Skip Navigation

Expanded Access to Highly Active Antiretroviral Therapy: A Potentially Powerful Strategy to Curb the Growth of the HIV Epidemic

  1. Viviane D. Lima1,4,
  2. Karissa Johnston2,3,
  3. Robert S. Hogg1,5,
  4. Adrian R. Levy2,3,
  5. P. Richard Harrigan1,4,
  6. Aranka Anema1 and
  7. Julio S. G. Montaner1,4
  1. 1British Columbia Centre for Excellence in HIV/AIDS
  2. 2Centre for Health Evaluation and Outcome Sciences, St. Paul's Hospital
  3. 3Department of Health Care and Epidemiology, University of British Columbia, Vancouver
  4. 4Department of Medicine, Faculty of Medicine, University of British Columbia, Vancouver
  5. 5Faculty of Health Sciences, Simon Fraser University, Burnaby, British Columbia, Canada
  1. Reprints or correspondence: Prof. Julio S. G. Montaner, Director, BC Centre for Excellence in HIV/AIDS, 667-1081 Burrard St., Vancouver, BC, Canada V6Z 1Y6 (jmontaner{at}cfenet.ubc.ca).
 
Next Section

Abstract

We developed a mathematical model using a multiple source of infection framework to assess the potential effect of the expansion of highly active antiretroviral therapy (HAART) coverage among those in medical need on the number of individuals testing newly positive for human immunodeficiency virus (HIV) and on related costs in British Columbia, Canada, over the next 25 years. The model was calibrated using retrospective data describing antiretroviral therapy utilization and individuals testing newly positive for HIV in the province. Different scenarios were investigated on the basis of varying assumptions regarding drug resistance, adherence to HAART, therapeutic guidelines, degree of HAART coverage, and the timing of HAART uptake. Expansion of HAART lead to substantial reductions in the growth of the HIV epidemic and related costs. These results provide powerful additional motivation to accelerate the roll out ofHAARTprograms aggressively targeting those in medical need, both for their own benefit and as a means of decreasing new HIV infections.

The continued growth of the HIV epidemic poses a formidable challenge, even in the developed world [1]. The incidence of HIV infection remains unacceptably high [2]. Despite their potential efficacy and obvious appeal, HIV prevention strategies have been only partially successful [3]. Unfortunately, neither a cure nor a highly effective preventive vaccine are anticipated within the foreseeable future [4, 5]. At the rate of growth of the epidemic, in addition to the increasing human toll, the costs attributable to care and management of this growing and aging population may become unsustainable, in part because of the success of therapy in keeping infected patients alive for long periods [6–8].

Highly active antiretroviral therapy (HAART) has led to a dramatic decrease in morbidity and mortality among individuals infected with HIV, and international guidelines widely recommend that HAART be used before overt immune deficiency is apparent [9, 10]. The main goal of HAART is to attain durable suppression of viral replication (i.e., plasma HIV-1 RNA levels <50 copies/mL) [9, 10]. In recent years, provision of HAART to those in need has become an increasingly important and feasible global priority [11]. The current strategy is focused on the roll out of HAART programs on the basis of strict medical need [12, 13]. Priority is therefore being given to the treatment of those at the highest risk of disease progression or death. However, HAART coverage remains suboptimal, even in resource-rich areas of the world [11]. For example, despite universal access to health services and antiretroviral therapy in the province of British Columbia (BC), Canada, it is estimated that only 50% of those medically eligible to receive HAART are currently taking therapy [6]. Logistical and funding limitations have typically hampered the ability of therapeutic programs to expand to hard-to-reach populations or to those at earlier stages of HIV infection, leading to avoidable HIV/AIDS-related morbidity, mortality, and associated resource utilization.

In adherent patients, HAART predictably decreases plasma HIV-1 RNA levels to below the levels of detection of currently available assays [14]. As a result, HAART leads to an immunological benefit characterized by an increase in CD4 cell count, reducing, hence, the risk of AIDS-related disease progression and death [15]. Of note, HAART-induced decreases in plasma HIV-1 RNA levels have been shown to be associated with similarly marked reductions in HIV-1 RNA levels in genital secretions in men and women [16, 17] and with a substantial reduction in the risk of mother-to-child HIV transmission when HAART is given to HIV-infected pregnant women [18–20]. More recently, HAART has been shown to be associated with a decrease in HIV transmission between serodiscordant heterosexual couples, despite continued exposure [21––23]. It is therefore reasonable to speculate that antiretroviral therapy could have an additional beneficial impact by reducing the spread of HIV [24].

Several early mathematical modeling studies raised the concern that any possible benefit of antiretroviral therapy on the spread of HIV could be readily offset by even modest increases in HIV risk behavior [25–30]. More recently, however, Abbas et al. [28] have revisited this issue, proposing a mathematical model that predicts that antiretroviral therapy should have both individual and public health benefits that increase with time and, importantly, with the proportion of infected persons treated.

We therefore conducted the present study to assess the potential effect of expanding HAART coverage on the number of individuals testing newly positive for HIV and on related direct treatment costs in British Columbia over the next 25 years, using varying scenarios regarding rates of emergence of HIV drug resistance, degree of adherence to HAART, shift in therapeutic guidelines, and degree of HAART coverage.

Previous SectionNext Section

Methods

The British Columbia Centre for Excellence in HIV/AIDS (BC-CfE) transmission model. The BC-CfE semideterministic dynamic transmission model (figure 1) was built on the basis of the different stages of the natural history of HIV infection, whereby lower CD4 cell counts and higher plasma HIV-1 RNA levels are independently associated with a higher chance of HIV transmission [31–33]. The model also takes into consideration different predominant sources of HIV transmission in British Columbia. This includes transmission from men who have sex with men (MSM), injection drug users (IDUs), and MSM/IDUs, allowing the evaluation of the individual and overall contribution of these sources to the epidemic. The model characterizes the natural history of HIV infection on the basis of 4 infectivity periods: susceptible, primary infection (⩽12 weeks since HIV infection), symptomatic phase defined by 4 plasma HIV-1 RNA strata (<3, ⩾3 and <4, ⩾4 and <5, and ⩾5 log10 copies/mL), and late stage (individuals experiencing opportunistic diseases) [31]. During any of these infectivity periods we considered death as a possible outcome. The model considers a CD4 cell count ⩽200 cells/mm3 to be the key eligibility criterion for starting therapy.

In this model, plasma HIV-1 RNA levels and CD4 cell counts are updated every 3 months until death, and patient trajectories follow 3 main pathways. In the first, patients never receive treatment, and the plasma HIV-1 RNA and CD4 cell count trajectories are assumed to follow the natural history of HIV infection until death [31]. The second represents patients receiving HAART (i.e., started therapy with a CD4 cell ⩽200 cells/mm3), and CD4 cell counts and plasma HIV-1 RNA levels are modeled using populational data collected through the BC-CfE Drug Treatment Program (BC-CfE DTP). The third pathway represents patients waiting to start treatment once their CD4 cell count reaches ⩽200 cells/mm3. Antiretroviral agents are distributed at no cost to all eligible HIV-infected British Columbia residents through theDTPunder the auspices of the BC-CfE [14, 34]. Antiretroviral therapy guidelines are regularly updated by the center and remain consistent with those of the International AIDS Society–USA [9].

We modeled the epidemic in 3 major high-risk populations: MSM, IDUs, and MSM/IDUs. Appendix A, which appears only in the electronic edition of the Journal presents all of the details for the building of the BC-CfE transmission model. The transmission parameters for each risk category are described at the beginning of appendix A. The assumptions for these parameters were based on an extensive review of the literature and on consultations with experts in the field. These assumptions were calibrated and validated using 2 sources of information: the number of individuals testing newly positive for HIV obtained from the British Columbia Centre for Disease Control (BC-CDC) [35, 36], and the past and current active number of patients enrolled in the BC-CfE DTP. Figure 2A and 2B demonstrates that the parametric assumptions are adequate to fit the empirical data (P > .5 for statistical difference).

Figure 1.
Figure 1.

The British Columbia Centre for Excellence in HIV/AIDS transmission model. Patient trajectories are followed through 3 pathways: (1) those who will never receive treatment, (2) those who receive treatment because their CD4 cell count is currently ⩽200 cells/mm3, and (3) those who will eventually start treatment once their CD4 cell count falls to ⩽200 cells/mm3. Individuals in pathway 2 randomly move among infectivity strata each time the model is updated. This diagram shows only the distribution of the population at a given point in time. The infectivity groups are classified as follows: susceptible; primary infection (I0); the symptomatic phase, defined as having an HIV RNA plasma viral load <3 log10 copies/mL (I1), ⩾3 and <4 log10 copies/mL (I2), ⩾4 and <5 log10 copies/mL (I3), or ⩾5 log10 copies/mL (I4); and late stage. The susceptible population for each risk group (men who have sex with men [MSM], injection drug users [IDUs], and MSM/IDUs) has a specific probability of becoming infected according to the parameters listed in appendix A.

Figure 2.
Figure 2.

A, Comparison of empirical (source: British Columbia Centre for Excellence in HIV/AIDS drug treatment program [DTP]) and estimated no. of individuals receiving antiretroviral treatment from 1993 through 2005. B, Comparison of empirical (source: British Columbia Centre for Disease Control [BC-CDC]) and estimated no. of individuals testing newly positive for HIV from 1993 through 2005.

Estimates of direct treatment costs. Cost estimates associated with HAART expansion were based on the BC-CfE DTP and on the current recommendations for treating a treatmentnaive patient with HIV/AIDS [9]. In 2006, the direct treatment cost to treat each person with first-line therapy was estimated to be Can$17,288.22 (source, BC-CfE DTP). To calculate the treatment cost in the years 2007–2030, we assumed a discounting of the future health rate equal to 3%, after considering what value of discounting is conventionally used in the literature [37]. To obtain the lifetime cost for treating each individual testing newly positive for HIV, we based our estimate on a life expectancy of 22.9 years at the age of 30 years [34].

Structure of the BC-CfE model. Appendix A presents the BC-CfE transmission model in detail. Briefly, the BC-CfE model is composed by 1 deterministic and 2 random components. The deterministic component estimates the number of individuals testing newly positive for HIV generated each time the model is updated, defined by the set of differential equations for each risk category. The 2 random components consist of 2 statistical modeling techniques to predict CD4 cell count and plasma HIV-1 RNA level trajectories and the probability of the emergence of HIV drug resistance. We used generalized additive mixed models (GAMMs) to predict CD4 cell count and plasma HIV-1 RNA level trajectories [38]. GAMMs are an extension of generalized linear models, which allow for fitting smooth nonparametric curves describing the association between outcomes and predictor variables.

The following predictors were investigated: age, sex, CD4 cell count, plasma HIV-1 RNA level (log10 transformed), presenting date, previous AIDS diagnosis, history of injection drug use, first HAART date, first HAART regimen, adherence, drug resistance, and death. The presenting date was defined as the date of the first (ever) plasma HIV-1 RNA level and CD4 cell count measurement before therapy initiation. Estimates of adherence to antiretroviral therapy were based on dispensed medications, also known as refill compliance. For this study, we limited our measurement of adherence to the first year of therapy, estimated by dividing the number of months of medications dispensed by the number of months of followup. This measure of adherence has been found to be independently associated with HIV suppression and survival among HIV-infected persons enrolled in the DTP [34, 39]. HIV drug-resistance genotyping data were based on assays attempted on all plasma samples collected during the first 30 months after the initiation of HAART for which the plasma HIV-1 RNA level was ⩾1000 copies/mL or by physician request. Samples were assigned to 1 of 4 resistance categories (nucleoside reverse-transcriptase inhibitor [NRTI], protease inhibitor [PI], or nonnucleoside reverse-transcriptase inhibitor [NNRTI], or lamivudine/emtracitabine resistance [3TC]) on the basis of a modification of the International AIDS Society–USA table, as detailed in D'Aquila et al. [40].

Previous SectionNext Section

Results

Predictors of population plasma HIV-1 RNA level and drug resistance. The main predictors of current plasma HIV-1RNA level were adherence and previous plasma HIV-1 RNA level, whereas the main predictors of the emergence of drug resistance were plasma HIV-1 RNA level, CD4 cell count, and adherence. The estimates of resistance by adherence level were derived from previously published data [41]. Of note, the current level of adherence in the BC-CfE DTP is estimated to be 78.5% [41]. In brief, individuals with adherence levels in the stratum of 80% to <95% were more likely to have any drug resistance (median probability, 0.44; interquantile range [IQR], 0.32–0.56), followed by 95% to 100% (median probability, 0.40; IQR, 0.27–0.51), 40% to <80% (median probability, 0.23; IQR, 0.16–0.31), and 0% to <40% (median probability, 0.19; IQR, 0.11–0.28) (figure 3). These predicted probabilities of resistance were further stratified by plasma HIV-1 RNA level, CD4 cell count, and exposure category (MSM, IDUs, and MSM/IDUs) and incorporated into the BC-CfE model.

Figure 3.
Figure 3.

Estimated probability of emergence of HIV drug resistance, stratified by adherence level (0% to <40%, 40% to <80%, 80% to <95%, and 95% to 100%).

Impact of the expansion of HAART coverage. The BCCfE model was used to estimate the potential impact of the progressive expansion of HAART coverage for those who need it on the number of individuals testing newly positive for HIV. Current HAART coverage in the province of British Columbia is estimated to be 50% of the individuals in immediate medical need (namely, those with symptomatic disease or a CD4 cell count ⩽200 cells/mm3) [6]. Additional HAART coverage rates modeled include 75%, 90%, and 100%. Our model suggests that, at the present level of coverage of 50% and with a stable level of adherence (presently estimated to be 78.5% on the basis of provincial refill compliance data [41]), we can expect to see an increase in the annual number of individuals testing newly positive for HIV from 421 to 462 between 2006 and 2030 (figure 4A and figure 5 [4 panels on the left-hand side]). Our model further suggests that an increase in HAART coverage to 75%, 90%, and 100% of those in medical need (with no change in the criterion used for HAART eligibility or in adherence level) would lead to a decrease in the annual number of individuals testing newly positive for HIV in the province of British Columbia of 37%, 54%, and 62%, respectively. Furthermore, if adherence were to decrease to a level below 40% (in the 50% coverage scenario), the number of individuals testing newly positive for HIV would be predicted to increase by 10% per annum. In contrast, if adherence were to increase to levels above 80% (in the 50% coverage scenario), the number of individuals testing newly positive for HIV would be expected to decrease by 0.3% per annum. As shown in the figure, the magnitude of this change would increase substantially with enhanced coverage rates.

Effect of changes in HAART initiation on HIV transmission. We further explored the potential effect of a relatively modest change in the initiation of HAART from a CD4 cell count ⩽200 cells/mm3 to a CD4 cell count ⩽350 cells/mm3, which is still within the range of current guidelines [9]. The change is dependent on the degree of coverage and the level of adherence. For example, initiating HAART at 350 CD4 cells/mm3 with a stable coverage rate of 50% and an adherence rate of 78.5% would be expected to be associated with an increase in the annual number of individuals testing newly positive for HIV from 419 to 432 between 2006 and 2030 (figure 4B and figure 5 [4 panels on the right-hand side]). In contrast, if the same change in therapy initiation is associated with an increase in HAART coverage to 75%, 90%, and 100%, the model predicts that the annual number of individuals testing newly positive for HIV will decrease by 40%, 58%, and 67%, respectively. Also, if adherence in the population were to increase to levels between 80% and 100% in the 50% coverage scenario, the number of individuals testing newly positive for HIV would be predicted to decrease by 41% to 67%, depending directly on the higher coverage rate.

Figure 4.
Figure 4.

Results from the British Columbia Centre for Excellence in HIV/AIDS transmission model of the effectiveness of highly active antiretroviral therapy (HAART) uptake scenarios and coverage rates on the no. of individuals testing newly positive for HIV from 2006 through 2030. In panels A and B, we considered the effect of immediate (1-year) HAART uptake by varying coverage rates (from 50% to 75%, 90%, or 100%) and the impact of drug resistance measured indirectly through adherence (78.5% [current level]), given current guidelines (CD4 cell count ⩽200 cells/mm3) (A) and new guidelines (CD4 cell count ⩽350 cells/mm3) (B). In panel C, we considered the effect of HAART uptake given current guidelines (CD4 cell count ⩽200 cells/mm3) and current adherence level (78.5%) and varying coverage rates from 50% to 75% and uptake scenarios (1, 3, or 6 years).

Figure 5.
Figure 5.

Results from the British Columbia Centre for Excellence in HIV/AIDS transmission model on the effectiveness of highly active antiretroviral therapy (HAART) on the no. of individuals testing newly positive for HIV.

The effect of changing therapy initiation and coverage rates on the cumulative averted number of individuals testing newly positive for HIV is shown in table 1. After 25 years, at the current adherence level a total of 3108 (29% of total new positive test results), 4776 (44%), and 5701 (53%) new positive HIV test results could be averted if the coverage rate were to increase from 50% to 75%, 95%, and 100%, respectively.

Impact of varying rates of HAART uptake. We further explored the impact of varying rates of expanded HAART uptake using our model. Maintaining the current guidelines (CD4 cell count ⩽200 cells/mm3) and adherence level (78.5%), we considered 3 uptake scenarios: 1 year (immediate), 3 years (midterm), and 6 years (long term). Figure 3C (as well as figure 6) shows the model predictions for the number of individuals testing newly positive for HIV in each of the scenarios. These scenarios produce a consistent pattern whereby faster expansion of HAART uptake is associated with a faster decrease in new positive HIV test results. For example, it is anticipated that it would take an additional 2 to 4 and 13 to 14 years for the 3- and 6-year uptake scenarios, respectively, to catch up with the 1-year uptake scenario with regard to the averted number of individuals testing newly positive for HIV.

Figure 6.
Figure 6.

Results from the British Columbia Centre for Excellence in HIV/AIDS transmission model on the effectiveness of different highly active antiretroviral therapy (HAART) uptake scenarios and coverage rates on the no. of individuals testing newly positive for HIV and direct treatment costs from 2006 through 2030.

Table 1.
Table 1.

Total no. of individuals testing newly positive for HIV and averted no. of individuals testing newly positive for HIV from 2006 through 2030.

Table 2.
Table 2.

Total averted no. of individuals testing newly positive for HIV from 2006 to 2030 associated with 1-, 3-, and 6-year highly active antiretroviral therapy (HAART) uptake and total and lifetime direct treatment cost savings for 1-year HAART uptake from 2006 to 2030.

Table A1.
Table A1.

Parameters for the differential equation for the MSM risk group.

Table A2.
Table A2.

Parameters for the differential equation for the IDU risk group.

Table A3.
Table A3.

Parameters for the differential equation for the MSM/IDU risk group.

Figure
Figure 3.
Figure 3.

Estimated probability of emergence of HIV drug resistance, stratified by adherence level (0% to <40%, 40% to <80%, 80% to <95%, and 95% to 100%).

Figure 5.
Figure 5.

Results from the British Columbia Centre for Excellence in HIV/AIDS transmission model on the effectiveness of highly active antiretroviral therapy (HAART) on the no. of individuals testing newly positive for HIV. In all graphs, we considered the effect of immediate HAART uptake. Given current guidelines (CD4 cell count ⩽200 cells/mm3) and new guidelines (CD4 cell count ⩽350 cells/mm3), we considered different scenarios by varying coverage rates (from 50% to 75%, 90%, or 100%) and the impact of drug resistance measured indirectly through adherence (0% to <40%, 78.5% [current], 80% to <90%, and 95%–100%).

Figure 6.
Figure 6.

Results from the British Columbia Centre for Excellence in HIV/AIDS transmission model on the effectiveness of different highly active antiretroviral therapy (HAART) uptake scenarios and coverage rates on the no. of individuals testing newly positive for HIV and direct treatment costs from 2006 through 2030. Given current guidelines (CD4 cell count ⩽200 cells/mm3) and current adherence level (78.5%), we considered different scenarios by varying coverage rates (from 50% to 75%, 90%, or 100%) and uptake scenarios (1, 3, or 6 years).

Results for direct treatment costs associated with HAART expansion. Next we present the predictions of the model for the excess direct costs for treating all HIV-positive individuals in British Columbia who were infected because of delaying HAART expansion. In general, immediate HAART uptake is costly in the short-term; however, it is also associated with decreased number of individuals testing newly positive for HIV. Table 2 shows the total averted number of individuals testing newly positive for HIV and the total and lifetime treatment saving costs associated with treating more individuals. As noted in this table, the longer it takes to expandHAARTcoverage, the more new people testing positive for HIV are observed (from 34 to 414, depending on the scenario).Wealso observed that if HAART expansion were done immediately, it would generate total and per capita lifetime treatment cost savings starting at Can$95 million and Can$368 thousand.

Previous SectionNext Section

Discussion

In the present study, we investigated the potential impact of scaling up HAART, as a strategy to decrease HIV load at the population level, on the spread of HIV. We considered different scenarios by varying guidelines, coverage rates, HAART coverage, and adherence. Our model consistently predicted that enrolling at least 75% of individuals clinically eligible for treatment would be associated with a substantial decrease in new HIV infections. Furthermore, we showed that there is a complex interaction between guidelines for the initiation of therapy (i.e., the size of target population), HAART coverage, and adherence as well as the emergence of resistance with respect to the relationship between HAART coverage and HIV incidence. In summary, our model shows that the impact of HAART expansion on HIV transmission is optimized by rapid entry and a high rate of coverage of medically eligible HIV-infected individuals and by higher adherence.

Previous studies have produced conflicting evidence regarding the impact of HAART on the spread of HIV [25–28, 30, 42–45]. These studies include mathematical simulation models [25–28, 42–45] and ecological studies [30]. Although most studies support the notion that increased use of HAART could lead to a decrease in HIV incidence, the magnitude of the effect remains poorly characterized. Furthermore, some of these studies have suggested that any potential gain derived from the expansion of HAART coverage on HIV transmission could be easily overwhelmed by even modest increases in high-risk behavior or decreased adherence to HAART [28, 42–44]. Our results suggest that these concerns, although still valid, may have been overestimated, largely because previous models tended to underestimate the effect of HAART on HIV transmission. Of note, in British Columbia there has been a substantial growth in the incidence of sexually transmitted diseases, including syphilis (from 0.5 to 6.8/100,000 population), gonorrhea (from 13.0 to 28.0/100,000 population from 1996 to 2005), and chlamydia (from 123.4 to 213.3/100,000 population from 1996 to 2005) with no overall impact on new HIV infections per year [32, 33, 46].

Our model and those already published in the literature are different in several ways [25–28, 30, 42–50]. The BC-CfE transmission model has a random structure that is comprised of advanced statistical methods using GAMMs, which allow the transmission probabilities (which highly depend on plasma HIV-1 RNA level and other clinical and behavioral parameters) to change over time. The structure of this transmission model allowed us to take into account the complexity of the process underlying the HIV epidemic without requiring elaborate assumptions about the distribution of plasma HIV-1 RNA level, CD4 cell count, and other clinical parameters over time. The model also considered different assumptions for the level of therapy adherence in a population, given that scaling up treatment can potentially increase the likelihood of suboptimal adherence, virologic failure, mortality, and drug resistance. In addition, the dynamic aspect of the BC-CfE model allowed us to adjust our estimates for changes in the prevalence of individuals carrying drug-resistant HIV variants in the population. Differently from other models, the BC-CfE model incorporated changes over time in important behavior parameters that are influential in the HIV epidemic of major risk groups in British Columbia (MSM, IDUs, MSM/IDUs), via the calibration of parameters on the basis of the data from BC-CDC and the BC-CfE DTP. Finally, this model is flexible and can produce short- and long-term predictions of the effect of expanded HAART coverage on the HIV epidemic.

Several unique features of the model have been described here. First, our initial model represents a theoretical exercise aimed to relate the impact of lowering HIV load at the population level, via improvements in HAART coverage, on HIV transmission. In a second stage, the model was optimized by adjusting infectivity parameters to match empirical data collected prospectively in the past between 1993 and 2005. Then, we used semideterministic mathematical modeling to predict the effect of HAART expansion over time, which allowed us to work with a small number of simple differential equations. Our model was built on the principle that expanded access to HAART will be used as a supplement, rather than replacement, of current HIV control programs, such as education about and promotion of safer sex, condom use, microbicides, needle exchanges, and methadone maintenance, to cite a few. Therefore, a major feature of this mathematical model is that the risk behavior-specific transmission parameters are not detrimentally influenced by the expansion of HAART coverage over the next 25 years. We also should emphasize that, for expansion of HAART coverage to be a successful complementary strategy to limit the HIV epidemic, it is necessary that HIV screening programs also be expanded in order to discover patients without a diagnosis who are unaware of their HIV status. Finally, our model is flexible enough to easily incorporate and access different treatment coverage rates, different adherence scenarios, and assumptions regarding any of the transmission parameters. Of note, the number of individuals testing newly positive for HIV was used as a surrogate for HIV incidence in our study, because the latter is not routinely determined by the provincial health authorities [35, 36]. Future work should be aimed to carefully characterize the relationship between expansion of HAART coverage, HIV load, and HIV incidence prospectively.

In conclusion, we have successfully simulated and validated the dynamics of the HIV epidemic in British Columbia, particularly as it relates to the impact of expanding HAART coverage and the emergence of individuals testing newly positive for HIV. Our results indicate that a higher HAART coverage consistently leads to a decrease in the number of individuals testing newly positive for HIV. Although increased use of HAART is associated with an increase in the prevalence of individuals carrying drug-resistant virus, as a result of varying degrees of incomplete adherence and, consequently, high HIV-1 RNA plasma viral load, this appears to be relatively small and has a limited impact on the overall strategy. We therefore conclude that expansion of HAART coverage should lead to a substantial reduction of the growth of the HIV epidemic and related direct treatment costs. Our model supports a powerful and as-of-yet little appreciated additive preventive value of expanding HAART coverage.

Previous SectionNext Section

Acknowledgments

We thank Benita Yip, Nada Gataric, Kelly Hsu, Elizabeth Ferris, Myrna Reginaldo, Marnie Gidman, and Peter Vann for their administrative assistance; the staff at the British Columbia Centre for Disease Control, the Providence Health Care Laboratory, British Columbia Centre for Excellence in HIV/AIDS Laboratory, and St. Paul's Virology Lab for their assistance in obtaining empirical data for the model; and Drs. Rolando Barrios, Thomas Lampinen, and Mark Tyndall for their input in some of the parametric assumptions.

Previous SectionNext Section

Footnotes

  • Potential conflicts of interest: P.R.H., R.S.H., and J.S.G.M. have received honorariums, travel grants to attend conferences, and research grants from pharmaceutical companies working in the area of HIV/AIDS. All other authors report no potential conflicts.

  • Financial support: Michael Smith Foundation for Health Research (Senior Scholar Award to R.S.H.); Canadian Institutes of Health Research (via the Emerging Team Grant Program—HIV/AIDS and the Capacity Building through Enhanced Operating Grants in HIV/AIDS Program).

  • Received August 28, 2007.
  • Accepted December 18, 2007.
Previous SectionNext Section

Appendix A

Previous SectionNext Section

The BC-CfE Semideterministic Dynamic Transmission Model

The transmission parameters for each risk category (MSM, IDU, and MSM/IDU) are as follows.

The BC-CfE model includes 1 deterministic and 2 random components. The deterministic component estimates the number of individuals testing newly positive for HIV generated each time the model is updated, defined by the set of differential equations for each risk category. All differential equations were solved in Maple (version 8) and later incorporated into R (version 2.4.1).

We used the notation that population variables are capitalized Latin letters, dimensionless coefficients are lowercase Latin letters, and dimensional quantities (e.g., transmission rates) are Greek symbols. The index j represents each of the infectivity periods in the model described in figure 1.

The differential equation for the MSM risk group is

Formula

where, for j = 0,…,5,

Formula

and

Formula

The parameters for this equation are shown in table A1.

The differential equation for the IDU risk group is

Formula

where, for j = 0,…,5,

Formula

and

Formula

The parameters for this equation are shown in table A2.

The differential equation for the MSM/IDU risk group is

Formula

where, for j = 0,…,5,

Formula

and

Formula

and

Formula

The behavioral parameters that the MSM/IDUHIV− group shared with the MSMHIV− and IDUHIV− groups were assumed to be the same. The other parameters shown in table A3 were obtained using the same methodology described in tables A1 and A2.

The 2 random components consist of 2 statistical modeling techniques to predict CD4 cell count and plasma HIV-1 RNA level trajectories and to predict the probability of the emergence of HIV drug resistance. The statistical analyses and the BC-CfE transmission model were implemented in R (version 2.4.1) and SAS (version 9.1.3, service pack 3).

We used GAMMs to predict CD4 cell count and plasma HIV-1 RNA level trajectories [38, 63]. GAMMs are an extension of generalized linear models that allow for fitting smooth nonparametric curves describing the association between outcomes and predictor variables. The covariance structure incorporated autocorrelated error terms to account for correlation between longitudinal observations from the same individual. A predictive logistic regression model was also developed for identifying those patient characteristics that were most influential in the development of any drug resistance during antiretroviral treatment [64]. A backward stepwise technique was used in the selection of covariates [65]. The area under the receiver operating characteristic curve was used to measure the model's ability to discriminate between those in whom resistance developed versus those in whom it did not.

We considered the following predictors: age, sex, CD4 cell count, plasma HIV-1 RNA level (log10 transformed), presenting date, previous AIDS diagnosis, history of injection drug use, first HAART date, first HAART regimen, adherence, drug resistance, and death. Presenting date was defined by the date of the first (ever) plasma HIV-1 RNA level and CD4 cell count measurement before therapy initiation. Estimates of adherence to antiretroviral therapy were based on dispensed medications, also known as refill compliance. For this study, we limited our measure of adherence to the first year of therapy, estimated by dividing the number of months of medications dispensed by the number of months of follow-up. This measure of adherence has been found to be independently associated with HIV suppression and survival among HIV-infected persons enrolled in the DTP [34, 39, 66]. Adherence was categorized as 0% to <40%, 40% to <80%, 80% to <95%, and ⩾95%. Plasma HIV-1 RNA measurements were recoded to range from 500 to 100,000 copies/mL only, to account for changes in the range of quantification reported over time.

HIV drug-resistance genotyping data were based on assays attempted on all plasma samples collected during the first 30 months after the initiation of HAART for which the plasma HIV-1 RNA level was ⩾1000 copies/mL or by physician request. Samples were assigned to 1 of 4 resistance categories (i.e., NRTI, PI, NNRTI, or 3TC resistance) on the basis of a modification of the International AIDS Society–USA table, as detailed in D'Aquila et al. [40]. Lamivudine/emtracitabine resistance was analyzed as a separate category because of the very common appearance of this mutation and the lack of cross-resistance conferred to other NRTIs. Because genotyping does not yield consistently successful results on samples with low plasma HIV-1 RNA levels, samples with plasma HIV-1 RNA levels <1000 copies/mL were not systematically genotyped and were assumed to have no drug-resistance mutations. In our models, the cumulative emergence of drug resistance in any of the 4 resistance categories described previously [39] was used as a dichotomous variable (yes vs. no).

The GAMMs to predict CD4 cell count and plasma HIV-1 RNA level trajectories included history of injection drug use, age, and sex as linear terms. TheGAMMpredicting CD4 cell count additionally included time since presenting date, time since therapy initiation, adherence, resistance, and previous CD4 cell count measurements as smooth terms. TheGAMMpredicting plasma HIV-1 RNA level also included these smooth terms as well as an additional smooth term for previous plasma HIV-1RNAmeasurements. For the logistic regression model, we used baseline values for all predictor variables. Figure 3 shows the estimated model-based probabilities of developing HIV drug resistance by adherence levels. This figure demonstrates a nonlinear relationship between adherence and drug resistance, with those individuals with adherence levels in the stratum 80% to <95% being more likely to harbor drug-resistant virus.

Previous SectionNext Section

References

  1. 51.
    1. Raboud JM,
    2. Boily MC,
    3. Rajeswaran J,
    4. O'sShaughnessy MV,
    5. Schechter MT
    . The impact of needle-exchange programs on the spread of HIV among injection drug users: a simulation study. J Urban Health 2003;80:302-20.
    MedlineWeb of Science
  2. 52.
    1. Archibald CP,
    2. Jayaraman GC,
    3. Major C,
    4. Patrick DM,
    5. Houston SM,
    6. Sutherland D
    . Estimating the size of hard-to-reach populations: a novel method using HIV testing data compared to other methods. AIDS 2001;15(Suppl 3):S41-8.
  3. 53.
    1. Ekstrand ML,
    2. Stall RD,
    3. Paul JP,
    4. Osmond DH,
    5. Coates TJ
    . Gay men report high rates of unprotected anal sex with partners of unknown or discordant HIV status. AIDS 1999;13:1525-33.
    CrossRefMedlineWeb of Science
  4. 54.
    1. Vittinghoff E,
    2. Douglas J,
    3. Judson F,
    4. McKirnan D,
    5. MacQueen K,
    6. Buchbinder SP
    . Per-contact risk of human immunodeficiency virus transmission between male sexual partners. Am J Epidemiol 1999;150:306-11.
    Abstract/FREE Full Text
  5. 55.
    1. San Francisco Department of Public Health
    . HIVStatistics and Epidemiology Section. The HIV epidemiology annual report 2002. [Accessed 31 August 2006]. Available at: http://www.dph.sf.ca.us.htm.
  6. 56.
    1. San Francisco Department of Public Health AIDS Office
    . San Francisco HIV behavioral risk factor telephone survey. 1999 April [Accessed 31 August 2006]. Available at: http://www.sfdph.org_brfs.pdf.
  7. 57.
    1. Xiridou M,
    2. Geskus R,
    3. de Wit J,
    4. Coutinho R,
    5. Kretzschmar M
    . Primary HIV infection as source of HIV transmission within steady and casual partnerships among homosexual men. AIDS 2004;18:1311-20.
    CrossRefMedlineWeb of Science
  8. 58.
    1. Strathdee SA,
    2. Martindale SL,
    3. Cornelisse PG,
    4. et al
    . HIV infection and risk behaviours among young gay and bisexual men in Vancouver. CMAJ 2000;162:21-5.
    Abstract/FREE Full Text
  9. 59.
    1. Baeten JM,
    2. Overbaugh J
    . Measuring the infectiousness of persons with HIV-1: opportunities for preventing sexual HIV-1 transmission. Curr HIV Res 2003;1:69-86.
    CrossRefMedlineWeb of Science
  10. 60.
    1. Kral AH,
    2. Lorvick J,
    3. Ciccarone D,
    4. et al
    . HIV prevalence and risk behaviors among men who have sex with men and inject drugs in San Francisco. J Urban Health 2005;82(1 Suppl 1):i43-50.
  11. 61.
    1. Weinhardt LS,
    2. Kelly JA,
    3. Brondino MJ,
    4. et al
    . HIV transmission risk behavior among men and women living with HIV in 4 cities in the United States. National Institute of Mental Health Healthy Living Project Team. J Acquir Immune Defic Syndr 2004;36:1057-66.
  12. 62.
    1. Long EF,
    2. Brandeau ML,
    3. Galvin CM,
    4. et al
    . Effectiveness and cost-effectiveness of strategies to expand antiretroviral therapy in St. Petersburg, Russia. AIDS 2006;20:2207-15.
    CrossRefMedlineWeb of Science
  13. 63.↵
    1. Hastie TJ ,
    2. Tibshirani RJ
    . New York: Chapman & Hall; 1990. Generalized additive models (monographs on statistics and applied probability).
  14. 64.↵
    1. Hosmer D Jr,
    2. Lemeshow S
    . 2nd ed. New York: John Wiley & Sons; 2000. Applied logistic regression.
  15. 65.↵
    1. Lima VD,
    2. Kopec JA
    . Quantifying the effect of health status on health care utilization using a preference-based health measure. Soc Sci Med 2005;60:515-24.
    CrossRefMedlineWeb of Science
  16. 66.↵
    1. Wood E,
    2. Montaner JS,
    3. Yip B,
    4. et al
    . Adherence and plasma HIV RNA responses to highly active antiretroviral therapy among HIV-1 infected injection drug users. CMAJ 2003;169:656-61.
    Abstract/FREE Full Text
Previous Section
 

References

  1. 1.↵
    1. Piot P
    . Why AIDS is exceptional. AIDS Matters 2005 [Accessed 10 March 2007]. Available at: http://www.aidsmatters.org/archives/100-Why-AIDS-is-Exceptional.html.
  2. 2.↵
    Joint United Nations Programme on HIV/AIDS. 2006 report on the global AIDS epidemic. 2006 [Accessed 10 March 2007]. Available at: http://www.unaids.org/en/HIV_data/2006GlobalReport/default.asp.
  3. 3.↵
    1. Gayle HD
    . Expanding access to HIV prevention. AIDS Res Ther 2006;3:2.
    CrossRefMedline
  4. 4.↵
    Joint United Nations Programme on HIV. Financing the expanded response to AIDS: HIV vaccine and microbicide research and development. 2005 [Accessed 10 March 2007]. Available at: http:\www.hivresourcetracking.org.
  5. 5.↵
    1. Chen LF,
    2. Hoy J,
    3. Lewin SR
    . Ten years of highly active antiretroviral therapy for HIV infection. Med J Aust 2007;186:146-51.
    MedlineWeb of Science
  6. 6.↵
    1. Montaner JS,
    2. Hogg R,
    3. Wood E,
    4. et al
    . The case for expanding access to highly active antiretroviral therapy to curb the growth of the HIV epidemic. Lancet 2006;368:531-6.
    CrossRefMedlineWeb of Science
  7. 7.
    Joint United Nations Programme on HIV. Resource needs for an expanded response to AIDS in low- and middle-income countries. 2006 [Accessed 10 March 2007]. Available at: http:\data.unaids.orgpub06\parresourceneedsreport_24jun05_en.pdf.
  8. 8.↵
    1. Kerr T,
    2. Kaplan K,
    3. Suwannawong P,
    4. Jurgens R,
    5. Wood E
    . The Global Fund to Fight AIDS, Tuberculosis and Malaria: funding for unpopular public-health programmes. Lancet 2004;364:11-12.
    CrossRefMedlineWeb of Science
  9. 9.↵
    1. Hammer SM,
    2. Saag MS,
    3. Schechter M,
    4. et al
    . Treatment for adult HIV infection: 2006 recommendations of the International AIDS Society- USA panel. Top HIV Med 2006;14:827-43.
    Medline
  10. 10.↵
    1. Department of Health and Human Services
    . Guidelines for the use of antiretroviral agents in HIV-1-infected adults and adolescents. 2006 [Accessed 29 May 2007]. Available at: http://AIDSinfo.nih.gov.
  11. 11.↵
    1. World Health Organization
    . Progress on global access to HIV antiretroviral therapy: an update on 3 by 5. 2006 [Accessed 5 December 2006]. Available at: http://www.who.int/hiv/3by5evaluation/en/.
  12. 12.↵
    1. World Health Organization
    . WHO case definitions of HIV for surveillance and revised clinical staging and immunological classification of HIV-related disease in adults and children. 2006 [Accessed 5 December 2006.]. Available at: http://www.who.int/hiv/pub/guidelines/en/.
  13. 13.↵
    1. World Health Organization
    . Antiretroviral therapy for HIV infection in adults and adolescents: towards universal access. 2006 [Accessed 5 December 2006.]. Available at: http://www.who.int/hiv/pub/guidelines/en/.
  14. 14.↵
    1. Hogg RS,
    2. Rhone SA,
    3. Yip B,
    4. et al
    . Antiviral effect of double and triple drug combinations amongst HIV- infected adults: lessons from the implementation of viral load-driven antiretroviral therapy. AIDS 1998;12:279-84.
    CrossRefMedlineWeb of Science
  15. 15.↵
    1. Mocroft A,
    2. Vella S,
    3. Benfield TL,
    4. et al.,
    5. EuroSIDA Study Group
    . Changing patterns of mortality across Europe in patients infected with HIV-1. Lancet 1998;352:1725-30.
    CrossRefMedlineWeb of Science
  16. 16.↵
    1. Cu-Uvin S,
    2. Caliendo AM,
    3. Reinert S,
    4. et al
    . Effect of highly active antiretroviral therapy on cervicovaginal HIV-1 RNA. AIDS 2000;14:415-21.
    CrossRefMedlineWeb of Science
  17. 17.↵
    1. Vernazza PL,
    2. Gilliam BL,
    3. Flepp M,
    4. et al
    . Effect of antiviral treatment on the shedding of HIV-1 in semen. AIDS 1997;11:1249-54.
    CrossRefMedlineWeb of Science
  18. 18.↵
    1. Guay LA,
    2. Musoke P,
    3. Fleming T,
    4. et al
    . Intrapartum and neonatal singledose nevirapine compared with zidovudine for prevention of motherto- child transmission of HIV-1 in Kampala, Uganda: HIVNET 012 randomised trial. Lancet 1999;354:795-802.
    MedlineWeb of Science
  19. 19.
    1. De Cock KM,
    2. Fowler MG,
    3. Mercier E,
    4. et al
    . Prevention of mother-tochild HIV transmission in resource-poor countries: translating research into policy and practice. JAMA 2000;283:1175-82.
    Abstract/FREE Full Text
  20. 20.↵
    1. Karon JM,
    2. Fleming PL,
    3. Steketee RW,
    4. De Cock KM
    . HIV in the United States at the turn of the century: an epidemic in transition. Am J Public Health 2001;91:1060-8.
    Abstract
  21. 21.↵
    1. Tovanabutra S,
    2. Robison V,
    3. Wongtrakul J,
    4. et al
    . Male viral load and heterosexual transmission of HIV-1 subtype E in northern Thailand. J Acquir Immune Defic Syndr 2002;29:275-83.
    MedlineWeb of Science
  22. 22.
    1. Musicco M,
    2. Lazzarin A,
    3. Nicolosi A,
    4. et al
    . Italian study group on HIV heterosexual transmission. Arch Intern Med 1994;154:1971-6.
    Abstract/FREE Full Text
  23. 23.↵
    1. Castilla J,
    2. Del Romero J,
    3. Hernando V,
    4. et al
    . Effectiveness of highly active antiretroviral therapy in reducing heterosexual transmission of HIV. J Acquir Immune Defic Syndr 2005;40:96-101.
    CrossRefMedlineWeb of Science
  24. 24.↵
    1. Hosseinipour M,
    2. Cohen MS,
    3. Vernazza PL,
    4. Kashuba AD
    . Can antiretroviral therapy be used to prevent sexual transmission of human immunodeficiency virus type 1? Clin Infect Dis 2002;34:1391-5.
    Abstract/FREE Full Text
  25. 25.↵
    1. Blower SM,
    2. Gershengorn HB,
    3. Grant RM
    . A tale of two futures: HIV and antiretroviral therapy in San Francisco. Science 2000;287:650-4.
    Abstract/FREE Full Text
  26. 26.
    1. Law MG,
    2. Prestage G,
    3. Grulich A,
    4. Van de Ven P,
    5. Kippax S
    . Modelling the effect of combination antiretroviral treatments on HIV incidence. AIDS 2001;15:1287-94.
    CrossRefMedlineWeb of Science
  27. 27.
    1. Velasco-Hernandez JX,
    2. Gershengorn HB,
    3. Blower SM
    . Could widespread use of combination antiretroviral therapy eradicate HIV epidemics? Lancet Infect Dis 2002;2:487-93.
    CrossRefMedlineWeb of Science
  28. 28.↵
    1. Abbas UL,
    2. Anderson RM,
    3. Mellors JW
    . Potential impact of antiretroviral therapy on HIV-1 transmission and AIDS mortality in resource-limited settings. J Acquir Immune Defic Syndr 2006;41:632-41.
    CrossRefMedlineWeb of Science
  29. 29.
    1. Bunnell R,
    2. Ekwaru JP,
    3. Solberg P,
    4. et al
    . Changes in sexual behavior and risk of HIV transmission after antiretroviral therapy and prevention interventions in rural Uganda. AIDS 2006;20:85-92.
    MedlineWeb of Science
  30. 30.↵
    1. Fang CT,
    2. Hsu HM,
    3. Twu SJ,
    4. et al
    . Decreased HIV transmission after a policy of providing free access to highly active antiretroviral therapy in Taiwan. J Infect Dis 2004;190:879-85.
    Abstract/FREE Full Text
  31. 31.↵
    1. Richman DD
    . Human immunodeficiency virus. London: International Medical Press; 2006.
  32. 32.↵
    1. Quinn TC,
    2. Wawer MJ,
    3. Sewankambo N,
    4. et al.,
    5. Rakai Project Study Group
    . Viral load and heterosexual transmission of human immunodeficiency virus type 1. N Engl J Med 2000;342:921-9.
    CrossRefMedlineWeb of Science
  33. 33.↵
    1. Mellors JW,
    2. Munoz A,
    3. Giorgi JV,
    4. et al
    . Plasma viral load and CD4_ lymphocytes as prognostic markers of HIV-1 infection. Ann Intern Med 1997;126:946-54.
    Abstract/FREE Full Text
  34. 34.↵
    1. Lima VD,
    2. Hogg RS,
    3. Harrigan PR,
    4. et al
    . Continued improvement in survival among HIV infected individuals following initiation of triple and boosted combination antiretroviral regimens. AIDS 2007;21:685-92.
    MedlineWeb of Science
  35. 35.↵
    1. British Columbia Centre of Disease Control
    . HIV statistics 2005. 2005 [Accessed 15 January 2007]. Available at: http://www.bccdc.org/content.php?item=5.
  36. 36.↵
    1. British Columbia Centre of Disease Control
    . HIV statistics 2004. 2004 [Accessed 15 January 2007]. Available at: http://www.bccdc.org/content.php?item=5.
  37. 37.↵
    1. Drummond MF,
    2. Torrance GW
    . Oxford: Oxford Medical Publications; 1998. Methods for economic evaluation of health care programmes.
  38. 38.↵
    1. Wood SN
    . New York: Chapman & Hall/CRC; 2006. eneralized additive models: an introduction with R.
  39. 39.↵
    1. Hogg RS,
    2. Bangsberg DR,
    3. Lima VD,
    4. et al
    . Emergence of drug resistance is associated with an increased risk of death among patients first starting HAART. PLoS Med 2006;3:e356.
    CrossRefMedline
  40. 40.↵
    1. D'sAquila RT,
    2. Schapiro JM,
    3. Brun-Vezinet F,
    4. et al
    . Drug resistance mutations in HIV-1. Top HIV Med 2002;10:21-5.
    Medline
  41. 41.↵
    1. Lima VD,
    2. Johnson K,
    3. Hogg RS,
    4. et al
    . Program and abstracts of the 14th Conference on Retroviruses and Opportunistic Infections (Los Angeles). Alexandria, VA: Foundation for Retrovirology and Human Health; 2007. Modeling the impact of increasing utilization of HAART on HIV incidence in settings with multiple sources of infection [poster 978].
  42. 42.↵
    1. Salomon JA,
    2. Hogan DR,
    3. Stover J,
    4. et al
    . Integrating HIV prevention and treatment: from slogans to impact. PLoS Med 2005;2:e16.
    CrossRefMedline
  43. 43.
    1. Boily MC,
    2. Bastos FI,
    3. Desai K,
    4. Masse B
    . Changes in the transmission dynamics of the HIV epidemic after the wide-scale use of antiretroviral therapy could explain increases in sexually transmitted infections: results from mathematical models. Sex Transm Dis 2004;31:100-13.
    MedlineWeb of Science
  44. 44.↵
    1. Gray RH,
    2. Li X,
    3. Wawer MJ,
    4. et al.,
    5. Rakai Project Group
    . Stochastic simulation of the impact of antiretroviral therapy and HIV vaccines on HIV transmission, Rakai, Uganda. AIDS 2003;17:1941-51.
    CrossRefMedlineWeb of Science
  45. 45.↵
    1. Blower SM,
    2. Aschenbach AN,
    3. Gershengorn HB,
    4. Kahn JO
    . Predicting the unpredictable: transmission of drug-resistant HIV. Nat Med 2001;7:1016-20.
    CrossRefMedlineWeb of Science
  46. 46.↵
    1. British Columbia Centre of Disease Control
    . 2005 STD annual report. 2005 [Accessed 15 January 2007]. Available at: http://www.bccdc.org/content.php?item=5.
  47. 47.
    1. Baggaley RF,
    2. Garnett GP,
    3. Ferguson NM
    . Modelling the impact of antiretroviral use in resource-poor settings. PLoS Med 2006;3:e124.
    CrossRefMedline
  48. 48.
    1. Baggaley RF,
    2. Ferguson NM,
    3. Garnett GP
    . The epidemiological impact of antiretroviral use predicted by mathematical models: a review. Emerg Themes Epidemiol 2005;2:9.
    CrossRefMedline
  49. 49.
    1. Cormick AW,
    2. Walensky RP,
    3. Lipsitch M,
    4. et al
    . The effect of antiretroviral therapy on secondary transmission of HIV among men who have sex with men. Clin Infect Dis 2007;44:1115-22.
    Abstract/FREE Full Text
  50. 50.↵
    1. Griffiths JD,
    2. Lawson ZF,
    3. Williams JE
    . Modelling treatment effects in the HIVepidemic. J Oper Res Soc 2006;57:1413-24.
    CrossRef
| Table of Contents

This Article

  1. J Infect Dis. (2008) 198 (1): 59-67. doi: 10.1086/588673
  1. AbstractFree
  2. » Full Text (HTML)Free
  3. Full Text (PDF)Free

Classifications

Share

  1. Email this article

Navigate This Article

Current Issue

  1. March 1, 2012 205 (5)
  1. Journal of Infectious Diseases
  1. Alert me to new issues

Most