Cost-effectiveness of obinutuzumab plus bendamustine followed by obinutuzumab monotherapy for the treatment of follicular lymphoma patients who relapse after or are refractory to a rituximab-containing regimen in the US

Abstract

Aims: Obinutuzumab (GA101, G) was approved in February 2016 by the US Food and Drug Administration to treat follicular lymphoma (FL) patients who relapsed after, or are refractory to (R/R), a rituximab-containing regimen (R/R-rituximab). In the GADOLIN trial, R/R-rituximab patients who received G plus bendamustine (B) followed by G-monotherapy (G + B) for up to 2 years had significantly improved progression-free survival and overall survival compared to patients receiving B-monotherapy. This study estimated the cost-effectiveness of G + B vs B-monotherapy for R/R-rituximab FL patients from a US payer perspective.

Materials and methods: Patient outcomes were simulated using a 3-state area under the curve model including progression-free survival, progressive disease, and death. This study used R/R-rituximab data from the National LymphoCare Study to extrapolate the GADOLIN trial’s refractory FL progression-free and overall survival data to a R/R-rituximab FL population. Drug utilization and adverse events were based on trial data, and costs were based on Medicare reimbursements and drug wholesale acquisition costs in 2016. Utility estimates were derived from published literature. Post-progression treatment costs were based on observed post-progression therapies in GADOLIN. Sensitivity analyses were conducted to assess model uncertainty.

Results: G + B resulted in an increase in quality-adjusted life years relative to B-monotherapy of 1.24 (95% CR = 0.61–1.87); the incremental total cost was $58,100 (95% CR = $54,500–$61,500). The incremental cost-effectiveness ratio was $47,000 per QALY gained, and, based on probabilistic simulations, there was a 98% probability that G + B was cost-effective at the $100,000 per QALY threshold.

Limitations and conclusions: This US-based analysis suggests that treatment with G + B compared to B-monotherapy is likely cost-effective in R/R-rituximab FL patients. Modeling a R/R-rituximab population based on a synthesis of GADOLIN and the National LymphoCare Study data introduces uncertainty in the analysis. However, the findings were robust to sensitivity analyses.

Keywords:

• Obinutuzumab
• follicular lymphoma
• rituximab-refractory
• decision model
• cost-effectiveness
• economic

JEL classification codes:

• I11
• I19

Introduction

Indolent non-Hodgkin’s lymphomas (iNHL) constitute approximately one third of all cases of non-Hodgkin’s lymphoma¹. Follicular lymphoma (FL) is the most common sub-type of indolent NHL, accounting for ∼22% of newly-diagnosed NHL cases². The National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) Program estimates the number of new cases of FL is 3.6 per 100,000 men and women per year³, and ∼20% of FL patients will experience disease progression within 2 years of first-line therapy and have poorer overall survival than patients who do not progress within 2 years⁴. While FL patient survival may exceed 10 years, the majority of patients present with advanced-stage disease, and transformation to a more aggressive lymphoma is possible⁵. With each FL relapse the time to next the progression event becomes shorter⁶.

The addition of rituximab to chemotherapy has been shown to extend progression-free survival (PFS) and overall survival (OS) in previously untreated patients^7–11. However, a growing population of patients progress during or after a rituximab based regimen; these patients have few subsequent treatment options and a poor prognosis. Bendamustine (B) has been shown in phase II studies to be efficacious in the treatment of rituximab-refractory indolent lymphoma; however, median PFS is limited to 7–9 months^12–14. Therefore, a medical need to improve upon the degree of and duration of treatment response for iNHL patients who have progressed on rituximab remains.

Obinutuzumab (GA101, G), a type II humanized anti-CD20 monoclonal antibody, was evaluated in the GADOLIN trial (NCT01059630), an open-label, multi-center, randomized phase III study to investigate the efficacy and safety of G combined with B followed by G monotherapy (G + B) vs B-monotherapy (B-mono), in patients with rituximab-refractory iNHL¹⁵. Eligible patients in GADOLIN were defined as those who failed to respond to, or progressed during, any previous rituximab-containing regimen (monotherapy or combined with chemotherapy), or progressed within 6 months of their last rituximab dose in the induction or maintenance treatment settings. In the primary analysis (data cut-off: 1 September 2014; 396 iNHL patients), median independent review committee-assessed PFS was not reached in the G-B arm (194 patients) and was 14.9 months in the B arm (202 patients), a 45% reduction in risk of progression or death (HR = 0.55; 95% CI = 0.40–0.74; p = .0001)¹⁵. Seventeen additional patients were enrolled after the data cut-off for the primary analysis; as of the data cut-off of 1 April 2016, G + B showed statistically significant improvement in investigator-assessed PFS in patients with FL (HR = 0.52; 95% CI = 0.39–0.69) and prolonged OS (HR = 0.58; 95% CI = 0.39–0.86) compared to B-mono, with a median follow-up of 31.2 months vs 21.1 months in the primary analysis¹⁶.

In February 2016, the US Food and Drug Administration (FDA) approved G + B for the treatment of patients with FL who relapsed after, or are refractory to (R/R), a rituximab-containing regimen (R/R-rituximab)¹⁷. Thus, G + B is indicated for use in the US in a more broadly defined patient population than the GADOLIN trial population. The objective of this study was to estimate the US-based cost-effectiveness of G + B vs B-mono in a R/R-rituximab FL population based on a combination of GADOLIN trial data and long-term survival data of patients with R/R-rituximab FL from the National Lymphocare Study.

Methods

Approach

We developed a three-state area under the curve (AUC, also called a partitioned survival model) framework, which included progression-free survival (PFS), progressive disease (PD), and death (Figure 1). The PD state included 2^nd-line treatments. The AUC approach enabled us to estimate the time in PFS, time in PD, and overall life expectancy using the PFS and OS curves from the GADOLIN trial. Patients in the PFS state could be either on or off treatment to account for patients who finish therapy but remain in the PFS state. The analysis was conducted from a US payer perspective, focused on direct medical care costs only, and used a lifetime horizon. We used a 3% discount rate for all future outcomes and costs¹⁸.

Figure 1. Decision model—three states. All patients start in the progression-free health state, where they either receive study drugs or not (on/off treatment) during each approximate 1-month cycle. Patients can subsequently move either to another health state or stay in the same state at the end of each subsequent analysis cycle. The possible transitions are indicated by arrows. Abbreviations. PFS, progression-free survival; PD, progressive disease.

Clinical trial results

The GADOLIN trial’s primary end-point was independent review committee (IRC)-assessed PFS, defined as the time from randomization to the first occurrence of progression, relapse, or death from any cause. Secondary end-points included PFS assessed by the investigator, OS, and safety among other end-points. The GADOLIN trial enrolled 396 patients; 194 patients were randomized to the G + B group and 202 to the B-mono group^15,16. All patients had histologically documented, CD20-positive, refractory iNHL, including FL of grades 1–3a (G + B = 80%, B-mono = 82%). Patients were mostly male (G + B = 56.9%, B-mono = 58.4%) and white (G + B = 88.2%, B-mono = 86.6%). The median age (62 years) and age range were similar for both groups. The treatment arms were well balanced with respect to other demographic factors. The average weight and height were 81.1 kg and 169.7 cm, respectively (body surface area = 1.92 m²)¹⁵.

After a median follow-up of 21.9 months, PFS by IRC was found to be statistically longer with G + B (median not reached) vs B-mono (median 14.9 months) (HR = 0.55, 95% CI = 0.40–0.74) at the pre-planned interim analysis¹⁴. At the 1 April 2016 data cut-off (median follow-up = 31.2 months), the investigator assessment showed statistically significant improvement in G + B PFS (HR = 0.52; 95% CI = 0.39–0.69) and prolonged OS (HR = 0.58; 95% CI = 0.39–0.86) compared to B¹⁶. The estimated median time to a PFS event for patients with FL in the B arm was 14.0 months (95% CI = 11.3–15.3), and 25.3 months (95% CI = 17.4–36.0) in the G + B arm. Fewer patients died in the G + B arm (25.5%) vs the B-mono arm (34.9%)¹⁶. The median duration of survival for FL patients had not been reached in the G + B treatment arm, while it was reached in the B arm (53.9 months (40.9–NA)).

Modeled population

The GADOLIN trial population was composed of rituximab-refractory iNHL patients, defined as those who failed to respond to, or progressed during, any previous rituximab-containing regimen (monotherapy or combined with chemotherapy), or progressed within 6 months of their last rituximab dose in the induction or maintenance treatment settings¹⁶. The trial enrolled primarily FL patients (81%), as well as patients with marginal zone lymphoma, small lymphocytic lymphoma, and Waldenström’s macroglobulinaemia¹⁵. G is indicated in the US for the treatment of patients with FL who relapsed after, or are refractory to, a rituximab-containing regimen. Therefore, our model specifically focused on the US label-indicated population of R/R-rituximab FL patients. Our approach to estimating outcomes in a R/R-rituximab FL population based on the GADOLIN trial’s refractory FL population is outlined below.

Survival curves

We utilized a 2-phase approach to model R/R-rituximab survival, combining the GADOLIN trial survival curves in the first phase with the R/R-rituximab population-adjusted extrapolations of long-term survival in the second phase. For the first phase, several parametric functions were fit to GADOLIN PFS and OS Kaplan-Meier data; the fitted distributions included exponential, Weibull, log-logistic, log-normal, gamma, and Gompertz. The parametric functions were then assessed for their goodness of fit to the data using the Akaike Information Criterion (AIC), as well as a qualitative assessment for face validity.

The proportional hazards assumption was insufficient to model the GADOLIN PFS curves, so we fit independent curves to the INV-assessed PFS for each arm. By AIC score, the Weibull function ranked fourth out of the six distributions for B-mono and fifth for the G + B arm; however, it was qualitatively judged to have the best face validity, and we selected this as the base-case function for both regimens. The proportional hazards assumption for the OS curves was upheld based on log cumulative hazard plotting and visual inspection. Therefore, we assumed that the OS treatment effect HR of G + B vs B-mono was applied throughout the GADOLIN follow-up period, and then assumed that G + B survival was equivalent to B-mono survival thereafter. The selected function for estimating the B-mono OS curve was also the Weibull, which yielded the lowest modeled incremental lifetime survival between the two arms. We explored the use of the other parametric curves for both arms in scenario analyses.

We then utilized long-term data on R/R-rituximab vs refractory patients from the National LymphoCare Study of FL patients to adjust the GADOLIN PFS and OS extrapolation curves to better reflect a R/R-rituximab population¹⁹. The National LymphoCare Study is a disease-specific, prospective registry that enrolled more than 2,700 patients with newly diagnosed FL from 2004–2007 from more than 200 practice sites in the US²⁰. Progressive disease (PD) within 6 months of rituximab therapy, including maintenance therapy (if any), was considered as rituximab-refractory²¹.

We first compared the R/R-rituximab and refractory FL populations from the National Lymphocare Study to assess survival differences, and verified that the R/R-rituximab population showed improved survival compared to a refractory population (Figure 2). Second, we estimated a HR between refractory and R/R-rituximab populations for both PFS and OS curves. Third, these HR were used to adjust the B-mono PFS and OS curves at a point where the National LymphoCare Study R/R-rituximab curves crossed the B-mono curves; the time points at which the curves crossed were 15 and 22 months for the PFS and OS curves, respectively. The R/R-rituximab vs refractory HR were applied from these points onward, thus we conservatively modeled refractory survival beyond GADOLIN’s observed B-mono median PFS before applying the R/R-rituximab FL survival benefit (Figures 3 and 4).

Figure 2. The National LymphoCare Study vs GADOLIN B-mono survival. GADOLIN trial and National LymphoCare Study curves were derived from separate analyses, but are presented together here for comparison; survival data from the two studies should not be interpreted as being concurrently observed or methodologically similar. The months at which the GADOLIN trial B-mono and LC R/R-rituximab curves cross are where the NLCS-derived hazard ratios are thereafter applied in the modeled population. The National LymphoCare Study analysis population included any patient who received a first-line rituximab-containing regimen, progressed during treatment, and received a second-line rituximab-containing regimen. The National LymphoCare Study relapsed patients progressed more than 6 months after the conclusion of first-line rituximab therapy; refractory patients progressed less than 6 months. Abbreviations. B-mono, bendamustine monotherapy from the GADOLIN trial of rituximab-refractory patients; OS, overall survival; PFS, progression-free survival; KM, Kaplan-Meier curve; R/R-rituximab, rituximab-relapsed or refractory; LC, National LymphoCare Study.

Figure 3. Modeled overall survival. Overall survival curves were adjusted to reflect a R/R-rituximab population at 22 months; unadjusted curves (LC-derived hazard ratio not applied) are also presented. The GADOLIN-observed G + B treatment effect was applied through 66 months assuming proportional hazards, then G + B OS is assumed to equal B-mono mortality thereafter. G + B and B-mono, obinutuzumab plus bendamustine followed by obinutuzumab monotherapy and bendamustine monotherapy, respectively, from the GADOLIN trial of rituximab-refractory patients. Abbreviations. OS, overall survival; KM, Kaplan-Meier curve; R/R-rituximab, rituximab-relapsed or refractory; LC, National LymphoCare Study.

Figure 4. Modeled progression-free survival. Progression-free survival curves were adjusted to reflect a R/R-rituximab population at 15 months; unadjusted curves (LC-derived hazard ratio not applied) are also presented. G + B and B-mono curves were modeled independently due to observed violation of the proportional hazards assumption. G + B and B-mono, obinutuzumab plus bendamustine followed by obinutuzimab monotherapy and bendamustine monotherapy, respectively, from the GADOLIN trial of rituximab-refractory patients. Abbreviations. PFS, progression-free survival; KM, Kaplan-Meier curve; R/R-rituximab, rituximab-relasped or refractory; LC, LymphoCare.

Regimen parameters

G was administered as an absolute (flat) IV dose of 1,000 mg for six cycles during induction on days 1, 8, and 15 of cycle 1; and on day 1 only of cycles 2–6; and every 2 months thereafter until progression or for up to 2 years (whichever was earlier). In the G + B regimen, B was administered IV at a dose of 90 mg/m²/day on days 1 and 2 of cycles 1–6 of each 28-day cycle. In the control arm, B was administered as monotherapy at a dose of 120 mg/m²/day on days 1 and 2 of cycles 1–6 of each 28-day cycle. We utilized the actual average dose per administration as recorded in the GADOLIN study (average amount per population), wherein the vial utilization is calculated based on the mean amount of drug administered per population and the observed number of administrations.

Adverse event parameters

We included all GADOLIN-reported grade 3–5 adverse events (higher grades indicating more severe AEs, 5 = death) that were at least 2% greater in the G + B regimen compared to the B-mono regimen. The frequency of each event was modeled as the total number of occurrences in the trial. This frequency was then multiplied by the average cost per event to derive a total adverse event cost for each regimen. Costs per adverse event were based on data from the Centers for Medicare and Medicaid Services (CMS) list of Medicare Severity-Diagnosis Related Groups (MS-DRGs) for the fiscal year 2016 (Table 1)²².

Table 1. Model parameters.

Parameter	Base case	Low value	High value	PSA distribution	Source
Patient characteristics
Average age	62.06				GADOLIN Phase III trial^15
Average body weight	81.09				GADOLIN Phase III trial^15
Average height	169.66				GADOLIN Phase III trial^15
Utilities
PFS health state	0.805	0.769	0.839	Beta	Wild et al.^24
Progression health state	0.618	0.506	0.724	Beta	Wild et al.^24
Progression Free Survival (PFS)
Parametric distribution for G + B	Weibull				GADOLIN Phase III trial^15
Parametric distribution for B-mono	Weibull				GADOLIN Phase III trial^15
HR adjustment (refractory vs R/R-rituximab)	0.796				Data on file^19
Overall Survival
G + B hazard ratio vs B	0.58	0.39	0.86	Log-normal	GADOLIN Phase III trial^15
Parametric distribution for B	Weibull				GADOLIN Phase III trial^15
Treatment effect duration	66 months (5.5 years)				GADOLIN Phase III trial^15
HR adjustment (refractory vs R/R-rituximab)	0.732				Data on file^19
Drug costs
Obinutuzumab 1,000 mg	$5,779.43				Analy$ource
Bendamustine 100 mg/mL	$2,700.20				Analy$ource
Bendamustine 25 mg/mL	$675.06				Analy$ource
Administration costs
Obinutuzumab loading dose (Cycle 1)	$262	$196	$327	Log-normal	CMS^22
Obinutuzumab subsequent cycles (2–6)	$163	$123	$204	Log-normal	CMS^22
Bendamustine dose	$136	$102	$170	Log-normal	CMS^22
Adverse event costs
Anemia	$6,499	$5,199	$7,799	Log-normal	CMS^22
Febrile neutropenia	$12,019	$9,615	$14,423	Log-normal	CMS^22
Infusion-related reaction	$5,683	$4,546	$6,819	Log-normal	CMS^22
Nausea	$3,462	$2,770	$4,154	Log-normal	CMS^22
Neutropenia	$6,499	$5,199	$7,799	Log-normal	CMS^22
Pneumonia	$7,824	$6,259	$9,389	Log-normal	CMS^22
Thrombocytopenia	$8,927	$7,142	$10,713	Log-normal	CMS^22
Post-progression treatment costs
Mean costs (G + B)	$7,125	$5,344	$8,906	Log-normal	GADOLIN Phase III trial^15, Analy$ource
Mean costs (B-mono)	$11,982	$8,987	$14,978	Log-normal	GADOLIN Phase III trial^15, Analy$ource

Quality-of-life parameters

We utilized the most widely cited utility values for FL health states in the health economics literature reported by Pattengell et al.²³ and Wild et al.²⁴ (Table 1). The Wild et al.²⁴ study included a total of 222 patients with FL from eight sites across the UK who completed the EQ-5D questionnaire. This study made no distinction between PFS on or off treatment, thus the same utility was applied to both sub-states. We did not include adverse-event related disutilities due to the low frequency of adverse events in both treatment arms.

Treatment cost parameters

We used drug wholesale acquisition costs, which were derived from the 2016 Analy$ource database. The drug cost per patient per administration was calculated using the average patient baseline characteristics, and administration instructions from the relevant clinical trial publications and PIs. Drug costs account for wastage, but not vial sharing. Intravenous infusion costs for the therapies were determined using rates from the 2016 Medicare reimbursement (Table 1)²².

Post-progression treatment costs

Salvage therapy costs for patients who progressed following G + B and B-mono were based on weighted cost of a weighted average of salvage therapies (immunotherapy with rituximab [R], chemotherapy with CHOP, and immunochemotherapy [R + CHOP]) observed post-progression in GADOLIN¹⁹. The model assumes that salvage therapy costs are different for G + B vs B-mono, based on the distribution of the type of regimens (e.g. chemotherapy, immunotherapy, immunochemotherapy) received by patients in the GADOLIN trial (Table 1).

Analysis

We calculated lifetime direct medical costs, quality-adjusted life years (QALY), and life years for the G + B and B regimens in a R/R-rituximab FL population. The incremental cost-effectiveness ratio (ICER) was calculated as the difference in costs divided by the difference in QALYs. We also performed scenario analyses to evaluate assumptions in our model, including (1) variation of the PFS and OS curve cut-offs (15 months and 22 months, respectively, in base case analysis) when The National LymphoCare Study-adjusted curves are applied; and (2) variation of the duration of treatment effect HR applied to the G + B OS curve. Finally, we performed one-way and probabilistic sensitivity analyses to assess the impacts of uncertainty in model parameter estimates on the results. In one-way sensitivity analysis, one parameter at a time was varied to its low and high value, while keeping all other parameters constant; in probabilistic sensitivity analysis, all model parameters were simultaneously, randomly varied according to an assigned probability distribution over 5,000 simulations, and 95% credible ranges (CR) were calculated for each model result.

Results

Base case results are presented in Table 2. G + B resulted in 5.54 (95% CR = 3.04–7.88) QALYs and 7.82 (95% CR = 3.78–11.54) life years, at a lifetime cost of $121,000 (95% CR = $119,400–$124,100). B-mono resulted in fewer QALYs (4.30, 95% CR = 2.15–6.52) and life years (6.44, 95% CR = 2.98–9.76), but at lower cost ($62,900, 95% CR = $60,600–$66,300). Comparing G + B vs B-mono, incremental differences in QALYs (1.24, 95% CR = 0.61–1.87) and cost ($58,100, 95% CR = $54,500–$61,500) resulted in an incremental cost-effectiveness ratio (ICER) of $47,000 (95% CR = $30,900–$94,700). The cost of treating progressive disease was lower for G + B ($5,500) vs B-mono ($9,600), due to the lower rate of progression conferred by G + B. However, the cost of adverse events was higher for G + B ($3,300) vs B-mono ($1,400), due to longer G + B PFS and, thus, greater exposure to therapy. The difference in cost was primarily driven by increased drug costs for G + B vs B-mono in the induction and maintenance periods ($112,200 vs $52,000) as well as increased administration costs ($1,800 vs $600).

Table 2. Base case model results.

	Cost	QALYs	Life years	ICER
G + B	$121,024	5.54	7.87
B-Mono	$62,929	4.30	6.439
Incremental	$58,096	1.24	1.43	$47,016

The incremental QALY gain for G + B predicted by the model was primarily driven by the difference in modeled OS curves, which was always positive due to the OS HR not drawing values greater than 1 from its probabilistic distribution. One-way sensitivity analysis showed that our results were most sensitive to the overall survival hazard ratio of G + B vs B, followed by drug costs, the cost of subsequent treatment in the progressive disease state, and utility values for each health state; however, the ICER was generally robust to changes in model parameters, staying below $100,000 per QALY gained in all one-way analyses (Figure 5). Probabilistic sensitivity analysis results are presented as a cost-effectiveness acceptability curve, which shows the Bayesian probability that the results are cost-effective at increasing willingness-to-pay per QALY thresholds (Figure 6). Probabilistic sensitivity analysis showed that G + B had 49% and 94% probabilities of being cost-effective vs B-mono at $50,000 and $100,000 willingness-to-pay thresholds, respectively. We also varied the parametric distributions used for the PFS curves (both arms) and OS curve (B only), which resulted in ICERs ranging from ∼$39,000–$50,000 (Table 3).

Figure 5. One-way sensitivity analysis. The tornado diagram shows the impact of variability in individual parameters on the ICER result.

Figure 6. Probabilistic sensitivity analysis. The cost-effectiveness acceptability curve (CEAC) shows that G + B becomes increasingly cost-effective as willingness-to-pay (WTP) per QALY gained increases. At $50,000 WTP and $100,000 WTP, G + B is cost-effective vs B-mono in 55% and 98% of probabilistic simulations, respectively.

Table 3. Results using different parametric curve fits. The Weibull distribution was chosen as the base case curve for both PFS (both arms) and OS (B arm only; G + B OS was derived through application of the OS HR to the B arm).

Curve fit	G + B PFS			B PFS			B OS
	AIC score	AIC rank	Resulting ICER	AIC score	AIC rank	Resulting ICER	AIC score	AIC rank	Resulting ICER
Exponential	437.9	4	$47,950	423.3	5	$47,905	605.4	5	$45,665
Weibull	439.7	5	$47,016	412.8	4	$47,016	607.0	4	$47,016
Log Normal	434.9	1	$39,125	397.8	2	$49,861	605.0	2	$39,220
Gamma	436.7	3	$40,320	399.4	3	$49,089	606.6	3	$40,388
Log Logistic	435.1	2	$39,998	390.4	1	$49,508	605.5	1	$39,799
Gompertz	439.9	6	$47,950	424.7	6	$47,227	607.4	6	$45,665

Variation of the cut-off times when we applied the National LymphoCare Study-derived hazard ratios to estimate a R/R-rituximab population had little impact on model results. When we turned off the National LymphoCare Study adjustments entirely, essentially modeling a refractory-only population, the QALY difference for G + B vs B-mono decreased to 1.12 QALYs, but with little change in incremental cost, leading to an ICER of $51,700. When we applied the National LymphoCare Study adjustment sooner (e.g. at 1 month), PFS and OS for both arms improved as expected; however, the incremental QALYs (1.18) and ICER ($49,100) were similarly unaffected. The changes in results were more pronounced when we varied the duration of the proportional hazards assumption in OS curves. Assuming a treatment effect duration of only 2.5 years decreased incremental QALYs (0.84) and increased the ICER ($68,900), whereas a treatment effect duration of 8 years led to increased QALYs (1.48) and a decreased ICER ($39,100).

Discussion

We estimated the cost-effectiveness of the GADOLIN trial’s G + B vs B-mono in FL patients who relapsed after, or were refractory to, a rituximab-based regimen in the US. The GADOLIN trial indicated that G + B is an important advance in the treatment of patients with FL who are no longer benefiting from rituximab. Our findings expand upon the trial results by highlighting the key health economic implications for stakeholders facing treatment, recommendation, or coverage and reimbursement decisions related to R/R-rituximab FL. G + B was estimated to increase both time in the progression free health state and OS vs the B-mono, but at additional cost. Nonetheless, the ICER for G + B vs B-mono falls within commonly used thresholds for cost-effectiveness, and this result was robust to model uncertainty.

Our results imply that, although G + B was studied in a refractory population, it is likely cost-effective in a refractory or relapsed FL population. We assumed that relapsed patients have better survival than refractory patients, thus we modeled greater long-term benefit vs a standard extrapolation of the refractory-only GADOLIN trial’s G + B treatment effect. A recent publication from the National Lymphocare Registry of patients with FL with a median follow-up of 8 years reported that patients refractory to rituximab-monotherapy treatment reported a shorter median time to PFS and OS compared to non-refractory patients²¹.

Previous studies of implied willingness to pay levels for cancer in the US suggest thresholds in the range of $100,000–$300,000 per QALY gained^25,26. Notably, even when we modeled a refractory-only FL population by turning off our adjustment to the population derived from the National LymphoCare Study, G + B remained cost-effective at a $100,000 per QALY threshold.

Our modeling approach was consistent with other economic models examining similar populations in the published literature. We identified 18 previous modeling studies of rituximab in FL patients through systematic literature review. Thirteen of these utilized a similar three state oncology model structure, with progression free, progressed, and death health states. In two instances the model structure differed: (1) Boland et al.²⁷, who used a 5-state transition model including three progression-free states (induction, maintenance therapy, and post-treatment), progressive disease, and death; and (2) Blommestein et al.²⁸, who used both health states and a time-to-next-treatment (TTNT) approach.

Our model has several limitations worth noting. Long-term modeling of a R/R-rituximab population based on the National LymphoCare Study introduces a degree of uncertainty in the analysis. Nonetheless, the adjustment had little impact on incremental results between treatment strategies, since both comparators received a survival benefit, and overall the incremental results were robust to various applications of the National LymphoCare Study survival curve extrapolation and parametric distribution approaches. Second, the GADOLIN trial showed prolonged OS in G + B patients, but the data continues to mature. Nonetheless, we believe that modeling the reported, statistically significant 95% confidence interval for the GADOLIN OS hazard ratio (HR = 0.58; 95% CI = 0.39–0.86) in probabilistic sensitivity analysis, and demonstrating the impact of the upper bound of this parameter in one-way sensitivity analysis, sufficiently demonstrates that G + B remains cost-effective. Uncertainty about long-term survival will be reduced as more follow-up data from GADOLIN become available.

Third, although our model utilizes a US payer perspective, the GADOLIN trial included patients from multiple European countries and Canada, in addition to US patients, and there may be differences in healthcare and underlying health in these populations compared to the US health system. Fourth, adverse events classification as well as post-progression treatment attribution in GADOLIN relies on investigator assessment, and our model assumes that reporting of these factors was accurate and consistent. Finally, the 3-state model structure does not allow for patients to become cancer-free following a transition to the progressive disease health state, even though subsequent therapy costs are applied in the progressed state; nonetheless, FL is typically considered to be a lifelong condition¹⁹.

In conclusion, our US-based analysis suggests that treatment with G + B with G monotherapy compared to B-mono is likely cost-effective compared to B-mono in patients with FL who relapsed after, or are refractory to, a rituximab-containing regimen. This economic analysis provides policy-makers with a framework for evaluating the clinical and economic value of obinutuzumab in the modeled patient population.

Transparency

Declaration of funding

The economic analysis was sponsored by Genentech, Inc.

Declaration of financial/other interests

AM, CR, and CB are employees of Genentech, Inc. TK is an employee of F. Hoffmann-La Roche Ltd. GFG and DLV are paid consultants of Genentech, Inc. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Acknowledgments

The authors would like to acknowledge Statisticians Aino Launonen (F. Hoffmann-La Roche Ltd) and Yong Mun (Genentech, Inc.) for their assistance with analysis of the data from the National LymphoCare Study.