| |
Can antiviral therapy for hepatitis C reduce the prevalence of HCV among injecting drug user populations? A modeling analysis of its prevention utility
|
| |
| |
Download the PDF here
"modest rates of hepatitis C treatment among active injecting drug users could effectively reduce transmission. Evaluating and extending strategies to treat hepatitis C among active injectors are warranted."
Journal of Hepatology (June 2011)
Natasha K. Martin12Corresponding Author Informationemail address, Peter Vickerman12, Graham R. Foster3, Sharon J. Hutchinson45, David J. Goldberg4, Matthew Hickman1
Background & Aims
Hepatitis C virus antiviral treatment is effective for individual patients but few active injecting drug users are treated. We considered the utility of antiviral treatment for primary prevention of hepatitis C.
Methods
A hepatitis C transmission model among injecting drug users was developed, incorporating treatment (62.5% average sustained viral response) with no retreatment after initial treatment failure, potential re-infection for those cured, equal genotype setting (genotype 1:genotype 2/3), and no immunity. In addition, we examined scenarios with varied treatment response rates, immunity, or retreatment of treatment failures.
Results
In the baseline scenario, annually treating 10 infections per 1000 injecting drug users results in a relative decrease in hepatitis C prevalence over 10years of 31%, 13%, or 7% for baseline (untreated endemic chronic infection) prevalences of 20%, 40%, or 60%, respectively. Sensitivity analyses show that including the potential for immunity has minimal effect on the predictions; prevalence reductions remain even if SVR is assumed to be 25% lower among active IDU than current evidence suggests; retreatment of treatment failures does not alter the short-term (<5years) projections, but does increase treatment gains within 20years; hepatitis C free life years gained from treating active injecting drug users are projected to be higher than from treating non-injecting drug users for prevalences below 60%.
Conclusions
Despite the possibility of re-infection, modest rates of hepatitis C treatment among active injecting drug users could effectively reduce transmission. Evaluating and extending strategies to treat hepatitis C among active injectors are warranted.
Introduction
Hepatitis C virus (HCV) is a blood-borne disease which causes significant morbidity and mortality worldwide [1]. Injecting drug use is the primary mode of transmission in developed countries [2], with 15-90% of injecting drug users (IDU) testing positive for HCV antibodies [3].
Current interventions for reducing HCV transmission focus on reducing the frequency of injecting and syringe sharing/unsafe injection [2], [4]. These may have reduced HCV transmission in some settings [5], [6], [7], but there is no evidence for substantial reductions in HCV prevalence.
HCV antiviral treatment with peginterferon alfa and ribavirin is the standard care for chronic HCV, with a 50-85% cure rate (sustained viral response, SVR) depending on genotype [8]. In many countries, including the US and UK, treatment is recommended for all patient groups (including IDU) and is considered cost effective [8], [9], [10], [11]. However, few current IDU have ever been treated (<3-4%) [12], [13] because physicians are reluctant to treat due to concerns about treatment compliance and re-infection [14], [15], [16]. However, the limited information available from studies suggests that current IDU exhibit similar response rates to treatment [17], [18], [19] and compliance [18], [20], [21] when compared to non- or ex-IDU. Furthermore, there are little data on re-infection after successful treatment, except from small scale studies which report low rates in the first year [22].
Discussion
Main findings
Our simplified HCV transmission model provides evidence that antiviral treatment at achievable rates may be an effective primary prevention tool for substantially reducing the prevalence of HCV infection, despite the persistent risk of re-infection. Even more substantial reductions in prevalence are possible if effective treatments become available for those who do not attain SVR with the current regimen. Reductions in average SVR (through more genotype 1 in the population, or lowered completion and compliance) reduces the expected reduction in HCV prevalence, whereas higher SVR (e.g. through more individuals with genotype 2/3 or improved treatment regimens) increases the impact of treatment. Re-infection will reduce the HCV free life years gained (HCV free LYG) from successfully treating an individual. However, because successful treatment also prevents a potential chain of HCV transmissions, our projections suggest more HCV free LYG are achieved from treating current IDU in populations with less than 60% chronic infection prevalence than from treating non- or ex-IDU.
Strengths and limitations
There are several limitations. First, the findings are based on model projections of the treatment effect instead of experimental evidence. Second, the use of a fixed treatment rate annually assumed that treatment could be sustained at the same rate despite a reduction in prevalence. This means that as prevalence decreases, a larger proportion of infected IDU will be treated each year although the number treated remains constant. In the short term, this may be a reasonable assumption, but would require increased efficiency in efficient and effective HCV testing and case finding to find those fewer infected IDU [29].
Third, the model assumed that all infected IDU have an equal probability of being treated, completing treatment and cure. In practice, barriers exist related to accessing IDU and ensuring they are referred to and remain in specialist care. Additionally, we assumed no difference in infection risk among non-immune IDU following spontaneous clearance or successful treatment. In reality, it is highly likely that there are heterogeneities in treatment presentation and completion (as well as in behavior and risk following treatment) both between different IDU and at different times during a person's injecting career. However, in the UK and many other developed countries, a high proportion of IDU are in opiate substitution treatment (~40%), and during this time IDU may continue to inject but at a lower injecting frequency or temporarily cease injecting. As most IDU will have periods of opiate substitution treatment, there will be multiple opportunities for HCV treatment among IDU in contact with treatment services, who may achieve similar SVRs to published findings (used in our projections). Equally, biological and behavioral heterogeneity in infection risk may reduce the impact of treatment for a specific baseline prevalence. For example, there may be an elevated transmission risk in the acute stage of infection [30], [31], injectors in their first year of injecting may be at a greater risk of becoming infected than at other times, and there may be a high risk group that is resistant to interventions that could promote HCV treatment (such as opiate substitution treatment). In addition, we have assumed average cessation and drug related death rates - whereas these may vary in the first year or at other times during an injectors' career [32]. Furthermore, it is possible, though not well documented, that those undergoing or exiting treatment may exhibit increased cessation rates. This increased cessation during or after treatment could decrease the potential impact of treatment, although it would still reduce prevalence and HCV transmission due to infected IDU being removed from the pool of active injectors. Unfortunately, there is insufficient evidence to parameterize any of these consequences or changes, which can only be incorporated once additional clinical evidence has been collected.
Fourth, although we explored SVR rates corresponding to scenarios with all genotype 1, all genotype 2/3, and mixed genotypes, we did not explicitly stratify the populations by genotype. Stratifying the population by genotype could refine our model predictions because we could incorporate the dynamic effect of treatment on the proportion of each genotype in a given population.
Fifth, the model used a deterministic approximation of the infection dynamics, and as such may not be reliable for very low prevalences where stochastic effects can have substantial effect. However, as HCV prevalence is currently extremely high, the use of a deterministic model is appropriate in the current circumstance.
Evidence from other studies
Only one other modeling study has considered the impact of HCV treatment on prevalence [33]. Although a similar model structure was used, our analysis produced more optimistic projections because our fixed number treatment term is a more realistic model of antiviral treatment capacity and delivery, in contrast to a treatment term which diminishes as the prevalence decreases. In addition, our analyses present a wider range of prevalence scenarios that have been observed in different sites and countries (over a wider range of follow-up times) rather than considering a single prevalence (>60%) scenario over a limited time period. A preliminary economic evaluation also suggests that HCV treatment may be a cost effective strategy [34].
Implications
To date, uptake of therapy remains low among active IDU, and is rarely encouraged. Many factors contribute to the low treatment rates in IDU including concerns about re-infection, the low priority of injectors for scarce treatment resources and the lack of evidence showing unequivocal benefits. Nonetheless, the available evidence suggests that IDU can be treated successfully [17], [18], [19], [20], [21], [35], especially when innovative approaches to delivering therapy are used [36]. Our model suggests, for the first time, that therapy could play a valuable role in controlling the HCV epidemic in IDU. However, the experimental or clinical data are limited and probably subject to considerable selection biases. The models outlined here require validation in clinical trials of the impact of therapy on disease prevalence among IDU. Such studies are complex as treatment only indirectly affects incidence through reductions in prevalence. This analysis emphasizes the importance of mathematical models for setting intermediate targets of changes in HCV prevalence for considering the effectiveness of treatment. Further modeling work is needed to examine the cost effectiveness of different treatment strategies; the benefits of combining HCV treatment with opiate substitution therapy or needle and syringe programmes; and to incorporate transitions between periods of injecting and drug treatment and extend the model to ex-IDU.
Results
Reduction in chronic prevalence
For an IDU population with 20% baseline chronic prevalence and treatment rates of 5, 10, 20, or 40 per 1000 IDU annually, the model predicts a 15%, 30%, 62%, and 72% reduction in chronic prevalence after 10years, respectively (Fig. 2). These reductions in prevalence are at most halved for 40% baseline prevalence, and quartered for 60% prevalence.
Fig. 3 shows how the impact projections could vary over time. For a 20% baseline chronic prevalence, annually treating 10 cases per 1000 IDU results in a 16% reduction in prevalence within 5years, doubling to 30% after 10years, and 57% after 20years (Fig. 3B). In contrast, if the baseline prevalence was 40%, then the same treatment rate reduces prevalence by 8% after 5years and only 22% after 20years. With a high 60% baseline prevalence, annually treating 10 cases per 1000 IDU may only achieve a 9% reduction in prevalence after 20years.
If the treatment rate was 5 per 1000 IDU annually (Fig. 3A) then substantial prevalence reductions (>20%) within 20years only occur in populations with low baseline prevalences (<25%). In contrast, higher treatment rates (>20 per 1000 IDU annually) would be required to result in similar prevalence reductions in high prevalence settings (>60%). Reductions in HCV incidence mirror the decrease in prevalence because fewer chronically infected IDU results in fewer new infections.
The prevalence reductions flatten off in the long-term (20years) or for higher treatment rates (20-40 per 1000 IDU annually) due to the persistence of a non-responder population (Fig. 3C and D).
Comparing impact with treating non-IDU
Fig. 4 suggests that for baseline prevalences under 60%, the IDU intervention results in more HCV free life years gained (LYG) per person treated than for the ex/non-IDU intervention scenario, given equal SVR rates. Only at very high prevalences (>60%) does the ex/non-IDU treatment intervention result in greater impact per treatment initiated, due to the high risk of re-infection in the IDU scenario at these prevalences. However, if treatment is 80% as effective in active IDU, greater impact is only seen with baseline prevalences less than 30% (Supplementary Fig. 1). Varying the treatment rate has minimal effect (data not shown).
Sensitivity and uncertainty analysis
Fig. 5 shows the results of the uncertainty analysis (based on an annual treatment rate of 20 per 1000 IDU) on the predicted reduction in prevalence, including the possibility of re-infection and lower SVR rates (parameter ranges from Table 1). Introduction of immunity has little impact on the projections (Fig. 5B). Overall, uncertainty increases as time progresses (±50% after 20years) and for higher treatment rates (not shown). The infection rate (ϖ) and exit rate (μ) are collinear and along with treatment SVR rate (α) account for the majority of this variability (Fig. 5B). For most scenarios, greater impact is achieved in populations with lower incidence paired with long injecting durations. In certain circumstances (low prevalence and long time scales or high treatment rates), longer injecting duration is associated with decreased treatment impact due to the persistence of the treatment failure population. In all scenarios, greater impact is seen for higher SVR rates.
Uncertainty around the SVR resulting from treatment (due to different genotype distributions or lower compliance to treatment) can increase or decrease projections by up to 27% over 20years with an annual treatment rate of 10-20 per 1000 IDU (Fig. 6). The greater the treatment magnitude or intervention timescale, the more pronounced the difference between the impact projections for each genotype.
Adapting the model such that treatment failures are eligible for retreatment (Fig. 7) results in no change to the short-term impact projections, and only changes the long term projections at either low baseline prevalence (<20%), or moderate prevalence (<30%) with high levels of treatment (20 per 1000 IDU per year).
Lastly, there is little difference in the HCV free LYG predictions in relation to varying SVR (provided SVR rates are equal in non- or ex-IDU) or if retreatment is introduced (data not shown).
|
|
| |
| |
|
|
|
