TA894 · STA

Axicabtagene ciloleucel for treating relapsed or refractory follicular lymphoma

Not recommendedTechnology appraisal committee CApril 2023

Source documents

Intervention

axicabtagene ciloleucel (not stated)
genetically modified autologous T cell immunotherapy · chimeric antigen receptor T-cell therapy · intravenous

Condition

relapsed or refractory follicular lymphomahaematology · relapsed_refractory

Comparators

NameType Established Committee preferred
rituximab with cyclophosphamide, doxorubicin, vincristine and prednisolone (chop)active drug
rituximab with cyclophosphamide, vincristine and prednisolone (cvp)active drug
rituximab with bendamustineactive drug
rituximab with lenalidomideactive drug
obinutuzumab with bendamustineactive drug
standard care (blended comparator after 3 or more treatments)standard of care
standard carestandard of careYes

Clinical trials

TrialDesignPhasePivotal
ZUMA-5single_armphase 2Yes
SCHOLAR-5observational
DELTAsingle_armphase 2
ZUMA-22RCT3Yes

Economic model

partitioned survival (company)
Time horizon: Not explicitly stated in chunk
Cycle length: Not specified in chunk

ICER

Above £100,000 · high uncertainty
Above £100,000 (axicabtagene ciloleucel vs standard care) · very_high uncertainty

Methodological decisions (17)

comparator selection

Company used blended comparator approach reflecting multiple treatment options after 3 or more systemic treatments, excluding rituximab monotherapy and best supportive care

Company: Blended comparator of rituximab combinations, rituximab with lenalidomide, and obinutuzumab with bendamustine is appropriate; rituximab monotherapy and BSC excluded as would be used in less fit patients

Committee: Company's blended comparator approach and treatments included were suitable for decision making; reflects clinical practice in NHS

ICER impact: uncertain_direction

comparator selection

Selection and composition of blended comparator for standard care after 3 or more treatments

Company: Used SCHOLAR-5 data to inform comparator effectiveness; restricted cohorts to people with follicular lymphoma after at least 3 prior treatments in line with anticipated marketing authorisation; excluded idelalisib, radioimmunotherapy, CVP and experimental treatments as not representative of NHS treatments

ERG: Concerned that comparative effectiveness from single-arm studies prone to bias due to lack of randomised comparators; noted ERG's concerns about SCHOLAR-5 data quality and generalisability

Committee: Noted ERG's concerns about lack of randomised comparators; concluded using SCHOLAR-5 data acceptable to inform comparative effectiveness despite concerns

ICER impact: uncertain_direction

cost assumption

Infusion, monitoring and hospitalisation costs for axicabtagene ciloleucel treatment

Company: Applied daily hospitalisation costs of £903 to mean duration of hospitalisation observed in ZUMA-5; aligned with previous CAR T-cell therapy appraisals in inpatient monitoring setting

ERG: Noted time-on-treatment curves not consistent with derived PFS and OS curves; highlighted company capped time on treatment at overall survival; noted time on comparator treatments remains uncertain due to limited data

Committee: Noted company's calculated costs lower than NHS England Cancer Drugs Fund tariff of £96,016 (2022-23) for CAR T-cell therapy delivery in people aged 19+ years; noted clinical experts expected intravenous immunoglobulin for longer than company's 12-month estimate; noted at first meeting that tariff was best available source for current NHS costs; noted lack of full transparency on tariff derivation; recognised need to explore potential issues of double counting or undercounting costs

ICER impact: increases

cost assumption

CAR T-cell therapy delivery costs including infusion, monitoring, hospitalisation, leukapheresis, conditioning chemotherapy, and bridging therapy

Company: Company used bottom-up costing approach with individual cost categories; estimated cost lower than NHS tariff; disagreed that NHS tariff (£96,016 for 2022-23) should be used, citing lack of transparency and potential double-counting

ERG: ERG agreed company's approach was underestimate but raised concerns about NHS tariff derivation methodology; agreed transparency would be beneficial

Committee: Committee agreed that company's revised cost of £41,101 was most appropriate, as it accounted for increased staffing needs while representing a reasonable projection avoiding the high infrastructure costs in the tariff

ICER impact: decreases

cure assumption

Long-term survivor assumptions in economic model; both company and ERG assumed proportion of people could be considered long-term survivors from future time point with zero risk of progression; long-term survivors assumed to have 9% higher probability of death than general population from year 5 onwards

Company: Base-case assumption was 25% of people receiving axicabtagene ciloleucel were long-term survivors; extrapolation applied from 5 years; non-long-term survivors followed Weibull distribution fitted to full ZUMA-5 dataset; based on clinical opinion and clinical validation; long-term survivors assumed to have utility equal to general population (revised to agree with ERG after consultation)

ERG: Agreed mixture-cure model not possible due to immature ZUMA-5 data; expected proportion of long-term survivors due to unique mechanism of action but proportion cannot be validated; used Weibull distribution with hazard of death adjusted to 1.2 times higher than general population matched for age and sex; explored scenarios with 1.5x and 2x adjustment and pessimistic generalised gamma scenario; believed long-term survivors unlikely to attain general population utility while experiencing elevated mortality risk; preferred utility decrement for long-term survivors

Committee: At first meeting, concluded it was uncertain if company's long-term survival assumptions appropriate given immature ZUMA-5 data and uncertainties in SCHOLAR-5; noted long-term survivor proportion assumption had little effect on cost-effectiveness results; noted company did not clearly present model predictions for long-term vs non-long-term survivors separately; after company provided stratified survival graph in response to consultation, took both company and ERG approaches into account but noted analyses did not resolve issues with long-term and non-long-term survivor assumptions; concluded ERG's approach of using utility decrement for long-term survivors more appropriate

ICER impact: decreases

cure assumption

Long-term survivor proportion and how survivors are modelled after treatment

Company: Company made assumptions about long-term survivor proportions in the model

ERG: ERG had concerns about survivor proportion assumptions

Committee: Committee considered long-term survivor proportion assumptions as a key source of uncertainty

ICER impact: uncertain_direction

indirect comparison method

Company used propensity-score-weighted indirect comparison with SMR weighting to adjust baseline imbalances between ZUMA-5 and SCHOLAR-5 studies

Company: Propensity-score weighting with SMR weighting improved comparability between studies and reduced bias in comparative effectiveness

ERG: SMR weighting application was not transparent; stronger assumptions needed for unanchored comparison; some key covariates (e.g. follicular lymphoma subtype/grade) excluded from weighting despite large standardised mean differences

ICER impact: uncertain_direction

indirect comparison method

Company used propensity-score-weighted indirect comparison (SMR weighting) to address baseline imbalances between ZUMA-5 and SCHOLAR-5; company applied unanchored indirect comparison using propensity-score weighting and matching methods

Company: Propensity-score-weighting approach based on NICE DSU Technical Support Document 18 and recent literature; focused on covariates strongly correlated with outcomes; excluded covariates with limited effect on sample size

ERG: Propensity-score weighting should adjust for all treatment effect modifiers and prognostic variables; failure to adjust for follicular lymphoma subtypes and other lower grade subtypes in SCHOLAR-5 may have biased results in favour of current fourth line care; generating comparative effectiveness from real-world data is challenging with limited sample sizes

Committee: Committee noted the company's approach and use of propensity-score-weighting method was highly uncertain at first meeting; acknowledged ERG concerns about excluded covariates (particularly follicular lymphoma subtype); recognised trade-off between adjusting for all variables and maintaining adequate sample size; noted company explored alternative methods including G-estimation and E-value with consistent results; concluded the adjustment method used is highly complex

ICER impact: uncertain_direction

model structure

Use of single-arm trial data with external control comparator data rather than randomised comparison

ERG: ERG noted that comparative effectiveness results from single-arm studies were prone to bias due to lack of randomised comparators

Committee: Using SCHOLAR-5 data to inform comparative effectiveness was acceptable in context of lack of data for this specific population

ICER impact: increases

model structure

Company used partitioned survival model with 3 health states (pre-progression, progressed, death) to estimate cost-effectiveness; committee questioned whether mixture-cure modelling approach should be used

Company: Mixture-cure model or spline model not possible due to immaturity of ZUMA-5 data

ERG: Model structure appropriate; partitioned survival model captured all relevant health states; some uncertainties noted in model assumptions such as long-term survivor assumptions

Committee: Committee concluded that the company's model structure was appropriate for decision making

ICER impact: negligible

stopping rule

Time on treatment for comparator treatments in the model

Company: Originally used median number of treatment cycles from product characteristics and fitted exponential distribution; assumed equal subsequent treatment costs in both arms; originally allowed treatment to continue beyond progression; after technical engagement, agreed to cap time on treatment at point of progression

ERG: Noted time-on-treatment curves not consistent with PFS and OS curves; highlighted that allowing treatment beyond progression while applying subsequent treatment costs may overestimate comparator arm costs; broadly satisfied with company's updated base case; noted time on comparator treatments remains uncertain due to limited data

Committee: Clinical experts explained treatment unlikely to continue beyond progression; noted time on comparator treatments had large impact on cost-effectiveness results; despite uncertainty, concluded would accept the approach for decision making in context of this appraisal

ICER impact: decreases

stopping rule

Time on treatment for comparator treatments in relation to disease progression

Company: Original model capped time on treatment at overall survival, allowing treatment to continue beyond progression

ERG: ERG noted time-on-treatment curves were not consistent with progression-free survival and overall-survival curves; highlighted clinical implausibility of treatment beyond progression

Committee: Committee agreed with ERG that treatment should be capped at point of progression rather than continuing beyond

ICER impact: decreases

survival extrapolation

Committee preference for parametric survival distributions for extrapolating progression-free and overall survival for standard care using SCHOLAR-5 data

Company: Company used gamma extrapolation in line with its original base case; did not clearly justify the choice of gamma distribution

ERG: ERG considered generalised gamma, log-logistic and log-normal distributions to provide the best statistical fits; noted estimates of survival extrapolation were highly uncertain

Committee: Committee preferred exponential distribution for progression-free survival and gamma distribution for overall survival; noted company did not justify gamma choice; noted removing DELTA cohort from SCHOLAR-5 had large effect on cost-effectiveness results

ICER impact: increases

survival extrapolation

Choice of parametric distributions for extrapolating overall survival and progression-free survival beyond trial data

Company: Company used exponential and other distributions in original model

ERG: ERG had concerns about curve selection

Committee: Committee preferred Weibull distribution for overall survival and generalised gamma distribution for progression-free survival in axicabtagene ciloleucel arm; gamma distribution for overall survival and exponential distribution for progression-free survival in standard care arm

ICER impact: uncertain_direction

treatment effect waning

Uncertainty about long-term treatment effect of axicabtagene ciloleucel due to immature survival data and lack of plateau in Kaplan-Meier curves

Company: Post-hoc analysis censoring allogeneic stem cell transplants at 24 months showed plateau and no positive impact of subsequent transplant on overall survival

ERG: Kaplan-Meier curves from 36-month data cut showed alignment with 5-year extrapolation but with heavy censoring from 3 years onwards; 24 months too early to determine impact of subsequent treatment; longer follow-up needed

Committee: While data indicates sustained benefit, long-term treatment effect is uncertain

ICER impact: increases

utility source

Source of health-state utility values for economic model; no health-related quality-of-life data collected in ZUMA-5 and SCHOLAR-5

Company: Originally used utility values from NICE technology appraisal on lenalidomide with rituximab for previously treated follicular lymphoma based on AUGMENT study; after technical engagement, agreed with ERG and updated base case to use Wild et al. (2006) utility values; assumed long-term survivors have utility equal to general population; clarified assumptions based on previous NICE CAR T-cell technology appraisal guidance

ERG: Concerned most people in AUGMENT study were at earlier disease stage and would have higher quality of life than people after 3+ treatments; preferred Wild et al. (2006) with EQ-5D data from relapsed or refractory follicular lymphoma patients; highlighted study limitations but considered it better reflected quality of life after 3 treatments; preferred utility decrement for long-term survivors; broadly agreed with company's updated base case but considered it unlikely long-term survivors would attain general population utility while experiencing elevated mortality risk

Committee: Concluded ERG's approach of using utility decrement for long-term survivors was more appropriate; noted source of utility values had small effect on cost-effectiveness results; recognised uncertainties; noted company's assumption of rebound to general-population utility for long-term survivors favoured axicabtagene ciloleucel

ICER impact: decreases

utility source

Whether to apply a utility decrement for long-term survivors after axicabtagene ciloleucel treatment

Company: Not explicitly stated but company's base case choices implied

ERG: Not explicitly stated

Committee: Committee preferred including a utility decrement for long-term survivors after axicabtagene ciloleucel treatment

ICER impact: increases

Evidence gaps

immature overall survivalMedian follow-up in ZUMA-5 was short and survival data immature; no plateau observed in Kaplan-Meier curves based on very few events at data cut-off
no direct comparisonNo evidence on effectiveness of axicabtagene ciloleucel directly compared with standard care; single-arm study design required use of external control data
single arm evidence onlyZUMA-5 is a single-arm study; comparator data sourced from retrospective SCHOLAR-5 study
single arm evidence onlyZUMA-5 is a single-arm study with no randomised comparators; comparative effectiveness derived from non-randomised SCHOLAR-5 study
short follow upRelatively short follow-up of ZUMA-5 trial; long-term treatment effect of axicabtagene ciloleucel is uncertain despite 36-month data cut
no direct comparisonNo head-to-head comparison between axicabtagene ciloleucel and standard care; indirect comparison via propensity-score-weighted method used
immature overall survivalImmature overall survival data from ZUMA-5 limits use of mixture-cure modelling; long-term survival assumptions cannot be fully validated
immature overall survivalOverall survival data from ZUMA-5 is immature and extrapolations from the model are uncertain
no direct comparisonNo direct comparative efficacy data available for axicabtagene ciloleucel compared with standard care
short follow upLimited available evidence to inform time on treatment for comparator treatments in people having treatment after 3 or more previous treatments
immature overall survivalImmature overall-survival data from ZUMA-5
no direct comparisonNo direct comparative efficacy data for axicabtagene ciloleucel compared with standard care

Commercial arrangement

simple discount pas · confidential · critical for recommendation

Special considerations

End of life considered (not met) Innovation acknowledged

Cross-references

comparator guidance — Previous appraisals for CAR T-cell therapies including axicabtagene ciloleucel for diffuse large B cell lymphoma and primary mediastinal large B cell lymphoma (TA421) and autologous anti-CD19-transduced CD3+ cells for relapsed or refractory mantle cell lymphoma (TA567)
same drug — Earlier NICE appraisal of axicabtagene ciloleucel for treating relapsed or refractory diffuse large B cell lymphoma and primary mediastinal large B cell lymphoma (referenced in context of infusion/monitoring/hospitalisation practice)