Polatuzumab vedotin with rituximab and bendamustine for treating relapsed or refractory diffuse large B-cell lymphoma

Company: Company used rituximab with bendamustine as comparator from trial GO29365

ERG: ERG agreed that rituximab with bendamustine was a reasonable proxy given lack of robust data on relative effectiveness of alternatives

Committee: Rituximab with bendamustine is a reasonable proxy for standard of care in NHS for relapsed/refractory DLBCL when transplant not an option, despite not being commonly used in UK

ICER impact: negligible

Company: Company conducted multivariable regression and propensity score weighted regression models to adjust for imbalances

ERG: ERG considered the company's adjustment methods appropriate with range of methods tested in sensitivity analyses

Committee: Company's adjustments for imbalances between treatment arms were appropriate

ICER impact: negligible

Company: Company used cure-mixture model assuming about two-thirds of those progression-free at 2 years were 'cured', with standardised mortality ratio of 1.41 compared to general population and no healthcare resource use after 3 years

ERG: ERG used standard independent parametric survival modelling without cure assumption, citing lack of plateau in Kaplan-Meier curve and implausible probabilistic results

Committee: Standard parametric survival modelling preferred over cure-mixture model due to insufficient evidence for cure assumption and implausible probabilistic results

ICER impact: decreases

Company: Cure-mixture model appropriate based on 2-year progression-free survival data from small trial (40 people per arm). Presented sensitivity analyses with varied cure rates.

ERG: Cure-mixture model not justified due to lack of plateau in K-M curves, implausible smoothed hazard plots, and that PFS may not be appropriate for estimating long-term remission. Cure rates not sufficiently justified.

Committee: Insufficient evidence to justify assuming a cured proportion from the outset. Cure rate estimates highly uncertain. Cure-mixture model not suitable for decision making due to lack of robust long-term evidence.

ICER impact: increases

Company: Initially used individual patient-level approach based on age distribution in trial. Updated to single-age cohort approach (69 years) in response to committee feedback.

ERG: Used single age cohort-based modelling in revised base case, consistent with methods for PFS and OS modelling.

Committee: Single-age cohort approach (69 years) preferred and appropriate for modelling background mortality.

ICER impact: negligible

Company: Company submitted trial evidence showing broadly reflective population in terms of age and previous treatments

ERG: ERG raised concerns about underrepresentation of non-white people and most patients having ECOG 0-1

Committee: Trial GO29365 is generalisable to the UK, with clinical experts confirming population is representative and ethnicity/ECOG status not major concerns

ICER impact: negligible

Company: Maximum of 6 cycles of treatment in line with licence and trial GO29365 protocol. No patients had more than 6 cycles; time to off-treatment curve reflects delayed doses.

ERG: Concerned that 5% of patients appeared to have more than 6 cycles based on K-M curve. Revised base case included drug costs for patients with delayed doses.

Committee: Company's assumption of 6 cycles appropriate as it reflects clinical practice and marketing authorisation. ERG's change to include delayed dose costs had small effect (<£2,000 per QALY) and not a key driver.

ICER impact: negligible

Company: Company initially used generalised gamma cure-mixture model, then updated to log-normal cure-mixture model

ERG: ERG used standard independent parametric survival modelling with generalised gamma, log-logistic, or log-normal distributions

Committee: Standard parametric approaches with generalised gamma preferred over cure-mixture distributions

ICER impact: uncertain_direction

Company: Initially used cure-mixture model; later presented scenario analysis using standard independent parametric survival model with generalised gamma distribution

ERG: Used standard independent parametric survival modelling with generalised gamma for PFS and either log-logistic or log-normal for OS. More standard approach that captures long-term survival.

Committee: Standard parametric survival modelling preferred over cure-mixture model. Revised standard parametric modelling with generalised gamma for PFS and either log-logistic or log-normal for OS is appropriate.

ICER impact: decreases

Company: Based on ZUMA-1 trial using EQ-5D-5L from small sample of patients with mixed histology lymphoma. Noted that chosen values produced most conservative ICER estimates.

ERG: Identified alternative utility sources but did not consider them any better than those used by the company.

Committee: Company had used best available data but considerable uncertainty remains. Utilities not a key driver of cost-effectiveness results. Committee did not endorse approach of basing utility values on ZUMA-1 trial data and expressed disappointment that no HRQoL data available from GO29365 trial.

ICER impact: negligible