Events following a small-scale anthrax attack.
Schmitt B, Dobrez D, Parada JP, Kyriacou DN, Golub RM, Sharma R, Bennett C. Responding to a Small-scale Bioterrorist Anthrax AttackCost-effectiveness Analysis Comparing Preattack Vaccination With Postattack Antibiotic Treatment and Vaccination. Arch Intern Med. 2007;167(7):655-662. doi:10.1001/archinte.167.7.655
Copyright 2007 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2007
In 2001, a small-scale bioterrorism-related anthrax attack was perpetrated via the US mail. The optimal future response may require strategies different from those required in a large-scale attack.
We conducted a cost-effectiveness analysis using Monte Carlo simulation during a 10-year time frame from a societal perspective to determine the optimal response strategy for a small-scale anthrax attack perpetrated against US Postal Service distribution centers in a large metropolitan area. Three strategies were compared: preattack vaccination of all US distribution center postal workers, postattack antibiotic therapy followed by vaccination of exposed personnel, and postattack antibiotic therapy without vaccination of exposed personnel. Outcome measures were costs, quality-adjusted life-years, and incremental cost-effectiveness. The probabilities for anthrax exposure and infection; vaccine and antibiotic benefits, risks, and costs; and associated clinical outcomes were derived from the medical literature and from bioterrorism experts.
Postattack antibiotic therapy and vaccination of exposed postal workers is the most cost-effective response compared with other strategies. The incremental cost-effectiveness is $59 558 per quality-adjusted life-year compared with postattack antibiotic therapy alone. Preattack vaccination of all distribution center workers is less effective and more costly than the other 2 strategies. Assuming complete adherence to preattack vaccination, the incremental cost-effectiveness compared with postattack antibiotic therapy alone is almost $2.6 million per quality-adjusted life-year.
Despite uncertainties about a future anthrax attack and exposure risk, postattack antibiotic therapy and vaccination of exposed personnel seems to be the optimal response to an attack perpetrated through the US Postal Service.
In October 2001, multiple government agencies and media organizations were simultaneously attacked with mailed envelopes containing Bacillus anthracis spores.1- 5 These attacks resulted in 11 cases of cutaneous anthrax, 11 cases of inhalational anthrax with 5 deaths, and 10 000 potentially exposed persons who were advised to undergo prophylactic antibiotic therapy for 60 days without anthrax vaccination.6,7 The 6 survivors of inhalational anthrax continue to experience marked reductions in their quality of life (QOL).8 These attacks prompted prolonged closure of many government and corporate facilities; caused substantial social disruption and economic loss; and highlighted systemic problems with distribution of antibiotics and vaccine, and low antibiotic therapy adherence rates.8- 11
The optimal response for similar future small-scale anthrax attacks is unclear. Published models have generated various recommendations for an optimal public health response,9,12- 15 including postattack antibiotic prophylaxis alone,9 postattack antibiotic prophylaxis combined with postattack vaccination,15 and preattack vaccination combined with both postattack antibiotic prophylaxis and vaccination.12 However, the vaccination efficacy assumed in all models was based on animal studies and does not necessarily reflect human response during an attack. Most models primarily focused on a large-scale airborne attack. Small-scale attacks may differ for several reasons. Adherence with recommended interventions will likely be lower than in a large-scale attack.12 Further, the targeted population for preattack strategies is smaller, with less total cost, and the likelihood of another postal attack is probably greater than a large-scale airborne attack.
We evaluated the cost-effectiveness of prevention and treatment strategies for small-scale bioterrorism-related anthrax attacks perpetrated through the US Postal Service (USPS), simulating the postal attacks of October 2001. Detection of the attack was assumed to occur similarly to the 2001 attacks (base case scenario). We then reanalyzed for early detection through autonomous detection systems recently installed by the USPS to detect anthrax spores.16
A Markov model was used to estimate the cost-effectiveness of response strategies to reduce morbidity and mortality from a small-scale anthrax attack during a 10-year time frame. A societal perspective with a 3% annual discount factor for costs of all outcomes was adopted.17 Results were expressed as incremental cost-effectiveness ratios (ICER), calculated as the incremental cost per additional quality-adjusted life-year (QALY). The willingness-to-pay threshold was set at $100 000 per additional year of life. Analyses were conducted using TreeAge Pro Healthcare software (version 2006; TreeAge Software, Williamstown, Mass). Sensitivity analyses were conducted using 1- and 2-way analyses, and Monte Carlo simulation. Model parameters and ranges are given in Table 1.
The USPS employed 351 470 field career employees in 2004, excluding delivery carriers, motor vehicle operators, and maintenance personnel.18 The base case analysis considers a postal service workforce of 350 000 persons aged 18 to 60 years. Workforce turnover was estimated at 10% annually, and all workers were assumed to have no previous exposure to anthrax.
Three response strategies were compared: preattack vaccination of all distribution center workers; postattack antibiotic therapy followed by vaccination of exposed personnel; and postattack antibiotic therapy without vaccination. Postattack vaccination alone was not modeled because it would not prevent early spore germination. The first strategy uses the cell-free anthrax vaccine absorbed given in a series of 6 inoculations, at 1 day, 2 and 4 weeks, and 6, 12, and 18 months. The efficacy of this approach has been demonstrated in several animal models.34- 36 The second strategy uses an appropriate antibiotic (ciprofloxacin; Bayer Pharmaceutical, Westhaven, Conn) for 60 days and 3 inoculations of anthrax vaccine absorbed at 1, 14, and 28 days in exposed individuals. The therapeutic efficacy as well as protection against repeat challenge provided by combining postexposure antibiotic therapy with vaccination has been demonstrated in animal models.28,37,38 The third strategy uses only the described antibiotic regimen for 60 days in exposed individuals.
In the 2001 postal attack, prophylactic antibiotic adherence was approximately 44%. Partial adherence to antibiotic therapy postattack was assumed, with 25% of treated persons completing 15, 30, 45, and 60 days of antibiotic therapy. Partial adherence (50%) was also assumed for preattack vaccination. A rapid attack response was assumed in the base case, with postattack antibiotic therapy distribution to all exposed persons by 6 days. Predicted inhalational anthrax infection rates for no vaccination and postattack vaccination using a 60-day regimen of oral antibiotic therapy were 34% and 29%, respectively.9 The predicted anthrax infection rate for preattack vaccination was 56%.39 Cutaneous infection rates were assumed to be equal to inhalational infection rates.20 Infection rates derived from simulations conducted by Brookmeyer et al9 under different assumptions of infectious dose, response time, and adherence to prophylactic therapy are given in Table 2. Sensitivity analyses were conducted under each scenario and for a substantially more rapid response to an attack.
Because the likelihood and nature of a future postal attack are unknown, scenario probabilities are based on the best subjective estimates of the investigative team (Table 1). An attack was assumed to occur with 10% annual probability (1 attack expected during the 10-year time frame) at multiple USPS distribution centers, exposing 2000 postal workers. A 1% infectious dose, described by Brookmeyer et al as low, was assumed in the base case analysis9 and is within the range observed in the 2001 attacks at the New Jersey and Washington, DC, postal facilities.40,41 Without intervention, 1% of exposed persons would become infected with anthrax.
Outcomes for the preattack and postattack vaccination strategies are shown in the Figure (4 more figures are available from the corresponding author on request). Patients with inhalational anthrax die without treatment, die after inpatient treatment, or survive after inpatient treatment. Inhalational anthrax mortality in the 2001 attack was 45% despite intensive medical care.20 The probability of being dead on arrival, dead after hospitalization, or surviving inhalational anthrax was estimated at 0%, 45%, and 55%, respectively. For cutaneous anthrax exposure, lack of treatment could result in severe disease with mortality as high as 20%, whereas with appropriate treatment, mortality is less than 1%.42,43 Consistent with the 2001 attack,20 we assumed that all persons with cutaneous anthrax exposure in our model were treated with oral antibiotics and survived. Survivors of both inhalation and cutaneous exposure to anthrax are assumed to have immunity to subsequent infection.
Treatment assumptions were made based on the 22 anthrax cases from the 2001 postal attacks.20 Inhalational anthrax survivors would be hospitalized for 14 days, receive parenteral antibiotic therapy for 7 days followed by oral ciprofloxacin therapy for 53 days, and receive biannual outpatient visits after recovery for the duration of the study. Patients who died were assumed to have been hospitalized for 3 days before death. Patients with cutaneous anthrax exposure would be treated as outpatients with oral antibiotic therapy.
Medical costs were estimated in 2005 US dollars and adjusted for inflation as needed using the medical care component of the Consumer Price Index.44 Hospital costs were based on mean daily costs for hospital stays for adults aged 18 to 64 years, derived from the Agency for Healthcare Research and Quality Hospital Cost and Utilization Project, Nationwide Inpatient Sample.33 Outpatient visit costs were based on 2005 Medicare physician charges for an established patient, 25-minute visit, using Current Procedural Terminology code 99214.31
The current recommendation for prophylactic antibiotic therapy is ciprofloxacin, 500 mg orally twice a day, for 60 days in unvaccinated anthrax-exposed persons.20 Costs for ciprofloxacin were based on the manufacturer's current average wholesale price.32 The current recommendation for prophylactic vaccination is 6 doses administered during 18 months and followed by annual boosters. Vaccination dose cost is $24.50, or $147.00 for a 6-dose preattack series and $73.50 for a 3-dose postattack series.30 Administration costs were estimated using Current Procedural Terminology code 90471. Sensitivity analyses varied all cost estimates (Table 1). Other disposable costs of vaccination such as transport, storage, and distribution were excluded.
Mild adverse reactions to oral antibiotic therapy or vaccination are defined as those treated in the outpatient setting with 1 physician visit. Severe adverse reactions require a 2-day hospitalization and 2 outpatient follow-up visits. The model assumed a 16% probability of a mild adverse reaction to the antibiotic therapy, a 0.5% probability of a severe adverse reaction, and no deaths from antibiotic prophylaxis.11 The model assumed that mild adverse reaction to the vaccine occurred in 0.05% of the population and that severe adverse reaction occurred in 0.04% of the population, based on reports of adverse events in US military personnel.24 Effects of potential adverse events are assessed in the sensitivity analyses.
Short-term adjustments in QOL were made for mild and severe adverse reactions to oral antibiotic therapy and vaccination and were based on utilities reported for similar health states.8,29 Estimation of the short-term QOL effect of inhalational anthrax exposure was made by selecting a value 1 SD below the mean utility reported for acute illness, to capture the QOL effect of having an illness with a high probability of death.29 Long-term adjustments were based on reports from anthrax survivors 1 year after infection.8 Inhalational and cutaneous anthrax survivors reported QOL that was 60% and 71%, respectively, of the normal value8 and were assumed to have utility in the years after infection of 0.6 and 0.71, respectively. Variations in QOL were explored in sensitivity analyses.
One-way sensitivity analyses were conducted for all variables, based on the ranges given in Table 1. Brookmeyer et al9 estimated the effect of simultaneous variations in infectious dose, adherence, and response time on the expected infection rates for the 3 strategies we evaluated. Two-way analyses based on adherence and response time assumptions were conducted separately for 1% and 10% inoculation rates (Table 2). Costs and QALYs for each of the 3 strategies were calculated for the 10 scenarios defined by these variables.
Another sensitivity analysis assessed longer (4 months) antibiotic therapy required for higher spore exposure. The USPS has recently deployed autonomous detection systems to promptly identify mailed anthrax.16 A sensitivity analysis was conducted assuming administration of postattack prophylaxis as rapidly as 12 hours after exposure. Using the algorithm of Brookmeyer et al,45 spore germination would occur in 2.4% of exposed persons by 12 hours. Sensitivity analyses estimated the minimal improvement in effectiveness conveyed by the addition of postattack vaccination to postattack antibiotic therapy required for combined postattack antibiotic therapy and vaccination to remain cost-effective.
Two analyses were conducted to model scenarios that would most favor preattack vaccination. The effect of a mass vaccination of postal workers before an attack on the probability of an attack targeted at postal workers was tested. In addition, a sensitivity analysis was conducted assuming a positive probability of attack but zero probability of adverse reaction to vaccination, slow attack response, no adherence to postattack antibiotic therapy, and complete adherence to vaccination.
Monte Carlo simulations of 1000 randomly selected observations were conducted by varying 36 variables simultaneously for the 10 scenarios defined by infectious dose, response time, and adherence to prophylactic treatment to test the sensitivity of the results for the entire range of possible parameter values. Uniform distributions were assumed for each variable (Table 1). Strategies that were both more costly and less effective or were more costly and had a higher ICER than a remaining strategy were excluded from the Monte Carlo simulations. The optimal strategy was identified based on a willingness-to-pay threshold of $100 000. The percentage of simulations in which each strategy was optimal was calculated.
Our base case analysis indicates that postattack antibiotic therapy without vaccination is the least costly strategy but that combined postattack antibiotic therapy and vaccination is the most cost-effective strategy (Table 3). Preattack vaccination is dominated by both remaining strategies, reflecting the higher infection rate in the reference case analysis that assumes partial adherence. Even if a preattack vaccination program entirely prevents any anthrax attack on the postal service and there is complete adherence to the preattack vaccination program, it costs nearly $2.6 million per QALY for postattack antibiotic therapy alone. In contrast, the incremental cost-effectiveness of combined postattack antibiotic therapy and vaccination vs postattack antibiotic therapy alone was $59 558 per QALY.
One-way sensitivity analyses indicate that postattack combined antibiotic therapy and vaccination is also cost-effective compared with postattack antibiotic therapy alone over the entire range of most model parameters (Table 4). However, postattack antibiotic therapy alone is optimal when the proportion of infected persons dying from anthrax is very low (<1.4%). Preattack vaccination is not cost-effective over the range of any of the model parameters, including vaccination cost. Although the cost-effectiveness of postattack combined antibiotic therapy and vaccination compared with postattack antibiotic therapy alone is relatively unchanged with increases in the probability of exposure (scope of the attack), preattack vaccination becomes cost-effective compared with postattack antibiotic therapy alone (ICER<$100 000) only with complete adherence to preattack vaccination and with a probability of exposure equal to or greater than 18.5%, or 64 750 of the 350 000 USPS employees.
Table 5 gives the ICER between postattack combined antibiotic therapy and vaccination compared with postattack antibiotic therapy alone for 10 scenarios defined by response time, adherence, and infectious dose. Postattack combined antibiotic therapy and vaccination remained cost-effective when antibiotic therapy adherence was partial or none, regardless of response time or infectious dose. Preattack vaccination was not cost-effective in any of the 10 scenarios; however, with a 10% infectious dose, complete adherence, and slow response time, postattack combined antibiotic therapy and vaccination was dominated by postattack antibiotic therapy alone; thus, the Monte Carlo simulations were conducted comparing preattack vaccination with postattack antibiotic therapy for this scenario.
Postattack antibiotic therapy and vaccination was supported in at least 50% of the Monte Carlo simulations when adherence was partial. In addition, postattack antibiotic therapy and vaccination was cost-effective when adherence to antibiotic therapy was complete (100%), provided the response time was rapid and the inoculation rate was 10%. Postattack antibiotic therapy alone was the optimal strategy in the remaining 3 scenarios: complete adherence to antibiotic therapy and 1% inoculation rate, regardless of response time, and complete adherence to antibiotic therapy and 10% inoculation rate, provided that the response time was slow. In no case was preattack vaccination cost-effective, including the 2 scenarios designed to favor preattack vaccination, that is, slow response time and no adherence to antibiotic therapy but complete adherence to preattack vaccination.
Infection rates with postattack strategies were reduced substantially to reflect what might be expected in a rapid (12-hour) response scenario with complete adherence. Postattack combined antibiotic therapy and vaccination was no longer cost-effective compared with postattack antibiotic therapy alone, with an ICER of more than $2.3 million per QALY.
Despite the uncertainties of an anthrax attack and the efficacy of human vaccination, postattack combined antibiotic therapy and vaccination of exposed persons seems to be the most cost-effective strategy after a small-scale anthrax attack. When adherence to postattack antibiotic therapy is high, the incremental benefit of postattack vaccination is small and postattack antibiotic therapy alone may be optimal, depending on response time and infectious dose. However, the overall adherence after the 2001 attacks on the USPS was 44%.12
Several factors should be considered when interpreting our findings. In small-scale attacks, the effect of low adherence to postattack antibiotic therapy and vaccination on lives saved is small (1 life) compared with a large-scale attack,13 in which the number of lives saved is substantial (85 000 lives) with complete adherence. Even with substantial reductions in vaccination cost and adverse events, postattack strategies are preferred. Consequently, new vaccines (eg, from VaxGen Inc, Brisbane, Calif) with reduced cost per dose and number of required doses would not alter the optimal response strategy for a small-scale postal attack.
Autonomous detection systems substantially lower expected infection rates in the postattack strategies by promptly identifying mailed anthrax.16 With the infectious dose observed in the 2001 postal attacks, early detection and rapid response (12 hours to initiation of treatment) reduces the added benefit of postattack vaccination to almost zero and changes the optimal response to postattack antibiotic therapy alone. Similar to large-scale attacks in which early recognition, antibiotic distribution, and adherence to antibiotic therapy all influence the optimal strategy,15 the choice between postattack strategies is affected by adherence to preattack and postattack prophylaxis and a prompt response. However, unlike a large-scale attack, preattack vaccination is never preferred in a small-scale attack with reasonable assumptions of exposure risk, infectious dose, and adherence, because of the expected low infection rate and the high total vaccine cost. Even assuming terrorists would not attack a worksite where the employees had been vaccinated, preattack vaccination resulting in zero probability of an attack is not cost-effective.
Two limitations should be mentioned. First, the duration of postattack antibiotic therapy may be inadequate for victims exposed to high spore concentrations. Expanding the duration of antibiotic therapy from 2 to 4 months would add to costs and likely reduce adherence. However, the recommended strategy would not change because the additional $688.00 per additional 60-day regimen is negligible compared with the costs of preattack vaccination. Second, our model evaluated a small-scale attack while the United States remains at risk of a large-scale attack. The preferred treatment strategy identified in this study, postattack combined antibiotic therapy and vaccination, is also recommended for a mass attack, although the implications of the sensitivity analyses in the 2 settings differ.15
While the military has determined that exposure risk is sufficient to warrant preattack vaccination in Iraq and Afghanistan, anthrax vaccination of civilian populations at risk for a small-scale attack is likely to be controversial because of safety concerns. Targeted preattack vaccination of US postal workers at highest risk may increase its cost-effectiveness. However, identification of postal distribution centers at risk of attack is impossible; this option is not cost-effective in any reasonable circumstances. Our results suggest that postattack antibiotic therapy alone may be preferred when response time is rapid (≤12 hours), which is feasible given the presence of the autonomous detection systems in postal offices. Development of a plan for rapid delivery of treatment (vaccination and antibiotic therapy) is critical for successful mitigation of a future attack.
Correspondence: Brian Schmitt, MD, MPH, Medicine and Neurology Service Line (111), Hines VA Hospital, Hines, IL 60141 (email@example.com).
Accepted for Publication: December 13, 2006.
Author Contributions:Study concept and design: Schmitt, Dobrez, Parada, Kyriacou, Golub, Sharma, and Bennett. Acquisition of data: Schmitt, Dobrez, Parada, Kyriacou, and Sharma. Analysis and interpretation of data: Schmitt, Dobrez, Parada, Kyriacou, Golub, and Bennett. Drafting of the manuscript: Schmitt, Dobrez, Parada, Kyriacou, and Bennett. Critical revision of the manuscript for important intellectual content: Schmitt, Dobrez, Parada, Kyriacou, Golub, Sharma, and Bennett. Statistical analysis: Schmitt, Dobrez, Kyriacou, and Golub. Obtained funding: Schmitt and Parada. Administrative, technical, and material support: Schmitt, Parada, and Sharma. Study supervision: Schmitt and Bennett.
Financial Disclosure: None reported.
Funding/Support: This study was supported by grant VA IIR 02-080-1 from the Department of Veterans Affairs.
Acknowledgment: We acknowledge Paul Yarnold, PhD, Steve Belknap, MD, Jessica Zagory, Nicole Cohen, MD, Mark Roberts, MD, and Adam Kahn, EMT-I, for their contributions to this project and their comments about the manuscript.