Antiretroviral treatment-based cost saving interventions may offset expenses for new patients and earlier treatment start

Authors


Abstract

Objectives

Combination antiretroviral therapy (cART) has become the main driver of total costs of caring for persons living with HIV (PLHIV). The present study estimated the short/medium-term cost trends in response to the recent evolution of national guidelines and regional therapeutic protocols for cART in Italy.

Methods

We developed a deterministic mathematical model that was calibrated using epidemic data for Lazio, a region located in central Italy with about six million inhabitants.

Results

In the Base Case Scenario, the estimated number of PLHIV in the Lazio region increased over the period 2012–2016 from 14 414 to 17 179. Over the same period, the average projected annual cost for treating the HIV-infected population was €147.0 million. An earlier cART initiation resulted in a rise of 2.3% in the average estimated annual cost, whereas an increase from 27% to 50% in the proportion of naïve subjects starting cART with a nonnucleoside reverse transcriptase inhibitor (NNRTI)-based regimen resulted in a reduction of 0.3%. Simplification strategies based on NNRTIs co-formulated in a single tablet regimen and protease inhibitor/ritonavir-boosted monotherapy produced an overall reduction in average annual costs of 1.5%. A further average saving of 3.3% resulted from the introduction of generic antiretroviral drugs.

Conclusions

In the medium term, cost saving interventions could finance the increase in costs resulting from the inertial growth in the number of patients requiring treatment and from the earlier treatment initiation recommended in recent guidelines.

Introduction

The widespread adoption in clinical practice of combination antiretroviral therapy (cART) has produced a substantial increase in the survival of persons living with HIV (PLHIV) [1, 2], and has also provided good value for money as shown in a number of cost-effectiveness studies [3, 4].

Longer survival has resulted in a rise in lifetime costs at the individual level [5, 6], as well as in the total annual costs of treating an HIV-infected population [7]. Costs for antiretroviral (ARV) drugs have also increased, to the point that they account for most (about 70%) of the overall direct costing budget [8-10]. Therefore, clinical prescription practices regarding the use of ARV drugs have become crucial determinants of the total costs for treating PLHIV in resource-rich countries [11].

Accumulated evidence suggests that starting treatment earlier in the course of HIV disease may reduce AIDS- and non-AIDS-associated morbidity and mortality [12, 13], and may decrease the risk of sexual transmission of infection, with a potential impact on the future course of the epidemic [14]. Based on this evidence, several guidelines now recommend also offering cART to patients with a lower level of immunosuppression [15, 16], and the strategy of treating all individuals diagnosed with HIV infection, regardless of CD4 cell count, has recently gained momentum [17]. The sustainability for health care systems, even in resource-rich countries, of the increasing cost of care that may be associated in the short to medium term with these changes in the policy of ARV use is yet to be investigated in detail.

A series of treatment strategies, including simplification to nonnucleoside reverse transcriptase inhibitors (NNRTIs) co-formulated in a single tablet regimen (STR) and to protease inhibitor/ritonavir-boosted (PI/r) monotherapy, have been proposed to curb the increase in treatment costs while retaining individual benefits [18-20], and some local health authorities in charge of health care budgets have already issued recommendations to reconcile national guidelines with the objective of ensuring affordable and high-quality care for PLHIV [21].

The objective of the present study was to analyse short/medium-term cost trends in response to the evolution of clinical practices in HIV care, in order to provide data useful for health policy makers in optimizing the allocation of scarce resources for managing HIV disease. To this end, we examined, using a mathematical model, the impact [22] of guidelines issued in 2011 on the costs of treating an HIV-infected population over the period 2012–2016 in an Italian region, with a focus on costs of ARV drugs.

Methods

Context

In Italy, care for PLHIV, including cART, is provided free of charge by the Regional Health Systems. Guidelines for the use of ARVs are issued periodically at a national level. In the 2010 revision, cART initiation was strongly recommended for individuals with a CD4 count < 350 cells/μL, while it was moderately recommended in selected conditions in those with a CD4 count between 350 and 500 cells/μL and not recommended in those with a CD4 count > 500 cells/μL. The 2011 guidelines expanded the indication for starting cART, which was strongly recommended for all PLHIV with a CD4 count < 500 cells/μL and, in selected conditions, also for those with a CD4 count above this threshold. cART initiation was also moderately recommended in the case of an elevated risk of secondary HIV transmission. Different regimens were also recommended for naïve and experienced patients [23]. Shortly afterwards, some regions issued local therapeutic protocols recommending the choice of less costly regimens among those identified as clinically equivalent in the national guidelines [24].

Model overview

To explore the impact of new treatment guidelines on the costs of treating a population of HIV-infected individuals, we used a deterministic mathematical model implemented in Microsoft Office Excel 2007. As usual in simulation studies, the assumptions on which the model is based determine the reliability of the conclusions reached. The model is structured in three stages that simulate the phases in the therapeutic path of the patients (Fig. 1) and are associated with different patterns of resource use. Transitions between stages were assumed to happen uniformly over time with a cycle length of 1 year.

Figure 1.

Diagram of the model structure. The figure displays the stages of the model that simulate the phases in the therapeutic path of HIV-infected individuals. cART, combination antiretroviral therapy.

We assumed that the infection is transmitted essentially through sexual behaviour, a mode of acquisition that at the moment accounts for nearly 80% of new HIV diagnoses in Italy [25]. Different rates of transmission (new infections per 100 infected individuals) were computed for each stage of the model according to the approach proposed by Marks et al. [26], with a lower rate for treated subjects to model the prevention effect of cART.

Newly infected subjects entered the model in the undiagnosed stage. At diagnosis, they progressed to the naïve stage and were classified as being in one of five states defined on the basis of the CD4 cell count (<200, 200–349, 350–499, 500–749 and ≥750 cells/μL), with changes of state thereafter regulated by transition probabilities.

One-year probabilities of cART initiation according to the level of CD4 count determined progression to the treated stage, where subjects remained until death. Patients could receive up to six sequential lines of cART and switch rates were based on specific average durabilities derived from the literature [6].

Model calibration

The epidemic model was calibrated using data for the period 2007–2010 for Lazio, a region with about six million inhabitants that is located in central Italy and accounts for 13–14% of the Italian cases reported to the National AIDS Surveillance System [25].

The number of individuals receiving care for HIV infection and their distribution according to age and gender at the beginning of 2007 were obtained from a survey conducted at Infectious Diseases Centers in Lazio (unpublished report from the Italian program for the evaluation of health care services for HIV/AIDS, 2007). Numbers and characteristics of individuals unaware of being HIV-infected at 1 January 2007 were estimated using a method proposed by the Working Group on Estimation of HIV Prevalence in Europe [27] and based on surveillance of simultaneous HIV/AIDS cases [28].

The Lazio Surveillance System provided the number of new diagnoses for the years 2007–2010, and the distribution of diagnoses by CD4 count was obtained from a regional multicentre study [29]. In the hypothesis of no change in the screening and testing guidelines, we estimated newly diagnosed subjects for 2012–2016 by projecting the number of new diagnoses per 100 individuals unaware of their infection and multiplying it by the number of undiagnosed subjects at the beginning of each year of the studied period.

The results of a biomarker test for recent infection, performed on a subgroup of newly diagnosed individuals, provided the basis for estimating the HIV incidence for the years 2007–2010 [30]. For 2011 onwards, incident cases were estimated by applying the transmission rates computed for the previous year to the number of infected individuals at the start of the year for each subpopulation.

For naïve individuals, transition probabilities between CD4 strata and CD4-specific probabilities of starting cART were estimated using data from an Italian cohort of HIV-infected patients who were antiretroviral-naïve at the time of enrolment [31].

Mortality rate ratios for PLHIV relative to the general population [32, 33], stratified for naïve and treated subjects, were applied to age- and gender-specific mortality rates for the general population of the Lazio region provided by the Italian Census Bureau to compute the mortality rates used in the model.

Model parameters and assumptions for the Base Case Scenario (BCS) are reported in Table 1.

Table 1. Model parameters and assumptions in the Base Case Scenario for the year 2011
VariableValueReferences
  1. cART, combination antiretroviral therapy; ARV, antiretroviral.
  2. *At-risk sex partners are those who are HIV-negative or of unknown serostatus. Subjects with viral load < 500 copies/ml.
HIV disease transmission and screening  
Relative risk of unprotected sex with at-risk* sex partners (aware compared with unaware group) (%)43.0[26]
Ratio of number of at-risk* sex partners (aware compared with unaware group)66.7[26]
New diagnoses as proportion of undiagnosed subjects (%)40.9Estimate
Relative risk of transmission for stably suppressed compared with untreated subjects (%)10.0[14]
CD4 count at HIV diagnosis (cells/μL) [median (interquartile range)]334 (124–529)[29]
cART treatment  
1-year probability of starting cART (%)45.4[31]
cART durability (years) [6]
First line6.7 
Second line4.3 
Third line4.2 
Fourth line4.0 
Fifth line2.6 
Proportion of stably suppressed treated subjects (%)70.0[50]
Average proportion of the year covered by ARV drug prescriptions (%)90.0[51]
Costs (2011 €)  
Average monthly cART costs746[34]
Average yearly costs per diagnosed subject [8]
Out-patient and laboratory tests1488 
Non-ARV drugs486 
Hospital admissions982 

Cost and utilization of health resources

We considered costs related to hospital admissions, ARV and non-ARV drugs, out-patient visits and laboratory tests from the diagnosis of infection onwards, as we assumed that the health resources and services used by PLHIV as a consequence of HIV infection prior to its diagnosis were negligible. Costs were computed from the perspective of the Italian National Health System (INHS) and reported in euro currency at 2011 unit costs.

The average annual cost for each cART regimen was computed on the basis of present standard daily dosages and unit prices of ARVs for hospitals of the INHS (see Supplementary Table S1). Then, based on the distribution of cART regimens resulting from the regional database of ARV drug use and from a survey conducted at the ‘Lazzaro Spallanzani’ National Institute for Infectious Diseases in Rome [34], we could compute the average cost for each line of treatment. We assumed that 90% of the year was covered by drug prescription to take into account patients dropping out of care and suboptimal treatment adherence.

For non-ARV costs, we used the average annual cost per patient for the year 2007 reported in an Italian study [8], which was inflation-adjusted to 2011 prices using the Consumer Price Index for health expenditure, and assumed to be constant over the period studied.

Budget impact analyses

The BCS represented the inertial scenario not taking into account the changes in national guidelines and local therapeutic protocols. Starting from 2012, we manipulated parameters in the model to investigate, first, the financial consequences of Italian 2011 recommendations concerning when to start cART. Secondly, we assessed the impact of selecting the least costly cART regimens among those recommended by the Lazio therapeutic protocol for naïve subjects. We also evaluated the potential savings related to treatment simplification strategies, the aim of which is to improve quality of life and adherence by reducing the pill burden or to decrease toxicities by sparing drugs. In particular, we simulated the switch from PI/r-based triple regimens to Atripla® (Bristol-Myers Squibb and Gilead Sciences Limited, Carrigtohill, Ireland) (efavirenz/emtricitabine/tenofovir) and to PI/r monotherapy. Finally, we simulated the impact on costs of the future availability of generic ARV drugs.

Results

A total of 14 414 individuals were estimated to be living with HIV at the start of the simulation (31 December 2011) in the Lazio region, of whom 1660 (11.5%) were estimated to be unaware of being infected (Table 2). Of the total population of PLHIV, 72.3% were male and nearly 40% were between 40 and 50 years old.

Table 2. Projected characteristics of the HIV-positive population and costs of care in the Lazio region in the Base Case Scenario for 2011–2016
 201120122013201420152016Average annual increase rate
  1. cART, combination antiretroviral therapy; ARV, antiretroviral.
  2. *New infections per 100 persons living with HIV.
HIV-infected subjects (n) (December 31)14 41414 97015 52516 07916 63117 1793.6%
On cART treatment (%)78.579.079.479.880.180.4 
Diagnosed and naïve (%)10.09.79.49.18.98.7 
Undiagnosed (%)11.511.311.211.111.010.9 
Female (%)27.727.627.527.427.327.2 
Distribution according to age (%)       
15–29 years11.611.511.511.411.411.4 
30–39 years24.823.422.121.120.219.5 
40–49 years37.136.135.133.932.831.7 
50–64 years22.123.925.627.128.429.4 
≥ 65 years4.45.05.76.57.28.0 
Prevalence (×100 000)247.5254.2260.7267.1273.2279.22.4%
New HIV infections (n)7277377477587687791.4%
Incidence rate (×100 000)12.612.612.712.712.712.80.3%
Transmission rate* (%)5.145.024.904.794.704.61−2.2%
HIV diagnoses (n)6736796866947027111.1%
Subjects starting cART (n)6566566586626686750.6%
Mortality rate (×1000)12.012.312.612.913.213.6 
Cost per diagnosed subject (€)10 07710 22810 35910 47310 57210 6611.1%
Total cost of care (million €)125.9133.1140.2147.1153.9160.55.0%
Components of annual cost of care (%)       
ARV drugs70.871.171.471.771.972.1 
Non-ARV drugs4.84.74.74.64.64.6 
Hospital admissions9.79.69.59.49.39.2 
Out-patient and laboratory tests14.714.614.414.314.214.1 

Base Case Scenario

In the BCS (Table 2), the estimated number of new infections showed a slightly increasing trend, from 727 in 2011 to 779 in 2016, and remained higher than the number of new diagnoses over the whole study period.

The number of PLHIV increased by 2765 individuals, while the subpopulation of cART-treated subjects increased by 2504 individuals, with an average annual increase rate (AAIR), respectively, of 3.6 and 4.1%. By the end of the study period, we estimated that 80.4% of the total HIV-infected population and 90.2% of diagnosed individuals were under treatment, compared with 78.5 and 88.7%, respectively, in 2011. The projected transmission rate declined from 5.14 to 4.61%.

The projected yearly cost of care of the HIV-infected population rose from €125.9 to €160.5 million (AAIR 5.0%) (Table 2). The increase in the estimated number of diagnosed individuals accounted for 74.6% of the growth in total costs, while the rise in the projected yearly cost per subject, from €10 077 to €10 661 (+5.8%) (Table 2), accounted for the remainder. The latter was a result of the rise in the proportion of treated subjects, as well as of the shift towards more costly cART combinations as they moved through subsequent lines of treatment.

The distribution of the components of projected cost per diagnosed subject was stable over the studied period, with ARV drugs accounting for an average of 71.5% of the cost (€7434), followed by out-patient and laboratory costs at 14.4%, hospital admissions at 9.4% and non-ARV drugs at 4.7%.

Early initiation of cART

To simulate expanded cART initiation criteria, as recommended by the 2011 Italian guidelines, we increased the 1-year probabilities of starting treatment to 85, 80, 30 and 20%, respectively, for CD4 count strata 201–350, 351–500, 501–750 and >750 cells/μL. As a consequence, the average 1-year probability of starting cART for naïve individuals in 2012 rose to 65.3% from 45.4% in the BCS, with 14 262 subjects estimated as being under treatment by the end of 2016, compared with 13 812 in the BCS (93.2% of diagnosed individuals vs. 90.2% in the BCS). At the same time, the estimated number of PLHIV declined from 17 179 to 17 152 individuals in 2016, as a result of the effect of earlier treatment initiation on transmission rates (4.55% in 2016 compared with 4.61% in the BCS).

The resulting economic impact was a rise of +2.3% in the average annual cost, from €147.0 to €150.3 million (Table 3).

Table 3. Projected average yearly costs of care for the HIV-infected population in the Lazio region, 2012–2016a
  1. ARV, antiretroviral; cART, combination antiretroviral therapy; NNRTI, nonnucleoside reverse transcriptase inhibitor; PI/r, protease inhibitor/ritonavir-boosted; STR, single tablet regimen.
  2. aEach scenario is evaluated separately and compared with the Base Case Scenario.
  3. bSimplification strategies were applied to stably suppressed subjects on first and second lines of treatment.
  4. cTenofovir,emtricitabine and efavirenz (Atripla®).
Base Case Scenario (million €)147.0  
Sensitivity analyses – million € (% change compared with Base Case Scenario)
Expanded cART initiation criteria150.3 (+2.3)  
 Proportion of naïve patients starting cART with an NNRTI-based regimen
 50.0%70.0% 
Choice of regimens for naïve subjects146.5 (−0.3)146.0 (−0.6) 
Simplification strategies from PI/r regimensProportion of subjects switchedb
 20.0%30.0%40.0%
Switch to NNRTI-based STRc146.4 (−0.4)146.1 (−0.6)145.9 (−0.7)
Switch to PI/r monotherapy146.0 (−0.6)145.6 (−0.9)145.1 (−1.3)
Use of generic ARVsProportion of branded ARVs substituted
 20.0%50.0%90.0%
30% discount compared with branded ARV146.0 (−0.6)144.6 (−1.6)142.7 (−2.9)
50% discount compared with branded ARV145.5 (−1.0)143.3 (−2.5)140.3 (−4.5)
70% discount compared with branded ARV145.0 (−1.4)141.9 (−3.4)139.9 (−6.1)

Cost-saving treatment strategies

With the objective of standardizing therapeutic decisions and reconciling their appropriateness with economic sustainability, Lazio Health Authorities, in June 2011, developed recommendations regarding the choice of cART regimen for naïve subjects.

For subjects starting cART, the choice of the least expensive alternative among those indicated as preferred by national guidelines was recommended. Based on administrative data collected in our region, we assumed in the BCS that 27% of HIV-infected naïve individuals would start cART with NNRTI-based regimens. Increasing this proportion to 50% or 70% from 2012 onwards, with a concomitant reduction in the proportion of patients starting PI/r-based regimens, resulted in a change in the estimated average yearly cost of care compared with the BCS of −0.3 and −0.6%, respectively (−€0.5 and −€1.0 million, respectively) (Table 3).

The guidelines also indicated which of the simplification strategies for long-term management of treated patients have the potential to reduce costs. Therefore, we modelled the switch in 2012 to STR or to PI/r monotherapy of a proportion ranging from 20 to 40% of stably suppressed subjects on first and second lines of treatment with a PI/r-based triple regimen. Simplification to STR meant a change in the estimated average yearly cost of care of between −0.4 and −0.7% (between −€0.6 and −€1.1 million), whereas the switch to monotherapy resulted in savings ranging from −0.6 to −1.3% (Table 3).

Cost impact of generic ARVs

Drugs with patent expiration dates within the studied period were nevirapine, lamivudine/zidovudine (Combivir® – ViiV Healthcare UK Limited, Brentford, Middlesex, UK) and efavirenz in 2013, and abacavir, lopinavir/ritonavir (Kaletra®, AbbVie Ltd, Maidenhead, UK) and emtricitabine, respectively in 2014, 2015 and 2016. For these drugs, including those present in branded co-formulations, we assumed rates of substitution by generic ARVs from 20 to 90% and discount rates of between 30 and 70%. Compared with the BCS, the estimated change in average yearly cost ranged from −0.6% (−€0.9 million) to −6.1% (−€9.0 million) (Table 3).

Finally, we simulated a new scenario combining the previous sensitivity analyses (Fig. 2). In addition to the expanded cART initiation criteria, we made the hypothesis that 50% of naïve subjects started cART with an NNRTI-based regimen; 30% of stably suppressed subjects on first and second lines of treatment with PI/r-based regimens in 2012 were switched to STR and 30% to monotherapy; 90% of non-co-formulated and 50% of co-formulated branded ARVs with expiring patents were substituted with generic drugs at a 50% discount. Despite the increase in costs of +2.3% attributable to early cART initiation, cost-saving measures resulted in a reduction of 3.1% in the projected annual cost of care (−€4.6 million) compared with the BCS.

Figure 2.

Impact of changes in treatment guidelines and of generic ARV drugs on average yearly costs of care in the Base Case Scenario, 2012–2016 (million €). cART, combination antiretroviral therapy; BCS, Base Case Scenario; NNRTI, nonnucleoside reverse transcriptase inhibitor; PI/r, protease inhibitor/ritonavir-boosted; STR, single tablet regimen.

a cART initiation strongly recommented for CD4 cell count <500 cells.

b 50.0% of naïve patients starting NNRTI-based regimens.

c Switch of 30% of suppressed subjects on first and second lines of treatment with Pl/r-based triple regimen to STR.

d Switch of 30% of suppressed subjects on first and second lines of treatment with Pl/r-based triple regimen to Pl/r monotherapy.

e 50% discount on branded drugs with 50% of substitution for co-formulated and 90% for non-co-formulated antiretroviral drugs.

Discussion

By using a simulation model based on data from an Italian region, we explored the financial impact of clinically and economically validated interventions in the context of projected epidemiological trends of the HIV epidemic.

By the end of 2011, we estimated that 88.7% of diagnosed PLHIV received cART, similar to the 88.2% reported for France [35] and 83.5% for the UK [36]. However, the estimated proportion of PLHIV who are unaware of their infection (11.5%) was lower than figures for France and the UK [37]. This is probably the result of the estimation method, based on simultaneous HIV/AIDS cases, which seems to produce estimates of PLHIV who are unaware of their infection substantially lower than methods based on prevalence surveys [27, 28, 38].

In our BCS, the total cost of treating the HIV-infected population of the Lazio region was estimated to grow over the period 2012–2016 at an annual rate of 5.0% as a result of the increases in the number of diagnosed PLHIV and in the cost per diagnosed subject (respectively +3.6% and +1.1% annually).

The projected increase in the proportion of naïve subjects starting cART in each year as a consequence of expanded cART initiation criteria (from 45.4% in the BCS to 65.3%) resulted in a rise of +2.3% in average annual costs over the studied period. Moreover, the time horizon of our model was too limited to observe appreciable cost savings related to reduction in transmission (just 27 infections averted at the end of the period). The impact of earlier treatment initiation on the medium-term budget, as well on HIV transmission, could be greater in the presence of a contemporary expansion in screening programmes [39].

Without taking into account the effects of cART on HIV transmission and using a model-based approach, Mauskopf et al. [40] found that starting cART with a CD4 count > 350 cells/μL compared with < 350 cells/μL increased the average yearly cost of treatment by 6.7%; this result, however, was sensitive to the sequence of cART regimens used in the simulation. With a similar approach, a more recent study [6] reported an increase in yearly costs ranging from 1.9 to 2.7% depending on the distribution of CD4 cell count at diagnosis.

Despite lower costs, NNRTI-based regimens have shown similar effectiveness compared with PI/r-based regimens as first-line cART [41]. Consequently, an increase in the proportion of naïve subjects starting treatment with NNRTI-based regimens, from 27% reported for the Lazio region in 2011 to 50 or 70%, consistent with other experiences [21], represents an option for cost containment.

STR and PI/r-based monotherapy have been proposed as the most promising [18, 20, 42] treatment simplification strategies in terms of clinical effectiveness and costs.

At present, uncertainties exist about the number of patients who could be eligible for such strategies and consequently about their financial impact. In a study for Germany [43], a saving of between 10.9 and 19.8% in the costs of ARV drugs could be achieved, in the hypothesis of switching, respectively, 50 and 90% of subjects on first-line treatment with PI/r, to NNRTI-based regimens. For the UK, Gazzard et al. [44] assumed that 40% of currently treated patients (about 18 000 individuals) would be eligible for the switch to monotherapy, with the potential to save up to £60 million per year. The main limiting factors for switching to the above-mentioned simplification regimes are contraindications to STR or lack of selection criteria for PI/r monotherapy [45-47].

We simulated the switching of at least 20% and up to 40% of virologically suppressed patients on first and second lines of treatment with PI/r-based regimens to both STR and PI/r monotherapy, corresponding to an overall proportion of treated PLHIV of between 4.7 and 9.5%.

Potential savings related to the introduction of ARV generics could be substantial. Making the assumption of a range of price reductions for generic drugs of between 30 and 70% and a rate of substitution of between 20 and 90%, we estimated an impact on average annual costs in the 2012–2016 period of between -0.6 and -6.1%, and between -0.9 and -8.6% when we take into account only ARV costs. These figures are lower than those reported by Stoll et al. for Germany [43] and by Walensky et al. for USA [48], as a result of a different mix of regimens at baseline.

However, there are questions concerning the possible increase in the pill burden for the patient as a result of the unbundling of branded co-formulated drugs [49]. Furthermore, pharmaceutical companies could introduce new drugs to replace those with an expiring patent. However, the substitution of branded with generic drugs, as well as the substitution of PIs with NNRTIs, could lead to greater competition in the ARV market, with the possibility of further reductions in prices. For this reason, in the combined scenario (Fig. 2), we assumed a 50% discount on generic ARVs, with different rates of substitution for co-formulated and non-co-formulated branded drugs of 50 and 90%, respectively.

Our study has several limitations, mainly as a result of the uncertainty associated with the assumptions used in the simulation model. First, we made the simplifying assumption of treating people living in the Lazio region as a closed population. Average non-ARV costs per subject used in the model were for a population with similar demographic characteristics, but we cannot exclude the possibility of differences in the type and amount of resources used. Moreover, we assumed these costs to be constant over the studied period. We did not take into account the introduction of new ARV agents, which could result in an underestimation of growth in ARV costs. Finally, the clinical data and the mix of ARV regimens used in the model reflected the situation in the Lazio region, which may be not representative of other settings.

However, some general conclusions can be drawn. The projected inertial growth rate of costs over the studied period was mainly driven by the increase in the number of PLHIV. Moreover, the additional impact on this trend of earlier cART initiation could be more than balanced by cost-saving treatment strategies and above all by the introduction of generic ARV drugs. Therefore, savings associated with these measures could finance the increase in costs consequent not only to earlier treatment initiation, but also to the inertial growth in the number of patients requiring treatment.

Acknowledgements

This work was funded by the AIDS Project of the Italian Ministry of Health (Grant 40H79), Ricerca Corrente INMI Spallanzani and an unrestricted research grant from Gilead Sciences.

Ancillary