Getting to know the unknown

Reports that say that something hasn't happened are always interesting to me, because as we know, there are known knowns; there are things we know that we know. There are known unknowns; that is to say, there are things that we now know we don't know. But there are also unknown unknowns—there are things we do not know we don't know. —Former United States Secretary of Defense, Donald Rumsfeld Whether this statement is wisdom or gibberish is a matter of debate, but it does, in many ways, illustrate the difficulties of managing emerging infections and their potential to impact blood safety. One of the known unknowns is extent of the risk to patients once an emerging agent has been recognized. This piece is critical, informing about the need for, and extent of, possible interventions. Increasingly, we are looking to formalized risk-based decision making, cost–benefit assessments, and proportional response to guide our decisions and all of these approaches require that we understand the size (as well as the severity) of the threat. Certainly, most decisions about the implementation of blood safety interventions have been made in the absence of adequate information. This shortcoming has been understandable in cases of dread diseases, such as variant Creutzfeldt-Jakob disease, but should probably not be the norm. The origins of residual risk estimation date back to the early days of testing for human immunodeficiency virus (HIV), when it became apparent that infections continued to occur, despite the implementation of antibody tests. Observational studies, based on lookback, defined the length of the infectious window period and allowed risk estimation through application of the well-known equation: risk equals incidence times window period.1 This approach has been applied to other infections for which testing is available, although the window period estimates may have been made by other means. A number of refinements have been made to these estimates over the years.2 Two things should be noted, however. First, the risk estimate itself is merely a statement of the likely number of donations that are collected during the at-risk window period—a measure that ignores the likelihood of transmission. Second, these estimates depend on the ability to assess incidence of new infections among donors—relatively easy to do with knowledge of donor dynamics and availability of test results, but not so easy in the absence of such information. The incidence–window period approach was first used for an emerging infection by the US Centers for Disease Control and Prevention (CDC), estimating the risk of a West Nile virus (WNV) RNA–positive donation during the initial outbreak of WNV in Queens, New York.3 To deal with the uncertainties of the data points, a Monte Carlo method was used to make the estimate, although the underlying method was essentially as outlined above. Incidence was derived from community data and the known period of viremia was used in place of the window period. Within a few weeks of the publication of this estimate, the first cases of transfusion transmission of WNV in the United States were reported.4 Subsequently, further estimates were developed and found to be very much in line with the results of routine donor testing for WNV RNA.5 The same approach has also been used to estimate the potential risk of chikungunya and dengue virus infection by transfusion.6, 7 The ability to develop estimates of risk for emerging infections is now widely available in the form of the EUFRAT (European Up-Front Assessment Tool) application, supported by the European CDC and developed by Oei and colleagues; details may be found in a recent publication.8 Given appropriate input data or assumptions, the tool is designed to estimate not only the risk (as defined above), but goes on to provide an estimate of the actual number of infected recipients that might be expected. In addition, the impact of potential interventions may be estimated, again given the availability of appropriate estimates of their efficacy. This tool is currently being validated but is already in use, as can be seen in this issue of TRANSFUSION.9 In this article, the EUFRAT tool has been applied to the recent outbreak of Q fever (ironically, itself named on the basis of absence of knowledge) in the Netherlands. This outbreak of human infection was carefully monitored through standard public health initiatives, thus generating baseline data. The outbreak was eventually brought under control by managing the source of the human infection, which was a result of intensive goat-farming practices. Additionally, Sanquin, the Dutch national blood system, implemented cautionary measures, including surveillance, donor management, and nucleic acid testing for the agent (Coxiella burnetii) in donations,10 plus limited lookback on potentially exposed recipients.11 In the areas of greatest incidence of infection, three donations among 1004 tested had detectable C. burnetii nucleic acids.10 These data are very much in line with the projections of numbers of infected donors derived from the EUFRAT model and the application of the original US CDC model. Moving beyond these estimates introduces a higher degree of uncertainty: indeed, there is a wide gap between the estimates of Oei and colleagues9 of the total number of infected blood recipients during the outbreak and observed transfusion transmissions—also, of course, an uncertain measure. In fact, there is really no basis for estimating the actual infectivity of the donations, so the choice of a worst-case value is probably justified. In this context, it is of interest to note that, in the United States, observed cases of transfusion-transmitted HIV are considerably lower than the numbers predicted by risk models. The EUFRAT application is clearly a valuable addition to the toolkit for managing emerging infections, but it must be recognized that it is only as good as its inputs, as is true of any predictive model. Key variables are unlikely to be apparent early in emergence of a novel infection: it is unclear whether the application of the model would have been of significant value during the first year of the Q fever outbreak. In this context, it should be noted that the first risk estimate for WNV appeared at the same time as the first recognized cases of transfusion transmission. An earlier framework for estimation was published from Canada, based on presumed properties of two different agents, but provides little realistic guidance for decision making.12 In other words, predictive risk models may not provide an accurate result for policy decisions in a timely fashion. However, it would be possible to use the EUFRAT tool to prospectively develop scenarios that would trigger the need for a given action. For example, in the case of an outbreak of chikungunya in the contiguous United States, a specific level of incidence, window period, and infectivity levels might be defined that, taken with the potential severity of the disease, would trigger the implementation of an appropriate intervention, based on a predicted number of transfusion-transmitted infections. For those of us in the United States, however, there is a very large unknown unknown. The blood collection system has entered a period of financial uncertainty based on a declining need for blood, while the hospital environment is seeking greater and greater economies. Blood safety interventions come at a cost that will vary by the agent and intervention, yet there seems to be no clear mechanism for effecting risk-based decisions in a timely fashion. The will is there, but the way forward is uncertain and the actual value of the tool in these conditions is moot. The author has disclosed no conflicts of interest.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.