Introduction
The US Public Health Service (PHS) recommends post-exposure prophylaxis (PEP) with antiretroviral agents for health-care workers who have been exposed to HIV [1]. For deep percutaneous exposure to HIV-infected or suspect blood, the US PHS recommends prophylaxis consisting of a 4-week regimen of triple combination therapy with two reverse transcriptase inhibitors, zidovudine (ZDV) and lamivudine (3TC), and the protease inhibitor indinavir. In a recent case-control study conducted by the Centers for Disease Control and Prevention (CDC), PEP monotherapy with ZDV was associated with a 79% reduction in the risk of seroconversion following percutaneous exposure to HIV-infected blood among health-care workers in the United States, the United Kingdom, and France [2]. The effectiveness of triple combination PEP is currently unknown.
The use of antiretroviral drugs for PEP in health-care workers has led some HIV/AIDS experts to question whether PEP would also be effective as a 'morning-after' regimen to prevent HIV infection resulting from sexual contact with an infected person or with someone of unknown HIV status [3-6]. PEP might be particularly appropriate in cases of sexual abuse or assault [7], or for use by HIV-discordant couples when accidentally exposed to HIV (e.g., when a condom breaks) [4]. Others have expressed concern that the availability of PEP could result in increased sexual risk-taking by some people, that it could facilitate the development of antiretroviral-resistant strains of HIV, or that patients might needlessly suffer PEP-related side-effects [4,6,8].
The effectiveness of PEP against sexually transmitted HIV has not been established. One potentially important difference between PEP for sexual and occupational exposures is that much longer delays between exposure and the initiation of chemotherapy are likely following sexual exposure. To prevent initial infection and rapid viral replication, the PHS guidelines recommend the initiation of PEP within hours of occupational exposure [1]; this recommendation is based, in part, on animal studies that document substantial decreases in effectiveness if PEP is delayed [3,9-11]. Delays in obtaining PEP may be unavoidable for many instances of suspected sexual exposure to HIV, thereby reducing the effectiveness of PEP. There are also potentially significant differences in the transmission dynamics of sexual (mucosal) and occupational (percutaneous) exposures, although host responses to HIV challenge are likely to be similar [8]. Finally, it is unknown whether adherence to the complicated antiretroviral regimens would differ in occupational and non-occupational settings.
Despite considerable uncertainty about its effectiveness, some physicians have apparently already begun prescribing PEP for their patients who report sexual exposure to HIV [8,12,13], leading commentators to call for the development of a 'rational policy on this issue' [8]. As a preliminary step toward such a policy, we have conducted an economic analysis of sexual PEP to assess whether it constitutes a cost-effective use of HIV prevention and treatment resources.
Economic analyses such as that provided here are needed to ensure that the limited resources available to combat HIV/AIDS are allocated wisely [14,15]. For example, a recent analysis has indicated that providing triple combination PEP to occupationally exposed health-care workers is cost-effective only when a substantial risk of infection would otherwise exist (e.g., deep percutaneous exposure to the blood of an HIV-infected patient), but not for lower risk exposures [16].
A preliminary examination of the cost-effectiveness of PEP following non-occupational exposure (including injection-associated exposure) suggests that similar conclusions are also reached in this situation [17]. In particular, it appears that prophylactic therapy with ZDV and 3TC would not be cost-effective for sexual exposures other than receptive anal intercourse, except perhaps when it is known with near certainty that the partner is infected [17]. However, this analysis assumed that PEP would be as effective for non-occupational as for occupational exposures, which may not hold. Moreover, by analogy to the occupational situation, this simplified analysis considered only the special case of a single HIV exposure, without a backdrop of continuing risk behavior. However, individuals who seek PEP following sexual exposure may do so on a repeat basis and may also face ongoing behavioral risk of HIV infection. It is important to consider how repeated episodes of PEP or a context of ongoing risk affects the results of the analysis. Finally, the effect of incorporating a protease inhibitor into the treatment regimen has not been fully addressed.
To examine these issues further, the present study reports the results of a cost-utility analysis of PEP following sexual exposure to HIV. Two scenarios are considered. The first assumes that the potential HIV exposure is an isolated event and that this exposure poses the only threat of infection to the individual (this scenario is similar to another scenario examined by us [17]). The second scenario examines the economic impact of repeat PEP and assesses how a backdrop of ongoing risky sexual behavior affects the economic efficiency of PEP. A range of potential PEP effectiveness values is examined to determine the sensitivity of the results to this critical parameter, and an incremental analysis is performed to determine how much more effective triple combination PEP with ZDV, 3TC, and indinavir would need to be in order to be more cost-effective than PEP with ZDV and 3TC alone.
Methods
Standard methods of cost-utility analysis were employed [18-20]. The analysis was conducted from the societal perspective, in accordance with the recommendations of the Panel on Cost-Effectiveness in Health and Medicine [21]; thus, all costs and consequences of PEP were included, regardless of who pays for or experiences them. The analytic horizon encompasses current (1996) costs and the present value of benefits accrued in the future (i.e., potentially averted HIV/AIDS medical care costs) discounted in the base-case analysis at a 3% annual rate (the impact of changing the discount rate is explored in the sensitivity analyses). Parameter estimates were obtained from the relevant literature and subjected to one-way, multiway, and threshold sensitivity analyses [19,21].
The main outcome variable in the present study is the cost-utility ratio, which is defined as the net program cost per quality-adjusted life year (QALY) saved. Symbolically, the cost-utility ratio is expressed as follows:
where C is the average cost of providing PEP to a patient, A is the probability that PEP would prevent the patient from becoming infected with HIV, T is the discounted lifetime cost of treating a case of HIV/AIDS, and Q is the number of discounted QALY saved by preventing someone from becoming infected with HIV [22,23]. PEP would be considered cost-saving if the numerator of this ratio was negative (i.e., if the cost of PEP was less than the cost of treating HIV/AIDS, adjusted for the probability that infection would have occurred in the absence of PEP).
Even if PEP is not cost-saving, funding PEP might constitute an acceptable utilization of scarce health-care resources, and hence be considered cost-effective. In the United States, health-care interventions with non-negative cost-utility ratios that are less than approximately US$ 40 000-50 000 per QALY saved are usually considered cost-effective, but those with ratios over US$ 180 000 per QALY are difficult to justify as cost-effective; interventions with intermediate ratios may or may not be cost-effective [23-26].
The analysis compared PEP provision with a 'no-program option'. In the base-case analysis PEP was assumed to consist of a 4-week regimen of triple combination therapy with ZDV, 3TC and indinavir, and was assumed to be as effective as ZDV monotherapy in the occupational setting (these assumptions were relaxed in the sensitivity analyses, in which dual combination therapy and a range of effectiveness estimates were examined). The analysis considered four distinct sexual act/role combinations (receptive and insertive anal intercourse, and receptive and insertive vaginal intercourse), which vary in the per-act probability that HIV would be transmitted, to determine the relative cost-effectiveness of treating these different routes of exposure. Transmission is only possible when the sexual partner is infected. The probability that a particular partner was infected was varied in the analysis from very low up to 1.0 (representing the case in which the partner was known to be infected).
In the first scenario, the potential HIV exposure was treated as an isolated incident. PEP, if successful, prevents the patient from becoming infected, hence the PEP program can be credited with averting an infection. In contrast, the second scenario assumed that PEP occurs within a context of continuing risk behavior, which may entail multiple episodes of PEP.
The main components of the cost-utility analysis are as follows: (i) estimation of the probability that the use of PEP would prevent the establishment of HIV infection, the consequent savings in future medical costs, and the number of QALY saved by preventing infection; (ii) calculation of the total cost of PEP, the cost per HIV infection averted, and the net cost (total cost minus future savings due to possible prevention of infection); and (iii) calculation of the cost-utility ratio (equation 1). In addition to the base-case analysis, various sensitivity analyses were performed to examine how variations in key parameter values affect the main results. Threshold analyses were also conducted to determine cut-off values above or below which PEP ceases to be cost-saving or cost-effective.
Parameter sources, base-case values, and ranges of values explored in the sensitivity analyses are listed in Table 1. The main steps of the analysis are described further below.
Transmission and prevention models
For an isolated, potential exposure to HIV (scenario 1), the probability that PEP prevents sustained infection in a previously uninfected individual is equal to the product,
of (i) the probability (π) that the sexual partner is infected; (ii) the per-contact HIV transmission probability (α) for a particular sex act/role combination; (iii) the percentage of patients (p) who complete the proposed 4-week regimen of antiretrovirals; and (iv) the effectiveness of chemoprophylaxis at preventing the establishment of sustained infection in the face of exposure to the virus (E). Incomplete therapy is assumed to be ineffective at preventing infection, but to incur a cost equal to some fixed proportion (e.g., 50%) of the cost of completed treatment (this assumption is reconsidered in Discussion). [Equation 2 can be derived as the difference between the probability of infection without PEP, πα, and the probability with PEP, (1 - E)pπα +(1 - p)πα.]
Repeated episodes of risky sex and subsequent PEP (scenario 2) can be described using a Bernoulli process model of HIV transmission, in which each act of intercourse followed by PEP is treated as a probabilistically independent event [27]. For n such episodes, the total cost of PEP is nK, where K is the cost of a single administration of PEP, and the probability that each of these acts of prophylaxis was successful is as follows:
Equation (Uncited)Image Tools
where m is the number of different sexual partners and j is the number of acts per partner. This model assumes varying degrees of independence amongst its many parameters. True independence is unlikely to hold in actual populations. For example, people who have many partners are likely to have partners who themselves have many partners, and the probability of transmission is likely to be higher for persons with many partners due to the higher probability of infection with other sexually transmitted diseases (STD). Thus, Sm may disproportionately underestimate the actual probability of transmission for high-risk individuals, which would lead to a corresponding underestimation of the cost-effectiveness of PEP for these individuals.
When j × m is fixed, Sm increases with the number of distinct partners, m; hence, Sm is maximal, and the analysis is the most favorable to PEP, when each partner is unique [27,28]. In this case,
(note that equation 5 is identical to equation 2 when n = 1).
If the patient engages in ongoing sexual risk behavior, then it may not be strictly accurate to view successful PEP as 'preventing' the patient from becoming infected with HIV; for those individuals whose risky behaviors substantially elevate their risk of infection, PEP might only delay infection rather than preventing it [23]. (However, as discussed further below, this is a somewhat controversial methodological point.)
If L denotes the patient's overall risk of becoming infected with HIV in an extended analytic window period (perhaps up to several years), excluding those episodes for which he or she receives PEP, then the probability that the individual did not become infected as a result of any of the exposures for which he or she did not receive PEP is 1 - L. In this case, whether PEP was successful determines whether the individual became infected. If the probability of successful PEP is as given in equation 4, then PEP prevents infection (in this strict sense) in
per cent of the cases. (It also delays infection in the L[1 - (1 -πα)n]pE per cent of patients for whom PEP was successful but who nevertheless became infected as a result of other risk behaviors.)
Transmission model parameter estimates
The probability that a particular sexual partner is HIV-infected (π) is varied in the analysis from 1.0 (certain infection) down to 0.02 (see below). Notice that the quantitative distinction between π = 1 and π < 1 is associated with a qualitative distinction with possibly important policy implications: in the former case it is known with certainty that the partner is infected, hence unprotected sex necessarily entails exposure, whereas in the latter case, exposure is uncertain. If the HIV status of a partner is completely unknown, then π may be estimated by the prevalence of HIV infection in the local population. Base-case probabilities of π = 0.18 and 0.02, respectively, were derived for anal and vaginal intercourse partners from estimates of the overall prevalence of HIV infection among men who have sex with men (MSM) and high-risk heterosexuals in the 96 largest US cities [29] (notice that this presumes homosexual anal intercourse).
Estimates of the per-contact HIV transmission probability (α) for different sexual act/role combinations were obtained from Katz and Gerberding's summary [8]; similar estimates have been reported by others [30]. The base-case transmission probabilities associated with receptive anal and vaginal intercourse are 0.02 (range explored in sensitivity analyses, 0.008-0.032) and 0.001 (range, 0.0005-0.0015), respectively. For insertive vaginal intercourse, the estimated base-case probability of transmission is 0.0006 (range, 0.0003-0.0009), and the probability for insertive anal intercourse is assumed to be similar [8]. As evident from these ranges, there is considerable uncertainty in the actual value of the transmission probability, which is affected by a number of factors, including the stage of disease of the infected person, the presence of STD in either partner, whether he or she is on effective antiretroviral therapy, and how rough the sex was (this may be especially relevant in cases of rape) [31].
The effectiveness of combination therapy in preventing the establishment of sustained viral infection following exposure to HIV is unknown. For present purposes, baseline estimates of PEP effectiveness were inferred from the CDC's multinational case-control study that compared health-care workers who received ZDV prophylaxis following percutaneous exposure to HIV-contaminated blood with those who did not [2]. Overall, this study documented a 79% reduction in seroconversion risk for health-care workers who received ZDV PEP (95% confidence interval, 43-94) [2]. In the base-case analysis, our study assumes that combination therapy following sexual exposure is as effective as ZDV monotherapy following occupational exposure. The impact of uncertainty in this parameter is assessed in the sensitivity analyses, in which PEP effectiveness is varied from 43 to 94% (the lower and upper bounds of the 95% confidence interval reported previously [2]).
Not all patients who initiate PEP will complete the full 4-week regimen [32]. Some patients will experience mild-to-moderate reactions to the drug therapy (and infrequently, more severe side-effects), causing them to prematurely discontinue PEP. The occurrence of side-effects in uninfected health-care workers receiving post-exposure ZDV monotherapy has been empirically characterized in clinical settings. Although ZDV is generally well-tolerated, side-effects such as gastrointestinal complaints, headaches, nausea and fatigue are experienced by one-third to three-quarters of health-care workers receiving ZDV prophylaxis [1,32-35]. Adverse effects of combination therapy in uninfected persons have not been well characterized. In HIV-infected adults, 3TC can cause headache, fatigue, gastrointestinal symptoms, and in rare cases, pancreatitis [1,36]. Indinavir can cause gastrointestinal symptoms, and less frequently, hyperbilirubinemia and kidney stones [37]. The impact of these potential side-effects on compliance with PEP regimens in sexually exposed patients is unknown.
In four domestic studies, completion rates for healthcare workers receiving ZDV PEP ranged from 59 to 76% [33]. In our analysis, a base-case completion rate of 69% was assumed [32] (this equals the rate observed in the CDC's occupational exposure surveillance program and thus matches the effectiveness data [2]), and was varied from 59 to 76% in the sensitivity analysis. Partial PEP was presumed to be ineffective and to incur a fixed proportion, varied from 0 to 100% with a base case of 50%, of the cost of the full treatment regimen [16].
Medical care cost savings and QALY saved
Recently revised estimates of the lifetime costs of medical care to treat a case of HIV disease and AIDS in the United States were used. HIV/AIDS-related medical care costs include the cost of medications for antiretroviral therapy, opportunistic infection prophylaxis and treatment, regular viral load and CD4 cell monitoring, hospital and or hospice charges, and so forth. Based on a careful review of the literature, Holtgrave and Pinkerton [38] estimated that standard HIV/AIDS care, in accordance with current recommendations regarding the initiation and sequencing of antiretroviral therapies [3], costed about US$ 195 188 in 1996 dollars discounted at 3% [38]. (Estimates of the cost of care prior to the introduction of combination therapy have been previously reported [39-43].) Therapy reduces the number of QALY that would otherwise be lost due to HIV infection. Nevertheless, a 26-year-old person with HIV who received the level of care envisioned in this medical care scenario would still lose 11.23 QALY (discounted at 3%); thus, 11.23 QALY are saved by preventing infection [38].
Two other medical care cost-QALY pairs are considered in the sensitivity analyses. In the low level of care scenario, lifetime medical care costs total US$ 87 045, and 13.18 QALY are lost when someone becomes infected; this medical care scenario assumes ZDV monotherapy, and therefore no longer represents an acceptable level of care but is included as a lower bound for comparison with previous analyses of HIV prevention interventions [23]. The high level of care scenario (which includes triple combination therapy from time of seroconversion) provides an upper bound cost estimate of US$ 296 844 and 9.34 lost QALY [38].
PEP cost parameters
The antiretroviral drugs in triple combination therapy are expensive and constitute the primary cost associated with PEP. Wholesale drug prices for a 4-week supply at recommended doses [1] are as follows: US$ 260 for ZDV (200 mg thrice daily); US$ 209 for 3TC (150 mg twice daily); and US$ 336 for indinavir (800 mg thrice daily) [44]. Total drug costs are therefore about US$ 469 for a 4-week regimen of ZDV and 3TC, and US$ 805 if indinavir is added. Katz and Gerberding [8] estimated the cost of PEP-associated laboratory work and office visits at between US$ 100 and 200. Taking the mean of US$ 150 [45], brings the total base-case cost of ZDV-3TC PEP to US$ 619 (US$ 955 for triple combination PEP).
As discussed above, in the base-case analysis, patients who prematurely discontinue therapy are assessed at a fixed percentage (50% in the base case) of the full treatment cost. The mean per-patient cost of providing PEP to an entire cohort (including some patients who might not complete therapy) is therefore,
Equation (Uncited)Image Tools
where K is the unadjusted cost of PEP (K = US$ 619 in the base case), p is the proportion of patients who complete the regimen, and y is the percentage of the total cost, K, that is assessed to patients who do not complete prophylaxis.
Three additional cost estimates were examined in the analysis in addition to the base-case estimate of US$ 619: (i) a lower bound estimate of US$ 400, as suggested by Katz and Gerberding [8]; (ii) a higher estimate of US$ 769 that includes an additional US$ 150 for more extensive laboratory work, and for monitoring and treating possible PEP-induced side-effects (side-effects requiring treatment have been relatively infrequent in health-care workers receiving ZDV PEP [8,32]); and (iii) an upper bound estimate of US$ 955 corresponding to triple combination PEP with ZDV, 3TC and indinavir.
Incremental analysis
Because the effectiveness of dual and triple combination PEP following sexual exposure is uncertain, it is difficult to compare the relative cost-effectiveness of these regimens. However, an incremental threshold analysis can be used to determine, for a given level of effectiveness of dual combination PEP, the minimum effectiveness of triple combination PEP necessary for the incremental cost-utility ratio to be less than some predetermined threshold (e.g., US$ 50 000 per QALY saved). If C2 and C3 represent the cost and A2 and A3 the effectiveness of dual and triple combination PEP, respectively, then the incremental cost-utility ratio equals,
Equation (Uncited)Image Tools
where and T and Q are defined as above (equation 1).
Results
The results of the base-case analysis of ZDV-3TC combination PEP (scenario 1) are shown in Table 2. In this analysis, which assumes that the probability that a partner is infected is the same as the prevalence of infection in the appropriate US population (MSM, or high-risk heterosexuals), PEP following receptive anal intercourse could be expected to prevent a total of nearly 20 HIV infections amongst a cohort of 10 000 MSM (including those who do not complete therapy), at a gross cost of about US$ 5.2 million, or US$ 267 000 (gross) per infection averted. The cost-utility ratio is just over US$ 6000 per QALY saved, making this a highly cost-effective intervention. Prophylaxis is not cost-effective for sex act/role combinations other than receptive anal intercourse (all cost-utility ratios exceed US$ 750 000 per QALY saved).
However, if it is known with certainty that the sexual partner is infected (π = 1 in equation 2), then PEP following receptive vaginal intercourse appears to be cost-effective by conventional standards (Fig. 1). In contrast, PEP following insertive exposures is marginally cost-effective at best. If there is sufficient uncertainty about the partner's HIV status, then PEP is probably cost-effective only for receptive anal intercourse (Fig. 1): the cost-utility ratio for receptive anal intercourse exceeds US$ 100 000 per QALY saved only when the probability that the partner is infected, π, is less than 0.04. Moreover, PEP is actually costsaving for receptive anal intercourse if π is greater than about 0.25. In contrast, for receptive vaginal intercourse the ratio exceeds US$ 100 000 per QALY saved whenever π is less than 0.73; for insertive intercourse the cost-utility ratio is greater than US$ 100 000 per QALY saved regardless of the value of π.
Table 3 illustrates the effect of varying the per-contact transmission probabilities for the different sex act/role combinations. Although the results are sensitive to variations in this parameter, it again appears that PEP is cost-effective under only two conditions: (i) following receptive anal intercourse, and (ii) when it is likely that the partner is infected. If the partner is known to be infected, PEP is cost-effective following insertive anal or vaginal intercourse at the upper bound of the range of transmission probabilities, but not at the lower bound; a similar comment applies to PEP after receptive vaginal intercourse. When the partner is unlikely to be infected, PEP is clearly not cost-effective for receptive vaginal, insertive anal, or insertive vaginal intercourse, and may not be cost-effective for receptive anal intercourse if the per-contact transmission probability is very small (e.g., π = 0.02, α = 0.008 in Table 3).
This observation is underscored by the results of the cost-threshold analysis, which indicate that even if PEP were completely effective and all patients completed the prescribed regimen (the best-case scenario), PEP after insertive anal intercourse would need to cost less than US$ 239 to be economically defensible (costutility ratio less than US$ 180 000 per QALY saved); the corresponding figures for receptive and insertive vaginal intercourse are US$ 44 and 26, respectively. These PEP cost estimates are much less than the base-case value of US$ 619 per patient and also less than the lower bound cost estimate of US$ 400 [8].
Because PEP is not especially cost-effective for exposure routes other than receptive anal intercourse when the partner's HIV status is unknown, the remaining analyses shall focus on four scenarios: receptive anal intercourse at the base-case partner infection probability (π = 0.18), and receptive anal, receptive vaginal, and insertive intercourse with a partner who is known to be infected (π = 1). The results of the one-way and multiway sensitivity analyses are presented in Table 4. Prophylaxis following receptive anal intercourse remains highly cost-effective for π = 0.18 and costsaving for π = 0 under all the conditions assessed. In particular, PEP is highly cost-effective for receptive anal intercourse at all of the effectiveness values examined; moreover, a threshold analysis indicates that the cost-utility ratio remains below US$ 50 000 per QALY saved provided that PEP is at least 28% effective.
The cost-utility ratios for receptive vaginal and insertive intercourse are sensitive to the assumed effectiveness of PEP, to the discount rate used to convert the cost-of-illness and QALY estimates into present value, and to the cost of PEP, but are not especially sensitive to the completion rate, the cost of incomplete therapy, or the particular medical care scenario (Table 4). In general, PEP following receptive vaginal intercourse with an infected partner costs about US$ 50 000-100 000 per QALY saved, hence is likely to be cost-effective. Prophylaxis following insertive intercourse, in contrast, generally costs over US$ 100 000 per QALY saved, and therefore is probably not cost-effective.
Incremental analysis
PEP for receptive anal intercourse remains highly cost-effective even when the additional cost of indinavir is included (the cost-utility ratio is under US$ 50 000 per QALY saved provided that PEP costs less than US$ 1750; Table 4). In contrast, triple combination PEP is only marginally cost-effective for receptive vaginal intercourse, and not cost-effective for insertive anal or vaginal intercourse. However, this analysis may be misleading because it assumes that triple combination PEP is no more effective than dual combination therapy (assumed to be 79% effective).
The results of the incremental threshold analysis indicate that for triple combination PEP to be incrementally cost-effective (at US$ 100 000 per QALY saved) relative to dual combination therapy would require that it be at least 9% more effective when used after receptive anal intercourse (assuming π = 0.18), and 31 or 52% more effective after receptive vaginal or insertive intercourse, respectively, with a partner known to be infected (e.g., if dual combination PEP is 79% effective for receptive anal intercourse, triple combination therapy would need to be at least 88% effective in order to be incrementally cost-effective). Because triple combination PEP is unlikely to be 30% more effective than dual combination PEP, it is probably not incrementally cost-effective for any sexual behavior other than receptive anal intercourse, regardless of the HIV status of the partner. Similarly, for triple combination PEP following receptive anal intercourse (π = 0.18) to be incrementally cost-saving relative to dual combination PEP would require it to be 59% more effective, an unlikely possibility; thus, the additional drug costs will not be offset by treatment savings.
Ongoing risk (scenario 2)
The effect of individuals undergoing repeated episodes of PEP appears to be minimal. For PEP following receptive anal intercourse (π = 0.18), the cost-utility ratio rises from US$ 6354 per QALY saved to US$ 6741 per QALY saved, a 6% increment, when the number of episodes increases from 1 to 10; the increase is essentially negligible (<1%) for prophylaxis after receptive vaginal or insertive intercourse with a known-infected partner.
Because the cost-utility ratio is not especially sensitive to repeated PEP, the following analysis assumed only a single episode of PEP. If an individual seeks PEP once, against a backdrop of ongoing HIV risk behavior, Fig. 2 illustrates how the cost-effectiveness of PEP decreases as the background risk level rises for receptive anal intercourse (π = 0.18), and receptive vaginal or insertive intercourse with a partner known to be infected. The cost-utility ratio increases rapidly at elevated levels of background risk, L (equation 6). However, for PEP after receptive anal intercourse, the ratio remains less than US$ 50 000 per QALY saved for risk levels up to 0.64, and less than US$ 100 000 per QALY saved for risk levels up to 0.79. For receptive vaginal intercourse with an infected partner the cost-utility ratio is less than US$ 100 000 per QALY saved when the background risk is less than 0.27 (for comparison, the risk of transmission arising from 100 unprotected receptive vaginal contacts with an infected partner is about 0.1). For insertive intercourse with an infected partner, the cost-utility ratio exceeds US$ 125 000 per QALY saved, regardless of the background risk level.
Discussion
The above cost-utility analysis suggests that prophylaxis following suspected sexual exposure to HIV is cost-effective compared with other health-related programs in the United States only for receptive anal intercourse or for receptive vaginal intercourse with a partner who is likely to be infected. (If the probability that the partner is infected is very small, however, PEP may not be cost-effective following receptive anal intercourse.) PEP after insertive vaginal or anal intercourse is probably not cost-effective, regardless of the partner's risk status. Thus, from a purely economic standpoint, PEP should be restricted to regular partners of infected persons (e.g., serodiscordant couples), to patients reporting unprotected receptive anal intercourse (including condom breakage), and possibly to cases where there is either a substantial likelihood that the partner is infected, there is reason to suspect that transmission would be unusually effective (e.g., the patient or partner has an STD or genital ulcers, or the sex was violent or traumatic), or in cases of sexual assault [7].
The analysis is somewhat sensitive to several parameters, most notably the uncertain effectiveness of PEP in preventing the establishment of infection in HIV-exposed individuals. Although it seems reasonable to suppose that the effectiveness of combination therapy would exceed that of ZDV monotherapy, in practice the effectiveness of PEP may depend critically on the elapsed time between exposure and the initiation of therapy, as suggested by animal models of PEP [46]. The PHS recommends that exposed health-care workers should initiate treatment within 1-2 h [1]; therapy is probably not effective if begun more than 24 h after exposure [47].
This uncertainty in the effectiveness of PEP has obvious implications for the cost-effectiveness of PEP. For example, if the success of PEP depends upon preventing early replication, then perhaps a 2-week course of combination therapy would be as effective as the full 4-week regimen, but at a much lower cost [3]. But, even if PEP were perfectly effective and the completion rate for therapy were 100%, PEP would still need to cost less than US$ 239, 44, or 26 for insertive anal, receptive vaginal, or insertive vaginal intercourse with a partner of unknown HIV status, respectively (assuming base-ease values for the remaining parameters) for it to be even marginally cost-effective (US$ 180 000 per QALY saved). Conversely, if the effectiveness of combination PEP following sexual exposure is much less than assumed in the base case (e.g., as a result of long delays between exposure and treatment), then prophylaxis after receptive vaginal intercourse with a partner known to be infected might no longer be cost-effective (e.g., if PEP was only 43% effective rather than 79%, then the cost-utility ratio would rise from US$ 68 000 to 140 000); in contrast, PEP following receptive anal intercourse remains highly cost-effective in the base case even at the lower effectiveness estimate.
Prophylaxis after receptive anal intercourse also remains cost-effective in the base case when much higher PEP costs are assumed, in contrast to PEP after receptive vaginal intercourse with an infected partner. In particular, PEP still appears to be cost-effective following receptive anal intercourse when the greater cost of triple combination PEP is considered, even when it is assumed that triple combination therapy is no more effective than dual combination PEP. Further analyses indicate that triple combination would need to be 9% more effective when used following receptive anal intercourse to be considered incrementally costeffective (US$ 100 000 or less per QALY saved) in relation to dual combination PEP.
A related issue concerns the base-case assumption that PEP is completely ineffective when the antiretroviral regimen is discontinued prematurely. It is possible that the benefit of PEP is fully realized in the first course of therapy. If prematurely halted PEP was completely effective (instead of completely ineffective, as assumed above), the base-case cost-utility ratios for receptive anal (probability, π, that partner is infected = 0.18), receptive vaginal (π = 0.02), insertive anal (π = 0.18), and insertive vaginal intercourse (π = 0.02) would decrease from about US$ 6354, 4.3 million, 774 000, and 7.1 million per QALY saved, to US$ 1004, 2.9 million, 528 000, and 4.9 million per QALY saved, respectively. (This analysis assumes, as before, that incomplete therapy incurs a cost equal to half that of full therapy.) Thus, in this scenario, PEP following receptive anal intercourse would be cost-saving, but the overall pattern of results would be little changed.
The analysis also assumes that all individuals receiving PEP are uninfected, and that PEP can therefore prevent them from becoming infected. This assumption will not hold in all instances. Treating already infected individuals is clearly not cost-effective from a purely preventive standpoint, although there may be important benefits associated with identifying infected persons and initiating (non-prophylactic) treatment. Because the probability that the patient was already infected (prior to the exposure for which he or she is seeking prophylaxis) is likely to be correlated with the overall prevalence of infection in the community, the above results may disproportionately overstate the costeffectiveness of PEP in high prevalence communities.
The above results suggest that PEP following receptive anal intercourse is cost-effective under most of the conditions examined here, particularly when the probability that the partner is infected is assumed to be the same as the overall estimated HIV prevalence among MSM in the United States. Although PEP remains costeffective (US$ 100 000 or less per QALY saved) for receptive anal intercourse provided that this probability is greater than 0.04, it may not be cost-effective when the probability that the partner is infected is small (e.g., in countries with a low prevalence of HIV among MSM). Similarly, PEP may not be cost-effective for women reporting anal intercourse exposures with low-risk (e.g., non-drug-injecting) strictly heterosexual men.
The impact of individuals receiving PEP on multiple occasions is comparatively small, amounting to less than a 6% change in the cost-utility ratio in all scenarios considered. However, PEP appears to be less cost-effective when considered in a context of ongoing HIV risk behaviors. Whether it is appropriate to incorporate exogenous risk into the analysis is controversial [23]. Although one may argue that any future HIV/AIDS-associated medical care savings from successful PEP are illusory if the patient subsequently becomes infected through additional exposures and therefore incurs these costs, a similar objection could be raised if the patient died from a heart attack 1 month after successful PEP. By definition, successful PEP prevents the patient from becoming infected at that time. Whatever happens in the future (HIV, a heart attack, or a motor vehicle accident) falls into the category of competing risks; such risks are often not explicitly incorporated into cost-effectiveness analyses.
Nevertheless, individuals who seek PEP repeatedly or who do so in a context of continuing sexual risk behavior would probably benefit from counseling or participating in a behavioral risk-reduction intervention, many of which are themselves highly cost-effective [15,48,49]. Studies to determine the most effective and cost-effective models for combining PEP with behavioral risk-reduction programs are needed.
Because the cost-effectiveness of PEP varies widely according to the type of exposure and the risk status of the partner, PEP 'on demand' for all who request it does not appear to be an economically efficient use of limited HIV prevention resources. If PEP was made available to all patients, regardless of the risk of the suspected exposure, the provision of PEP to individuals with low-risk exposures would diminish the overall cost-effectiveness of the program. Conversely, it seems impractical to restrict PEP to patients who report having been exposed to HIV through receptive anal intercourse, while simultaneously denying PEP to patients who were exposed through other routes. Moreover, barring access to effective treatment is ethically dubious.
A related limitation of the above analysis concerns the accuracy of the per-contact transmission probabilities used to characterize the various sex act/role combinations. These probabilities are not known with certainty, and the probable ranges for different act/role combinations may overlap (Table 3) [30,31]. Moreover, numerous patient and partner factors are believed to affect the risk of transmission, including stage of disease, viral load, genetics, and facilitation by STD [31]. These factors should be considered when making PEP decisions with individual patients.
Even greater uncertainty surrounds the probability of transmission for oral sex. Although this probability has not been carefully quantified, it presumably is much lower than that associated with anal or vaginal intercourse. Although PEP following oral sex is not explicitly examined above, PEP is probably less cost-effective for oral sex than for insertive anal or vaginal intercourse; hence, the results obtained for these transmission modalities can serve as a guide.
Because economic resources to prevent the spread of HIV are limited, society cannot afford to fund inefficient programs. A number of HIV prevention interventions, including needle exchange, behavioral risk reduction, and perinatal prophylaxis, have been shown to be cost-effective or even cost-saving [10]; many of these programs report cost-utility ratios of less than US$ 10 000 per QALY saved [15,48,49]. The economic analysis presented above implies that expanding such programs would constitute a more efficient use of limited HIV prevention resources than the provision of PEP on demand.
There are, of course, other important considerations besides cost-effectiveness that should influence whether PEP is offered. PEP may not be appropriate or feasible for all patients, due to the complex dosing requirements of combination antiretroviral therapies and possible drug interactions with both prescription and over-the-counter medications. There is also a theoretical risk that PEP could facilitate the development of antiretroviral-resistant strains of HIV when prescribed to an already infected patient or when it fails to prevent infection in a previously uninfected individual [4,6,8]. Conversely, as a form of early therapy, PEP might benefit the patient even if it is unsuccessful in preventing the establishment of infection.
Amongst the benefits of PEP that are harder to quantify is the alleviation of psychological distress associated with possible exposure. This aspect might be especially valuable to survivors of sexual assault, many of whom worry that their assailant might have infected them with HIV [7]. Similarly, PEP could provide backup protection for HIV-discordant couples who regularly practice safer sex but who are accidentally exposed when a condom breaks. Conversely, the availability of PEP might be seen by some people as condoning unsafe sex [6,8], much as the availability of antibiotic treatment for gonorrhea resulted in decreased public anxiety about this once-feared pathogen [50]. Regardless of whether PEP is offered, persons presenting with anxiety over possible exposure to HIV should receive appropriate psychological and HIV riskreduction counseling.
Public health decision makers, physicians, and other health-care professionals will need to balance myriad concerns when deciding whether and when PEP should be provided. However, the results of the economic analysis presented above support only a limited role for PEP in the prevention of HIV.
References
1. Centers for Disease Control: Update: provisional Public Health Service recommendations for chemoprophylaxis after occupational exposure to HIV. MMWR 1996, 45:468-472.
2. Centers for Disease Control: Case-control study of HIV seroconversion in health-care workers after percutaneous exposure to HIV-infected blood - France, United Kingdom, and United States, January 1988-August 1994. MMWR 1995, 44:929-933.
3. Carpenter CCJ, Fischl MA, Hammer SM, et al.: Antiretroviral therapy for HIV infection in 1996: recommendations of an international panel. JAMA 1996, 276:146-154.
4. American Health Consultants: CDC: avoid preventive therapy for sexual risk. AIDS Alert 1996, 11:75-76.
5. Gerberding JL: Prophylaxis for occupational exposure to HIV. Ann Intern Med 1996, 125:497-501.
6. Macready N: Investigators study morning-after AIDS treatment [news]. Lancet 1997, 349:780.
7. Gostin LO, Lazzarini Z, Alexander D, Brandt AM, Mayer KH, Silverman DC: HIV testing, counseling, and prophylaxis after sexual assault. JAMA 1994, 271:1436-1444.
8. 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 1997, 336:1098-1100.
9. Centers for Disease Control: Public Health Service statement on management of occupational exposure to human immunodeficiency virus, including considerations regarding postexposure use. MMWR 1990, 39 (RR-1):1-14.
10. Mathes LF, Polas PJ, Hayes KA, Swenson CL, Johnson S, Kociba GJ: Pre- and postexposure chemoprophylaxis: evidence that 3′-azido-3′-dideoxythymidine inhibits feline leukemia virus disease by a drug-induced vaccine response. Antimicrob Agents Chemother 1992, 36:2715-2721.
11. Martin LN, Murphey-Corb M, Soike KF, Davison-Fairburn B, Baskin GB: Effects of initiation of 3′-azido-3′-dideoxythymidine (zidovudine) treatment at different times after infection of rhesus monkeys with simian immunodeficiency virus. J Infect Dis 1993, 168:825-835.
12. Gorman C: If the condom breaks. Time 1997, 23 June.
13. Zuger A: 'Morning after' treatment for AIDS. New York Times 1997, 10 June.
14. Holtgrave DR: Cost analysis and HIV prevention programs. Am Psychol 1994, 49:1088-1089.
15. Holtgrave DR, Qualls NL, Graham JD: Economic evaluation of HIV prevention programs. Annu Rev Public Health 1996, 17:467-488.
16. Pinkerton SD, Holtgrave DR, Pinkerton HJ: Cost-effectiveness of chemoprophylaxis after occupational exposure to HIV. Arch Intern Med 1997, 157:1972-1980.
17. Pinkerton SD, Holtgrave DR, Bloom FR: Postexposure treatment of HIV. N Engl J Med 1997, 337:500-501.
18. Drummond MF, Stoddart GL, Torrance GW: Methods for the Economic Evaluation of Health Care Programmes. New York: Oxford University Press; 1987.
19. Haddix AC, Teutsch SM, Shaffer PA, Duñet DO (Eds): Prevention Effectiveness: A Guide to Decision Analysis and Economic Evaluation. New York: Oxford University Press; 1996.
20. Weinstein MC, Stason WB: Foundations of cost-effectiveness analysis for health and medical practices. N Engl J Med 1977, 296:716-721.
21. Gold MR, Siegel JE, Russell LB, Weinstein MC (Eds): Cost-Effectiveness in Health and Medicine. New York: Oxford University Press; 1996.
22. Pinkerton SD, Holtgrave DR: Assessing the cost-effectiveness of HIV prevention interventions: a primer. In Handbook of Economic Evaluation of HIV Prevention Programs. Edited by Holtgrave DR. New York: Plenum Press; 1998 (in press).
23. Pinkerton SD, Holtgrave DR: A method for evaluating the economic efficiency of HIV behavioral risk reduction interventions. AIDS Behav 1998 (in press).
24. Ganiats TG, Wong AF: Evaluation of cost-effectiveness research: a survey of recent publications. Fam Med 1991, 23:457-462.
25. Laupacis A, Feeny D, Detsky AS, Tugwell PX: Tentative guidelines for using clinical and economic evaluations revisited. Can Med Assoc J 1993, 148:927-929.
26. Fabian R: The Qualy approach. In Valuing Health for Policy: An Economic Approach. Edited by Tolley GL, Kenkel D, Fabian R. Chicago: University of Chicago Press; 1994:118-136.
27. Pinkerton SD, Abramson PR: The Bernoulli-process model of HIV transmission: applications and implications. In Handbook of Economic Evaluation of HIV Prevention Programs. Edited by Holtgrave DR. New York: Plenum Press; 1998 (in press).
28. Pinkerton SD, Abramson PR: Evaluating the risks: a Bernoulli process model of HIV infection and risk reduction. Eval Rev 1993, 17:504-528.
29. Holmberg SD: The estimated prevalence and incidence of HIV in 96 large US metropolitan areas. Am J Public Health 1996, 86:642-654.
30. Mastro TD, de Vincenzi I: Probabilities of sexual HIV-1 transmission. AIDS 1996, 10 (suppl A):S75-S82.
31. Royce RA, Seña A, Cates Jr W, Cohen MS: Sexual transmission of HIV. N Engl J Med 1997, 336:1072-1078.
32. Tokars JI, Marcus R, Culver DH, et al.: Surveillance of HIV infection and zidovudine use among health care workers after occupational exposure to HIV-infected blood. Ann Intern Med 1993, 118:913-919.
33. Henderson DK: Postexposure chemoprophylaxis for occupational exposure to human immunodeficiency virus type 1: current status and prospects for the future. Am J Med 1991, 91:322S-329S.
34. Puro V, Ippolito G, Guzzanti E, et al.: Zidovudine prophylaxis after accidental exposure to HIV: the Italian experience. AIDS 1992, 6:963-969.
35. Schmitz SH, Scheding S, Voliotis D, Rasokat H, Diehl V, Schrappe M: Side-effects of AZT prophylaxis after occupational exposure to HIV-infected blood. Ann Hematol 1994, 69:135-138.
36. van Leeuwen R, Katlama C, Kitchen V, et al.: Evaluation of safety and efficacy of 3TC (lamivudine) in patients with asymptomatic or mildly symptomatic human immunodeficiency virus infection: a phase I/II study. J Infect Dis 1995, 171:1166-1171.
37. Deeks SG, Smith M, Holodniy M, Kahn JO: HIV-1 protease inhibitors: a review for clinicians. JAMA 1997, 277:145-153.
38. 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 Hum Retrovirol 1997, 16:54-62.
39. Gable CB, Tierce JC, Simison D, Ward D, Motte K: Costs of HIV+/AIDS at CD4+ counts disease stages based on treatment protocols. J Acquir Immune Defic Syndr Hum Retrovirol 1996, 12:413-420.
40. Guinan ME, Farnham PG, Holtgrave DR: Estimating the value of preventing a human immunodeficiency virus infection. Am J Prev Med 1994, 10:1-4.
41. Hellinger FJ: The lifetime cost of treating a person with HIV. JAMA 1993, 270:474-478.
42. Hurley SF, Kaldor JM, Gardiner S, Carlin JB, Assunção RM, Evans DB: Lifetime cost of human immunodeficiency virus-related health care. J Acquir Immune Defic Syndr Hum Retrovirol 1996, 12:371-378.
43. Pinkerton SD, Holtgrave DR: Lifetime costs of HIV/AIDS medical care [letter]. J Acquir Immune Defic Syndr Hum Retrovirol 1997, 14:380-381.
44. Editorial: New drugs for HIV infection. Med Lett Drugs Ther 1996, 38:35-37.
45. Li WR, Wong JB: Postexposure treatment of HIV. N Engl J Med 1997, 337:499-500.
46. Black RJ: Animal models of prophylaxis. Am J Med 1997, 102 (suppl 5B):S39-S44.
47. Gerberding JL: Management of occupational exposures to blood-borne viruses. N Engl J Med 1995, 332:444-451.
48. Holtgrave DR, Pinkerton SD: The cost-effectiveness of small group and community-level interventions. In Handbook of Economic Evaluation of HIV Prevention Programs. Edited by Holtgrave DR. New York: Plenum Press; 1998 (in press).
49. Holtgrave DR, Pinkerton SD: The economics of HIV primary prevention. In Handbook of HIV Prevention. Edited by Peterson JL, DiClemente RJ. New York: Plenum Press; 1998 (in press).
50. Brandt AM: No Magic Bullet: A Social History of Venereal Disease in the United States. Oxford: Oxford University Press; 1985.
© 1998 Lippincott Williams & Wilkins, Inc.