Evaluating different adoption scenarios for TIL-therapy and the influence on its (early) cost-effectiveness

A Delphi method was used to systematically generate consensus on themes related to the adoption of TIL therapy to subsequently incorporate these themes in the adoption scenarios. Figure 1 shows the six steps used to draft the scenarios [13, 14].

First, relevant themes that could influence the adoption of TIL therapy were identified through: brainstorming with internal experts, reviewing the literature on TIL therapy and research developments in treating advanced melanoma, and scanning ongoing clinical trials investigating TIL therapy. Second, the identified themes were discussed during semi-structured interviews with stakeholders in the TIL study process at the NKI to identify their expectations on these themes for the coming years [11]. They were allowed to add new themes and were specifically asked to describe likely “what if” scenarios for the coming five and 10 years [13]. The details on these semi-structured interviews are described in a previous publication [11]. In the third step, the results of the interviews were discussed with the direct research group (ML, VR, WvH), where the final themes were chosen to incorporate in the first (pilot) set of adoption scenarios. In step four, this first set of adoption scenarios (15 scenarios and two questions) was piloted in an expert group consisting of lab members, health insurers, clinicians, researchers, a representative of a patient association, a board member of the Dutch Immunotherapy Working Group for Oncology (WIN-O), and policy advisers. In the fifth step, the set was adapted according to their given feedback which resulted in the final set of scenarios. This set consisted of 15 adoption scenarios and 5 questions on, for example, minimal effectiveness, patients’ and clinicians’ attitudes towards TIL therapy (Table 1).

The adoption scenarios and questions were included in a web-based questionnaire (Supplement 1) and were shared among a larger group of experts to evaluate the likelihood of the scenarios happening in the coming 5 years. To reach international clinical experts, flyers regarding the questionnaire were distributed at the congress of the European Cancer Congress Organization (ECCO) in Amsterdam (January 2017) after melanoma-related sessions. Additionally, the questionnaire was emailed to the scientific and clinical network of our internal experts, by which we invited 119 international experts; all were reminded after 1 month.

The questionnaire consisted of three parts. Part one introduced the TIL therapy and the RCT that is currently ongoing. Part two evaluated the characteristics of the respondent (years of experience with TIL therapy and years of experience with melanoma care, their position, and their self-reported level of expertise with TIL therapy) [15]. The third part contained the 15 adoption scenarios and the five questions, as listed in Table 1. In this part, the respondents indicated the likelihood of the scenarios occurring in the coming 5 years from 0 to 100%. 0% indicates that the scenario will not occur within 5 years, and 100% indicates that the scenario will occur within 5 years. This method is similar to the method used in a publication focusing on the adoption of Next Generation Sequencing [16]. Table 1 lists the names of the scenarios which are used in the following sections to refer to the specific scenarios.

The results of the scenarios incorporated in the cost-effectiveness model are expressed by the Incremental Cost-Effectiveness Ratio (ICER), Net Monetary Benefit (NMB), and the probability of TIL therapy being cost-effective. The ICER is a deterministic statistic calculated by dividing the difference in costs by the difference in Quality Adjusted Life Years (QALYs) for TIL therapy and Ipilimumab. An ICER, negative (less costly, more effective) and/or below a certain threshold (Willingness To Pay (WTP)), in this study €80,000, would mean that TIL therapy is favored over Ipilimumab. The WTP of €80,000 is the informal ceiling ratio in the Netherlands for diseases with the highest symptom burden [21]. As internationally different WTP thresholds are used, a second WTP threshold was used in evaluating the NMB: £30,000 (€34,821; April 2019), which is the WTP threshold used in the United Kingdom [22]. A two-way sensitivity analysis evaluates the effect of various levels of two parameters on the ICER. We varied the 1-year progression-free survival rate and the costs of TIL in a two-way sensitivity analysis.

Both NMB and probability of being cost-effective are probabilistic statistics in which uncertainty surrounding the input parameters is taken into account by randomly drawing parameter values from the parameter distributions, using Monte Carlo simulations with 1000 iterations. The NMB was calculated using the WTP ratios and the following formula per iteration: (incremental QALYs x WTP) - incremental costs. A mean NMB ≥ €0 indicates that TIL therapy is cost-effective compared to ipilimumab, given the chosen threshold.

To calculate the probability of TIL therapy being cost-effective, the NMB was calculated over different thresholds, ranging from €0 to €80,000 in steps of €1000. An NMB value of ≥€0 is cost-effective, which is indicated with 1, an NMB value <€0 is not cost-effective, which is indicated with 0. This was done for all the iterations in the Monte Carlo simulation per threshold. A mean of this binary value was calculated per threshold which shows the probability of TIL being cost-effective compared to ipilimumab at that threshold. Finally, the mean of these average probability scores gives the probability of TIL therapy being cost-effective in a WTP range of €0 - €80,000.

Twenty-nine respondents, mainly clinicians (76%; 24% other), completed the web-based questionnaire between January and October 2017. The majority of respondents originated from the Netherlands (n = 14), fifteen experts originated from other countries, namely Belgium, Denmark, Germany, Israel, Italy, Poland, Portugal, Spain, UK, the US, and Australia. Most respondents described themselves as familiar (52%), expert (28%), or a former expert (10%) with TIL therapy and had on average 2.7 years of experience with TIL treatment. Table 3 shows the characteristics of the respondents.

The mean and median likelihood of each of the scenarios is presented in Table 4 and Fig. 2. Large variability was seen in the expected likelihood of the scenarios suggesting that respondents are uncertain about the future developments surrounding TIL therapy (Fig. 2). On average, most of the scenarios scored a likelihood of around 50% (46–54%). Two scenarios scored a likelihood of ≥55%: “Combination therapy” (57%) and “Automatic TIL production” (58%). Four scenarios were thought to be less likely: “Biomarker” (37%), “TCR therapy” (32%), “Low-cost competition” (30%), and “Less IL-2 treatment” (36%). Finally, the likelihood of the “Base case” in the coming 5 years was estimated at 54%. The results of the questions related to the adoption of TIL are listed in Supplement 6.

Using the cut-off value of ≥55%, “Combination therapy” and “Automatic TIL production” were labeled as “likely”. Using the stratified results based on the level of expertise, “Base case” and “3^rd line treatment” were also labeled as “likely” (Table 4). The expert panel verified the likelihood of those four scenarios. Based on the literature review (step four in Supplement 2), “TIL more effective” and “Less IL-2 treatment” were labeled as “likely” as several studies described potential opportunities to increase the effectiveness of TIL therapy and studies are investigating an IL-2 decreasing dose scheme to lower the intensity of the treatment [17]. The other scenarios or topics were not described in the recent literature review. The experts evaluated (step five) “Clinicians unconvinced”, “Low-cost competition”, “TIL production outsourced” as “likely” and the scenarios “Competition”, “Biomarker”, “TCR therapy”, “Patients unconvinced” and “Influence by companies” were labeled as “unlikely”. No scenario was solely labeled as “unlikely” based on the score from the survey. The arguments for labeling these scenarios as “likely” or “unlikely” are described in Supplement 5. As the base case scenario already evaluates the effect of using TIL therapy as a second-line therapy, the scenario: “2^nd line treatment” was not incorporated in the cost-effectiveness analysis because it would show the same results. Eventually, scenarios resulting in no implementation of TIL therapy e.g. “Patients unconvinced” and “Clinicians unconvinced”, regardless of their likelihood, were not incorporated in the cost-effectiveness model as this results in an analysis comparing ipilimumab to ipilimumab.

Additionally, the potential combinations of scenarios were drafted and incorporated in the cost-effectiveness model. Three combinations were made related to research developments including “TIL more effective”, “Automatic TIL production” and “TIL production outsourced”. Besides, three other combinations were defined incorporating the scenario “combination therapy”, and the scenarios: “Automatic TIL production”, “less IL-2” and “Low-priced competition”,

Figures 3 and 4 show the NMB and probability of TIL therapy being cost-effective. Four out of six adoption scenarios showed a positive NMB: “TIL more effective”, “Low-cost competition”, “Less IL-2 treatment” and “Automatic TIL production”, and a high probability of being cost-effective. Even when the total costs of the comparator (ipilimumab) are reduced with 20%, TIL therapy had a 55% chance to be cost-effective (“Low-priced competition”). In contrast, “Combination therapy” showed a negative NMB with an ICER of €151,520 per QALY based on the first clinical results [19], and when the production of TILs is outsourced, TIL therapy had a 0% likelihood to become cost-effective (“TIL production outsourced”). All the results from the cost-effectiveness analysis are presented per scenario in Supplement 4. Figure 5 shows the results of the two-way sensitivity analysis and incorporated scenarios. This graph shows for instance that the effectiveness should improve substantially when TIL production is being outsourced or TIL therapy is combined with another therapy.

The combination of “TIL more effective” and “Automatic TIL production” showed a positive NMB as it combined the two most favorable scenarios for TIL therapy (more effective and less expensive). The other two combinations related to research developments showed that a slight improvement for TIL therapy in response rates does not outweigh the extra costs when TIL production is commercialized (0% probability of TIL being cost-effective), which holds when TIL therapy becomes more automatic (11% probability of TIL being cost-effective). The combinations that focused on the combination of TIL therapy with a different therapy, showed a negative NMB in all combinations of scenarios. A potential reduction of the costs of TIL therapy seems however to have the highest impact on the probability of being cost-effective (from 12% in the base case to 28% in the combination of “combination therapy” and “TIL production outsourced”).

The results of the cost-effectiveness analysis showed in most of the preferred scenarios a high probability for TIL therapy to become cost-effective (55–99%) and they identify aspects that could facilitate the wide adoption of TIL therapy. For example, as the scenario “Automatic TIL production” showed the highest probability for TIL therapy to become cost-effective (99%), Research and Development should focus on optimizing the production process to facilitate potential upscaling and adoption of TIL therapy. In contrast, the scenarios showing a reduced chance for TIL therapy to become cost-effective, identify crucial contextual factors that should be considered when deciding to adopt TIL therapy. For example, outsourcing of the production of TILs may at first be expected to overcome known ATMP barriers as (i) inadequate financial support for the required investments, and manufacturing costs; (ii) a lack of regulatory knowledge, and (iii) challenging to upscale the production and (iiii) to comply with Good Manufacturing Practices [9, 23‐26]. However, as a result of commercial pricing levels, assuming that the costs will be at least 3 times as high, this scenario resulted in a 0% probability for TIL therapy to become cost-effective. Following our analysis, assuming a WTP threshold of €80,000, the production costs of TIL may only increase 1.5 times (~€53,000) to be cost-effective compared to ipilimumab. Within this scenario, it should be kept in mind that the estimation of the commercial costs is uncertain. Especially because our assumption was based on the manufacturing costs in an academic setting and literature showed that commercial prices are mostly linked to what society would pay instead of its expected added value or of the actual manufacturing costs [27]. However, although the estimation is uncertain, the conclusion related to commercialization remains the same: Outsourcing could facilitate the implementation of TIL therapy, however, pricing agreements should be made with the commercial party to ensure cost levels that remain within the willingness to pay range of cost-effectiveness. In the US setting, TIL can be expected to be licensed and filed for FDA approval within the coming years, these insights are especially of interest to guide reimbursement decisions by insurance companies. An interesting scenario is the “Combination therapy” which showed a 12% probability of being cost-effective (ICER of €151,520), revealing that in this case either the treatment costs should decrease or the efficacy has to improve considerably. By automatizing the production of TILs, the probability of being cost-effective increased only to 28%. Therefore, when a combination of TIL therapy and for instance a certain checkpoint inhibitor seems promising, agreements on pricing with pharmaceutical companies for the combination therapy are necessary to remain within cost-effectiveness ranges.

The results on the questions related to future developments suggest that the adoption of TIL therapy may be hampered by the attitude of patients and clinicians. First of all, clinicians nowadays seem unconvinced to apply TIL therapy because of its perceived complexity and treatment intensity unless the therapy shows a 1-year survival rate of at least 61.3% (CI:55.2–67.5). Secondly, only a small proportion of the eligible patients seemed to be aware of TIL therapy as a treatment option. As the attitude of stakeholders and especially clinicians, is a known barrier for implementation of ATMPs, a pro-active information strategy in anticipation of this attitude is crucial when deciding to diffuse TIL therapy [24].

Another barrier that ATMPs face in the translational pathway is the rapidly evolving field of immunotherapy [8, 9]. We therefore compared our results with the most recent developments described in the literature and most of the “likely” scenarios still seem to be in line. For example, several trials investigate a combination of TIL therapy and other targeted therapies: pretreatment with ipilimumab followed by TIL and IL-2 (NCT01701674) [19] or pretreatment with vemurafenib followed by TIL and IL-2 and followed by vemurafenib (NCT01659151) [28]. Besides, several trials investigate or investigated the effectiveness of a lower dose of IL-2 treatment [17, 29] (NCT02354690), and finally, research groups evaluate the optimal process of producing TILs, aiming to improve the efficacy of TIL therapy e.g. by enriching T cell products with neo-antigens [17, 30].

Some developments found in literature, however, were not incorporated. For instance, several studies evaluate different lymphodepleting preparative regimens such as total body irradiation (TBI) in combination with chemotherapy [17, 31]. It is currently unclear whether such a regimen would be applied soon, but this scenario could influence the cost-effectiveness as TBI (requiring autologous PSC support) would significantly increase the costs. Besides, a very likely scenario that is not incorporated in this analysis, is the use of TIL therapy in other tumors such as renal cell cancer, ovarian cancer, and colorectal cancer [32‐35] (NCT01174121). This scenario should be kept in mind as it may facilitate the adoption of TIL by positively influence the clinicians’ attitude as clinical experience and exposure increases, and production costs may decrease. Furthermore, a recent literature review highlighted several potential agents (e.g. TIM3, GITR, OX40) that could be promising in treating advanced melanoma in the future [36]. Those agents are currently subject to the first phase I and II studies to evaluate their safety and efficacy [36]. Therefore, our study might have underestimated the likelihood of the competition scenario. However, available data on the efficacy and possible costs of those alternatives is too preliminary to incorporate these results in the cost-effectiveness analysis. When these agents are proven to be safe, effective, and more effective compared to TIL therapy, those new treatments could hamper the adoption of TIL therapy.

The main strength of the present study is that we systematically drafted future scenarios (qualitatively) with internal and external experts, using a Constructive Technology Assessment framework [11, 41] and using multiple Delphi rounds. This provides a comprehensive insight into the potential future developments that could influence TIL adoption and provides research and development teams with valuable information to anticipate possible future developments. Since the landscape of immunotherapy for melanoma is continuously developing, the expectations of the experts were compared to the most recent literature reviews and ongoing clinical trials to select the “likely” scenarios and discuss our results. The main limitation is obviously, that the scenarios may not even keep up with actual developments. Other limitations are related to the early nature of this analysis. For example, to simulate the combination therapy, the input for the model was based on a first observational study in which only 13 patients were enrolled that received the combination of TIL therapy and ipilimumab [19]. Additionally, the chosen cut-off value of ≥55% to evaluate scenarios as “likely” could be questioned due to the high uncertainty surrounding the likelihood scores. However, since the expert opinions and recent literature verified that the “likely” scenarios based on the cut-off value were “likely”, a different cut-off value is not expected to have altered our conclusions. Furthermore, the cost-effectiveness analyses were conducted from a Dutch perspective similar to the original model by Retèl et al. 2018. The costs for both TIL therapy and ipilimumab are expected to differ between countries [42] which limits the generalizability of the cost-effectiveness results in different settings. The generalizability may also be limited by the fact that mainly experts from European countries completed the questionnaire. However, by verifying the likelihood results with the most recent literature, the identified crucial contextual factors are expected to hold also in other countries because similar (financial) challenges are expected regarding e.g. outsourcing and providing a combination of therapies.