Postexposure prophylaxis (PEP), which refers to treatment with antiretroviral medications soon after a suspected HIV exposure, remains controversial. The US Public Health Service (USPHS) recommends PEP for health-care workers and others who have experienced a high-risk ‘occupational’ exposure to HIV . However, the USPHS ‘cannot definitively recommend for or against antiretroviral agents in (sexual and other non-occupational transmission risk) situations because of the lack of efficacy data’ . Nevertheless, evidence of the effectiveness of PEP in the occupational setting  and efficacy in animal model systems  has prompted physicians in North America and abroad to consider the use of PEP for persons potentially exposed to HIV through sexual or needle-sharing risk behaviors [5–9].
Several commentators have expressed concerns over the relatively high cost of PEP, viewed as an HIV-prevention intervention, especially in comparison with interventions that help clients to modify the behaviors that place them at risk of HIV infection [5,6,10–14]. Previous, model-based evaluations of the cost-effectiveness of PEP indicate that PEP would be cost-effective only under a very limited set of circumstances, such as following receptive but not insertive anal intercourse with a high-risk partner [13,15].
In contrast, a recently published, empirically based economic analysis of a large-scale non-occupational PEP feasibility study implemented in San Francisco  demonstrated that this program, which provided PEP to a mix of clients with varying exposure risks, was cost-effective . However, the HIV epidemic in San Francisco is unique in many respects. Would PEP appear as cost-effective in other US cities? To address this question, a city-by-city analysis was conducted of the cost-effectiveness of hypothetical PEP programs in the 96 largest US metropolitan statistical areas (MSA). The analysis combined epidemiological and MSA population size estimates from a published study with information about the distribution of exposure types, PEP completion rate, proportion of source partners known to be HIV infected, and PEP program costs obtained from the San Francisco PEP program to evaluate the cost-effectiveness of PEP in the 96 MSA.
Estimates from the literature were combined with data from the San Francisco PEP program to calculate cost–utility ratios (CUR) for hypothetical PEP programs in each of 96 MSA. The CUR is the ratio of net PEP program costs to the total number of quality-adjusted life years (QALY) saved by the program. It is given by (C − AT)/AQ: where T and Q are the HIV-related medical care costs and QALY, respectively, that are saved per averted case of HIV infection; A is the number of HIV infections averted by PEP; and C is the cost of the PEP program . An HIV prevention intervention with a negative CUR is considered cost-saving in that it can prevent an HIV infection for less than the estimated lifetime cost of HIV/AIDS-related medical care, thereby producing net societal savings.
Holtgrave and Pinkerton's estimates [18,19] were used for the lifetime cost of HIV-related medical care and the QALY saved by preventing someone from becoming infected. They estimated that someone who acquired HIV at 32 years of age (the median age of participants in the San Francisco program) would lose 9.31 QALY, discounted at a 3% annual rate. The corresponding lifetime cost of HIV/AIDS-related medical care is $223 072 in discounted year 2000 dollars.
Postexposure prophylaxis program costs
The potential cost of providing PEP in each of the 96 MSA was calculated based on data from a previously published cost analysis of the San Francisco PEP program . This cost analysis was conducted from the societal perspective  and, therefore, included all identifiable costs, including program-related costs to clients, such as time and transportation for clinic visits. All costs were expressed in year 2000 dollars.
Program costs were divided into fixed costs, which do not depend on the size of the PEP program, and variable (per-client) costs. Separate variable cost estimates were derived for patients who completed the PEP regimen versus those who did not. The total cost of the program in a particular MSA was calculated using the formula
where N is the number of PEP clients in the MSA (see following section); r is a cost of living adjustment factor (described below); y is the proportion of clients who completed therapy in the San Francisco PEP study (77.8%), F is the fixed program cost calculated for San Francisco ($13 726.76); S is the per-client screening cost ($40.12); Mc, Lc, and Kc, respectively, are the mean variable costs associated with laboratory work ($26.00), antiretroviral medications ($597.72: 96.7% of clients were prescribed a two-drug regimen and 3.5% a three-drug regimen), and other clinical services ($523.74) for program completers; and Md, Ld, and Kd, respectively, are the mean variable costs associated with laboratory work ($14.77), antiretroviral medications ($454.52), and other clinical services ($240.30) for non-completers.
The year 2000 All-Urban Consumers/All Items component of the Consumer Price Index (CPI) was used as to determine the cost of living adjustment factor (r; see above) for each of the 96 MSA by dividing the corresponding CPI value by the value for San Francisco . Where available, an MSA-specific CPI value was used in these calculations (36 of 96 MSA, including San Francisco); otherwise a regional (west, south, northeast, midwest) value was applied. Nationwide wholesale prices  were used to estimate drug costs; therefore, these costs were not CPI adjusted.
Estimated postexposure prophylaxis program size, prevalence of infection, and distribution of exposures
Potential PEP clients were divided into four distinct risk groups based on reported exposure type: men who have sex with men (MSM), injection drug users (IDU), male and female heterosexuals, and persons exposed to HIV through non-occupational needlestick injuries or other routes, such as bites and assaults involving mucousal exposure to potentially infected bodily fluids. To estimate the number of potential PEP clients in each of these groups, it was assumed that the same proportion of each group would access PEP in each MSA as they had in the San Francisco PEP study (Table 1). Specifically, the number of PEP clients in MSA i who belong to risk group j was estimated using the equation:
where Ntotal(i,j) is the total number of at-risk members of risk group j in MSA i and NPEP(SF,j) is the number of persons in the risk group who accessed PEP in the San Francisco study . Holmberg's  estimates of the number of at-risk MSM, IDU, and heterosexuals in each MSA (including San Francisco) was used to estimate Ntotal(i,j) for these groups. The number of at-risk persons in the overall population – which subsumes the MSM, IDU, and heterosexual subgroups – was obtained by subtracting Holmberg's estimate of the number of HIV-infected MSM, IDU, and heterosexual persons from his estimate of the total MSA population size.
Holmberg's estimates of the number of HIV-infected MSM, IDU, and heterosexuals in each MSA, Ninf(i,j), were used to estimate the prevalence of HIV infection in these groups,
The prevalence of infection in the overall population was calculated by dividing the number of HIV-infected MSM, IDU, and heterosexuals by the total MSA population size.
To extrapolate the findings of the San Francisco PEP study to other MSA, it was assumed that the distribution of PEP-prompting exposures (e.g., receptive anal intercourse, injection-related exposure) in each MSA would mimic the distribution observed in San Francisco (Table 1), that risk group membership could be inferred from exposure type (e.g., a man reporting vaginal intercourse belongs to the heterosexual group), and that the source partner would belong to the same group as the PEP client (with one exception: because discarded syringes are the likely cause of most non-occupational needlesticks, the prevalence of infection among IDU was used instead of the overall population prevalence to estimate the prevalence of infection among needlestick exposure sources; see Table 1).
Number of infections averted by postexposure prophylaxis
The number of HIV infections averted by a PEP program is the difference between the number of infections that would be expected with and without the program. In the absence of PEP, the number of HIV infections expected among N potentially exposed persons equals Npα, where p is the probability that the exposure source is HIV infected (see below) and α is the per-exposure transmission probability (Table 1). It was assumed that PEP can be effective only if the antiretroviral regimen was completed. For such a ‘completer,’ PEP reduces the risk of sustained HIV infection from pα to pα(1 − E), where E denotes the effectiveness of PEP. Therefore, the number of infections averted by a PEP program that reaches N clients is
where y, as above, is the proportion of PEP clients who completed therapy. The number of infections averted was calculated separately for each risk/exposure category in a particular MSA and then summed across categories to determine the total number infections averted in that MSA.
In the base-case analysis, y was set to 77.8%, which is the proportion of clients in the San Francisco PEP program who completed therapy . The probability that an exposure source was infected was estimated either as p = x + (1 − x)π or p = z + (1 − z)π, where π = πi,j is the risk group-specific HIV prevalence estimate for MSA i, as described above, and x and z denote, respectively, the proportion of clients (43.4%) and the proportion of completers (46.5%) in the San Francisco program who reported that their partner was infected with HIV (the first of these estimates was used in calculating the risk of infection for persons who did not complete PEP and the number of infections expected in the absence of PEP, whereas the latter estimate was used for PEP completers). (Thirteen clients in the San Francisco study discontinued PEP either because they were found already to be HIV positive or because the source partner tested HIV negative during the course of the study. These non-susceptible clients were included in the cost calculations but excluded from the effectiveness analysis.)
The effectiveness of PEP using combination antiretroviral therapy for non-occupational exposures to HIV is not known. It was assumed that combination therapy following sexual or injection-related exposure is only as effective as zidovudine monotherapy following occupational exposure, which was shown to reduce the risk of infection by 81% in a multinational case–control study of health-care workers percutaneously exposed to HIV-contaminated blood  therefore the value of E was 81% in the base-case analysis.
To determine whether the results were sensitive to variations in the input parameters, key parameters were varied, one at a time, in univariate sensitivity analyses. Further, to gain a deeper understanding of the factors associated with variations in PEP cost-effectiveness across MSA, a series of correlational analyses were conducted associating the inverse of the CUR with various population characteristics, including the prevalence of infection among MSM, IDU, and heterosexuals; the sizes of the MSM, IDU, and heterosexual subpopulations; the proportions of persons in the total, at-risk, or HIV-infected population belonging to these groups; and the proportions of MSM, IDU, or heterosexual persons in the ‘high-risk’ subpopulation (comprising MSM, IDU, and heterosexual individuals) relative to the number of persons in the total, at-risk, or HIV-infected population. (Taking the inverse of the CUR helped to linearize the ratio because the main disparity between MSA was in the effectiveness estimate, rather than the cost estimate. Compare CUR = (C − AT)/AQ as used here with the alternate expression, CUR = (C/A − T)/Q.)
Figure 1 illustrates the distribution of CUR obtained for the hypothetical PEP programs in 96 MSA. The ratios ranged from $4137 per QALY saved for San Francisco to $39 101 per QALY saved for the Stockton–Lodi MSA, which was one of only two with a CUR greater than $30 000 per QALY saved (Table 2). The mean and median CUR were $15 728 and $15 367 per QALY saved, respectively and 74% of the CUR values were between $10 000 and $20 000 per QALY saved. The per-client cost of PEP ranged from $1037 (Houston) to $1481 (Stockton–Lodi), with a mean of $1135.
Combined across the 96 MSA, the hypothetical PEP programs would serve 19 154 clients at a total cost of approximately $21.7 million (Table 2). The programs would avert a total of 63.9 HIV infections, primarily among MSM, who made up 75.9% of PEP clients but accounted for 96.1% of the averted infections. The combined CUR for the 96 MSA was $12 567 per QALY saved.
Across MSA, PEP was cost-saving (CUR < 0) for male–male and male–female receptive anal intercourse exposures (Table 3). The CUR for needle-sharing exposures and for needlesticks were $97 867 and $159 686 per QALY saved, respectively. The CUR exceeded $380 000 per QALY saved for all other types of exposure.
The results of the sensitivity analyses are displayed in Table 4. Only one of the parameter manipulations resulted in an overall, MSA-wide CUR greater than $60 000 per QALY saved, which often is considered a cut-off for cost-effective medical and public health interventions . Reducing the per-exposure transmission probability for receptive anal intercourse from 0.02 to 0.008 increased the CUR to $61 352 per QALY saved, whereas setting this probability to 0.032 decreased the CUR to < 0 (cost-saving). (Individually varying the other transmission probabilities within plausible ranges produced < 4% deviation from the base-case CUR; these results are omitted from Table 4.) The results were moderately sensitive to the effectiveness of PEP and somewhat sensitive to the proportion of persons who completed the PEP regimen, the proportion of PEP clients with known-infected source partners, and to the lifetime cost of medical care and lost QALY associated with a case of HIV infection. The cost of living adjustment had little effect on the CUR.
Although the overall CUR across the 96 MSA was not especially sensitive to the prevalence of infection in the MSA (halving the prevalence in each MSA increased the CUR to $17 713 whereas increasing it by 50% decreased the CUR to $7859), several prevalence-related measures were moderately to strongly correlated with the inverse of the CUR values obtained for individual MSA. In particular, the proportion of the ‘high-risk’ MSM, IDU, heterosexual subpopulation classified as HIV-infected MSM was strongly correlated with the inverse CUR (r2 = 0.75; Fig. 2 and Table 2), as was the proportion of the overall MSA population so classified (r2 = 0.72). The only other measure that accounted for more than half the variance in the inverse CUR was the proportion of the total population classified as MSM (r2 = 0.55). The prevalence of infection among MSM was moderately correlated with the inverse CUR (r2 = 0.41). Notably, weak associations (r2 = < 0.01) were obtained for the prevalence of infection among IDU and among heterosexuals.
The results of this analysis suggest that PEP with antiretroviral medications following sexual or injection-associated exposures to HIV could be a cost-effective adjunct to existing HIV prevention efforts. The CUR of hypothetical PEP programs for the 96 largest MSA in the United States was less than $20 000 per QALY saved for all but two. The ratios obtained compared favorably with other medical and public health interventions [25–27] including many behavioral risk reduction interventions to prevent HIV transmission .
The cost-effectiveness of PEP could be enhanced through better targeting to persons with high-risk exposures [15,29,30]. Targeting could take many forms. The present analysis generally confirmed the findings of previous, model-based estimates of the cost-effectiveness of non-occupational PEP, which suggested that PEP would be cost-effective only in limited circumstances, such as following receptive anal intercourse with a partner at high risk of infection, and possibly following other high-risk exposures with a partner known to be infected . Overall, across the 96 MSA, PEP was cost-saving for receptive anal intercourse and possibly cost-effective for needle sharing and non-occupational needlestick injuries, but of questionable economic value for all other types of exposure. Likewise, PEP was highly cost-effective for MSM, less so for IDU and high-risk women, and probably not cost-effective for general population exposures or heterosexual men. If PEP was restricted to clients with known HIV-infected partners (43.4% of clients in the San Francisco PEP study ), it would be cost-saving overall and in 91 of the 96 MSA. With regard to targeting at the city level, PEP was most cost-effective in MSA in which a large proportion of the population were MSM and in which the prevalence of HIV among MSM was high.
However, individual or group-level targeting may be difficult in clinical practice [5,8,15]. In the San Francisco PEP study, potential clients completed a personal risk assessment during screening and decided for themselves whether PEP was an appropriate option after discussing the ‘pros and cons’ of PEP with a counsellor and the prescribing physician. Consequently, although the program was intended for persons with significant exposure risks, a small number of lower-risk clients also were included in the study. It is important to note that, despite the inclusion of lower-risk clients, the San Francisco PEP program was highly cost-effective .
There are a number of sources of uncertainty in the present analysis. The results were most sensitive to the effectiveness of PEP and the per-exposure transmission probability for receptive anal intercourse. The effectiveness of PEP using combination antiretroviral therapy for non-occupational exposures to HIV has not been established. Although animal models and growing clinical experience suggest that PEP would be biologically active against sexual and injection-associated exposures, until the effectiveness of PEP in non-occupational settings has been demonstrated, the results of the present analysis should be viewed with caution.
The per-exposure transmission probabilities are another important source of uncertainty. The transmission probabilities used in the present analysis do not reflect variations in infectiousness over the course of HIV disease, interpersonal variability, or potential reductions from the use of highly active antiretroviral therapy. Additional studies of HIV infectivity are needed to lessen uncertainty in this key parameter.
The source of the MSA-specific HIV prevalence and population size estimates  was published in 1996 and, therefore, these estimates reflect the epidemiological conditions that existed in the first half of the 1990s. Although the overall cost-effectiveness results appear robust despite uncertainty in these parameters, the estimates for particular MSA should be accepted with caution. For instance, the present analysis produced a CUR of $4137 per QALY saved for San Francisco, which is substantially smaller than the ratio ($14 449 per QALY saved) obtained in a prior cost-effectiveness analysis of the San Francisco PEP program . The discrepancy between these estimates results from differences in the HIV prevalence estimates used in the analyses and from differences in the modeling approaches.
Finally, the analysis employed sample-level averages for several key parameters. The analysis did not take into account individual-level differences such as the ages of clients (which would affect the number of QALY lost to infection ) or whether or not a particular client completed the PEP regimen or was known to have been exposed to HIV by an HIV-positive partner. Instead, the median age of completers in the San Francisco PEP study was used to derive an average value for the number of QALY lost to infection, and the mean proportions of clients completing therapy and having a known HIV-positive source partner were used in the analysis. Further, the analysis assumed that the distribution of client exposure groups in each MSA would mimic that observed in the San Francisco study. Actual implementation of PEP services in a given locale is apt to differ from the San Francisco experience, resulting in different completion rates, HIV-positivity rates, and distributional characteristics.
The present analysis does not capture all of the potential benefits of PEP. Post-exposure programs provide an excellent opportunity to access persons who have recently engaged in high-risk activities [11,16]. Comprehensive PEP programs that incorporate risk reduction counselling and HIV antibody testing can capitalize on this critical time to help clients to adopt safer behaviors . Moreover, PEP programs provide an opportunity to help clients to reduce their risk for other sexually transmitted diseases through counselling, testing, and immunization (e.g., against hepatitis A and B).
Different intervention strategies address different HIV prevention goals, whether increasing clients’ knowledge of their HIV serostatus, changing behaviors that place clients at risk, or decreasing the likelihood of transmission from HIV-infected persons to their sexual or needle-sharing partners. PEP – which is intended to lessen the risk of HIV infection after exposure to the virus already has occurred – is unique among existing HIV-prevention strategies. Provided that PEP is clinically effective in preventing infection after non- occupational exposures, the results of this analysis suggest that PEP also would be cost-effective, with a cost-effectiveness ratio similar to that of existing behavioral interventions to reduce HIV transmission. Consequently, we believe that PEP should receive serious consideration as a key component of an overall strategy to reduce the incidence of new HIV infections in the United States.
The authors thank Ralph Resenhoeft, Janice Sherman, and Keane Weinreich for their assistance.
Sponsorship: This research was supported by investigator grants R01-MH55440 (S. D. P.), K02-MH01919 (S. D. P.), K24-MH64384 (J. O. K.), and R01-AI42523 (T. J. C.), and center grants P30-MH52776 (Center for AIDS Intervention Research), P50-MH42459 (Center for AIDS Prevention Studies), and P30-MH59037 (UCSF-Gladstone Center for AIDS Research) from the National Institutes of Health. Additional support was provided by the UCSF AIDS Research Institute and the University of California University-wide AIDS Research Program (CC97-0962 and CC99-SF-001).
1. Centers for Disease Control and Prevention
. Public health service guidelines for the management of health-care worker exposure to HIV and recommendations for postexposure prophylaxis.MMWR Morb Mortal Wkly Rep
2. Centers for Disease Control and Prevention
. Management of possible sexual, injecting-drug-use, or other nonoccupational exposure to HIV, including considerations related to antiretroviral therapy.MMWR Morb Mortal Wkly Rep
3. Cardo DM, Culver DH, Ciesielski CA, Srivastava PU, Marcus R, Abiteboul D, et al.A case–control study of HIV seroconversion in health care workers after percutaneous exposure.N Engl J Med
4. Black RJ. Animal studies of prophylaxis.Am J Med
5. Baylis F, Ginn D. Expanding access to PEP: ethical and legal issues.AIDS Publ Policy J
6. Desmond NM, Coker RJ. Should preventative antiretroviral treatment be offered following sexual exposure to HIV? The case for.Sex Transm Infect
7. Katz MH, Gerberding JL. Postexposure treatment of people exposed to the human immunodeficiency virus through sexual contact or injection-drug use.N Engl J Med
8. Laporte A, Jourdan N, Bouvet E, Lamontagne F, Pillonel J, Desenclos JC. Post-exposure prophylaxis after non-occupational HIV exposure: impact of recommendations on physicians’ experience and attitudes.AIDS
9. Rey D, Bendiane MK, Moatti JP, Wellings K, Danziger R, MacDowall W, et al.Post-exposure prophylaxis after occupational and non-occupational exposure to HIV: An overview of the policies implemented in 27 European countries.AIDS Care
10. Kahn JG, Pinkerton SD, Paltiel AD. Postexposure prophylaxis following HIV exposure.JAMA
11. Katz MH, Gerberding JL. The care of persons with recent sexual exposure to HIV.Ann Intern Med
12. Low-Beer S, Weber AE, Bartholomew K, Landolt M, Oram D, Montaner JSG, et al.A reality check: The cost of making post-exposure prophylaxis available to gay and bisexual men at high sexual risk.AIDS
13. Lurie P, Miller S, Hecht F, Chesney M, Lo B. Postexposure prophylaxis after nonoccupational HIV exposure.JAMA
14. Pinkerton SD, Holtgrave DR. Prophylaxis after sexual exposure to HIV.Ann Intern Med
15. Pinkerton SD, Holtgrave DR, Bloom FR. Cost-effectiveness of post-exposure prophylaxis following sexual exposure to HIV.AIDS
16. Kahn JO, Martin JN, Roland ME, Bamberger JD, Chesney M, Chambers D, et al.Feasibility of postexposure prophylaxis (PEP) against human immunodeficiency virus infection after sexual or injection drug use exposure: The San Francisco PEP study.J Inf Dis
17. Pinkerton SD, Martin JN, Roland ME, Katz MH, Coates TJ, Kahn, JO. Cost-effectiveness of post-exposure prophylaxis following sexual or injection drug exposure to human immunodeficiency virus.Arch Intern Med
18. Pinkerton SD, Holtgrave DR. A method for evaluating the economic efficiency of HIV behavioral risk reduction interventions.AIDS and Behavior
19. Holtgrave DR, Pinkerton SD. Updates of cost of illness and quality of life estimates for use in economic evaluations of HIV prevention programs.J Acquir Immune Defic Syndr
20. Gold MR, Siegel JE, Russell LB, Weinstein MC (eds). Cost-Effectiveness in Health and Medicine
. New York: Oxford University Press; 1996.
21. US Department of Labor, Bureau of Labor Statistics.
22. Medical Economics. Drug Topics Red Book
. Montvale, NJ: Medical Economics Co.; 2001.
23. Holmberg SD. The estimated prevalence and incidence of HIV in 96 large US metropolitan areas.Am J Public Health
24. Owens DK. Interpretation of cost-effectiveness analyses.J Gen Intern Med
25. Graham JD, Corso PS, Morris JM, Segui-Gomez M, Weinstein MC. Evaluating the cost-effectiveness of clinical and public health measures.Annu Rev Public Health
26. Stone PW, Teutsch S, Chapman RH, Bell C, Goldie SJ, Neumann PJ. Cost-utility analyses of clinical preventive services: published ratios, 1976–1997.Am J Prev Med
27. Tengs TO, Adams ME, Pliskin JF, Safran DG, Siegel JE, Weinstein MC, et al.Five-hundred life-saving interventions and their cost-effectiveness.Risk Anal
28. Pinkerton SD, Johnson-Masotti AP, Holtgrave DR, Farnham, PG. Using cost-effectiveness league tables to compare interventions to prevent sexual transmission of HIV.AIDS
29. Greub G, Gallant S, Zurn P, Vannotti M, Bürgisser P, Francioli P, et al.Spare non-occupational HIV post-exposure prophylaxis by active contacting and testing the source person.AIDS
30. Postma MJ, Bos JM, de Jong-van den Berg LTW, Tramarin A, van Bergen JEAM. HIV post-exposure prophylaxis: enhancing its pharmaco-economic profile by discriminate prescribing.AIDS
31. Roland ME, Martin JN, Grant RM, Hellmann NS, Bamberger JD, Katz MH, et al.Postexposure prophylaxis for human immunodeficiency virus infection after sexual or injection drug use exposure: Identification and characterization of the source of exposure.J Inf Dis
32. Martin JN, Roland ME, Neilands TB, Krone MR, Bamberger JD, Kohn RP, et al.Use of post-exposure prophylaxis against HIV infection following sexual exposure does not lead to increases in high-risk behavior.AIDS
33. Gerberding JL. Management of occupational exposures to blood-borne viruses.N Engl J Med
34. Kahn JG. Are NEPs cost-effective in preventing HIV infections?
In: Public Health Impact of Needle Exchange Programs in the United States and Abroad
. Edited by Lurie P, Reingold AL. Berkeley, CA: School of Public Health, University of California; 1993:473–509.
35. Royce RA, Seña A, Cates W Jr, Cohen MS. Sexual transmission of HIV.N Engl J Med
36. Vittinghoff E, Douglas J, Judson F, McKirnan D, MacQueen K, Buchbinder SP. Per-contact risk of human immunodeficiency virus transmission between male sexual partners.Am J Epidemiol
37. Puro V. Post-exposure prophylaxis for HIV infection.Lancet
Keywords:© 2004 Lippincott Williams & Wilkins, Inc.
post-exposure prophylaxis; HIV; prevention; cost-effectiveness; cost–benefit