This posting was written by Dr. Michael Montague, staff scientist at the J. Craig Venter Institute. It is difficult to briefly sum up Michael’s qualifications. However, I will refer back to a conversation I had with Dr. Tara O’Toole in 2006. Tara told me that the pace of change in biotechnology was advancing at incredible speed “because the people who speak in ones and zeroes are working with the people who speak in A, G, C, T.”
Dr. Michael Montague speaks both languages.
During a recent visit to the Venter Institute, Michael and I discussed the importance in the early detection of a bio-event, either man-made and naturally occurring. Michael proposed an idea that I thought was worthy of discussion and debate.
Michael and I look forward to your comments.
Randy Larsen
Toward Continuous Real-Time Bio-Surveillance
by
Dr. Michael Montague
If the United States was to come under attack from a biological weapon, how would we know it? How soon would we know it? If the attack was of the classical type (airborne release of anthrax spores) then the first anyone on the national level would know about it might be an increase in hospitalizations in a defined geographic region, but the early symptoms of many diseases are very similar: fever and a headache. It might be some time before we actually knew what had happened. This is terrifying enough, but what if the attack was a little less conventional? What if the biological weapon’s delivery system was not a single airborne plume, but rather based in part upon contagion? What if the contagion had a long incubation period before it presented symptoms? What if it was initially distributed in a geographically non-continuous manner such as by terrorists flying from airport to airport?
To recognize such a scenario, not just as an epidemic, but as a malicious attack, and to do so in enough time to have any chance of mitigating the impact of such an assault would require that the US have real-time capacity to sample and analyze the medley of infectious agents traveling through the civilian population.
What that means is that such a bio-surveillance effort must be part of the standard operating procedure of the US government in peacetime, and it must be engaged in continuously for decades to come. It does nobody any good to only initiate such an effort after a biological attack. It must be emphasized that this is an effort to collect epidemiological data not clinical data, although such an effort will have a huge and positive impact on clinical medicine as well. (Clinical data is concerned with the health of individual patients, epidemiology is concerned with populations of people.) The difference is that epidemiological data can sustain a relatively high noise level and still be useful. That is to say, uncertainty about any given data-point might be high, without making the trends across all data points less detectable.
This sort of capacity exists only in the most embryonic and minimal form today. (The CDC requires the reporting of hospitalizations and the diagnosis of certain specific diseases, but the process could not be called real-time, nor does it provide a constant stream of high-resolution data that could be mined to model epidemiology on emerging contagions as they happen).
What is required is an ongoing effort to amass and correlate bio-surveillance data concerning the day-by-day health of a statistically significant sampling of the US population (both sick and apparently healthy). In this article, I suggest a plan by which this can be incrementally achieved without raising privacy issues, without technology that is not already well within our reach, and without unreasonable expense.
There are several things that must be done to create the needed bio-surveillance capability. Data from a representative sampling of healthy individuals must be collected, also from a sampling of sick individuals. This must be done with broad-spectrum tests that can detect the presence or absence of thousands of pathogens at once. Lastly, an infrastructure to centrally collect and analyze this data must be created. Attempting to achieve all of these objectives in a single initiative would be a very difficult and expensive task. Consequently, this plan breaks the problems apart addressing them incrementally, with each stage leveraging the preceding stages. These incremental stages are:
1. Create a centralized database of blood-donation data that is updated daily. The only identifying information attached to each sample would be a five-digit zip code of the donating individual and another five-digit zip code of the donation site. Because, this is insufficient information to identify the donating individual, there should be little or no privacy concerns. It should be emphasized that this is data that is already being collected. Thus, the only cost is in the reporting and analysis infrastructure.
2. Expand the sources of data that are being fed into the database to include test-data (again stripped of identification information) made on blood that is not being donated but rather being drawn for diagnostic purposes at large clinics and hospitals. Again, at this stage we are just collecting into the centralized database test-data for tests that would have been performed anyway. So, the only additional cost is in the reporting and analysis infrastructure… an infrastructure that is largely already in place from the first stage.
3. In situations where many blood samples are processed in the same facility at the same time, such as large blood-banks and hospitals, small amounts of each sample could be pooled daily and subjected to a more broad spectrum series of tests that could identify a wider range of pathogens. These are procedures that would not have otherwise been performed, but because these assays would be performed on pools of samples, rather than individual samples, the number of tests performed would be very small (one per day per testing center). This would keep testing costs very low while simultaneously providing a broad detection capability.
4. Sponsor, at the national level, the use of more broad-spectrum tests for individual blood screening. The technology for such tests exists today, but because demand is low, the cost per sample is high. Because it would be done at the national level, as part of what would be at this stage an already existing national bio-surveillance effort, such tests could be procured in much large quantities. With mass production, the per-unit cost of any manufactured product goes down to the point where the addition of such broad spectrum tests to the processing of each individual sample would be a minor addition of cost to the medical procedures that were drawing the samples in the first place.
What kinds of information will be produced at each stage of this plan? How will it be useful will it be, and what limitations will it suffer?
It is the first stage of the plan where the most limitations apply. One of the obvious objections to a bio-surveillance system based in-whole or in-part upon blood donations is that sick people are not allowed to donate. However, this is not as significant a problem as it seems.
Today when people donate blood, they are asked a series of questions to determine their health. If they pass that questionnaire, a cursory examination is performed (including taking both temperature and blood pressure), and a small sample of blood is taken. If they pass the examination, the red-cell count is determined to see if it is safe to give blood. If so, a donation is taken and later subjected to tests to detect the presence of Transfusion Transmittable Infections(TTIs). Typically TTIs that are screened for in the US are Hepatitis B, Hepatitis C, HIV1, HIV2, HTLV, HCV, HCMV, West Nile Virus, Syphilis, and Chagas, but others such as Malaria might be added depending on location and perceived risk [1-7]. While we won’t get the TTI data from people who are not permitted to donate, the preliminary data, from the examination, the questionnaire, and the red-cell count is also information that could be sent, anonymously, to the database. This is potentially very useful information. For example, if the percentage of potential donors with low-grade fevers increased by 30% in a city it could be a crucial warning sign that might present several days before more sever symptoms of an epidemic.
Also, many tests to detect disease in blood are immune system assays that test, not for the presence of the disease, but rather the presence of the human antibody to the disease. Such assays test whether the subject has ever been exposed to the disease. A person’s immune system might have fought the disease off, but still have the antibody. Thus, it is possible to use such tests indirectly to sample the people surrounding the subject.
It is worth repeating that such data would not be analyzed on an individual level but an epidemiological level. Just as meteorologists compare the weather in the current year to previous years, so this data will become more valuable as a historical context is accumulated.
In the second stage of the plan, the limitation of donated samples being skewed toward healthy adults will largely disappear, because data from non-donated blood tests will also be fed into the database. Because, at this stage, we would still be working entirely with results from tests that are already being performed, we would still have very little pathogen-specific data, and very spotty data at that, since unlike blood donations, different tests are performed on different patient samples. However, much information can be inferred from this sort of data, for example alanine transaminase assays were routinely used to test for liver function in blood-donations before reliable hepatitis assays were available[1].
Once we are in a position to start pooling blood samples and performing tests on the pool (the third stage of the plan), we will start to have the sort of data that could make a huge difference in characterizing a biological event as it happens, because we will have geographically targeted, pathogen-specific, sampling of the larger population. A good example of the sorts of tests that could be performed on pooled samples would be multiplex PCR, followed by micro-array detection of amplified products. Such an assay could test for the presence or absence hundreds to thousands of blood-born pathogens simultaneously.[8, 10, 13] Such tests have also been successfully performed for the simultaneous detection of multiple upper-respiratory pathogens based upon nasal lavage samples rather than blood.[14]
Ultimately, the goal is to have that sort of presence or absence data acquired as standard procedure for every individual blood test performed in America, and to have that data anonymized and sent to the central database on a daily basis. Once such a database exist, and data is already being sent to it, and that kind of data is already being used (stages 1-3), the only remaining barrier to achieving this goal is to reduce the testing cost such that broad spectrum screening assays can be performed in the millions without adding unacceptably to the cost of medical procedures that these assays are attached to. This is quite possible. In addition to the sort of PCR plus micro-array detection tests described above, ELISA on a chip assays have already been developed that can test for many antibodies simultaneously.[9-12] There is nothing about these technologies that fundamentally has to be expensive. Making this sort of product cheaper is something that mass production has been doing successfully for centuries.
With this kind of capability, it would be possible to build an immune system for the nation that would recognize a contagion as it entered the population and allow for response to such a contagion potentially before anybody started showing symptoms. It would be possible to ask the question: Is this disease spreading in a natural or malicious way? Building this sort of bio-surveillance capacity would not just protect us from epidemics, natural and malicious. It would also greatly increase the safety of our blood supply by coordinating and bringing more resources to bear on blood screening. Lastly, and perhaps the most transformative aspect of this plan, it would open the door for a much needed shift in diagnosis away from symptoms and toward more objective data. Such a revolution in the amount and quality of diagnosis information available to doctors would have a huge impact on almost all medicine.
References:
[1] Screening Donated Blood for Transfusion-Transmissible Infections Recommendations (WHO)
Click to access ScreeningTTI.pdf
[2] Patient information: Blood donation and transfusion (up to date)
http://www.uptodate.com/patients/content/topic.do?topicKey=~cmR1zC2BBaaikJ
[3] Update: West Nile Virus Screening of Blood Donations and Transfusion-Associated Transmission — United States, 2003 (CDC)
http://www.cdc.gov/mmwr/preview/mmwrhtml/mm5313a1.htm
[4] Blood donation (wikipedia)
http://en.wikipedia.org/wiki/Blood_donation
[5] Blood donation (wiki Doc)
http://wikidoc.org/index.php/Blood_donation
[6] Global blood safety and availability (WHO)
http://www.who.int/mediacentre/factsheets/fs279/en/index.html
[7] Nucleic Acid Amplification Technology (NAT) Testing Blood Donors (Southeastern Community Blood Center)
http://www.scbcinfo.org/publications/bulletin_v1_n4.htm
[8] Diagnosis On A Chip (Berkeley)
http://coe.berkeley.edu/labnotes/1003/boser.html
[9] Makers of an ELISA on chip technology (Ostendum, associated with the University of Twente, in the Netherlands)
http://www.ostendum.com/
[10] Lab on a Chip (Journal)
http://www.rsc.org/Publishing/Journals/lc/Index.asp
[11] Optical chip detects blood molecules (including fertility homones and genes) (labonline)
http://www.labonline.com.au/articles/37238-Optical-chip-detects-blood-molecules-?topic_id=1270
[12] Microfluidic platforms for lab-on-a-chip applications (Lab on a Chip, Review article 2007)
Click to access Review-MicrofluidicPlatformforLOC-LOC2007.pdf
[13] The Biolog Microbial ID System
http://www.biolog.com/microbialid/?utm_content=mmontagu%40jcvi.org&utm_source=VerticalResponse&utm_medium=Email&utm_term=Biolog%26%2339%3Bs%20Revolutionary%20Gen%20III%20Microbial%20ID%20System%20Identifies%20Over%202%2C200%20Species%20of%20Bacteria%20and%20Fungi%20&utm_campaign=Fast%203%20step%20Microbial%20IDcontent
[14] Microarray-based detection and genotyping of viral pathogens
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC137777/
Biosecurity Blog 
Norman Anderson and his Viral Defense Foundation (http://www.viraldefense.org/bibliography_nga.htm) have proposed a similar screening mechanism focused on viruses. It is certainly true that we live in a time where you just get infected and have no idea what you have, unless you are so sick you end up in the hospital… and even then you may never know what landed you there. A good intermediate step between our current situation and the future proposed here, though, would be identifying the illnesses that make people very sick, and ideally offering more options for treating a viral infection.
Great article. In addition to the clear benefits to domestic surveillance, an effective national system could conceivably provide insight into questions on global surveillance.
Considering the interconnected nature of the world and the role of the US as a central global player, changes in US domestic data, combined with other relevant information, could reveal details on disease activity elsewhere in the world.
Excellent article. One of the problems with early diagnosis/detection of infectious diseases is that often the agent is not present in high enough numbers in the blood stream. For example, Bacillus anthracis, after inhalation, is engulfed by dendritic cells in the lung and rushed off to the lymph nodes, and is thus less accessible to blood-based detection during those crucial early hours. A solution to this problem is host-based detection of infection: microarray analysis at both the DNA and protein levels should be able to detect specific changes in the host immune response, to identify unique host response “signatures” and thus differentiate among infections. There are data supporting this approach, although they are slow in coming. But a database such as that described by Dr. Montague would be ideal for screening for such signatures, permitting earlier detection and intervention.
Nancy Connell quite correctly points out that many pathogens are not easily detectable in the blood, particularly early in the infection.
The kind of protein based detection that she puts forward as a possible solution is a technology that is promising, but less mature than DNA or immuno-detection methods. In order for it to be effective as part of the kind of national surveillance strategy outlined above, it would need to be compatible with broad-spectrum screening, and be sensitive enough to function with out an uninfected negative control from the same patient. The first of these requirements is fairly reasonable, as Nancy Connell points out, protein microarrays already exist, so a single protein chip could conceivably monitor for hundreds or even thousands of different host factors at the same time. The second requirement sounds more difficult to me since the host-factors being monitored will most likely exist in both healthy and infected patients. That is to say, instead of detecting for the digital signal of presence/absence of a blood factor, one would need to measure the analog signal of the relative amounts of such factors.
It is not clear to me that there is a need for protein level detection in order to gain access to such host-based data. Baochuan Lin et. al. 2003 have suggested that this sort of data-set could be derived via cDNA based host gene expression profiling:
http://www.clinchem.org/cgi/content/full/49/7/1045
Regardless of whether it was DNA, or protein based, or both, I fully agree Nancy Connell’s point that host factors represent a promising data source that might be used to build a national bio-surveillance effort.
Gigi Gronvall pointed out that Normon Anderson’s Viral Defense Foundation has suggested similar efforts. I have had the pleasure of talking to Normon Anderson, and greatly approve of his general strategy. However, he has focused upon microscopy combined with analysis of virus capsid size and shape as a detection method. While suitable for pooled samples of large volume (as might come from a slaughter house) such methods are ill suited for following human pathogens. (However, we must not ignore the dangers of agricultural pathogens being used as bioweapons as well).
Norman Anderson says:
September 23, 2010 at
In reference to Dr. Montague’ comments:
Global Screening for Human Viral Pathogens has been described in detail in http://www.cdc.gov/ncidod/eid/vol9no7/03-0004.htm. Briefly the procedure includes nationwide routine collection of both normal and diagnostic sera, shipment of these to centralized laboratories for analysis, pooling of excess diagnostic sera now routinely discarded (amounting to hundreds of liters per week), recovery of the viral loads from
100 liter batches using the K-II centrifuge (which I developed in collaboration with the Gas Cenrifuge program at Oak Ridge), further concentration and purification using microgradient ultracentrifuges recently developed, shotgun sequencing of the concentrates, and data reduction and sequence reconstruction. Assuming 1 ml per originall donor, and ten centrifuges in parallel, one million cases can be scanned per run, assuming sufficient collection. The total viral load is concentrated down to less than the average volume of one original sample, hence the final concentration can approach that in each original viremic donor. In general viremia peaks with fever, hence the importance of obtaining samples on admission, as can be done here. The methods apply equally well to urine, and modifications apply to tissues. A complete concentration run takes less than three days. The objective is a running inventory of all viruses in circulation in man. The K-II centrifuge is used world wide to purify commercial vaccines, especially those for influenza. Much of this work depends on studies done on marine viruses. Organization of this work is proceeding in China.