as rapid for building protection. For typical wind speeds of 5 to 10 meters per second, each kilometer of the detector's standoff distance from the actual target area results in an additional 3 minutes to take action.
In an outdoor release scenario, many of the potential agents can be treated with postexposure prophylaxis, presumably initiated by detect-to-treat systems, which can often provide effective alternatives to a detect-to-warn system. Collective protection systems can also provide safe interior zones to maintain critical functions. While the committee agrees that detect-to-treat systems will likely be the foundation of installation defense against outdoor releases in the near term, the employment of Phase 1 concepts may enable detect-to-warn capability for larger outdoor attacks and many facility attacks. This could add value in several areas, including the following:
Even partially effective detect-to-warn systems can enable response options that might avoid or reduce exposure to organisms engineered for antibiotic resistance or to other agents (e.g., toxins) for which no prophylaxis exists.
Initial detect-to-warn systems will provide options for that portion of the population that is contraindicated for prophylaxis.
Some detect-to-treat warning may enable protective responses even in areas (particularly interior spaces) where more complete collective protection measures are not implemented. This could provide partial, but much less costly, defense of a much larger population.
A nonspecific detection component will provide some capability against those agents not included in the few to tens of pathogens addressed by specific detect-to-treat assays.
If Phase 1 is skipped, the first detect-to-warn capability will be delayed for at least 5 years until rapid identifiers become available.
Because of the above considerations, additional systems analysis is needed to better understand the cost/benefit trade-offs associated with nonspecific detection for base protection. These studies should consider both standoff and point detection and should examine a range of plausible scenarios and concepts of operations to determine the portions of the threat space that could be addressed by nonspecific detectors and what additional response options are enabled by the earlier detection that may be afforded by standoff technologies.
TOP-LEVEL TECHNICAL FINDINGS AND RECOMMENDATIONS
The phased implementation strategies suggested above reflect the committee's best judgment as to the path forward that is most likely to lead to success. However, the committee recognizes that technologies that appear less applicable today may experience breakthroughs in the future, and that totally unforeseen technologies may emerge. Thus, it has chosen to group its top-level technical findings and recommendations into two categories: the most probable path and a technology watch list. The most probable path consists of those technologies whose currently demonstrated capabilities provide the basis for a reasonably well understood path to desired detect-to-warn capabilities. The technology watch list consists of promising technologies that have yet to demonstrate one or more critical features before a clear path emerges for detect-to-warn applications.
Most Probable Path
The committee finds that protection of buildings and military installations from biological attack requires the careful integration of detection capabilities with response options and procedures. Therefore, the committee recommends that military planners take a systems approach to facilities protection.
The committee finds that a successful detect-to-warn system requires that the local bioaerosol background be well understood. Therefore, the committee recommends that local aerosol backgrounds and their sources be characterized using the same methods that detectors would use. Within buildings where detectors are to be placed, steps should be taken to reduce these backgrounds.
The committee finds that the greatest disadvantage of using rapid, nonspecific detectors such as bioaerosol detectors is their potentially high false alarm rate at very low levels of detection. Therefore, the committee recommends that the false alarm rate of bioaerosol detectors be characterized in relevant facility environments as a function of detection threshold. Research should be supported on additional spectral and physical signatures and improved algorithms and techniques to further decrease the false positive rates.
The committee finds that structure-based assays appear to have the greatest potential for identifying biological agents with the speed, sensitivity, and specificity required for detect-to-warn applications.
Therefore, the committee recommends that research be supported that would lead to an improved structure-based detector. The goal of this program should be a system with very low false alarm rates and a 2-minute or less overall detection time.
Although a detect-to-warn system has its highest impact if it can initiate responses within approximately 1 minute of an attack, even response times on the order of 5 to 15 minutes can be useful. The committee finds that technologies that provide confirmation of the attack and identify the organisms involved will serve a vital function in the overall defensive architecture. Therefore, the committee recommends that research be continued on the development of an integrated, fully automated PCR system, including sample collection, preparation, and analysis.
The committee finds that while prototype instruments for standoff detection of biological agents have been developed and tested, there is no currently fielded capability for such standoff detection, nor is there a clear concept of operations for the use of such systems. Therefore, the committee recommends that a clear concept of operations be developed for standoff detection in support of base protection and, if appropriate, that the development of a hybrid infrared/ultraviolet laser-induced fluorescence system be expedited for these applications.
Technology Watch List
The committee finds that mass spectrometry has the potential to identify biological agents based on a biofingerprint matching method and has the potential to do so with limited reagent consumption. Therefore, the committee recommends that the use of laboratory mass spectrometry be investigated to better understand the performance of biofingerprinting in complex mixtures of naturally occurring microorganisms and other background contaminants. This should be done with parallel development of improved sample preparation methods.
The committee finds that the biological smoke alarm concept offers intriguing potential for rapid detection. This concept uses networked, low-cost, semiselective detectors distributed throughout the rooms in a building. Therefore, the committee recommends that research be conducted to develop and characterize the performance of low-cost arrays of semiselective sensors that can be used as a biological smoke alarm for triggering low-regret response measures.
The committee finds that ribosomal RNA assays might be capable of biothreat agent identification in one to several minutes. This approach, with a major development effort, could avoid the time-consuming amplification cycles of many nucleic acid sequencing assays. Therefore, the committee recommends that the potential and the limitations of rRNA detection for rapid identification of pathogens be explored.
The committee finds that function-based sensors are one of the few promising candidates for detecting unknown hazardous agents—that is, agents that had not been anticipated. Their response time is inherently tied to the time it takes an agent to have a physiological effect on sentinel organisms or tissues. For certain chemical agents and toxins, this effect can be very rapid, but for bacteria and viruses, it can take much longer. These longer response times for bacteria and viruses make it unlikely that function-based sensors will play a significant role in detect-to-warn applications for these agents, but they could nevertheless play a valuable detect-to-treat role in the overall biodetection architecture. Therefore, the committee recommends that studies be conducted to better understand the role of function-based sensors in overall biodetection architectures and to provide goals to focus research and development activities on those areas for which function-based sensors have the highest leverage.