Economic impact of generic antiretrovirals in France for HIV patients’ care: a simulation between 2019 and 2023

According to the World Health Organization, 36.5 million people worldwide were living with HIV in 2017 [1]. In France, people living with HIV (PHIV) were estimated to be 172,700 in 2016 [2]. Among them, 76% were receiving antiretroviral (ARV) treatments. The most recent figures from compulsory health insurances estimate the direct cost of HIV at €1.3 billion in 2017 [3]. ARV treatments accounted for 71% of these costs. Over the past 20 years, HIV has become a chronic disease due to successful HIV therapeutics [4]. The economic burden of HIV care will increase over the coming years because PHIVs on active antiretroviral regimens are now expected to have a close to normal lifespan [5, 6].

The European Medicines Agency (EMA) defines a generic drug as one that contains the same active substance(s) as the reference medicine and is used at the same dose(s) to treat the same disease(s). However, the drug name, the appearance (such as colour or shape) and packaging can be different from those of the reference medicine [7]. In France, a GD is defined as “a copy of an original medicinal drug whereby production and marketing are made possible by the expiry of the patent covering the innovator product” [8]. Additionally, the French Public Health Code defines a GD as “a specialty which is essentially similar and presents the same qualitative and quantitative composition of active ingredients, the same dosage form and bioequivalence as the original product” [8].

The last Association for Accessible Medicines (AAM) report shows that $265 billion were saved in 2017 for all payers thanks to GDs, and the 10-year savings were estimated at $1.19 trillion [9]. A 2014 report by the Intercontinental Marketing Services Institute for Healthcare Informatics shows that GDs provided savings of €100 billion a year for health systems in Europe [10]. Reports from the French Economic Committee for Health Products (CEPS) based on retrospective national databases of drug reimbursement statistics show that between 2010 and 2014, the use of GDs allowed the French health system to save €7 billion [11].

In the coming years, ARV generics launched on the market in 2017 (e.g. [ABC + 3TC]) as well as other molecules coming off patents (e.g., TDF) will provide an opportunity to improve efficiency in the context of resource allocation constraints. Two European studies estimated the five-year economic savings resulting from ARV generics at €187 million in Italy and €1.25 billion in the UK [12, 13]. These studies used population-based methods.

As an alternative to the usual population-based strategies, the agent-based approach aims to pinpoint the natural behaviour of a system from data [14, 15]. This behaviour can then be simulated to gain insight into the future of the system in a low-cost and time-saving manner. The main simulation lines are split into two phases: First, a group of patients (real or virtual patients [16]) is considered. Second, the behaviour of each patient’s outcomes over time and under predefined scenarios is simulated using predictive models commonly referred to as execution models [17]. There is a wide range of applications for agent-based models (ABM) [18], especially in Biology [19] and Economics [20], while they are fewer in the field of medical research [21, 22].

In France, robust real-life data are available to implement ABM to simulate potential savings from the imminent arrival of generic ARVs on the French market [23]. Our hypothesis is that these soon-to-be available generic ARVs will generate significant cost savings for the French National Health Insurance (FNHI), which entirely supports the care costs of PHIVs in the context of the Long-term Diseases scheme. The aim of this study is to estimate the economic impact on the FNHI, of the entry of generic ARVs on the French market from 2019 to 2023, according to different scenarios.

This budget impact model using an ABM, highlights the fact that the potential economic savings between 2019 and 2023 due to the introduction of generic ARVs are significant and mostly driven by the penetration rate of generics. They range from €309 million to €1.5 billion as penetration rates vary between 10 and 70%, from €894 million to €993 million as the time between patent and generic MAD varies between 10 and 15 years, and from €965 million to €993 million as generic NPUs vary between 40 and 50% of patent ARVs NPUs.

Only two studies were found in which models were used to estimate the economic impact of new generic antiretrovirals. Restelli et al. [12] developed a BIM to forecast the rates of use of brand and generic ARVs and their impact on the Italian National Health Service budget from 2015 to 2019. They estimated the five-year economic savings at €187 million. They used expert opinions to drive their model and to develop scenarios according to the introduction of generic drugs or new brand drugs. However, they did not consider the generic and brand drug costs and changes in penetration rates over time, the variability in generic drug marketing authorisation dates, or the patients’ individual characteristics and their ability to switch from one treatment to another several times. Contrary to this method, the one developed in our study puts individuals at the heart of predictions. Instead of learning, predicting, and applying ARV consumption rates to a population, regardless of individual specificities, it rather uses data from medical histories to learn and predict the changes in each individual’s characteristics and treatments. Therefore, this way of assigning treatments over time is closer to the constrained allocation of ARVs observed in real-life management of HIV. Hill et al. [13] estimated the economic impact of generics at €1.25 billion, using a comparison between two scenarios. In their basic case scenario, every patient was assigned to brand-name antiretrovirals while in the second scenario, they all switched to generic drugs. They used a UK Collaborative HIV Cohort database to estimate the proportions of patients taking each drug and the British National Formulary database to estimate drug costs. This study provides deterministic results without a sensitivity analysis. Nevertheless, this study was an oral presentation, and therefore, the limited amount of available information prevents us from making a more thorough comparison with our work.

The agent-based method proposed here in the context of health economics research has four major advantages. First, it makes it possible to integrate many more parameters in the prediction of cost savings, especially individual parameters together with their correlation structures, which makes the predictions more realistic. Second, the effect of time can be examined through longitudinal models. Third, the precision of predictions can be assessed by incorporating randomness into the dynamics of the system. This precision can be evaluated or illustrated by means of bootstrap confidence intervals derived from the distribution of the predictions obtained through several simulation runs. It is a precious tool in this context to compare and identify the main sources of randomness. Fourth, the individual behaviour of patients can be examined and therefore the conclusion can be modulated in terms of population together with individuals. The main drawback is usually modelling and comes from the choice of models and the performance of the calibration of such models. The choice of models is driven by the data and the clinical input on the disease. For this study, we benefited significantly from the help of the Dat’Aids scientific committee, comprised of experts in the management of HIV-infected patients.

This study has several limitations that must be discussed. The time required to clean the Dat’Aids database, design the ABM, and implement it was much more than we had imagined. Consequently, by the time we were able to produce the first results, we were near the end of the period during which we had initially planned to carry out simulations. Therefore, as the ABM was built for predictive purposes, we decided to change the simulation period to 2019–2023. We are aware that doing so with a model trained on 2015 data presents major drawbacks. First, we were unable to consider the ARVs that entered the French market after 2015 as we lacked the necessary data. This could lead to an overestimation of the cost savings as these new patent ARVs will not be genericised during the course of the study. Second, the baseline characteristics and treatments of the simulated patients were taken from the 2015 Dat’Aids data. Overcoming this limitation would entail retraining and running the model on a more recent extraction of the Nadis cohort, which, unfortunately, we were unable to do within a reasonable time frame. However, we were able to perform analyses on a 2019 Nadis extraction, and found that the distribution of the PHIVs’ characteristics did not differ from those of PHIVs included in the 2015 extraction. We also estimated that 18.9% of the PHIVs included in the 2019 Nadis extraction had treatment regimens containing an ARV that entered the French Market between 2016 and 2019. Approximately 83.5% were on an STR containing tenofovir alafenamide (TAF), and this was 15.8% of the PHIVs included in the 2019 Nadis extraction. The proportion of patients that switches from tenofovir disoproxil (TDF) to TAF can be expected to increase in the future as TAF has a lower toxicity than TDF, but we were unable to evaluate the share of patients that would have made that switch by 2023. Still, in light of these findings, using the 2015 PHIV data as a baseline for the simulations that start in 2019 is not likely to affect the model outcomes. In addition, we gathered the exact information on MADs, NPUs, and penetration rates for every generic ARV that entered the French market between 2015 and 2019. To stay as consistent as possible with the study period, the version of ARVs consumed by each PHIV in the cohort was randomly selected between patent and generic product according to the penetration rates of 2019.

We did not consider the changes in the value of the euro across the study period. In fact, drug NPUs in France are negotiated with the government. Therefore, it seemed improbable that they would be impacted by fluctuations in the euro. However, we accounted for changes in drug NPUs by applying a 3.8% yearly discount based on rates observed from 2013 to 2018.

This study only focused on ARV costs. It does not include all resources consumed by PHIVs such as other direct medical and non-medical costs, indirect or informal costs. The results presented here are based on the assumptions underlying the execution models which are detailed in the associated section in supplementary materials. Please note that we did not consider the ability of some patients to break their Single Tablet Regimen (STR) (i.e., switch from a one-pill combination of several medicines to several pills) as this would have resulted in a much higher complexity in the algorithm. Such a consideration ensures that when the generic version of a medicine that is also part of a combination is available, but the combination itself is not, patients are prevented from breaking it up to take the generic. Therefore, resulting cost savings may be underestimated. In addition, we conducted no analysis on the efficacy of generic drugs or their impact on health as we considered both the efficacy and the safety to be similar between brand-name and generic drugs. Walensky et al. [32] and Sweet et al. [33] used simulation models to compare both cost and efficacy between STRs based on a foundation of emtricitabine and tenofovir (FTC/TDF) and their multiple-tablet counterparts, including generics when available and exchanging lamivudine for emtricitabine (e.g., EFV/TDF/FTC vs. generic EFV + TDF + generic lamivudine). Both studies demonstrated a higher efficacy of brand-name STRs compared to generic-based multi-tablet regimens (gMTRs), mainly as a result of poorer adherence to gMTRs than to STRs because of the pill burden and a lower efficacy of lamivudine compared to emtricitabine. However, their cost evaluation differs significantly from ours as only a few brand-name STRs were studied, the focus was on the breaking of STR, and in both studies, the two scenarios that were compared were everyone taking a brand-name STR or everyone taking a gMTR.

In France, the care costs of HIV-infected patients are entirely supported by the FNHI in the context of the Long-Term Diseases scheme. In a context of resource constraints, it is therefore of paramount importance to develop and encourage alternative therapeutic strategies to lower health expenses without lowering the quality of patient care. With an acknowledged similar effectiveness and lower cost compared to their brand-name counterparts, generic drugs meet both the above criteria. This study highlights significant potential savings due to the introduction of new generic ARVs in France, ranging from €309 million to €1.5 billion as the penetration rate of generics varies between 10 and 70%. As shown by the sensitivity analyses, the penetration rate is the most important parameter that drives economic savings. In fact, these savings range from €894 million to €993 million as the time between patent and generic MADs varies between 10 and 15 years, and from €965 million to €993 million as generic NPUs vary between 40 and 50% of patent ARV NPUs. French health care authorities must encourage healthcare professionals to prescribe generic ARVs as well as patient compliance with generics. Such results should be taken into consideration by policy makers and encourage them to keep promoting the use of GDs. As shown in the discussion section, those potential savings remain limited by the increasing number of STRs being developed by pharmaceutical firms. In fact, although this could provide more savings, breaking STRs to be able to benefit from generic ARVs could result in poorer adherence to treatment [34]. A possible option to address this issue would be to scale the NPUs of STRs containing ARVs with already available generics based on these GD NPUs. These estimates of savings may also help decision makers to anticipate future choices for the funding of secondary prevention, specifically to fund the Pre-Exposure Prophylaxis (PrEP) [35‐37] or for the early screening of hidden HIV population such as migrants for example. They could also be used to provide headway for innovation, giving health care payers the possibility to reimburse innovative and expensive new medicines, especially in the field of infectious disease.