Cost-effectiveness of a potential vaccine for Coccidioides immitis.

Coccidioidomycosis, a systemic fungal infection, affects Americans living in the Southwest. We evaluated the cost- effectiveness of a potential vaccine against Coccidioides immitis. Using a decision model we developed, we estimate that among children, vaccination would saved 1.9 quality-adjusted life days (QALD) and $33 per person. Among adults, screening followed by vaccination would save 0.5 QALD per person and cost $62,000 per quality adjusted life year gained over no vaccination. If the birth cohort in highly endemic counties of California and Arizona were immunized in 2001, 11 deaths would be averted and $3 million would be saved (in net present value) over the lifetime of these infants. Vaccination of adults to prevent disseminated coccidioidomycosis would provide a modest health benefit similar in magnitude to other vaccines but would increase net expenditures. Vaccination of children in highly endemic regions would provide a larger health benefit and would reduce total health care expenditures.

. Areas in the United States endemic for Coccidioides immitis. Cross-hatching indicates the heavily disease-endemic area; single hatching indicates the moderately disease-endemic area. Reprinted with the author's permission from Kirkland (1). of a detailed cost-effectiveness analysis of a potential vaccine against C. immitis.

Data and Methods
We used a decision model to evaluate the health and economic consequences of withholding the vaccine, screening and vaccinating only those susceptible to infection (screening/vaccination), or vaccinating all eligible persons. Taking a societal perspective, we calculated the incremental costeffectiveness of these strategies (formula in Appendix I, online; URL: http://www.cdc.gov/ncid/eid/vol7no5/barnato_ appendix1.htm), and we discounted both costs and benefits at an annual rate of 3%. We performed one-way sensitivity analyses on all the model variables, as well as a three-way sensitivity analysis of the most sensitive variables, and bestand worst-case scenarios.
We based the estimates for input variables on the literature whenever possible (Appendix II Table). Our base-case estimates represent our judgment about the best estimate from the literature and discussion with experts. The ranges for costs represent variation by 25% above and below the base-case estimate, except where otherwise specified. The ranges for sensitivity analyses on quality-of-life estimates represent the 25th and 75th percentiles of patients' assessments of quality of life (15), except where otherwise specified. The authors had complete scientific and editorial independence from the funding agencies.

Decision Model
We developed a Markov model ( Figure 2) with Decision Maker software (version beta 0.99.11.14.0a, 2000, S. Pauker, et al., Boston, MA) (16,17) to track hypothetical cohorts of patients who either did not receive the C. immitis vaccine, received it, or received it only if their skin test was negative. Patients who received the vaccine were at risk for pain at the injection site, mild to moderate fever, and anaphylaxis without death. Vaccinated patients were at decreased risk for extrapulmonary dissemination after primary infection. Patients neither immune from previous infection nor successfully vaccinated were at risk for C. immitis infection and serious sequelae. All patients were at risk for dying from other causes (18). We followed cohorts until death, with a Markov cycle length of 1 month.

Patient Population
We used epidemiologic and demographic data from two California counties (Kern and Tulare) and eight Arizona counties (Cochise, Gila, Graham, Maricopa, Pima, Pinal, Santa Cruz, and Yavapai) as proxies for the population features of highly endemic regions (2,19; Internal Revenue Service, unpub. data). We used two cohorts representing the weighted average age and prior probability of naturally acquired immunity among children (ages <17) and adults (ages 18 to 65) in 10 highly disease-endemic counties. New residents with no natural immunity were added to these cohorts. Children had an average age of 8.85 years, and 14.5% were naturally immune; adults had an average age of 39.51 years, and 47.5% were naturally immune.

Clinical Manifestations
Smith's classic studies in military recruits stationed in disease-endemic regions during World War II established that 60% of infected persons have no symptoms and 40% have a flulike illness (20). Asymptomatic infection and symptomatic infection with recovery confer probable lifelong immunity with a positive delayed-type hypersensitivity skin-test reaction to coccidioidin or spherulin. However, a fraction of per- Figure 2. Schematic representation of the decision model and decision model subtrees. The square node represents a decision to use one of the three strategies: no vaccination, vaccination of susceptible persons identified through a screening skin test, or vaccination of all persons. Circles represent chance nodes. After a strategy is chosen, the patient enters a Markov tree (denoted by a rectangle containing circles connected by an arrow). The Markov tree represents clinical events that can occur during each 1-month period as a patient is followed until death. Subtrees show events that may occur to patients during a 1-month cycle. Dissemination subtree: The site of extrapulmonary dissemination can be to the brain (meningitis) or elsewhere (e.g., bone). The outcome of dissemination can be death from other causes, death from disseminated coccidioidomycosis, or survival. If the patient survives, he or she can survive in a disabled or nondisabled state. Each month thereafter, the patient is at risk for relapse. Chronic subtree: Each month a patient with chronic pulmonary coccidioidomycosis can die from other causes, remain infected, die from chronic pulmonary infection, or be cured and rendered immune. Immune subtree: On any given month, a patient who is immune to infection can die of other causes or remain alive and immune to coccidioidomycosis. *At risk for dissemination relapse.
Emerging Infectious Diseases sons do not have simple self-limited disease and instead more serious illness develops, such as respiratory failure, chronic pneumonia, and extrapulmonary dissemination (20)(21)(22)(23). We present more clinical detail in Appendix I, online, only (at URL:www.cdc.gov/eid/v7n5/barnato-appendix1.htm) and summarize probability estimates, costs, and utilities in the Appendix II Table.

Vaccine and Skin-Test Characteristics
Animal trials of the vaccine rely on intraperitoneal or intranasal challenge with large loads of C. immitis. The outcome measure is survival or death from disseminated disease. Because the mechanism of action of the putative vaccine is not known, we conservatively assumed that vaccination prevents extrapulmonary dissemination but not primary infection. We evaluated this assumption in a sensitivity analysis.
We assumed that all vaccinated persons comply with 3 doses in 1 year and all develop an antibody response. We assumed that the vaccine reduced the probability of dissemination by 75%, with lifelong duration, based on experience with hepatitis B vaccine, a widely used subunit vaccine. We examined the effects of compliance with each strategy and the effect of waning immunity.
Using data from hepatitis B vaccine as a proxy, we assumed that 25% of patients would experience mild side effects such as local arm pain, fever, and nausea and that 1 in 600,000 patients would experience nonfatal anaphylaxis requiring hospitalization (24).

Quality of Life
Our model included intermediate health states that may be associated with decrements in quality of life. To adjust for such decrements, we included quality adjustments in our model (Appendix II Table). Because there are no quality-oflife studies for coccidioidomycosis, we used proxy health state utilities in published studies or clinical judgment. All illness utilities were multiplied by age-specific "healthy" utilities to account for the age at which an illness was contracted. We assumed that primary pulmonary infection causes a substantial heath status decrement only in those who are sick enough to be diagnosed. Among those with primary pulmonary disease that goes undiagnosed, we assigned a health state utility equal to well for this illness episode, but accounted for lost time from work due to illness (see Costs, below). For the short-term vaccine side effects, we capture quality-of-life impact as a decrease in utility of onetenth of a quality-adjusted day. The base case and ranges for these quality-of-life adjustments reflect our judgment about the impact of these episodes on patients. We evaluated the effects of these utility assumptions in sensitivity analyses.

Direct Medical Costs
All costs are presented in 2000 U.S. dollars (Appendix II Table). For costs of inpatient care, we used cost-adjusted charges from the 1996 Nationwide Inpatient Sample of the Healthcare Cost and Utilization Project (31). For outpatient services, we used the 2000 Medicare national physician fee schedule (32). For outpatient services not listed in the Medicare fee schedule, we used reimbursement received by Kern County service providers (R. Talbot, pers. comm.) We assumed that patients severely disabled by coccidioidal meningitis would require home support and used the average payment per Medicare home health beneficiary as a proxy for this cost (33).

Vaccine and Other Costs
In our base-case analysis, we assumed the cost of the vaccine was $180. We chose this value for the base case because it is the reimbursed fee for the three-shot hepatitis B series, an existing subunit vaccine. We examined a broad range of vaccine costs ($100 to $400) in sensitivity analyses.
For circumstances in which quality-of-life changes do not capture the inconvenience of illness or medical care, we included time costs, as noted in the online Appendix I.

Prevention of Illness and Death
We calculated the number of cases of disseminated disease and deaths per 100,000 persons that would be prevented over the lifetime of each cohort if children and adults were immunized, assuming rates of vaccine coverage of 40%, 60%, 80%, and 100% (Table 1). For example, if 60% of children were vaccinated, 93 cases of dissemination, 64 cases of disability, and 7 deaths would be averted and $2 million saved per 100,000 population. If 60% of adults underwent screening followed by vaccination, 38 cases of dissemination, 24 cases of disability, and 2 deaths would be prevented at a cost of $5.4 million per 100,000 population.

Costs and Effectiveness
We calculated the effectiveness, costs, and cost-effectiveness of each strategy in each population, expressed in Table  2 as the per-person cost and health benefit. Among children, vaccination saved 1.9 quality-adjusted life days (QALD) and $33 per person. Among adults, screening followed by vaccination saved 0.5 QALD per person and cost $62,000 per QALY gained over no vaccination. For adults, the incremental gain from vaccinating all persons compared with screening followed by vaccination contributed an additional 0.05 QALD at a cost of $235,000 per QALY gained.

Sensitivity Analyses
We performed one-way sensitivity analyses over the ranges of all input variables listed in the table in Appendix II and on critical assumptions, such as the vaccine mechanism of action and target population. Among children, vaccination was no longer cost-saving at the lowest ranges of vaccine efficacy, infection rate, dissemination rate, long-term care cost for severe disability, and medical follow-up cost after nonmeningeal dissemination and chronic pulmonary disease; the vaccine also was not cost-saving in children when we used the highest ranges of vaccine cost, meningitis mortality, emigration, and discount rate among children. The counter-intuitive effect of meningitis deaths is due to the decrease in survivors subject to long-term care costs from post-meningitis disability. Among adults, sensitivity analyses that changed the cost-effectiveness ratio by >$40,000 per QALY gained included infection rate, vaccine effectiveness, discount rate, vaccine cost, dissemination rate, emigration rate, and office visit time. Only the vaccine cost changed the preferred strategy among adults. Below $106 per 3 doses, vaccination was preferred over screening/vaccination, with a cost-effectiveness ratio of $30,000 per QALY gained.

Vaccine Duration
Our base-case analysis assumed lifetime immunity after vaccination, but vaccine protection may wane. If vaccine protection waned to zero in 15 years, vaccination saved 0.82 QALD per child and cost $47,300 per QALY gained, and screening/vacccination saved 0.28 QALD per adult and cost $165,500 per QALY gained.

Vaccine Mechanism of Action
If the vaccine prevented primary infection rather than dissemination alone, vaccination saved 2.26 QALD and $46 per vaccinated child over no vaccination. Screening/vaccination saved 0.66 QALD per adult, costing $46,500 per QALY saved.

Three-Way Sensitivity Analysis of Infection Rate, Vaccine Effectiveness, and Cost
Our base case assumed an infection rate of 2% per year, vaccine effectiveness of 75%, and a vaccine cost of $180. However, the vaccine may have some use in areas of lower endemicity. Also, because the vaccine does not yet exist, its cost and effectiveness are unknown. We present the effects of varying costs and effectiveness of the vaccine under two conditions: 0.5% infection rate per year and 2% per year (our base case of highly endemic regions) (Figure 3). In our base case for children, the $180 vaccine remains cost-saving down to an effectiveness of 65% and costs <$50,000 per QALY until vaccine effectiveness drops below 30%, confirming that the vaccine need not be highly effective to be cost-effective in

Discount Rate
Among children, vaccination saved 5.95 QALD and $563 over no vaccination when costs and benefits were not discounted. At a discount rate of 5%, vaccination was no longer cost-saving at $27,600 per QALY gained. For adults, changing the discount rate from 0% to 5% changed the cost-effectiveness ratio of screening/vaccination compared with no vaccination from $30,800 to $155,000 per QALY gained.

Dissemination Rate
Our base-case analysis used a dissemination rate of 0.38% based on the assumption that only blacks and Asians had higher rates of dissemination than whites. We evaluated a range of 0.25% to 0.55% to determine the effect of assum-ing all non-whites (including Hispanics) had rates of dissemination equal to that of whites (0.25%) or that of blacks (3.4%). Among children, vaccination saved 2.71 QALD and $143 per child vaccinated at an overall dissemination rate of 0.55%, and 1.22 QALD at a cost of $15,300 per QALY gained at the 0.25% dissemination rate. Among adults, varying the dissemination rate from 0.25% to 0.55% changed the costeffectiveness ratio of screening/vaccination over no vaccination from $25,400 to $125,000 per QALY gained.

New Residents
If one considers new residents separately, vaccination saved 2.18 QALD and $75 per immigrant child vaccinated, and saved 1.13 QALD per immigrant adult vaccinated, costing $13,500 per QALY gained.
We present best-and worst-case scenarios in Appendix I (online).

Discussion
We evaluated the usefulness of a potential vaccine against C. immitis in highly disease-endemic regions. Vaccination was cost-saving among children. A screening/vaccination strategy cost $62,000 per QALY gained among adults. Although the increase in quality-adjusted life expectancy from an immunization program is modest for one individual patient, this aggregates to an important number of illnesses and deaths prevented (Table 1). Furthermore, the life expectancy gains are comparable with gains from other immunizations. Vaccination for coccidioidomycosis (universally among children and after screening in adults) saved 0.14 to 0.47 life days and 0.53 to 1.86 QALD per person over no vaccination. In comparison, vaccination against pneumococcal bacteremia among elderly people saves 1.2 QALD per person vaccinated (35); infant vaccinations against measles, mumps, rubella, and pertussis each save 2.7, 3, 0.3, and 3.3 life days, respectively (36). To further place benefits in perspective, if the birth cohort in highly endemic counties of California and Arizona were immunized in 2001, 11 deaths would be averted and $3 million would be saved (in net present value) over the lifetime of these children.
The only previous analysis of the cost-effectiveness of a vaccine against coccidioidomycosis drew different conclusions. An IOM report considered the costs and benefits of research and development into a vaccine for C. immitis and found that the immunization of infants in endemic regions and immigrants of any age would cost >$100,000 per QALY (14). Our analysis evaluated a different question: If a vaccine were currently available, would it be cost-effective to immunize people in highly endemic regions? Although the IOM report acknowledged that the committee used rough estimates and a simplified model, the primary reason that it reached a different conclusion was that the cost of vaccine development was included in the IOM model. Vaccine development costs will influence the sale price of the vaccine. Project directors of the Valley Fever Vaccine Project estimate that research and development through phase III clinical trials, supported largely by public and private philanthropic sources, will cost $28 million (G. Rutherford, pers. comm.). This figure is 10 times lower than that used by the IOM in its analysis. To recoup $28 million over 5 years by immunizing 60% of the 90,000 annual birth cohort in highly endemic regions of California and Arizona, the vaccine would have to be priced at $100 per vaccinated person. This is less than the $180 vaccine cost assumed in our base-case analysis.
Sensitivity analyses found that the vaccine would be cost-effective in children under assumptions of waning immunity and would be even more effective and cost-saving in children and cost-effective in adults if the vaccine prevented primary infection. In highly endemic regions, vaccination of children is cost-effective even at relatively low levels of vaccine effectiveness if the vaccine is priced at $180. A $180 vaccine had to be >85% effective for screening/vaccination to cost <$50,000 per QALY in adults. Paradoxically, an even higher annual infection rate would make a screening/vaccination strategy less favorable in adults because most of them would have already acquired natural immunity by age 40. At an annual infection rate one-fourth of that seen in the Central Valley of California, such as might be seen in southern California, the only cost-effective strategy would be to immunize children with an inexpensive and highly efficacious vaccine. If dissemination rates are higher than our base case, as might be seen in the elderly and those with chronic diseases, screening/vacccination becomes cost-effective for adults.
The vaccine is much more effective in persons without naturally acquired immunity. A vaccination strategy that targeted new residents in highly endemic regions would exploit easily obtainable risk factor information. However, such a policy might be unacceptable to members of the population excluded by such a strategy. Finally, because vaccination is a preventive intervention that can take years before accruing a health benefit (in contrast to acute health-care interventions), our results were highly sensitive to discount rate.
Our model has limitations. The ecology of endemic regions has changed substantially with urbanization since Smith's study in the 1940s (20). Furthermore, his studies documenting infection and dissemination were conducted in the military, which may not be representative of general population exposures. Our base-case analysis used the demographics in highly endemic areas to estimate a composite dissemination rate for a multiracial population. We did not model race independently, nor did we model exposure risk. Workers with high levels of dust exposure may have higher rates of infection and more to gain from early immunization.
Our model assumed that all age groups without prior immunity were at equal risk for adverse health outcomes and had identical health-care costs. However, illness severity and costs may be higher among the elderly. Reported cases of coccidioidomycosis during Arizona's 1991 to 1995 epidemic occurred disproportionately among older adults (11). The study did not discern whether this was due to reporting differences, immune-suppressing coexisting conditions, a preponderance of new immigrants with higher risk for infection, or an independent age-related phenomenon. The death rate from primary pulmonary infection is as high as 26.8% among persons over 65 (37), much higher than our composite basecase estimate of 0.5%. The Kern County data on which we based our cost assumptions revealed that hospitalization, the greatest health-care expenditure associated with acute illness, was 4.2 times more likely among people >50 years old. Thus, we may have underestimated the potential effectiveness and cost savings of the vaccine among older adults.
Finally, our analysis did not model the effects of an immunization program in immunocompromised patients. Persons with compromised cell-mediated immunity, including those with AIDS, malignancy, or therapeutic medical immunosuppression, have higher rates of serious illness. Many of these patients may have reactivation of previous infection; it is unclear whether vaccination would be useful in this clinical scenario. This is an area for future research as results from clinical trials become available.
An uncommon disease in a national context, coccidioidomycosis has substantial ramifications for several regions where epidemics have produced reported annual case rates as high as 15 per 100,000 and substantial economic impact. Diseases affecting <200,000 persons annually (orphan diseases) require more incentives for research and development. Our findings may contribute to the policy decisions for vaccine development and distribution for C. immitis (39). Our analysis suggests that a vaccine against C. immitis would have substantial public health benefit. An update of this cost-effectiveness analysis can be performed when the results of human vaccine trials become available.