FormalPara Key Points for Decision Makers

The introduction of highly active antiretroviral therapy (HAART) was found to be cost-effective in Danish healthcare. Despite the increase in annual costs per patient from pre-HAART to the latter periods, the number of hospital days per patient decreased due to effective treatment.

The study emphasizes the necessity of retrospective economic evaluations for informed decision-making by societies, policymakers, and clinicians.

To bridge the current and prevailing life expectancy gap for people living with human immunodeficiency virus (HIV) and minimize side effects, further advancements in HAART are imperative.

1 Introduction

Globally, human immunodeficiency virus (HIV)/acquired immunodeficiency syndrome (AIDS) has caused over 39 million HIV/AIDS-related deaths to date and more than 36 million people are currently living with HIV [1,2,3].

The first breakthrough HIV drug, the anti-retroviral therapy (ART) azidothymidine marketed in 1987 under the brand name Retrovir, provided temporal improvements in the excess mortality caused by HIV [3, 4]. Being the most expensive drug in the world at the time, however, its use was limited to only symptomatic patients experiencing opportunistic infections [5, 6]. In 1995, protease inhibitors were introduced, and a year later came non-nucleoside reverse transcriptase inhibitors (NNRTI) [7, 8], paving the way for highly active antiretroviral therapy (HAART) [7, 11]. Drug prices remained among the world’s highest, however, and the early HAART period was marked by international debate about patent laws and drug prices, questioning whether the international community accepted too high prices of vital medicine [12, 13].

In this period, countries did not have national health technology assessment (HTA) organizations like today doing appraisals of new pharmaceuticals and judging whether pharmaceutical prices are ‘fair’ or not [14]. The methodology of health economic evaluation with quality-adjusted life-years (QALY) and national threshold values, which is used to perform such appraisals today, was not developed at the time [15]. Instead, there was an expressed need for standardization of economic methods to reduce variation in methods and estimates of costs per patient and cost-effectiveness [16].

Despite existing health economic studies [5, 14, 17,18,19,20,21,22,23,24], the cost-effectiveness of the introduction of HAART remain unclear. To our knowledge, none of the previous studies has conducted an ex-post cost-utility analysis according to international guidelines for health economic evaluation in the same way as analyzes are presented to decision-makers in national HTA bodies today.

Today, more than 30 HIV antiretroviral drugs are available. Modern-day health economic evaluations of the costs and effects of various HAART combinations have demonstrated value for money of many newer improvements in HIV treatment strategies [19, 21, 22]. Nonetheless, these analyses are prospective and aim to determine whether new treatment combinations as incremental improvements compared to already implemented treatments would be cost-effective and financially advantageous in the future. Ex-post (historic) economic evaluations take place after the implementation of the initiatives and are retrospective in format. They can be used to measure impact, policy learning, accountability, and provide feedback on future ex-ante evaluations [25, 26]. Ex-post evaluation allows you to assess the economic consequences of past actions and evaluate the actual costs and the effects on the patients [25]. Thereby we can assess whether the introduction of HAART has been worth the money as we understand cost-effectiveness today.

The Danish public healthcare sector constitutes a useful setting for ex-post health economic evaluations. Since 1968, a unique personal identifier has been assigned to all residents at birth or upon immigration by the Civil Registration System (CRS) [27]. This identifier allows unambiguous data linkage at an individual level and tracks changes in vital status and migration for the entire population with daily updates. In an international comparison, Denmark has a relatively large number of high-quality data files containing information about individuals’ health and healthcare consumption, and a long research tradition due to the possibilities available from linking such information [28]. With a modern public healthcare system providing healthcare free of charge to all citizens, register-based analyses have been shown to provide unbiased cost estimates [28, 29].

The objective of this study was to estimate the costs and cost-effectiveness of introducing HAART in Denmark based on real-world evidence for the three treatment eras pre-HAART (1985–1995), early HAART (1996–2005), and late HAART (2006–2017). The primary measure of health effectiveness was QALY, and secondary outcome was survival in years after diagnosis.

2 Methods

We conducted a cohort study using Danish clinical and administrative registries and methods described in Jespersen et al (2021) and Ehlers et al (2022) [30, 31]. The results from the cohort study were used as input in the cost-effectiveness analysis in accordance with international guidelines for health economic evaluations [15, 32, 33]. In the health economic evaluation, we compare the costs, mortality, and QALYs of the average patient diagnosed and treated in each of the three eras using real-world evidence. The results are presented as the incremental cost-effectiveness ratio (ICER) from pre-HAART to early HAART and from early HAART to late HAART.

2.1 Study Sample and Setting

In Denmark, the National Health Service is tax-funded, offering all citizens free access to hospitals and outpatient specialty clinics as well as free hospital-dispensed medicines for all citizens [29]. HIV medicine is given to patients in need of care without any co-payment.

We used real-world evidence (RWE) from a study sample identical to the population in a previous study by Jespersen et al. [30]. The InfCare HIV database, the Danish National Patient Registry (DNPR), and the Danish HIV Cohort Study (DHCS) were used to identify PLHIV in the Central Denmark Region above the age of 18 with a confirmed first diagnosis of HIV during the years 1985–2017. The population is described in detail elsewhere [30]. In brief, 1043 PLHIV were included (Table 1). Overall, the median age at diagnosis was 36.6 years (IQR 29.3–45.5) differing from 33 years in pre-HAART to 39.5 years in late HAART. A higher proportion of PLHIV in late HAART was born in Western Europe compared to pre- and early HAART and, overall, 71.6% were of the male gender.

Table 1 Patient characteristics of the population from the Central Denmark Region in the three treatment eras of pre-HAART (1985–1995), early HAART (1996–2005), and late HAART (2006–2017)
Full size table

2.2 Estimating Survival Using Real-World Evidence

Utilizing the CRS, DNPR, DHCS, and the InfCare HIV database, cumulative survival in the first 9 years following diagnosis was calculated for each treatment era similar to Jespersen et al. [30]. Lifetime parametric extrapolation of survival probabilities was conducted in Stata (version SE17.0) using published extrapolation techniques [34,35,36] and calibrated to match long-term survival for PLHIV from the three eras based on published literature [10, 37,38,39,40,41,42] and estimations from a clinical expert (Supplementary Material S1). Choice of distribution was made by statistical verification using the Akaike information criterion (AIC), Bayesian information criterion (BIC), published literature [9, 10, 37, 39, 41], and visual inspection by Danish clinical experts. Ultimately, the exponential distribution was chosen for pre-HAART and the Weibull distribution for early and late HAART. Furthermore, the survival probabilities were adjusted using the relative survival method to account for the improvement over the whole period from 1985 to 2017 in life expectancy in the gender-matched Danish population [43]. The base case includes adjusted life expectancy but results without adjustment are also shown as well as a short time horizon reflecting only real-world evidence (RWE) without lifetime extrapolation.

2.3 Estimating Costs Using Real-World Evidence

The cost data and methods were previously reported by Ehlers et al. (2022) for late HAART [31] but extended to include the whole period from 1985 to 2017 with a healthcare sector perspective. In brief, the patient’s personal identifier (the CPR number) was used to link patients with information on healthcare consumption from the Danish National Patient Registry (DNPR), the Danish National Prescription Registry, and the National Health Insurance Service Register. Annual cost per patient was estimated for each of the three eras for the index year, incremental years, and the last year of life. Costs were adjusted to 2017 and converted from Danish Kroner (DKK) to Euro (€) following Statistics Denmark and the consumer price index [44]. Discounting of 3.5% for the first 35 years and 2.5% for the following years were applied in the base case according to Danish guidelines [45, 46], and analysis without discounting was also conducted.

Contacts with Danish hospitals are registered in the National Patient Register with information on the length of stay as well as unique linkages to individual-level information of the diagnosis-related group (DRG) and relevant Danish reimbursement tariffs designed to equal the average cost per Danish patient [28, 47]. Danish DRG tariffs include the costs of free hospital-dispensed medicines such as ART, but prices of hospital medicine are confidential, thus, cannot be estimated individually. The Danish National Prescription Registry was used to estimate expenditures of prescription medicine for selected noncommunicable diseases at the individual patient level using the pharmacy selling prices, however, only a subsample of the most relevant prescription drugs was included. All costs associated with primary care services including general practitioners and physiotherapy were obtained through the National Health Service Register.

For pre-HAART, Danish registers did not contain any information on unit costs, but only information on hospital days and visits. To estimate unit costs we assumed the same relative change in unit costs from pre-HAART to early HAART as reported by Beck for UK [19].

2.4 Quality of Life Estimates

Quality of life was estimated on the basis of the World Health Organization (WHO) Global Burden of Disease (GBD) study [48] using disability-adjusted life-years (DALY) disability weights as estimates of disutility. For each of the health states in the model, index year, incremental years, and last year of life, Danish clinical experts were consulted to help choose the relevant disutility values for PLHIV in each of the three periods, pre-HAART, early HAART, and late HAART. Disutility values were subtracted from a baseline utility from the Danish population norms to reflect the quality of life for Danish PLHIV [49]. QALYs were discounted similarly to costs according to Danish guidelines.

2.5 Comparing Cost, Survival, and QALY

We constructed a simple decision-analytic Markov model to summarize inputs from the cohort study and estimate the total cost per PLHIV in the three treatment eras and assess the value of medical innovation in HAART. The Markov model simulated the impact of the HIV pandemic if it were introduced today reflecting three scenarios: pre-HAART without treatment, early HAART with the first generation of treatments, and late HAART with highly effective treatments (Supplementary Material S3). The model was constructed in TreeAge Pro (Healthcare, version 2022) and allowed for disease prediction in two mutually exclusive health states: HIV and death. All patients began in the HIV health state at the age of 36, corresponding to the time of HIV diagnosis (Table 1). The model simulated a monthly cycle length with half-cycle correction over a 65-year time horizon during which all patients were expected to die. The outputs of the model were used to calculate pairwise incremental cost-effectiveness ratios (ICERs), specifically comparing early HAART to pre-HAART and late HAART to early HAART. The British National Institute for Health and Care Excellence (NICE) threshold of £20,000–30,000 approximately equal to €23,000–34,000 was used for reference (conversion rate 1.13, 13 February 2023) due to the absence of an official willingness-to-pay (WTP) threshold of cost per incremental QALY in Denmark.

2.6 Statistical Analyses

All statistical analyses on costs and survival from Danish registries were performed in the SAS statistical software package v.9.4. Lifetime extrapolation of survival data was performed in Stata (version SE/17.0) and comparisons of costs, LY, and QALY were estimated in TreeAge Pro Healthcare (version 2022, R2.0).

Deterministic one-way sensitivity analyses were conducted with a range of ±20% of the mean input value to reflect uncertainty (Supplementary Material S4) and a probabilistic sensitivity analysis (PSA) with 10,000 second-order Monte Carlo iterations to assess the uncertainty around the cost-effectiveness estimate (Supplementary Material S5). Uncertainty was included using a beta distribution based on the standard deviation (SD) assigned to utility values and a gamma distribution reflecting standard error (SE) to costs calculated directly from RWE (Table 2). Furthermore, a cost-effectiveness acceptability curve was generated (Supplementary Material S6).

Table 2 Disutility inputs for the three treatment eras of pre-HAART (1985–1995), early HAART (1996–2005), and late HAART (2006–2017) based on the WHO Global Burden of Disease study[48]
Full size table

3 Results

3.1 Costs

The total annual costs per person living with HIV (PLHIV) increased with the introduction of HAART for the index year but decreased in the incremental years and the last year of life (Table 3).

Table 3 Average annual costs and hospital days for PLHIV in the three treatment eras of pre-HAART (1985–1995), early HAART (1996–2005), and late HAART (2006–2017)
Full size table

3.2 Comparing Cost, Survival, and QALY

The total lifetime discounted (undiscounted) cost for an average PLHIV was €91,010 (€128,981) in pre-HAART, €103,130 (€199,062) in early HAART, and €126,317 (€254,964) in late HAART (Table 4). The lifetime projection of QALY (and LY) showed an average QALY per PLHIV of 4.56 (8.1), 13.35 (17.9), and 16.49 (20.4) for pre-, early, and late HAART respectively. Undiscounted QALY (and LY) were 6.44 (11.6) for pre-HAART, 24.98 (34.0) for early HAART, and 32.08 (40.4) for late HAART. Discounted (undiscounted) life expectancy for the Danish general population in 2006–2017, corresponding to late HAART, was 22.3 (44.9) LYs (Supplementary Material S2).

Table 4 Base-case results and sensitivity analyses from the Markov model comparing the three treatment eras
Full size table

The ICERs using discounted costs and QALYs showed that early HAART was cost-effective compared with pre-HAART at €1378 per QALY, and late HAART was cost-effective compared with early HAART at €7385 per QALY. The results were not sensitive to alterations in the time horizon, methods for estimating life expectancy, or discounting (Table 4).

The PSA showed a >99,9% likelihood of late HAART being cost-effective at €23,000 per QALY, i.e., the lower range of the threshold value. Early HAART had the highest probability of being cost-effective at a threshold range of approximately €1400–€7300 per QALY.

4 Discussion

In this first study to perform an ex-post cost-effectiveness analysis of HAART using individual-level register data we showed that the introduction of HAART was cost-effective in Danish healthcare and, in some scenarios, even seemed to be a truly disruptive innovation, i.e., HAART being both cheaper and more effective as in the scenario with a 9-year time horizon. The result showed that annual costs per patient with HIV increased from pre-HAART to early HAART and late HAART, however, because of the effective treatment the number of hospital days per patient also decreased.

Several limitations and challenges arise when conducting retrospective cost-effectiveness analyses, especially when dealing with RWE going several decades back in time, in the present study back to 1985 [50]. Due to limitations in data availability, we only had access to DRG tariffs for individual hospital costs from the year 2004. However, we were able to use data on the number of hospital days per patient for the entire period to estimate average costs per patient with HIV in the pre-HAART period. This provided a reasonable estimate of average costs as hospital budgets and personnel were linked to the number of hospital beds during that time [51].

Observational analyses may also be susceptible to bias when comparing patient populations in different time periods, which can lead to inaccuracies in the estimations of the impact of the intervention, i.e., the treatment effects of early and late HAART. The challenges of controlling for all relevant developmental factors were handled by adjusting for the increase in life expectancy of the general population. We did not correct for potential changes in socioeconomic status, ethnicity, the presence of comorbidities, and improved living conditions in the general population, which may influence the outcomes being compared. Other limitations include that we did not investigate the impact of CD4 cell count on the cost-effectiveness [28, 29]. Furthermore, there is a large uncertainty in the estimation of QALY gain using the WHO GBD disability values for PLHIV as estimates of disutility [15]. Estimating the gain in utility over a period of 30 years is difficult as preferences for health and instruments for measuring health-related quality of life have changed significantly over the last 30 years [15, 49]. We believe our assumptions about disutility are conservative in the sense that getting the HIV diagnosis in the pre-HAART and early HAART days may have felt like a “death sentence,” and our estimates mainly capture the difference in utility due to the state of the HIV disease and the availability of antiretroviral therapy [48]. As assumed in our study, published meta-analyses and results from the SMART trial present utility data showing an increase in quality-of-life for PLHIV over the three periods [52,53,54,55,56,57].

Furthermore, comparing treatment strategies over different time periods using RWE is contingent upon the fact that the healthcare system in general is changing over time. Thus, the increased effectiveness of HIV treatment over time cannot be solely related to specific breakthrough drugs or even competencies in HIV departments, rather it relies on the entire healthcare sector and society’s knowledge and competencies that have evolved. As such, ART was undoubtedly a central part of the disruptive innovation, but the increased effectiveness also relates to the way drugs were used as well as the healthcare system’s ways of dealing with HIV. This is the strength of ex-post health economic evaluation; it captures cost and effectiveness of treatments in the way they have been used in clinical practice. Furthermore, advances in other therapeutic areas during the whole period from pre- to late HAART have also affected mortality and have therefore contributed to the findings [30].

The findings from our study are generally in accordance with other health economic research, and it seems to ‘sum up’ the existing economic evidence in a coherent conclusion about cost-effectiveness.

For the pre-HAART period, the lack of a standardized methodology for health economic analysis gave a large variation in estimates of cost per patient with HIV (from 2000 to 70,000 US dollars per patient per year) [16]. Most of these earlier cost estimates showed patient costs that were higher than our estimates, presumably because they represented selected patient groups receiving treatment, rather than a population-based average like our study. We found no indication that our figures from the pre-HAART period were higher than other European healthcare systems, but like other studies from that era, we confirmed that the last year of life was very expensive [58].

Other health economic analyses from the early HAART period also showed the apparent paradox, that despite the increasing cost of pharmaceuticals, the total cost per patient with HIV did stabilize or even drop after the introduction of triple therapy [19, 22, 59]. In the late HAART, several cost-effectiveness analyses were published, all pointing to the further implementation of cost-effective new treatment possibilities for HIV. Summing up late HAART CUAs, it does not seem strange that our overall picture of the entire period also turns out to be cost-effective. At the same time, the same trend continues during late HAART with a continuous substitution of bed days with more effective medicine; however, in this period also at a slightly increasing annual cost per patient with HIV. Again, our study is population-based showing the average for all PLHIV, where only a few have received the most expensive treatments. Further, most PLHIV have not received specialized HIV drugs immediately after diagnosis until the result of the START INSIGHT study, demonstrating that early initiation of ART is beneficial for nearly all HIV-infected patients, regardless of the CD4+ count, were implemented in Danish clinical practice at the end of the late HAART period [60].

To our knowledge, only a single study has attempted to analyze with similar objectives as ours, but using modeling of cost instead of RWE data on cost. Perez-Elias et al. (2021) analyzed survival data from a cohort of PLHIV who started ART between 1985 and 2016 in Spain and compared their health outcomes and projected healthcare costs to those of a similar group of PLHIV who did not receive ART. The study found that ART has been a highly effective intervention for people living with HIV in Spain, with significant clinical and economic benefits. ART was associated with significant improvements in life expectancy and quality of life and a significant reduction in the incidence of AIDS-related illnesses. The study concluded that ART was cost-effective over the long term, as the healthcare costs associated with HIV care were lower for people receiving ART compared with those who did not [61].

Our findings suggest that societies, policymakers, and clinicians on an overall level have made cost-effective decisions when investing in the uptake of new treatment modalities for HIV. While breakthrough HIV drugs might have been among the most expensive in Danish healthcare, this analysis indicates that they have nevertheless been used cost-effectively. Historically, HIV has gradually changed in nature from a complicated, expensive disease, to a simpler, more manageable disease with a significantly improved prognosis [62, 63]. The introduction of ART led to a shift in the way that HIV was managed in health care, with a greater emphasis on prevention, early detection, and treatment, and not with a focus on managing symptoms of the disease [19, 64, 65]. Our analysis only includes innovation until 2017, and further improvements in the treatment of HIV have been implemented since then [66, 67].

This ex-post evaluation of the treatment of HIV highlights the importance of historical analyses. Future research should investigate the limitations of the single health technology evaluation approach usually applied to support decision-making in healthcare as the historical analysis points to the existence of important synergies between new technologies. Over the whole period, more than 30 effective HIV medicines emerged, and the treatments of noncommunicable diseases significantly contributed to the improved survival and quality of life in PLHIV. This development was not foreseen in earlier economic evaluations of antiretroviral therapies [5, 14, 17,18,19,20,21,22,23,24]. Future research should consider developing new methods for incorporating synergies and future innovation in single ex ante health economic evaluations.

5 Conclusion

We conclude that the introduction and implementation of HAART in Danish healthcare were cost-effective. Ex-post economic evaluations provide opportunities for measuring impact, policy learning, and feedback on future ex ante evaluations. Current and prevailing challenges in the treatment of HIV require further development of HAART to continue to close the gap in life expectancy and quality of life for PLHIV compared with the general population while minimizing costs.