TA865 · STA

Nivolumab with fluoropyrimidine- and platinum-based chemotherapy for untreated unresectable advanced, recurrent, or metastatic oesophageal squamous cell carcinoma

Recommended with restrictionsNovember 2022

Recommended for routine use only when pembrolizumab is unsuitable. Only for adults whose tumour cells express PD-L1 at 1% or more. PD-L1 testing for nivolumab and pembrolizumab suitability should be done concurrently.

Source documents

Intervention

nivolumab (Opdivo)
monoclonal antibody · blocks the programmed cell death-1 receptor (PD-1) · intravenous

Conditions

unresectable advanced, recurrent, or metastatic oesophageal squamous cell carcinomaoncology · metastatic
oesophageal squamous cell carcinomaoncology · metastatic
untreated unresectable advanced, recurrent, or metastatic oesophageal squamous cell carcinomaoncology · metastatic

Comparators

NameType Established Committee preferred
fluoropyrimidine-based and platinum-based chemotherapy aloneactive drugYes
pembrolizumab plus fluoropyrimidine- and platinum-based chemotherapyactive drugYesYes
pembrolizumab plus chemotherapyactive drugYes
chemotherapy alonestandard of careYes

Clinical trials

TrialDesignPhasePivotal
CheckMate 648RCTPhase 3Yes
KEYNOTE-590RCTPhase 3Yes

Economic model

partitioned survival (company)
Time horizon: lifetime
Cycle length: not specified

ICER

Confidential (PAS-dependent) (nivolumab plus chemotherapy versus pembrolizumab plus chemotherapy) · high uncertainty
£30,000–£50,000 (nivolumab plus chemotherapy vs chemotherapy alone (when pembrolizumab unsuitable)) · moderate uncertainty

Methodological decisions (18)

comparator selection

Chemotherapy alone versus pembrolizumab plus chemotherapy as appropriate comparators depending on PD-L1 suitability.

Company: Maintained pembrolizumab was recommended too recently to be considered standard of care; chemotherapy as main comparator

Committee: Chemotherapy is comparator when only nivolumab is suitable (PD-L1 ≥1% and CPS <10); pembrolizumab plus chemotherapy is comparator when both are suitable (PD-L1 ≥1% and CPS ≥10)

ICER impact: uncertain_direction

cost assumption

Calculation of treatment costs adjusted for delayed or missed doses. Company weighted relative dose intensity by time on treatment; ERG excluded time on therapy from base case.

Company: Time on therapy relevant for capturing relative dose intensity weighted by time each patient spent on treatment

ERG: Time on therapy not relevant; omitted from base case

ICER impact: uncertain_direction

cost assumption

Treatment cost adjustment methodology relative to dose intensity

Company: Adjusted treatment costs by relative dose intensity weighted by time on treatment

ERG: Did not weight relative dose intensity by time on treatment

ICER impact: increases

crossover adjustment

In-trial switching to anti-PD-L1 therapies (including nivolumab and pembrolizumab) observed in CheckMate 648. Company assumed no switching adjustment; ERG preferred adjustment using NICE DSU technical support document 16 methods.

Company: Few people switched to anti-PD-L1 therapy in trial; switching adjustment methods place large demands on limited data creating more uncertainty; base case excludes adjustment

ERG: Switching evidence present and difference between trial and clinical practice switching proportions suggests bias; adjustment preferred to avoid ICER underestimation

ICER impact: increases

equivalence assumption

Clinical expectation of immunotherapy 'class effect' between nivolumab and pembrolizumab

Committee: Clinical experts maintained effectiveness is almost the same in other cancers and expect consistency in oesophageal squamous cell carcinoma, but noted comparing across different PD-L1 definitions was 'risky' in terms of validity

ICER impact: negligible

indirect comparison method

Nivolumab plus chemotherapy indirectly compared with pembrolizumab plus chemotherapy using data from CheckMate 648 (nivolumab subgroup) and KEYNOTE-590 (pembrolizumab). No direct trial comparison exists.

Company: Compared populations from CheckMate 648 for whom both nivolumab and pembrolizumab were suitable, generating time-varying HRs

ERG: Preferred including people from CheckMate 648 for whom pembrolizumab was suitable to maintain maximum trial comparability. Generated time-varying HRs favoring pembrolizumab across all time points.

Committee: Acknowledged complexity and that no definitive evidence of superiority of one treatment over the other had been presented. Wide credible intervals crossed 1.

ICER impact: uncertain_direction

indirect comparison method

Company and ERG used different approaches to ITC. Company included CheckMate 648 people suitable for both drugs and KEYNOTE-590 people suitable for pembrolizumab, using log-normal distribution. ERG included only CheckMate 648 people suitable for pembrolizumab, using log-logistic distribution.

Company: Log-normal extrapolation and inclusion of people from CheckMate 648 for whom both nivolumab and pembrolizumab were suitable

ERG: Log-logistic parametric extrapolation and inclusion of only people from CheckMate 648 for whom pembrolizumab was suitable to maintain maximum trial comparability

ICER impact: uncertain_direction

indirect comparison method

ITC methodology for comparing nivolumab (from CheckMate 648) with pembrolizumab (from KEYNOTE-590). Company and ERG differed on which patient populations to include and which baseline/scaling approach to use.

Company: Included outcome data from people in CheckMate 648 for whom both nivolumab and pembrolizumab were suitable, and from KEYNOTE-590 for whom pembrolizumab was suitable. Scaled survival using chemotherapy as baseline. Used log-normal distribution for OS extrapolation.

ERG: Included people from CheckMate 648 for whom only pembrolizumab was suitable. Different distribution and baseline used for scaling pembrolizumab survival.

ICER impact: uncertain_direction

mortality assumption

Modelling of background mortality in overall survival prediction. Company and ERG disagreed on whether to include additional background mortality rate beyond predicted overall survival.

Company: Background mortality included additional to predicted overall survival, with minimal impact on results

ERG: This approach double-counts mortality; alternative methods should prevent implausibly low hazard rates

ICER impact: negligible

mortality assumption

Including all-cause mortality additional to predicted OS mortality

Company: Included all-cause mortality additional to predicted OS mortality

ERG: Not explicitly stated as different in this section

ICER impact: uncertain_direction

population generalisability

Comparability of CheckMate 648 (oesophageal squamous cell carcinoma only) with KEYNOTE-590 (squamous and adenocarcinoma, oesophagus and gastro-oesophageal junction)

ERG: Agreed subgroup appeared comparable but noted conclusions limited due to restricted characteristics presented and data availability issues

Committee: Noted uncertainty about comparability of the 2 trials used in ITC

ICER impact: uncertain_direction

survival extrapolation

Disagreement on approach to model overall survival. Company used semi-parametric method (Kaplan-Meier plus log-normal extrapolation from 6.9 months) while ERG used fully parametric log-logistic extrapolation.

Company: Semi-parametric approach better reflects changing hazard after 20 months with clear inflection point at 6 months

ERG: Parametric log-logistic extrapolation appropriate, with no clear inflection point in nivolumab arm and reasonable correspondence to observed landmarks

ICER impact: uncertain_direction

survival extrapolation

Parametric distribution choice for extrapolating progression-free and overall survival data from CheckMate 648.

Company: Semi-parametric extrapolation of PFS and OS data with log-normal distribution for OS

ERG: Different parametric extrapolation approach with modifications to PFS and OS modelling

ICER impact: increases

treatment effect waning

Company assumed no treatment waning of nivolumab effect; ERG included waning from 2.5-4 years after starting therapy and 6 months after stopping.

Company: Evidence of robust and durable treatment effect lasting beyond discontinuation; any waning assumed at 5 years after starting therapy

ERG: Treatment effect increase over time implausible; waning from 2.5-4 years is appropriate

ICER impact: increases

treatment effect waning

Whether nivolumab treatment effect wanes over time

Company: Excluded nivolumab treatment waning from base case

ERG: Included waning of nivolumab treatment effect between 2.5 and 4 years

ICER impact: increases

treatment sequencing

Adjustment to overall survival for subsequent therapy/switching

Company: Excluded any adjustments to OS for subsequent therapy

ERG: Same assumption as company in comparison with chemotherapy alone

ICER impact: negligible

utility source

Source of utility values and treatment-specific versus progression-based approaches. Company used average utility from both trial arms; ERG used treatment-specific utilities.

Company: Treatment-specific utilities not appropriate; use progression-based utilities independent of treatment

ERG: EQ-5D scores higher for chemotherapy than nivolumab plus chemotherapy; treatment-specific utilities more appropriate

ICER impact: decreases

utility source

Source and type of health-state utilities used in the model

Company: Using progression-based utilities independent of treatment received

ERG: Included treatment-specific progression utilities in the model

ICER impact: decreases

Evidence gaps

no direct comparisonIndirect comparison between nivolumab and pembrolizumab based on CheckMate 648 and KEYNOTE-590 with wide credible intervals crossing 1, uncertainty about trial comparability particularly regarding PD-L1 measurement methods and population differences
short follow upUncertainty about long-term overall survival extrapolation beyond trial follow-up, with disagreement between company semi-parametric and ERG parametric approaches
immature overall survivalTreatment effect waning assumptions uncertain, with ERG assuming waning from 2.5-4 years while company assumed lifelong effect
no direct comparisonNo direct RCT comparison between nivolumab and pembrolizumab; comparison made via indirect treatment comparison using CheckMate 648 and KEYNOTE-590 trials

Commercial arrangement

simple discount pas · confidential · critical for recommendation

Special considerations

End of life criteria met Severity modifier applied Cancer Drugs Fund eligible Equality issues raised

Cross-references

TA737comparator guidance — Pembrolizumab with platinum and fluoropyrimidine based chemotherapy for untreated advanced oesophageal and gastro-oesophageal junction cancer
TA737precedent — Semi-parametric extrapolation of overall survival and treatment waning approaches discussed; switching adjustment not included
comparator guidance — Nivolumab for HER2 negative advanced gastric, gastro-oesophageal junction or oesophageal adenocarcinoma; 5-year treatment waning assumption considered plausible