TA779 · STA

Dostarlimab for previously treated advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiency

Recommended for Cancer Drugs FundFebruary 2022

Recommended for use within the Cancer Drugs Fund only. Conditions in the managed access agreement must be followed.

Source documents

Intervention

dostarlimab (Jemperli)
anti-PD-L1 monoclonal antibody · programmed cell death ligand 1 (PD-L1) inhibitor · intravenous

Conditions

advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiencyoncology · metastatic
advanced or recurrent endometrial cancer with high microsatellite instability or mismatch repair deficiencyoncology · relapsed_refractory

Comparators

NameType Established Committee preferred
paclitaxel monotherapyactive drugYes
doxorubicin monotherapyactive drugYes
combination chemotherapyactive drugYes
hormone therapyactive drugYes
current clinical managementbest supportive careYes
chemotherapy (various options)active drugYes

Clinical trials

TrialDesignPhasePivotal
GARNETsingle_armPhase IIYes
KEYNOTE-775RCTunknown

Economic model

partitioned survival (company)
Time horizon: long-term extrapolation beyond trial data
Cycle length: not specified

ICER

Confidential (PAS-dependent) (dostarlimab vs current clinical management) · very_high uncertainty
£30,000–£50,000 (dostarlimab vs current clinical management) · high uncertainty

Methodological decisions (11)

comparator selection

Choice between real-world evidence (RWE) GARNET-like subgroup from NCRAS registry vs KEYNOTE-775 chemotherapy arm as comparator for dostarlimab

Company: RWE registry subgroup should be used as primary evidence to evaluate comparative efficacy due to large sample size, alignment to GARNET patient characteristics, and real-world representation

ERG: RWE subgroup has serious generalisability concerns due to unknown biomarker prevalence, missing ECOG data, histology differences, and differences in prior therapy lines. KEYNOTE-775 would be a more reliable comparator as a biomarker-matched RCT

Committee: KEYNOTE-775 may provide better source of comparative data. RWE subgroup is not a robust comparator for GARNET due to unknown prevalence of high MSI/MMR deficiency, differences in histology (more endometrioid in GARNET), and variations in number of prior treatment lines

ICER impact: uncertain_direction

indirect comparison method

Validity of matching-adjusted indirect comparisons (MAIC) between GARNET and RWE/other chemotherapy studies

Company: MAIC scenarios (based on expert-identified prognostic variables and statistically significant variables) show dostarlimab benefits persist after adjustment, suggesting minimal differences between cohorts

ERG: Systematic differences remain that cannot be adequately adjusted by statistical methods, including methodological issues with data collection, case definition, and selection. Small sample size studies and poorly reported patient characteristics in chemotherapy comparisons severely limit validity

Committee: MAIC scenarios cannot be considered robust sources of evidence due to high levels of uncertainty. Better data needed for robust analysis of comparative efficacy

ICER impact: increases

indirect comparison method

Validity and reliability of matching-adjusted indirect comparisons (MAIC) between GARNET trial and UK real-world evidence (RWE) cohort to estimate comparative efficacy of dostarlimab versus current clinical management

Company: Two MAIC scenarios show dostarlimab superiority in overall survival; minimal differences between GARNET and RWE-matched populations

ERG: Systematic differences remain between cohorts due to methodological issues in data collection, case definition and selection; systematic biases cannot be adjusted for by statistical methods; supplementary endometrioid-only MAIC still highly uncertain due to remaining population differences

Committee: Agreed with ERG that MAIC scenarios had high levels of uncertainty and could not be considered robust source of evidence; concluded better data needed for robust comparative efficacy analysis

ICER impact: uncertain_direction

indirect comparison method

Robustness of matched-adjusted indirect comparison (MAIC) between GARNET and real-world evidence for current clinical management

Company: Used MAIC to indirectly compare dostarlimab to current clinical management

ERG: Expressed concern about uncertainty in MAICs and differences between GARNET and RWE populations

Committee: Committee agreed there was too much uncertainty in the MAICs to be confident about robustness of any ICERs from the economic model

ICER impact: increases

population generalisability

Generalisability of RWE GARNET-like subgroup to GARNET population

Company: The RWE subgroup is closely aligned to GARNET patient characteristics

ERG: Multiple differences limit generalisability: no biomarker data in NCRAS, ECOG performance status missing for ~50% of RWE patients, histology differences (lower proportion of endometrioid in RWE), and RWE had only 1 prior platinum treatment vs GARNET allowing 1 or more prior treatments

Committee: Concerns about matching populations are significant. Clinical experts noted differences in number of prior therapy lines would be key prognostic factor. Discrepancy often exists between RCT and RWE evidence with lower observed efficacy in RWE

ICER impact: increases

stopping rule

Treatment discontinuation estimates for dostarlimab. GARNET trial contained no stopping rule; company elicited clinical expert opinions on time to treatment discontinuation based on numbers remaining at risk.

Company: Presented numbers at risk to clinical experts to obtain range of TTD estimates

ERG: Modelled TTD curve bears no resemblance to numbers at risk; presenting numbers at risk treats data cut-off as discontinuation event, seriously biasing estimates; prefers more conservative value at upper end of range

Committee: Agreed with ERG that elicitation exercise was poorly constructed and results unreliable; concluded more conservative value preferred by ERG is appropriate

ICER impact: increases

survival extrapolation

Choice of parametric curve for extrapolating overall survival beyond trial data. Company preferred generalised gamma, ERG preferred Weibull. Both models associated with substantial uncertainty due to immature data.

Company: Generalised gamma model is most clinically plausible and aligns with clinical expert feedback; provides good fit to Kaplan-Meier data and longer-term survival estimates

ERG: Generalised gamma is overly optimistic and predicts survivors beyond 30 years implausibly; Weibull is preferred due to concerns about bias in clinical expert elicitation

Committee: Committee noted not enough evidence to prefer either curve; true survival lies within wide range of possibilities

ICER impact: increases

survival extrapolation

Choice of parametric curve for overall survival extrapolation

Company: Generalised gamma curve

ERG: Weibull curve preferred; noted uncertainty about whether Weibull or generalised gamma appropriate

Committee: Committee concluded not enough evidence to prefer either Weibull or generalised gamma; true survival lies within wide range of possibilities

ICER impact: uncertain_direction

treatment effect duration

Time to treatment discontinuation (TTD) values and stopping rules elicitation

Company: Company conducted TTD and stopping rules elicitation exercise; proposed values toward upper end of range

ERG: ERG viewed company elicitation exercise as poorly constructed with unreliable results; preferred more conservative values

Committee: Committee concluded more conservative value preferred by ERG is appropriate given high degree of uncertainty

ICER impact: increases

treatment effect waning

When and how the treatment benefit from dostarlimab diminishes after discontinuation. Company assumed maintenance of full treatment effect for extended duration beyond discontinuation; ERG assumed waning begins immediately at treatment discontinuation.

Company: Referenced previous NICE appraisals accepting 3-5 years treatment effect duration when people stop immunotherapy after 2 years

ERG: No evidence to support this waning profile in this population; prefers immediate waning beginning at treatment discontinuation with declining effect over longer period

Committee: Preferred ERG's more conservative approach given immaturity of data

ICER impact: increases

treatment effect waning

Whether dostarlimab treatment benefit wanes over time

Company: No modelling of treatment waning

ERG: Both treatment waning and treatment discontinuation should be modelled

Committee: Committee agreed with ERG's preference to model both treatment waning and treatment discontinuation

ICER impact: increases

Evidence gaps

single arm evidence onlyGARNET is a single-arm study with no comparator arm, making it difficult to compare dostarlimab with current treatments
immature overall survivalData from GARNET is very immature with limited follow-up time
no direct comparisonDostarlimab has not been directly compared with other treatment options
no uk dataComparator data from real-world evidence and other trials may not be biomarker-matched or representative of UK population
single arm evidence onlyGARNET is a single-arm trial with no matched control arm; comparative efficacy based on indirect comparisons with real-world evidence cohort
immature overall survivalOverall survival data in GARNET are immature with insufficient events relative to participants; follow-up period too short to accurately predict long-term survival
no direct comparisonNo randomised controlled trial directly comparing dostarlimab with conventional chemotherapy treatments
immature overall survivalGARNET overall survival data is immature; additional survival data needed to reach decision on cost effectiveness
no direct comparisonAbsence of matched trial population for current clinical management; reliance on different real-world population for comparator data
short follow upUncertainty about long-term outcomes beyond 2 years given immature data

Commercial arrangement

simple discount pas · confidential

Special considerations

End of life criteria met Cancer Drugs Fund eligible Innovation acknowledged

Cross-references

comparator guidance — KEYNOTE-775 trial comparing pembrolizumab plus lenvatinib with paclitaxel or doxorubicin monotherapy in MMR deficient advanced endometrial cancer