de Vos, Anneke S.a; Prins, Mariab,c; Coutinho, Roel A.a,d; van der Helm, Jannie J.b; Kretzschmar, Mirjam E.E.a,d
Much attention has recently been given to the concept of treatment as prevention (TasP) [1–3]. As combination antiretroviral therapy (cART) lowers viral load, it not only completely changes the prognosis of those infected with HIV, but it also lowers their likelihood of infecting others . From modeling studies describing the sexual spread of HIV, some have even postulated that sufficient treatment uptake would lead to elimination of the virus . Others however have warned against overly optimistic expectations .
So far, the possible benefit in lowered HIV incidence of TasP has been less rigorously studied for treatment of injecting drug users (IDU), although IDU represent about 10% of those infected with HIV worldwide . Most studies that model the spread of HIV among IDU have included treatment only of those with low CD4+ cell counts, reflecting the current treatment guidelines which are based on health benefits of those treated rather than population benefits . Few studies so far have examined a test and treat strategy among IDU; providing immediate treatment for those testing positive independent of CD4+ cell count .
We previously introduced a model of the IDU population in Amsterdam to investigate the possible reasons for the decline in HIV and HCV prevalence and incidence in this city . We found that taking into account the strongly changed demography within the Amsterdam IDU population over the years was of great importance for explaining the observed patterns of disease spread. In particular, strong heterogeneity in risk behavior together with decreasing population size could explain much of the decline in HIV prevalence.
Demographic changes will also strongly impact on the possibility for lowering HIV incidence by uptake of cART. Compared with other populations for which treatment as prevention has been investigated, IDU populations may be less stable; in certain regions of China, Russia and India relatively recently established populations of IDU are found , on the contrary, in many parts of Europe injecting drug use is currently in decline . Compared with sexual spread, spread of HIV by use of contaminated injecting equipment is very efficient, so that very high levels of HIV prevalence can occur within IDU populations .
cART has been prescribed to IDU in Amsterdam since it first became available from around 1996. Here, we investigate the impact that treatment has likely had within this city using an adapted version of our earlier model. Extrapolating from Amsterdam, we also investigate how a strategy to test and treat IDU would impact on HIV incidence in populations with different demographic developments. For a stable, increasing or decreasing IDU population, we estimate the total incidence over the first 30 years of an HIV epidemic, as influenced by the rate of starting treatment for HIV. We also consider the influence of the elapsed time between the first HIV cases entering the population and the implementation of the test and treat regimen.
We adapted an individual based model, which was developed to describe demographic changes of both HIV and hepatitis C infection dynamics within the IDU population of Amsterdam . Most demographic parameters for this model were estimated from the Amsterdam Cohort Studies (ACS) among IDU. Recruitment for this study has been ongoing since 1985, and took place at methadone outposts, the weekly STD-clinic for drug-using prostitutes and by word of mouth. In principle every 4–6 months ACS participants were interviewed and provided blood-samples, from which HIV and (retrospectively) HCV-status were determined . Total IDU population size in Amsterdam was based on back calculations from IDU participating in methadone maintenance treatment .
Individuals entered the model at the start of their injecting-career. Subsequently, they could stop injecting but also relapse, acquire HIV and move out of Amsterdam or die, depending on their age and injection status, which were updated each month. To allow comparison between our model and the ACS data, we modeled ACS recruitment, thereby creating a model ACS-cohort within the modeled population. Modeled individuals were either high-risk, both borrowing from and lending syringes to many others, or low-risk throughout their injecting-career. Reliable information on risk behavior is very difficult to obtain, not just for IDU in Amsterdam but in general. Therefore, we based the absolute sharing rates, relative size of the behavioral sub-groups, and also an increased likelihood to share with someone of similar risk, on the model fit to ACS data of HIV and HCV prevalence and incidence  (as briefly summarized in Appendix Figure 1, http://links.lww.com/QAD/A466). We assumed no change in risk-behavior associated with either HIV-diagnosis or start of treatment .
The HIV acute stage was assumed to last 2 months , during which infectiousness was 10-fold that of chronic infection . Infectiousness was lowered 90% in either stage by use of cART, based on data from sexual transmission studies . HIV-induced mortality in our model depended on age and time since HIV infection, and is based on IDU-specific data from the CASCADE Collaboration . As most of this data was from hospital cohorts, we assumed that from the broad introduction of cART in 1997 onwards most participants in need of treatment received it. We therefore based mortality of those on cART in our model on this data from 1997. Those not on cART experienced mortality based on the CASCADE data from before 1997.
For our baseline scenario, cART uptake was fit to data on antiretroviral use by the HIV-infected IDU in the ACS. We assumed that those starting treatment remained on treatment. Until recently cART was usually initiated only with the onset of clinical symptoms, or when an individual's CD4+ cell count dropped below 350 cells/μl , which approximately corresponds to HIV infection more than 6 year duration . The fit between model and actual cART use was achieved by assuming a 1/96 (≈0.01) probability per month to start cART for all IDU with more than 6 year HIV infection from 1996 onwards. That is, including those never treated, the average individual started cART at 14 years from infection (see Fig. 1).
We considered four counter-factual scenarios for Amsterdam. We modeled cART never having been introduced and cART introduction from the start of the HIV epidemic among IDU in 1980. We also investigated an increased cART uptake rate (from 1996 treatment of all IDU >6 year HIV-infected) and treatment independent of CD4+ cell count (from 1996 a 0.01 monthly starting-rate for all HIV-infected IDU) (Table 1). Note that since cART use was updated at the end of each month, time from infection to lowered infectiousness was at least 1 month for individuals, also in this last scenario.
We then adapted the model demography to investigate benefits of a test and treat strategy for IDU in a more general setting. We first considered a stable IDU population: four new IDU entered each month, resulting in approximately 2000 IDU at equilibrium. We considered an increasing population (with initially 200 IDU) and a decreasing population (with no new IDU entering). We also considered a stable population in which all IDU shared syringes equally. An epidemic was initiated by entering 20 HIV-infected IDU. We determined the cumulative number of infections within the first 30 years of the epidemic, as influenced by the monthly cART starting probability and the time elapsed between the first HIV cases and test and treat implementation.
We additionally investigated the impact of changing behavior of (equivalent to isolating) HIV-infected IDU; in the above described stable population setting, we combined testing and treatment of HIV-infected individuals with assuming that the treated individuals fully ceased all syringe sharing with other IDU.
The model was implemented in Mathematica version 7. To account for the stochastic nature of events in our model, each scenario was repeated 50 times, and results were averaged. We considered a decline from the initial 20 prevalent cases at the end of the 30 years in more than 50% of model runs as indicating the eventual elimination of HIV.
Uptake of cART among participants in the ACS was well described by assuming a 0.01 probability of starting cART per month for all IDU HIV-infected more than 6 years (approximating the criterion of a CD4+ cell count < 350 cells/μl) (Fig. 1a). HIV prevalence within the ACS was fit by assuming two risk subgroups of IDU: as those sharing more syringes were infected and died of HIV first, population risk decreased, resulting in a decline in HIV prevalence over time (Fig. 1b). The risk behavior parameters were additionally informed by the HIV and HCV incidence and HCV prevalence data (see Appendix Figure 1, http://links.lww.com/QAD/A466).
In our baseline scenario, with all model parameters as best informed by the data, we estimated that around 2400 of all 8600 ever-IDU in Amsterdam became infected, and 1400 IDU died of HIV from the start of the epidemic up to 2010 (Table 1). To gain insight in the role of cART, we next examined several counter-factual scenarios for Amsterdam. We estimated that had cART not been available, mortality would have been 16% higher, which would have led to a decreased HIV prevalence compared with the baseline prevalence (Fig. 1). Conversely, had cART been available and uptake similar from the start of the HIV epidemic, we estimated that HIV-related mortality would have been 13% less. Although cART lowers infectivity of individuals, the increased survival it enables (resulting in higher HIV-prevalence) has a compensatory effect on incidence: we estimated a 2% decrease in overall incidence only if cART had been available from 1980 compared with its availability from 1996 at baseline.
Similarly, if from 1996 all HIV-infected IDU infected more than 6 years could have been found and treated immediately, 24% of HIV deaths could have been avoided, but prevalence would have been higher, and incidence would have dropped by only 6%. In this scenario, in total approximately three times as much cART was used compared to within the baseline scenario. When keeping the probability for finding a HIV-infected IDU at 0.01 each month, but treating independent of CD4+ cell count, only one and a half times the baseline cART would have been used. This would have lowered incidence by 4%, but mortality by only 6%.
When treating from low CD4+ cell count lowering of infectiousness by cART is largely offset by the lengthening of infectious life-time, hence prevention is clearly best served by early diagnosis and immediate treatment thereafter. We therefore, for various demographic settings, explored the effects of a test and treat strategy. We find that by such a strategy many new infections can be prevented; in a stable population of IDU, about half of all new infections within a 30-year period were avoided when IDU started treatment on average 1 year after becoming infected (corresponding to testing IDU on average once every 2 years) (Fig. 2a). HIV could even be eventually eliminated if IDU started treatment on average within 1.6 months after infection (corresponding to testing IDU once per 3.2 months, results not shown).
As HIV spreads quickly within the high-risk subpopulation, HIV incidence peaked very rapidly. Therefore, the test and treat practice had by far the largest impact when implemented soon after introduction of HIV into the population; initiating the test and treat strategy 2 years after the first HIV-cases are diagnosed prevented only about two-thirds, and after 5 years about half of the cumulative number of infections over the first 30 years of the epidemic, even when all IDU started treatment 1 month after becoming infected. When the population of IDU was increasing over time, later introduction of the test and treat intervention still prevented a larger part of new infections (Fig. 2b). Conversely, with the IDU population declining (as it was for Amsterdam) later intervention had even less impact (Fig. 2c).
These results depend strongly on the heterogeneity of risk within the population. The risk distribution assuming that 31% of IDU taking 10 times the risk of remaining IDU, and that 70% of contacts occur preferentially with IDU of similar risk-level, was based on fitting our model to ACS data of HIV and HCV prevalence and incidence . Assuming a homogenous population with respect to risk-behavior, the scope for a timely test and treat intervention was much enhanced, since initial HIV spread was much slower in such a population (Fig. 2d).
In the stable heterogeneous population, without intervention 48% of all new cases were due to infection from acutely (< 2 month) infected IDU (68% of the new infections within the first 5 years, 28% of infections from 5 to 30 years after the start of the epidemic) (results not shown). In the homogenous and the increasing IDU population 41%, but in the decreasing IDU population an even larger fraction, 57% of infections, was due to acutely infected IDU. This explains the rapid uptake of cART needed to eliminate HIV.
In a risk heterogeneous population, to increase efficiency, intervention might be targeted by risk level (Fig. 3). Incidence was lowered almost as much when limiting testing for HIV to only the 31% high risk IDU, compared to testing all IDU. Note however that total cART used in this scenario is also nearly equal to that when testing all IDU, since high risk IDU are much more likely to be infected than low risk IDU.
Here we assume that HIV-diagnosis alone does not change risk-behavior. As an alternative intervention, HIV-infected tested IDU might receive counseling or they might be isolated in order to stop them from lending syringes to other IDU. Under most testing scenarios, we found such intervention to be much less effective than cART usage in lowering HIV incidence (Fig. 4). When one specific IDU stops lending out syringes, those that would have borrowed from this individual will borrow syringes from others instead. Many of the HIV-infected IDU are likely not yet diagnosed and treated, especially those acutely infected who are most infectious. Isolating HIV-infected IDU compared to test and treat would then replace borrowing from HIV-infected IDU using cART with borrowing from more infectious IDU, thereby increasing population incidence. Only at low HIV prevalence, for example in a scenario with very frequent and timely testing, isolating HIV-infected IDU could be more beneficial compared with cART use by the tested IDU.
We adapted a model based on the Amsterdam IDU population in order to study the effects of cART on HIV incidence. From examining counter-factual scenarios, we concluded that use of cART only very slightly decreased incidence in Amsterdam. As IDU started treatment only at low CD4+ cell counts, the effect of lowered infectiousness on transmission was mostly offset by increased survival, which increases the time they could infect others. Clearly for cART to lead to large population benefits, treatment should be started independently of CD4+ cell count.
We therefore considered the possible benefits of a test and immediate treat strategy in a more generalized population of IDU. We found that such an intervention could prevent many new cases. For achieving a substantial reduction of incidence however, it is important that the test and treat strategy be implemented soon, within a few years of the first HIV cases entering an IDU population. Especially in IDU populations that are declining in size, most of the incidence will otherwise have already occurred at the moment of initiating the intervention.
There is ongoing debate as to whether the high infectiousness of individuals early after infection severely compromises the potential for treatment as prevention . In our model, almost half of all new infections were caused by individuals within the acute stage, despite our relatively conservative choice of a 10-fold increased infectiousness . This large proportion in our model is explained by the high HIV-transmissibility by injecting but a relatively short infectious-life-time of IDU (due to their relatively high death rate and cessation of injecting), and our focus on the early stage of the HIV epidemic. In our model elimination of HIV could occur, but only if the average time from infection to starting cART was within the 2-month period of acute infection.
Especially where harm reduction programs are already established that allow for regular contact with IDU, frequent testing for HIV infection may be achievable. However, only the actual implementation of a test and treat strategy will determine the successful uptake rate achievable within the IDU setting. For our model, we made the optimistic assumption that all IDU were willing to start treatment early, and that they would also be reasonably adherent, despite limited personal health benefits. In reality, this will pose a serious challenge [23–25].
With low adherence, viral load and thereby infectiousness remain higher , which could also boost the development of cART-resistant HIV-strains . It would therefore be prudent to monitor adherence, preferably through viral load measurements . Somewhat lower rates of cART adherence [28,29], correlated with lowered rates of viral suppression , have been reported for active IDU compared with ex-IDU or non-IDU, although a meta-analysis did not show increased risk of resistance to treatment within this group . Adherence by IDU may be enhanced by for example peer counseling, mobile-phone alarms, or directly observed treatment .
Direct reliable information on risk behavior is rare. Here we based risk parameters on the model fit to HIV as well as HCV incidence and prevalence in Amsterdam , but we may have over- or underestimated the extent of the heterogeneity in this population, and different populations will have different risk distributions. In a more homogenous population initial spread of HIV will be slower, so that even at lower uptake rates later implementation of a test and treat strategy could still have great effects. A completely risk-homogenous population of IDU seems unlikely however, and this scenario should therefore be interpreted also as a warning against overly optimistic expectations that result from modeling studies that ignore population risk heterogeneity.
The test and treat strategy could be made more efficient if directed mostly at those with highest risk behavior. On the contrary, high risk IDU might be the ones most difficult to reach and least likely to adhere to treatment, in which case our analysis represents an overestimate of the benefits of cART provision. For a realistic cost–benefit analysis of different implementation strategies, information on uptake stratified by risk behavior would be required.
Where there is a low coverage of harm reduction programs, instead of or in addition to providing them with cART, HIV-infected IDU could be targeted for counseling, or they might even be isolated forcefully by imprisonment, to stop them from lending out their used syringes to other IDU. We found that even if HIV-infected IDU would stop lending out syringes completely it would have less impact than treating them with cART at all but very low HIV prevalence.
In conclusion, we have shown that provision of cART to all HIV-infected IDU could lead to great reductions in HIV incidence. Previous modeling studies on the use of TasP have pointed to the optimistic prospect for complete elimination of the sexual spread of HIV [5,33]. We conclude that among IDU however, only unrealistically high uptake rates from very early after infection would allow cART provision by itself to eliminate HIV. In accordance with previous modeling studies on IDU, we therefore recommend an integrated approach of several different harm reduction strategies [8,9,34].
In countries confronted with a sudden rise of injecting drug users, rapid implementation of harm reduction services combined with regular testing for HIV and prompt treatment may avoid the spread of HIV. Behavioral intervention should be directed at those not yet infected; by injecting with fewer used syringes, IDU lower their risk of infection of HIV, but also their risk of other blood-borne infections such as HCV. Possibly, efficiency of these interventions and treatment as prevention could be enhanced by targeting specific risk-behavior sub-groups within the IDU community .
This study was conducted at the Utrecht Centre for Infection Dynamics (UCID). We thank the CASCADE collaboration in EuroCoord for use of data on HIV-induced mortality among IDU. Also we would like to thank the Amsterdam Cohort Studies (ACS) on HIV infection and AIDS, a collaboration between the Amsterdam Health Service, Academic Medical Center of the University of Amsterdam, Sanquin Blood Supply Foundation, University Medical Center of Utrecht, and the Jan van Goyen Clinic.
Authors’ contributions: All authors contributed to the study conception, as well discussion of the results and revision of the manuscript. J.J.vdH. and A.S.dV. performed ACS cART uptake data analysis. M.E.E.K. and A.S.dV. devised the model analysis. A.S.dV. implemented the model and prepared the initial manuscript draft.
This work was supported by ZonMW, the Netherlands organisation for health research and development . The ACS are part of the Netherlands HIV Monitoring Foundation and financially supported by the Center for Infectious Disease Control of the Netherlands National Institute for Public Health and the Environment (RIVM).
Conflicts of interest
There are no conflicts of interest.
1. Malhotra B. The strategic use of antiretrovirals for treatment and prevention of HIV Infection: 2nd expert panel meeting Meeting Report. WHO/HIV/2013.1.
2. HIV Modelling Consortium Treatment as Prevention Editorial Writing Group HIV treatment as prevention: models, data, and questions: towards evidence-based decision-making. PLoS Med. 2012; 9:e1001259
3. Eaton JW, Johnson LF, Salomon JA, Bärnighausen T, Bendavid E, Bershteyn A, et al. HIV treatment as prevention: systematic comparison of mathematical models of the potential impact of antiretroviral therapy on HIV incidence in South Africa. PLoS Med. 2012; 9:e1001245
4. Donnell D, Baeten JM, Kiarie J, Thomas KK, Stevens W, Cohen CR, et al. Heterosexual HIV-1 transmission after initiation of antiretroviral therapy: a prospective cohort analysis. Lancet. 2010; 375:2092–2098.
5. Granich RM, Gilks CF, Dye C, De Cock KM, Williams BG. Universal voluntary HIV testing with immediate antiretroviral therapy as a strategy for elimination of HIV transmission: a mathematical model. Lancet. 2009; 373:48–57.
6. Kretzschmar ME, Schim van der Loeff MF, Birrell PJ, De Angelis D, Coutinho RA. Prospects of elimination of HIV with test-and-treat strategy. Proc Natl Acad Sci USA. 2013; 110:15538–15543.
7. Bruce RD. Is it time for treatment as prevention among people who inject drugs?. J Acquir Immune Defic Syndr. 2013; 63:545–547.
8. Degenhardt L, Mathers B, Vickerman P, Rhodes T, Latkin C, Hickman M. Prevention of HIV infection for people who inject drugs: why individual, structural, and combination approaches are needed. Lancet. 2010; 24:285–301.
9. Kato M, Granich R, Duc Bui D, Tran HV, Nadol P, Jacka D, et al. The potential impact of expanding antiretroviral therapy and combination prevention in Vietnam: Towards elimination of HIV transmission. J Acquir Immune Defic Syndr. 2013; 63:e142–e149.
10. de Vos AS, van der Helm JJ, Matser A, Prins M, Kretzschmar ME. Decline in incidence of HIV and hepatitis C virus infection among injecting drug users in Amsterdam: evidence for harm reduction?. Addiction. 2013; 108:1070–1081.
11. Grassly NC, Garnett GP. The future of the HIV pandemic. Bull World Health Organ. 2005; 83:378–382.
12. EMCDDA Annual report on the state of the drugs problem in Europe. 2012; Lisbon:ISBN/ISSN, 1609–6150.
13. Termorshuizen F, Krol A, Prins M, van Ameijden EJ. Long-term outcome of chronic drug use: the Amsterdam Cohort Study among drug users. Am J Epidemiol. 2005; 161:271–279.
14. Matser A, Urbanus A, Geskus R, Kretzschmar M, Xiridou M, Buster M, et al. The effect of hepatitis C treatment and HIV coinfection on the disease burden of hepatitis C among injecting drug users in Amsterdam. Addiction. 2012; 107:614–623.
15. Smit C, Lindenburg K, Geskus RB, Brinkman K, Coutinho RA, Prins M. Highly active antiretroviral therapy (HAART) among HIV-infected drug users: a prospective cohort study of sexual risk and injecting behaviour. Addiction. 2006; 101:433–440.
16. Pilcher CD, Joaki G, Hoffman IF, Martinson FE, Mapanje C, Stewart PW, et al. Amplified transmission of HIV-1: comparison of HIV-1 concentrations in semen and blood during acute and chronic infection. AIDS. 2007; 21:1723–1730.
17. Cohen MS. Amplified transmission of HIV-1: missing link in the HIV pandemic. Trans Am Clin Climatol Assoc. 2006; 117:213–224.
18. Attia S, Egger M, Müller M, Zwahlen M, Low N. Sexual transmission of HIV according to viral load and antiretroviral therapy: systematic review and meta-analysis. AIDS. 2009; 17:1397–1404.
19. Babiker A, Bhaskaran K, Darbyshire J, Pezzotti P, Porter K, et al. CASCADE Collaboration Determinants of survival Decline of HIV and HCV among IDU in Amsterdam following HIV-1 seroconversion after the introduction of HAART. Lancet. 2003; 18:1267–1274.
20. Kuritzkes DR. HAART for HIV-1 Infection: zeroing in on when to start: comment on ‘Timing of HAART Initiation and Clinical Outcomes in Human Immunodeficiency Virus Type 1 Seroconverters’. Arch Intern Med. 2011; 171:1569–1570.
21. Pantaleo G, Graziosi C, Fauci AS. The immunopathogenesis of human immunodeficiency virus infection. Fauci N Engl J Med. 1993; 328:327–335.
22. Cohen MS, Dye C, Fraser C, Miller WC, Powers KA, Williams BG. HIV treatment as prevention: debate and commentary: will early infection compromise treatment-as-prevention strategies?. PLoS Med. 2012; 9:e1001232
23. Emamzadeh-Fard S, Fard SE, SeyedAlinaghi S, Paydary K. Adherence to antiretroviral therapy and its determinants in HIV/AIDS patients: a review. Fect Disord Drug Targets. 2012; 12:346–356.
24. McNairy ML, Cohen M, El-Sadr WM. Antiretroviral therapy for prevention is a combination strategy. Curr HIV/AIDS Rep. 2013; 10:152–158.
25. Lambers FA, Stolte IG, van den Berg CH, Coutinho RA, Prins M. Harm reduction intensity-Its role in HAART adherence amongst drug users in Amsterdam. Int J Drug Policy. 2011; 22:210–218.
26. Wagner BG, Blower S. Universal access to HIV treatment versus universal ’test and treat’: transmission, drug resistance & treatment costs. PLoS One. 2012; 7:e41212
27. Montaner JS. Treatment as prevention: toward an AIDS-free generation. Top Antivir Med. 2013; 21:110–114.
28. Malta M, Magnanini MM, Strathdee SA, Bastos FI. Adherence to antiretroviral therapy among HIV-infected drug users: a meta-analysis. AIDS Behav. 2010; 14:731–747.
29. Lucas GM. Substance abuse, adherence with antiretroviral therapy, and clinical outcomes among HIV-infected individuals. Life Sci. 2011; 88:948–952.
30. Weber R, Huber M, Rickenbach M, Furrer H, Elzi L, Hirschel B, et al. Uptake of and virological response to antiretroviral therapy among HIV-infected former and current injecting drug users and persons in an opiate substitution treatment programme: the Swiss HIV Cohort Study. HIV Med. 2009; 10:407–416.
31. Werb D, Mills EJ, Montaner JS, Wood E. Risk of resistance to highly active antiretroviral therapy among HIV-positive injecting drug users: a meta-analysis. Lancet Infect Dis. 2010; 10:464–469.
32. Wolfe D, Carrieri MP, Shepard D. Treatment and care for injecting drug users with HIV infection: a review of barriers and ways forward. Lancet. 2010; 376:355–366.
33. Hontelez JA, Lurie MN, Bärnighausen T, Bakker R, Baltussen R, Tanser F, et al. Elimination of HIV in South Africa through expanded access to antiretroviral therapy: a model comparison study. PLoS Med. 2013; 10:e1001534
34. Marshall BD, Paczkowski MM, Seemann L, Tempalski B, Pouget ER, Galea S, Friedman SR. A complex systems approach to evaluate HIV prevention in metropolitan areas: preliminary implications for combination intervention strategies. PLoS One. 2012; 7:e44833
35. De Vos AS, Kretzschmar ME. The efficiency of targeted intervention in limiting the spread of HIV and hepatitis C virus among injecting drug users. J Theor Biol. 2013; 21:126–134.