Scholarly article on topic 'Potential market size and impact of hepatitis C treatment in low- and middle-income countries'

Potential market size and impact of hepatitis C treatment in low- and middle-income countries Academic research paper on "Economics and business"

0
0
Share paper
Academic journal
Journal of Viral Hepatitis
OECD Field of science
Keywords
{""}

Academic research paper on topic "Potential market size and impact of hepatitis C treatment in low- and middle-income countries"

JOURNAL OF VIRAL HERATIT1S

Journal of Viral Hepatitis, 2016, 23, 522-534 doi:10.1111/jvh.12516

Potential market size and impact of hepatitis C treatment in low- and middle-income countries

M. E. Woode,1,2 M. Abu-Zaineh,1,2,3 J. Perriens,4 F. Renaud,4 S. Wiktor5 and J.-P.

Moatti1,2,3 1INSERM, UMR_S 912, Sciences Economiques & Sociales de la Santé et Traitement de l'Information Médicale (SESSTIM), Marseille, France; 2UMR_S 912, IRD, Aix Marseille Université, Marseille, France; 3Aix-Marseille School of Economics, Marseille, France; 4Department of HIV and Viral Hepatitis, World Health Organization, Geneva, Switzerland; and 5Global Hepatitis Program, World Health Organization, Geneva, Switzerland

Received May 2015; accepted for publication January 2016

SUMMARY. The introduction of direct-acting antiviral agents (DAAs) has made hepatitis C infection curable in the vast majority of cases and the elimination of the infection possible. Although initially too costly for large-scale use, recent reductions in DAA prices in some low- and middle-income countries (LaMICs) has improved the prospect of many people having access to these drugs/medications in the future. This article assesses the pricing and financing conditions under which the uptake of DAAs can increase to the point where the elimination of the disease in LaMICs is feasible. A Markov simulation model is used to study the dynamics of the infection with the introduction of treatment over a 10-year period. The impact on HCV-related mortality and HCV incidence is assessed under different financing scenarios assuming that the cost of the drugs is completely paid for out-of-pocket or reduced through either subsidy or drug price decreases. It is also assessed under different diagnostic

and service delivery capacity scenarios separately for low-income (LIC), Iower-middle-income (LMIC) and upper-middle-income countries (UMIC). Monte Carlo simulations are used for sensitivity analyses. At a price of US$ 1680 per 12-week treatment duration (based on negotiated Egyptian prices for an aII oraI two-DAA regimen), most of the peopIe infected in LICs and LMICs wouId have Iimited access to treatment without subsidy or significant drug price decreases. However, peopIe in UMICs wouId be abIe to access it even in the absence of a subsidy. For HCV treatment to have a significant impact on mortaIity and incidence, a significant scaIing-up of diagnostic and service deIivery capacity for HCV infection is needed.

Keywords: deveIoping countries, direct-acting antiviraI agents (DAA), hepatitis C, markov simulation, universal access.

INTRODUCTION

The World Health Organisation (WHO) estimates that between 130 and 150 million people globally have chronic hepatitis C virus (HCV) infection [1]. After contracting HCV, between 55% and 85% develop chronic HCV [2-4]. As a result, about 704 000 people die, each year, of

Abbreviations: CTP, capacity to pay; DAA, direct-acting antiviral agents; GDP, gross domestic product; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; HICs, high-income countries; HIV, human immunodeficiency virus; LaMICs, low- and middle-income countries; LICs, Iow-income countries; LMICs, Iower-middIe-income countries; OOPE, out-of-pocket expenditure; SIR, susceptible-infected-removed; SVR, sustained viroIogicaI response; UMICs, upper-middIe-income countries; WDI, worId deveIopment indicators; WHO, world health organization.

Correspondence: Maame Esi Woode, PhD. INSERM-IRD-UMR 912 (SESSTIM), 23 Rue Stanislas Torrents, 13006 Marseille, France. E-mail: maame-esi.woode@inserm.fr

HCV-related liver diseases [5]. HCV is found worldwide with most (72%) infections occurring in middle-income countries (MICs), 13% in low-income countries (LICs) and the remaining 15% in high-income countries (HICs) [6,7]. The spread of HCV is largely explained by the use of unsafe injection equipment, and for this reason, persons who inject drugs and populations exposed to nonsterile injection and other invasive medical equipment in the healthcare setting are at greater risk. As the infection is also sexually transmitted, people at high risk of acquiring sexually transmitted infections (STI) and human immunodeficiency virus (HIV) [7] are also at risk of HCV infection.

Most people with HCV are not aware of their status: >50% in HICs, >10% in LICs and Iower-middle-income countries (LMICs). Those in touch with health services because of HIV or because they are symptomatic are more likely to be diagnosed [8-10].

To prevent the spread of HCV, WHO recommends actions reducing exposure to HCV in heaIthcare settings (such as ensuring safety of injections) and impIementing

© 2016 The Authors. The Journal of Viral Hepatitis Published by John Wiley & Sons Ltd. This is an open access articIe under the terms of the Creative Commons Attribution-NonCommerciaI License, which permits use, distribution and reproduction in any medium, provided the originaI work is properIy cited and

is not used for commerciaI purposes.

harm reducing interventions for drug users [11]. Until recently, the use of antiviral treatment to curtail the extent and impact of the disease was not given high priority, as the existing treatments with interferon and ribavirin are expensive, relatively toxic and not very effective. The advent of all oral combination treatment with direct-acting antivirals (DAAs) in 2014 dramatically changed this, as it simplified treatment, reduced side effects and increased cure rates to approximately 95% [12]. This made eliminating HCV transmission possible for the first time.

Direct-acting antiviral agents are currently too expensive for governments worldwide to deliver on their promise to cure and eliminate the disease. Nevertheless, price decreases for HCV drugs have already been announced for some DAAs in a few LICs, and in the future voluntary licensing will increase the availability of generic DAAs [12]. This offers some hope that universal access to HCV treatment might be possible. We modelled the affordability and medical eligibility conditions under which HCV treatment can be optimally used to reduce HCV-related mortality and incidence in low- and middle-income countries (LaMICs).

MATERIALS AND METHODS

A susceptible-infected-removed (SIR) epidemic model was used to construct a compartmental Markov model assessing the evolution of the HCV epidemic [13-15]. Re-infection rates can be high among people who inject drugs (PWID), but PWID only represent approximately 5.4% of all persons with HCV infections [16, 17]. Re-infection rates are much lower among non-PWID [18,19]. In the absence of relevant data on mixing and the rate of reinfection in treated populations, re-infection is assumed not to occur in the model and that the population mixes homogeneously, that is to say, they randomly come into contact with HCV infection. The outcomes are the number of individuals receiving treatment, the number of HCV-related deaths and the disease incidence following the introduction of treatment.

The model is built in two parts. The first, represented by the lower line of events in Fig. 1, simulates the natural history of the disease while the second, represented by the 2 upper lines, simulates the effect of treatment strategies. Both parts follow a yearly cycle.

The population is assumed to be composed of seven groups: susceptible, acutely infected, immune (including the cured), chronically infected and never treated (comprising 5 subgroups F0, F1, F2, F3 and F4, in which the number following the 'F' represents the stage of fibrosis development), those with decompensated cirrhosis or with hepatocellular carcinoma (HCC), and those who experience HCV-related death.

Given the lack of initial incidence data, a steady-state population incidence is assumed; hence, the numbers

within each subpopulation remain unchanged [20]. The yearly incidence is calculated as

P -)(-)

'1 - P DR

where I the incidence rate, P is the prevalence rate and DR the disease duration (40 years). This is a necessary assumption that implies a more or less stable population, despite the fact that we allow for HCV and non-HCV-related mortality as well as immigration and emigration, which would imply a not so stable population. Further clinical studies on the incidence of the disease in developing countries are needed for this assumption to be relaxed.

The annual rate of natural population growth, obtained from the World Bank Population Projection [21], is used to estimate the size of the population each year. Given the homogeneity and no re-infection assumptions, the rate at which individuals join the susceptible population is equal to the birth rate.

Table 1A presents the initial fibrosis distribution of those chronically HCV-infected, based on prevalence data for Egypt [22]. Due to absence of data on HCV-related decom-pensated cirrhosis and HCC distribution, we assume that the ratio of people with HCV-related decompensated cirrhosis and HCC to those with chronically infected HCV is approximately 1% and 0.5%, respectively. The transition probabilities between disease states are given in Table 1B. Irrespective of infection status, individuals are assumed to exit from all groups at the natural death rate. The outcome of introducing treatment at each fibrosis stage is represented by the annual number treated (T0 to T4), which represents by the number of people cured and the number of people treated but not cured (F0 to F0). The latter are assumed to progress to decompensated cirrhosis and hepatocellular carcinoma with the same rate as untreated patients.

Treatment discontinuation and retreatment is not allowed for in the model. Therefore, once treatment is accepted, individuals follow through to the end, and if they do not achieve sustained virological response (SVR) after treatment has ended, they are not retreated. People are assumed to be cured (i.e. remain free from re-infection) when they reach SVR. The SVR rates used are summarized in Table 1C.

Each country cohort was analysed separately over a 10-year period using prevalence rates estimated by Gower et al. [23]. Results were aggregated by subgroup of LIC, LMIC and UMICs, as classified by the World Bank in 2015 [24].

Given that in LaMICs, only a small percentage of those with chronic HCV would be diagnosed; we assume, in the absence of data, an initial diagnosis rate (the number of persons newly diagnosed) of 1% for LICs, 3% for LMICs and 5% for UMICs. We also assume that the number of individuals diagnosed increases by a factor of 1.2 in

Immune i

?0 ?1 ?2 ?3 P4

Decompensated Cirrhosis Liver Related Mortality

HCC Related Mortality

Fig. 1 A simplified Markov model.

subsequent years. This estimate of 1.2 is based mainly on resuIts observed for HIV studies where demand for CD4 testing was expected to increase by a factor of 1.1-2.2 per year from 2013-2018. Demand for viral load testing was expected to increase by a factor of 1.2-4.4 per year for the same time frame. The demand for HIV testing was expected to increase by a factor of 1.1-1.5 per year [25].

EIigibiIity for treatment is assumed to depend on the fibrosis stage. In the base case scenario (based on the 2014 WHO recommendations [26]), only patients with 'advanced fibrosis and cirrhosis (Metavir stages F3 and F4) are assumed to be eIigibIe for, and offered, treatment.

Not everyone diagnosed and eIigibIe for treatment starts it. Several factors, including individuals' capacity to pay (CTP), severity of the disease (stage of disease) and trust in the remedies influence treatment uptake. The first factor considered in the model is the users' CTP. Data on the share of GDP heId by each income quintiIe from the WorId Development Indicators (WDI) are used as proxies for CTP. This is then divided by the number of individuaIs in each income group to estimate the income per capita per quin-tiIe. GDP is assumed to increase over time, at rates obtained from the World Economic Outlook Database for April (2014). A moving average is used to estimate the growth rates for the years after 2019. It is assumed that individuaIs can afford the treatment if its cost does not exceed 40% of their CTP, which corresponds to the conventional definition of the threshoId of catastrophic heaIth expenditure [27,28].

It is further assumed that a 10% decrease in prices Ieads to an increase in demand of about 2% for each income group provided they have sufficient income to afford the treatment. This is in Iine with previous studies, which found that the price eIasticity of demand for heaIthcare is about -0-2 (95% CI (-0.04 to -0-75)) [29].

After sofosbuvir - one of the DAAs currentIy in use - was given market authorization by the US Food and Drug Administration [12], negotiations between Gilead Sciences and the government of Egypt followed, in which a reduced price US$ 840 for a 12-week course of sofosbuvir was agreed [30]. If Iedipasvir, another DAA, were made avaiIabIe for the same price, a 12-week treatment with two DAAs wouId cost approximately US$ 1680. We assumed this to be the cost of a 12-week duaI drug DAA treatment course in our modeI.

As peopIe infected with genotypes 2 and 3 require a 24-week treatment course [31,32], which therefore costs twice as much, we adjusted the treatment cost in each country to take into account country-specific genotype distribution, from Messina and colleagues [33] (data not shown - available in supplemental Tables S1 and S2). Akin to the fall in the gIobaI price of ARV medications observed from 2003 to 2011, the average cost of treatment is assumed to faII at a rate of 3.8% each year [34].

As heaIth systems in deveIoping countries are underfunded and understaffed [35], we assumed that, in the initial year, the heaIth system wouId be capabIe of taking care of onIy 5% of those diagnosed. Not aII of those diagnosed wouId demand treatment. In the baseIine scenario, the capacity of the heaIth system is assumed to increase over time by a factor of 1.2 each year. For instance, if in 2014, the health system can provide the treatment to onIy 5 of 100 infected individuals, then in 2015 it can take care of 6(1.2*5) individuals and 7.2(1.2*6) individuals in 2016. If the capacity is lower than the number of those diagnosed, those offered treatment, and those abIe to pay for treatment, then those with a higher fibrosis stage are considered first for treatment.

Scenarios presented

The first scenario presented in this articIe is the 'base case' scenario. In keeping with the more conservative

Stage Distribution 95 CI%

(a) Initial fibrosis distribution in Markov model

F0 0.170 0.15-0.19 [41]

F1 0.350 0.32-0.39

F2 0.220 0.20-0.24

F3 0.140 0.13-0.15

F4 0.120 0.11-0.13

Decompensated Cirrhosis 0.010 Assumption

HCC 0.005

Transitions Probabilities 95% CI Source

(b) Annualized transition probabilities

Acute -> Spontaneous Recovery 0.250 0.150-0.500 [4]

Acute -> F0 0.750 0.500-0.850 [42]

F0 -> F1 0.117 0.104-0.130

Fl -> F2 0.085 0.075-0.096

F2 -> F3 0.120 0.109-0.133

F3 -> F4 0.116 0.104-0.129

F3 -> Decompensated 0.012 0.010-0.014 [43,44]

F3 -> HCC 0.011 0.009-0.013 [43,45]

F4 -> HCC 0.030 0.020-0.040 [43,45]

F4 -> Decompensated 0.040 0.030-0.050 [43,44,46,47]

Decompensated -> HCC 0.014 0.011-0.017 [43,47]

Decompensated -> Liver Related Mortality 0.130 0.100-0.160 [43,47]

HCC -> HCC Related Mortality 0.430 0.340-0.510 [43,47]

SVR (%) Duration

(c) Sustained virological response rate SVR

Genotype 1 100 12 Weeks

Genotype 2 93 24 Weeks

Genotype 3 93 24 Weeks

Genotype 4 100 12 Weeks

Genotype 5 100 12 Weeks

Genotype 6 100 12 Weeks

Exogenous Capacity

Diagnosis/ rate of to pay Price

Treatment Subsidy Initial detection Absorption Absorption price (as % of elasticity

Scenarios groups rates (%) diagnosis factor capacity factor reduction Income) of demand

(d) Other assumptions

Baseline F3 + F4 0 1% LIC, 1.2 5% 1.2 3.8% <40% 0.2, that is 10%

Scenario 2 F0 to F4 0 3% LMIC, decrease in

Scenario 3 F3 + F4 25 5% UMIC price leads

50 to a 2%

75 increase in

100 demand

Scenario 4 F3 + F4 80 3 3

Scenario 5 F0 to F4 80

Sensitivity F0 to F4 0 1.2 1.2

Analysis

The detection factor is the factor by which the number of people diagnosed increases over time.

The absorption factor is the factor by which the number of infected individuals catered for by the health sector increases over time.Source: [48,49].

version of the 2014 WHO recommendations, only those with fibrosis stages 3 and 4 are eligible for the treatment. It assumes that only 1%, 3% and 5% of HCV-infected patients are diagnosed in LICs, LMICs and UMICs, respectively, with the proportions increasing over time by a factor 1.2 per year. Only 5% of those diagnosed and eligible are offered treatment in year 1, increasing by a factor of 1.2 each subsequent year, to take into account health systems' increased capacity to offer the service. Treatment costs are assumed to be US$ 1680 for a 12-week treatment course (adjusted for the genotype distribution in the infected population) and the number of people accepting it increases 2% when the cost decreases by 10%.

The second scenario has the same assumptions, with the exception that, instead of limiting treatment to patients with fibrosis stages 3 and 4, all patients with chronic infection are eligible for treatment.

In the third scenario, a 25%, 50%, 75% and 100% subsidy of the cost of drugs is introduced, but eligibility is the same as that for the baseline scenario.

In the fourth scenario, people with fibrosis stages 3 and 4 are eligible for treatment, but the ability to diagnose and the capacity of the health system to offer services is allowed to increase by a factor of 3 (does not mean the number of patients are tripling but rather that the health system is able to take care of a maximum of 3 times the people they had in charge the year before), while drug costs are reduced (through subsidy or price decrease) by 80%, which corresponds to an initial price of US$ 400 for a 12-week course of dual combination treatment, which equals the upper boundary of their reported minimum production cost [36].

The fifth and final scenario has the same assumptions as the fourth scenario, except that all patients with chronic infection are eligible for treatment. A full description of the various assumptions can be found in Table 1a-d and in

Fig. 2.

Sensitivity analysis

A Monte Carlo simulation was carried out for the scenario in which individuals with fibrosis stages 0 to 4 are eligible for treatment, and in which the initial rate of diagnosis and the initial health system capacity are set at 5% with an increment factor of 1.2. The transition probabilities and SVR rates were allowed to randomly vary between their 95% confidence intervals.

RESULTS

Of the 132 LaMICS in the World Bank Classification, 116 were included in the model (see supplementary Table S2). Sixteen countries, which represent an estimated 1.78% of all persons with HCV infection, were not included because of missing financial data. As the total number of HCV-infected people in those countries is very small, their omission has minimal impact on the results. Full results are available in supplementary Tables S4-S6 and Figure S1.

Low-income countries

Results for LICs are shown in Table 2 and summarized in Fig. 3. The uptake of HCV treatment in LIC is predicted to be very limited in the base case scenario: 7081 cases treated over a 10-year period. Not surprisingly, the impact on the number of HCV-related deaths and incidence is also limited.

Expanding treatment eligibility from fibrosis stages 3 and 4 to all people with chronic HCV (scenario 2) does not change this picture, as demand is limited both by the health system's ability to provide HCV services and as the cost of drugs remains unaffordable for the great majority of patients.

The effect of reducing out-of-pocket expenditure (OOPE), which is modelled in scenario 3 with no measures to improve the health system's diagnostic and service delivery capacity, is almost linear - approximately doubling for

i. The detection factor is the factor by which the number ofpeople diagnosed increases over time.

ii. The absorption factor is the factor by which the number of infected individuals catered for by the health sector increases over time.

Fig. 2 Assumptions timeline.

Number treated

Difference compared with

Total baseline

HCV-related deaths

Percentage difference compared with baseline Total (%)

New HCV infections

Percentage difference compared with baseline Total (%)

Baseline F3 + F4 OOPE 7081 - 955 487 0.00 2 777 177 0.00

Scenario 2 F0 to F4 OOPE 7081 - 955 487 0.00 2 777 177 0.00

Scenario 3 F3 + F4 OOPE reduced by 25% subsidy 37 686 30 604 954 639 -0.09 2 776 334 -0.03

F3 + F4 OOPE reduced by 91 708 84 627 951 343 -0.43 2 773 260 -0.14

50% subsidy

F3 + F4 OOPE reduced by 139 975 132 893 943 990 -1.20 2 767 643 -0.34

75% subsidy

F3 + F4 OOPE reduced by 145 357 138 276 942 413 -1.37 2 766 847 -0.37

100% subsidy

Scenario 4 F3 + F4 OOPE reduced by 80% subsidy with improved diagnosis and strengthened health system 3 619 668 3 612 587 775 786 -18.81 2 559 553 -7.84

Scenario 5 F0 to F4 OOPE reduced by 80% subsidy with improved diagnosis and strengthened 9 497 902 9 490 820 773 084 -19.09 2 236 468 -19.47

health system

Differences are calculated as a subtraction of the baseline result from the scenario results (Scenario Result—Baseline Result).

The percentage differences are calculated as a ratio between this difference and the baseline result ((Scenario Result—Baseline Result)/Baseline Result).

A negative (percentage) difference indicates better outcomes in the scenario compared with the baseline. Calculations made over a 10-year period.

every 25% reduction until OOPE is reduced by 75%. When there is no OOPE for the drugs, the uptake is predicted to increase by a factor 1.04, compared with reducing OOPE by 75%. However, even with no OOPE, the number of deaths and incidence would be reduced by only 1.3 7% and 0.3 7%, respectively, over a 10-year period.

It is predicted that the effect of alleviating health system barriers, illustrated in scenario 4 for people with fibrosis stages 3 and 4, and for all people with chronic HCV in scenario 5, for a level of OOPE reduced by 80%, is a dramatic one. The model predicts that approximately 18.81% to 19.09% of the deaths and 7.84% to 19.47% of the new cases in the base case scenario would be avoided.

Lower-middle-income countries

Results for LMICs are shown in Table 3 and summarized in Fig. 3. In the baseline scenario, almost 787 000 people would access treatment. Offering the treatment to all people with chronic HCV, with payment of the full cost of the drugs by the patients, as shown in scenario 2, is seen to

almost increase the uptake by a factor of 1.05. However, this has only a minimal impact on the number of deaths, and on incidence, which are predicted to decrease by only 0.00% and 0.01%, respectively, over the 10-year projection horizon of the model.

The outcomes of progressively decreasing OOPE for the drugs are shown in scenario 3, which illustrates that the number of people accessing treatment remains limited as long as OOPE equals or exceeds 75% of the drug costs. When there is no OOPE for the drugs, the impact on the number of deaths and the incidence increases to 2.10% and 0.51%, respectively.

As with the scenarios in LICs, the effect of alleviating health system barriers, illustrated in scenario 4 for people with fibrosis stages 3 and 4, and for all people with chronic HCV in scenario 5, for a level of OOPE reduced by 80%, is predicted to be dramatic. The number of HCV-related deaths decreases by 39.95% when treatment is offered to people with fibrosis stages 3 and 4, and by 40.08% when it is offered to all people with chronic HCV; the incidence decreases by 15.08% and 33.91%, respectively.

i. Results of subsidised model do not vary greatly from the baseline model and so are excluded.

ii. Calculations made over a 10-year period.

Fig. 3 Model outcomes.

Upper-middle-income countries

Results for UMICs are shown in Table 4 and summarized in Fig. 3. Even in the baseline scenario, in which treatment is not subsidized, there is significant uptake of HCV treatment: it would exceed 1 million people over a 10-year period when offered to those with fibrosis stages 3 and 4 (Scenario 1) or to all those with chronic HCV (Scenario 2).

Reducing the level of OOPE, as illustrated in Scenario 3, shows no impact on uptake, because at the price levels assumed in the model, the majority of people with fibrosis stage 3 or 4 and offered treatment would be able to access it.

The effect of alleviating health system barriers, illustrated in scenario 4, for people with fibrosis stages 3 and 4, and for all people with chronic HCV in scenario 5, for a level of OOPE reduced by 80%, is again predicted to be dramatic.

Over a 10-year period, the number of HCV-related deaths would decrease by 46.38%, when treatment is offered to people with fibrosis stages 3 and 4, and by 46.40% when it is offered to all people with chronic HCV. Incidence would decrease by 17.10% and 39.10%, respectively.

Sensitivity analysis

The Monte Carlo simulation results indicated that our results are robust to changes in transition probabilities and SVR rates, with the average number of the never-treated patients not varying significantly from the values obtained using the initial transition probabilities and SVR rates. Thus, the uncertainties associated with these groups of variables do not create very large discrepancies. Please refer to Table 5 for summary results and Tables S7-S9 in the supplementary appendix for full results.

Number treated

Difference compared with baseline

HCV-reIated deaths

Percentage difference compared with baseline Total (%)

New HCV infections

Percentage difference compared with baseline Total (%)

Baseline Scenario 2

Scenario 3

Scenario 4

Scenario 5

F3 + F4 OOPE F0 to F4 OOPE

F3 + F4 OOPE reduced by

25% subsidy F3 + F4 OOPE reduced by

50% subsidy F3 + F4 OOPE reduced by

75% subsidy F3 + F4 OOPE reduced by

100% subsidy F3 + F4 OOPE reduced by 80% subsidy with improved diagnosis and strengthened health system F0 to F4 OOPE reduced by 80% subsidy with improved diagnosis and strengthened health system

787 535 -826 872

929 743

1 120 194

1 154 281

1 154 281

39 337

2 672 514 2 672 464

0.00 0.00

142 209 2 658 480 -0.53

332 659 2 627 802 -1.67

366 746 2 616 535 -2.09

366 746 2 616 261 2.10

13 103 802 12 316 267 1 604 713 -39.95

31 423 231 30 635 696 1 601 333 -40.08

7 832 116 7 831 083

0.00 0.01

7 819 898 -0.16

7 798 079 -0.43

7 792 132 -0.51

7 792 160 -0.51

6 651 065 -15.08

5 176 569 -33.91

Differences are calculated as a subtraction of the baseline result from the scenario results (Scenario Result - Baseline Result).

The percentage differences are calculated as a ratio between this difference and the baseline result ((Scenario Result - Baseline Result)/Baseline Result).

A negative (percentage) difference indicates better outcomes in the scenario compared with the baseline. Calculations made over a 10-year period.

DISCUSSION

This article assesses the effect of different assumptions on the number of people likely to be treated with DAAs for HCV in LaMICs and examines the related impact on HCV-related mortality and incidence.

The baseline model and the scenario expanding the eligibility for DAAs to all patients with chronic HCV demonstrate that at the lowest prices agreed for Egypt and Pakistan, and in the absence of full subsidy of the cost of DAAs, the uptake of DAA will be very limited in LICs and LMICs. Consequently, HCV-reIated mortality and incidence are not substantially decreased. Limiting OOPE for the drugs - by either subsidizing their cost or producing them much Iess expensiveIy - was predicted to not have a major impact on the number of deaths and incidence in LICs and LMICs, unless more than 75% of their full cost was covered. However, even when there is no OOPE for the drugs, the predicted impact on death and incidence was Iimited.

AIong with subsidizing or reducing the cost of the drugs, improving the health system's diagnostic and care delivery capacity is a criticaI requirement to reduce HCV-reIated mortaIity and incidence.

At the price IeveIs assumed in our modeI, decreasing OOPE for the drugs did not appear to affect the uptake in UMICs. It remains to be seen, however, whether UMICs wiII be abIe to access DAAs at the prices assumed in our modeI, as at Ieast one company producing DAAs announced that it will negotiate 3-tier pricing system [37]. This suggests that UMICs might faII in the middIe tier, where prices wiII IikeIy be higher. Furthermore, in severaI UMICs, HCV infection is prevaIent in peopIe who inject drugs, a popuIation who might have Iess discretionary CTP, and for whom treatment would need to be subsidized if HCV is to be contained.

As illustrated by scenarios 4 and 5, the greatest impact on treatment uptake and incidence wouId resuIt from subsidizing treatment, whiIe at the same time increasing the

Number treated HCV-related deaths New HCV infections

Percentage Percentage

Total Difference compared with baseline Total difference compared with baseline (%) Total difference compared with baseline (%)

Baseline F3 + F4 OOPE 1 377 901 - 1 652 121 0.00 5 038 466 0.00

Scenario 2 F0 to F4 OOPE 1 377 901 - 1 652 121 0.00 5 038 466 0.00

Scenario F3 + F4 OOPE reduced by 1 377 901 - 1 652 115 0.00 5 038 467 0.00

3 25% subsidy

F3 + F4 OOPE reduced by 1 377 901 - 1 652 115 0.00 5 038 467 0.00

50% subsidy

F3 + F4 OOPE reduced by 1 377 901 - 1 652 115 0.00 5 038 467 0.00

75% subsidy

F3 + F4 OOPE reduced by 1 377 901 - 1 652 115 0.00 5 038 467 0.00

100% subsidy

Scenario F3 + F4 OOPE reduced by 8 999 414 7 621 514 885 819 -46.38 4 176 820 -17.10

4 80% subsidy with improved

diagnosis and strengthened

health system

Scenario F0 to F4 OOPE reduced by 20 630 038 19 252 138 885 569 -46.40 3 068 177 -39.10

5 80% subsidy with improved

diagnosis and strengthened

health system

Differences are calculated as a subtraction of the baseline result from the scenario results (Scenario Result - Baseline Result).

The percentage differences are calculated as a ratio between this difference and the baseline result ((Scenario Result - Baseline Result)/Baseline Result).

A negative (percentage) difference indicates better outcomes in the scenario compared with the baseline. Calculations made over a 10-year period.

Table 5 Results from Monte Carlo simulations

Mean Std. Dev. Min Max

Low-income countries Number Treated HCV-Related Death New HCV Infections Lower-middle-income countries Number Treated HCV-Related Death New HCV Infections Upper-middle-income countries Number Treated HCV-Related Death New HCV Infections

11 881 756 956 937 2 820 206

32 221 747 1 663 036 5 386 147

20 609 983 915 902 3 113 891

1 178 740 116 708 259 111

3 716 982 224 659 627 012

3 468 214 164 414 565 713

7 930 590 580 847 1 988 242

21 406 995 997 583 3 707 795

11 419 303 537 534 1 933 251

15 883 385

1 518 637 3 708 681

43 558 825

2 577 542 7 668 104

29 264 170 1 542 838 5 236 299

Results for Scenario 5 presented.

10 000 simulations were carried out in Excel 2010.

Calculations made over a 10-year period.

health system's diagnostic and delivery capacity. Moreover, comparison of those 2 scenarios suggests that to have a major impact on incidence, treatment shouId be offered to everyone with chronic HCV. WhiIe this wouId cost more in terms of drugs and service deIivery, it wouId faciIitate access to treatment, because the cost and compIexity of fibrosis staging wouId no Ionger create a bottIeneck. Assuming that treatments are pan-genotypic and, therefore, do not require genotyping, if treatment is offered to aII peo-pIe with chronic HCV, aII one wouId need to initiate treatment is a diagnosis of chronic HCV, which couId be estabIished by documenting persistent viraI repIication with HCV core antigen tests or a nucleic acid test (PCR).

The question then is whether the gIobaI community is willing to finance global scale-up of access to DAAs. For LICs, about US$ 844 million would be needed to cover the cost of the 80% subsidy for our scenario 5, in which all those with chronic HCV are eIigibIe for treatment, for the first 5 years of the model. Approximately, US$ 13 billion wouId be required to cover the fuII drug costs over the entire 10-year period. In LMICs, the cumuIative cost of treating all with chronic HCV would be approximately US$ 9 billion by year 5 and US$ 49 billion by year 10. These costs are of the same order of magnitude those of other donor-funded disease priorities such as HIV (about US$ 8.1 billion in 2013 alone [38,39] and would be considerably Iess if heaIth systems couId not scaIe-up their diagnostics and service deIivery capacity as quickIy as we assumed in our modeI, or if eIigibiIity for treatment were Iimited to fibrosis stages 3 and 4, or indeed if the cost of DAA's decreased more quickIy than we assumed in our modeI, as some claim is possible [36]. As the cost of treating HCV infections wouId appear manageabIe, it shouId be noted that severaI LMICs, incIuding Egypt, Pakistan and Mongo-Iia, have decided to start HCV treatment programmes.

It is interesting to compare our resuIts with those of Obach et al. [40] whose objective was to maximize Life Years Saved given a fixed and limited number of treatment sIots. Our articIe, on the other hand, Iooks at the combination of treatment costs and eIigibiIity criteria that wouId resuIt in the greatest number of persons treated and thus maximaIIy reduce mortaIity and incidence. Their resuIts impIy that individuaIs at Fibrosis stages F3 and F4 shouId be given a priority to save the most Iife years. We find that treating onIy individuaIs at fibrosis stage F3 and F4 may save lives, but that the greatest effect on both mortality and incidence is obtained onIy when those at the earIy stages are aIso taken into consideration.

SeveraI study Iimitations must be acknowIedged. First, in spite of the fact that HCV infections are known to dis-proportionateIy affect persons who inject drugs and men who have sex with men, homogenous mixing is assumed, giving both these two groups and the generaI popuIation the same probabiIity of becoming infected. In addition, due to a Iack of data from other countries, we

used the distribution of fibrosis stages based on data from Egypt. It is unknown to what degree the distribution of fibrosis stages differs in different countries. Sensitivity anaIysis resuIts presented in TabIe 5 indicate that the resuIts do not vary greatIy when parameters are aIIowed to vary at the same time within their respective confidence intervaIs.

Second, onIy drug costs were considered: costs of increasing heaIth system's diagnostic and service deIivery capacity were ignored. There is a need to admit that there is some uncertainty about the future cost of DAAs. WhiIe some argue that drug costs wiII decrease faster than assumed [30], the real drug costs may be higher than those we projected. In the Iatter case, it may take Ionger to attain universaI access to HCV medications. We did not specificaIIy address whether for Iow-income popuIations within LMICs and UMICs there is a need to buiId addi-tionaI financiaI protection schemes (Iike income suppIe-mentation) or whether specific service deIivery investments such as outreach services are needed for most vuInerabIe popuIations. Both wouId increase the cost of HCV treatment.

Third, in view of their major impact on modeI outcomes, it is aIso criticaI to note that the absence on data forced us to choose arbitrary IeveIs of constraint in diagnostic and service deIivery capacity.

Fourth, there is a great deaI of uncertainty about the epidemioIogy and naturaI history of HCV infection. WhiIe it is estimated that HCV-reIated worIdwide mortaIity is about 704 000 persons/year [5], our estimate of HCV-reIated deaths is sIightIy higher. This is IargeIy due to the transition probabiIities used in our modeI [41]. In turn, this suggests that either the mortality obtained by the Global Burden of Disease study [5] is underestimated or that the transition probabiIities are on the high side. Without better information on country-specific naturaI history of HCV infection and better assessment of its impact on mor-taIity, it is not possibIe to suggest a better way of deaIing with this issue. However, even if the impact on the number of deaths in our modeI were overestimated, the tendency of decreasing death rates with increasing access to treatment wouId stiII be confirmed.

FinaIIy, the modeI assumes that aII patients who can afford treatment wouId start if offered, regardIess of any side-effects or the stage of their infection, and that cure rates wiII be high. AIthough, new DAAs are shown to have Iimited side-effects, patients' acceptabiIity of treatment may vary depending on other unobservabIe factors. Not aII service providers wouId compIy with recommendations to offer aII chronicaIIy infected patients' treatment if they are not convinced that it confers benefit to their patients' weII-being. And in reaI Iife, the cure rates might be Iower than those reported in cIinicaI triaIs. In the absence of reIiabIe data on these attrition rates, we refrain from specuIating how substantiaI they might be.

In spite of its limitations, our model provides strong suggestions that DAAs can play an important role in controlling the HCV epidemic and gives a first assessment of what is required to realize their potential in LaMICs. It also provides a strong indication that in LICs and LMICs, subsidies or much lower prices will be needed to make DAAs affordable. Furthermore, it underlines that it is critical to strengthen the ability of the health system to diagnose HCV infection and manage treatment to maximize its potential impact on the HCV epidemic. The latter will also help prevent the spread of HCV. It is, therefore, paramount that infected people and their health service providers combine their advocacy and technical skills to mobilize the political momentum and health system inputs needed to expand access to DAAs. Finally, one should realize that treatment is only one side of the story: the prevention of HCV infections, mainly by securing injection safety, ensuring a clean blood supply, and implementing harm reduction interventions targeting people who inject drugs, are also required.

the 'Investissements d'Avenir' French Government program, managed by the French National Research Agency (ANR).

COMPETING INTEREST

The authors declare that they have no competing interests. AUTHORS' CONTRIBUTION

Maame Esi Woode and Mohammad Abu-Zaineh were involved in literature search, figures, study design, data collection, data analysis, data interpretation and writing. Joseph Perriens involved in literature search, study design, data interpretation and writing. Francoise Renaud involved in study design, data interpretation and writing. Stefan Wiktor involved in study design and initial proof-reading of draft versions. Jean-Paul Moatti involved in study design, data analysis and data interpretation.

FUNDING

This work was initiated with a grant from WHO to the ANRS. It was completed thanks to the support of the A*MIDEX project (no. ANR-11-IDEX-0001-02) funded by

ACKNOWLEDGEMENTS

We are grateful for helpful comments and suggestions on earlier drafts of the manuscript by Patrizia Carrieri, Arnaud Fontanet, Daniel Wolfe and Els Torreele.

REFERENCES

1 Lavanchy D. The global burden of hepatitis C. Liver Int 2009; 29: 74-81.

2 Deuffic S, Buffat L, Poynard T, Val-leron AJ. Modeling the Hepatitis C Virus Epidemic in France. Hepatology 1999; 29(5): 1596-1601.

3 Kamal SM. Acute hepatitis C: a systematic review. Am J Gastroenterol 2008; 103: 1283-1297.

4 Lehman EM, Wilson ML. Epidemic hepatitis C virus infection in Egypt: estimates of past incidence and future morbidity and mortality. J Viral Hepatitis 2009; 16(9): 650-658.

5 Mohsen N, Haidong W, Rafael L. GBD 2013 Mortality and Causes of Death Collaborators. Global, regional, and national age-specific all-cause and cause-specific mortality for 240 causes if Death 19902013: a systematic analysis for the global burden of disease study 2013. Lancet. 2015; 385: 117171.

6 Lavanchy D. Evolving epidemiology of hepatitis C Virus. Clin Microbiol Infect 2011; 17: 107-115.

7 Medicins du Monde. New treatments for hepatitis C virus: strategies for achieving universal access. [Online]. Available from: http://hepcoalition. org/IMG/pdf/web_daas_strategies_for_ achieving_universal_access_en.pdf (accessed March 2014).

8 Cornberg M, Razavi H, Alberti A et al. A systematic review of hepatitis C virus epidemiology in Europe, Canada and Israel. Liver Int 2011; 31: 30-60.

9 Kershenocich D, Razavi HSAJ, Bessone F, Coelho H, Dagher L et al. Trends and projections of hepatitis C virus epidemiology in Latin America. Liver Int 2011; 31(2): 18-29.

10 Sievert W, Altraif I, Razavi H et al. A systematic review of hepatitis C virus epidemiology in Asia, Australia and Egypt. Liver Int 2011; 31: 61-80.

11 WHO. Hepatitis C Fact Sheet no. 164. [Online].; 2014 [cited 2015 03 4. Available at: http://www. who.int/mediacentre/factsheets/fs164/ en/ (accessed March 2015).

12 Pawlotsky J. New Hepatitis C therapies: the toolbox, strategies, and

challenges. Gastroenterology 2014; 146: 1176-1192.

13 Shepard CW, Finelli L, Alter MJ. Global epidemiology of hepatitis C Virus infection. Lancet Infect Dis 2005; 5: 558-567.

14 Elbasha E. Model for Hepatitis C Virus Transmissions. Math Biosci Eng 2013; 10(4): 1045-1065.

15 Hajarizadeh B, Grebely J, Dore GJ. Epidemiology and natural history of HCV infection. Nat Rev Gastroenterol Hepatol 2013; 10: 553-562.

16 Mohd Hanafiah K, Groeger J, Flax-man AD, Wiersma ST. Global Epidemiology of Hepatitis C Virus Infection: New Estimates of Age-Specific Antibody to HCV Prevalence. Hepatology 2013; 57(4): 1333-13342.

17 Nelson PK, Mathers BM, Cowie B et al. Global Epidemiology of Hepatitis B and Hepatitis C in People who Inject Drugs: Results of Systematic Reviews. Lancet 2011; 378(9791): 571-583.

18 Averhoff FM, Glass N, Holtzman D. Global burden of hepatitis C: consid-

erations for healthcare providers in the United States. Clin Infect Dis 2012; 55(suppI 1): S10-s15.

19 Hill AM, Saleem J, Heath KA, Simmons B. Effects of sustained virologi-caI response (SVR) on the risk of Iiver transpIant, hepatoceIIuIar arcinoma, death and re-infection: meta-anaIysis of 129 studies in 23,309 patients with hepatitis C infection. Hepatology 2014; 4(suppI): 218A-219A.

20 Pearce N. Effect measures in prevalence studies. Environ Health Per-spect 2004; 112: 1047-1050.

21 World Bank Group. Population Estimates and Projections. [Online].; 2014 [cited 2014 05 01. Available at: http://data.worIdbank.org/data-cataIog/popuIation-projection-tabIes (accessed May 2014).

22 EI-Kamary SS, Mohamed MM, EI-Raziky M et al. Liver fibrosis staging through a stepwise anaIysis of non-invasive markers (Fibrosteps) in patients with chronic hepatitis C infection. LiverInt 2013; 33: 982-990.

23 Gower E, Estes C, BIach S, Razavi-Shearer K, Razavi H. GIobaI epi-demioIogy and genotype distribution of the hepatitis C virus infection. J Hepatol 2014; 61: S45-57.

24 WorId Bank Group. Country and Lending Groups. [OnIine].; 2014 [cited 2014 05 01. Available at: http://data.worIdbank.org/about/ country-and-Iending-groups (accessed

May 2014).

25 WHO. HIV diagnostic tests in Iow-and middIe-income countries: Forecasts of global demand for 20142018. Geneva: WorId HeaIth Organization; 2015.

26 WHO. WHO GuideIines for the Screening, Care and Treatment of Persons with Hepatitis C Infection. Geneva: WorId HeaIth Organization; 2014.

27 Xu K, Evans DB, Kawabata K, Zer-amdini R, KIavus J, Murray CJL. HousehoId catastrophic heaIth expenditure: a muIticountry anaIy-sis. Lancet 2003; 362: 111-117.

28 Wagstaff A, van DoorsIaer E. Catastrophe and impoverishment in paying for heaIth care: with appIications to Vietnam 1993-1998. Health Econ 2003; 12: 921-934.

29 Liu S, ChoIIet D. Price and Income EIasticity of the Demand for HeaIth Insurance and HeaIth Care Services: A CriticaI Review of the Literature. Washington DC: Mathematica PoI-icy Research, Inc., 2006.

30 WHO. Database on procurement of HIV and hepatitis products. [OnIine].; 2015 [cited 2015 02 09. Available at: http://www.who.int/hiv/pub/amds/ hiv-hep-procurement-database/en/ (accessed February 2015).

31 Lok AS, Grdiner DF, Lawitz E. Pre-Iiminary study of two antiviraI agents for hepatitis C genotype 1. N Engl J Med 2012; 366(3): 216-224.

32 Jacobson IM, Gordon SC, KowdIey KV, Yoshida EM. Sofosbuvir for hepatitis C genotype 2 or 3 in patients without treatment options. N Engl J Med 2013; 368(20): 1867-1877.

33 Messina JP, Humpherys I, FIaxman A et al. GIobaI distribution and prevaIence of hepatitis C virus genotypes. Hepatology 2015; 61: 77-87.

34 WHO. GIobaI Price Reporting Mechanism Database. [OnIine].; 2014 [cited 2014 05 01. AvaiIabIe at: http://apps.who.int/hiv/amds/price/ hdd/ (accessed May 2014).

35 Kruk ME. Emergency preparedness and pubIic heaIth systems Iessons for deveIoping countries. Am J Prev Med 2008; 34(6): 529-534.

36 HiII A, Khoo S, Fortunak J, Simmons B, Ford N. Minimum costs for producing Hepatitis C direct-acting antiviraIs for use in Iarge-scaIe treatment access programs in deveI-oping countries. Clin Infect Dis 2014; 58(7): 928-936.

37 GiIead Sciences. Hepatitis B and C Treatment Expansion Fact Sheet. [Online].; 2015 [cited 2015. Avail-abIe at: http://www.giIead.com/re-sponsibiIity/deveIoping-worId-access/ viraI%20hepatitis (accessed March 2015).

38 Jennifer K, Adam W, Eric L, UNAIDS. Financing the Response to HIV in Low- and MiddIe-Income Countries: InternationaI Assistance from Donor Governments in 2013. CaIifornia: UNAIDS & The Henry J. Kaiser Family Foundation; 2014.

39 Jayasekera CR, Barry M, Roberts LR, Nguyen MH. Treating hepatitis C in

lover-income countries. N Engl J Med 2014; 370(20): 1869-1871.

40 Obach D, Yazdanpanah Y, Esmat G et al. How to optimize hepatitis C virus treatment impact on life years saved in resource-constrained countries. Hepatology 2015; 62(1): 31-39.

41 Hagan L, Sulkowski MS, Schinazi RF. Cost analysis of Sofosbuvir/Ribavirin Versus Sofosbuvir/Simeprevir for genotype 1 Hepatitis C virus in inter-feron-ineligible/intolerant individuals. Hepatology 2014; 60: 37-45.

42 Thein HH, Yi Q, Dore GJ, Krahn MD. Estimation of stage-specific fibrosis progression rates in chronic hepatitis c virus infection: a meta-analysis and meta-regression. Hepa-tology 2008; 48: 418-431.

43 Coffin PO, Dcott JD, Golden MR, Sullivan SD. Cost-effectiveness and population outcomes of general population screening for hepatitis C. Clin Infect Dis 2012; 54: 1259-1271.

44 Dienstag JL, Ghany MG, Morgan TR, Di Bisceglie AM, Bonkovsky HL, Kim HY. A prospective study of the rate of progression in compensated, histo-logically advanced chronic hepatitis C. Hepatology 2011; 54: 396-405.

45 Bruno S, Zuin M, Crosignani A et al. Predicting mortality risk in patients with compensated HCV-induced cirrhosis: a long-term prospective study. Am J Gastroen-terol 2009; 104: 1147-1158.

46 Liu S, Cipriano LE, Holodniy M, Owens DK, Goldhaber-Fiebert JD. New protease inhibitors for the treatment of chronoic hepatitis C: a cost-effectiveness analysis. Ann Intern Med 2012; 156: 279-290.

47 Fattovich G, Giustina G, Degos F et al. Morbidity and mortality in compensated cirrhosis type C: a retrospective follow-up study of 384 patients. Gastroenterology 1997; 112: 463-472.

48 Sulkowski M, Gerdiner D, Rodriguez-Torres M et al. Daclatasvir plus Sofosbuvir for Previously Treated or Untreated Chronic HCV Infection. N Engl J Med 2014; 370(3): 211-221.

49 UNITAID. Hepatitis C Medicines and Diagnostics in the Context of HIV/ HCV Co-Infection: A Scoping Report. Geneva: UNITAID; 2013.

SUPPORTING INFORMATION

Additional Supporting Information may be found in the online version of this article:

Figure S1. Complete outcomes of model.

Table S1. Country groups and genotype distribution.

Table S2. Average treatment prices at 2014 (in US$).

Table S3. Average cost of treatment per person.

Table S4. Complete model outcomes for low-income countries.

Table S5. Complete model outcomes for Iower-middle-income countries.

Table S6. Complete model outcomes for upper-middle-income countries.

Table S7. Monte-Carlo simulation

results for low-income countries.

Table S8. Monte-Carlo simulation results for lower-middle-income countries.

Table S9. Monte-Carlo simulation results for upper-middle-income countries.