Cost-effectiveness of Venetoclax plus Obinutuzumab versus Chlorambucil plus Obinutuzumab for the First-Line Treatment of Adult Patients with Chronic Lymphocytic Leukemia – an extended societal view

Objectives: Efficacy of venetoclax plus obinutuzumab (VenO) compared to chlorambucil plus obinutuzumab (ClbO) for treatment-naïve CLL adult patients with coexisting medical conditions was investigated in CLL14 (NCT02242942). Our aim was to evaluate the cost-effectiveness of VenO versus ClbO for these patients from a Dutch societal perspective. Methods: A three-state partitioned survival model was constructed to evaluate the costeffectiveness of VenO. The outcome of the analysis was the incremental cost-effectiveness ratio (ICER) with effectiveness measured in quality-adjusted life-years (QALYs) gained. Uncertainty was explored through deterministic and probabilistic sensitivity analyses (DSA & PSA), scenario analyses, and value of information analysis (VOI). Results: The base case resulted in a discounted ICER -49,928 EUR/QALY gained (with incremental negative costs and positive effects). None of the ICERs resulted from DSA and scenario analyses exceeded the chosen willingness-to-pay threshold of 20,000 EUR/QALY, and Jo urn al Pr e-p roo f


1.
Introduction: Chronic lymphocytic leukemia (CLL) is one of the most common types of leukemia in adults and especially in the elderly. 1 For those above the age of 80 years, the annual incidence increases to more than 30 per 100,000 person. 1 While CLL remains incurable, 2, 3 the disease can often be successfully managed with chemotherapeutic and immunotherapeutic agents for many years. 2  Despite the substantial improvement in PFS outcomes gained from treating with ClbO, there remains an unmet need for novel chemotherapy-free and fixed-duration 1L therapies with more acceptable and manageable safety profiles and improved clinical outcomes. 5 Consequently, the combination of venetoclax and obinutuzumab (VenO), a first chemotherapy-free, fixed-duration (i.e. 12 cycles) combination regimen, was proposed. 5 Recently, both efficacy and safety of VenO were investigated in the CLL14 study (NCT02242942), 6 a multicenter, randomized, open-label, phase 3 trial. In comparison with ClbO, VenO demonstrated statistically significant superior PFS (hazard ratio [HR], 0.31; p-value < 0.0001) in treatment-naïve CLL patients with coexisting medical conditions. 7 While the EMA issued marketing authorization for VenO in this population in 2020 8 , many European Member States base their decision to reimburse novel treatments on a formal Health Technology Assessment (HTA). In these assessments, the therapy's cost-effectiveness plays a vital role for J o u r n a l P r e -p r o o f the decision-making process. Economic evaluation studies can provide the necessary information by combining several sources of evidence (i.e. on treatment effects and costs). 9, 10 However, current information on the cost-effectiveness of VenO when compared to ClbO are only available from one non-European study, three conference abstracts, and two national assessment reports (one of which is in Dutch). [11][12][13][14][15] While the conference abstracts reveal only little on the employed methodology, most outcomes of the assessment report of the United Kingdom (UK) National Institute for Health and Care Excellence (NICE) are redacted due to commercial or academic confidentiality. This lack of transparency poses a challenge to the evaluation and comparison of these economic models, which heavily limits reproducibility. 16,17 In addition, several methodological choices of the available publications remain unclear and certain aspects were not studied. For instance, none of the economic evaluations fully considered future costs (both medical and non-medical), although their inclusion is often recommended. 18,19 Our aim is to add to the existing body evidence by performing and reporting on a comprehensive

Model Structure:
The design of our model structure was based on previously published models, which in turn were informed by the clinical pathway and clinical expert opinion. 11,15 More specifically, the model comprises three health states: progression-free (PF), progressed disease (PD), and death. 26 Figure   1 visualizes the model structure and the possible transitions between health states. At any given time, modeled patients could only occupy one of the three health states. Patients were initially treated with either treatment option (VenO or ClbO) in the PF state. At the end of each 28-day model cycle, patients either remained progression-free (i.e. stayed in PF), progressed (i.e. moved to PD), or died (i.e. moved to the state of death). Once patients progressed, they received subsequent treatment lines or died. Death was an absorbing health state and with a chosen lifetime horizon (i.e. 29 years), all patients eventually end and remain there. In this way, we also captured long-term effects and costs of the therapies of interest.

Model Inputs:
Since we did not have access to individual patient-level data (IPD), pseudo-IPD was created from discounted with 1.5% to account for the effect of differential timing. 18 All effectiveness parameter values are presented in Table 1.
In terms of health-care costs, we included costs for drugs (acquisition and administration), adverse events (AEs), subsequent treatments, routine care, and follow-up activities, as well as future medical costs. Regarding societal costs, we included costs for travel, and informal care. In scenario analyses, we also accounted for the impact of future non-medical costs. Since the average age of the modelled population (i.e. 71 years) was well above the current Dutch pension age (i.e. 66.3 years), 32 we did not include costs of productivity losses.
Drug dosing schedules were based on the planned dose derived from the CLL14 protocol. 33 Prices for drug acquisition (1L and subsequent treatments) were taken from the Dutch official medicine database (i.e. Zorginstituut Nederland 34 ). In scenario analyses, we also accounted for the impact of the so-called "patent cliff", meaning that prices of CLL therapies decrease following patent expiration. Particularly, venetoclax, obinutuzumab, and Ibrutinib were modeled to go off-patent in May 2030, 35 in November 2024, 36  Assumptions on resource use for travel and informal care were based on the literature, 15 and valued with standard unit prices from the Dutch costing manual. 46 Future medical costs were included using the iMTA PAID tool (version 3.0), which is available online (https://imta.shinyapps.io/PAID3code//). 49, 50 More information on the nature of these future costs can be found in Appendix 1.
All costs in this study were expressed in Dutch 2020-euros and prices of earlier years were indexed to 2020 with the pertinent consumer price index (CPI). 51 Cost outcomes were half-cycle J o u r n a l P r e -p r o o f corrected and discounted with 4.0%, following the Dutch EE guideline. 18 All resource use and cost parameter values are summarized in Table 1.

Statistical Analyses:
In the base-case analysis, we calculated the incremental cost-effectiveness ratio (ICER) in QALY Details on the model validation process can be found in Appendix 1.

Base-Case Analyses:
Based on the Latimer selection process, the log-logistic and exponential distribution were  (Figures 3 and 4).
The model estimated an average of 9.31 QALYs for VenO, and 8.06 QALYs for ClbO (averaged, discounted, and HCC results). All these outcomes are summarized in Table 2.
J o u r n a l P r e -p r o o f

Uncertainty Analyses:
The top ten influential parameters determined through the OWSA are depicted in Appendix 2 ( Figure 5). This sensitivity analysis demonstrated that varying the utility value at PFS state after receiving the 1L treatment was the most influential factor for the ICER.  These results are mainly driven by the extended PFS period following VenO. The sensitivity analyses demonstrated the robustness of these results. Furthermore, all explored scenarios including the consideration of future non-medical costs and the patient-cliff impact rendered VenO dominant with the chosen WTP threshold. Additionally, our VOI analyses results indicated that additional research is not recommended because our EVPI value is substantially lower than the threshold. 62 In other words, the cost-effectiveness conclusion of VenO in treatment-naïve CLL patients is robust based on currently available evidence.

Comparison with other Studies:
Although different, our model results are in line with previous studies examining the costeffectiveness in the given setting. However, we were able to note the differences in incremental QALYs gained between our analyses (1.25) and the NICE report (0.365). We hypothesize a reason to this variance as follow.
The utility values used in both analyses were derived from different sources. The difference in elicitation of utility scores might have yielded discordance between the QALYs gained observed between the two reports.

Strengths and Weaknesses:
Although our study is not the first to estimate the cost-effectiveness of VenO, it is the first to adopt an extended societal perspective by incorporating future medical and non-medical costs (in scenario analyses). For EEs performed under a US or Dutch perspective, it is suggested to J o u r n a l P r e -p r o o f consider future medical costs. 18,19 While the US guidelines recommend the inclusion of future non-medical costs as well, 19 the Dutch guideline does not mention its inclusion specifically (yet). 18 Our study, is the first to fully include both components in the analysis for this setting. In practice, future costs are often excluded from CEAs. 60 With our analysis, we bridge this gap, which could potentially be used as a reference point for future EEs.
In addition, we made our model and data sources openly accessible. To date, very few of decision models are made available due to lack of a standard model repository, 61 or due to the confidentiality of data. Nevertheless, the urgency of having these models available to all stakeholders such as policy makers, health authorities, industry sponsors, academicians, and others is increasing. 61 Furthermore, the availability of these models allows the research community to validate and even reuse the model with different data, 61  This study has several limitations 2 . First, owing to lack of IPD from the clinical trial, our study could not examine the heterogeneity of the study population. Thus, subgroup analyses could not be performed to further understand differences in the ICERs. Having access to IPD or real-world evidence will be desirable for specific subgroup analyses to better recommend the drugs of interest.
Second, some of the utility decrements due to AEs in our model were based on those of AEs