Lead team presentation - Dapagliflozin, in combination with insulin, for treating type 1 diabetes - NICE
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Public observer slides – Part 1 (contains no CIC data) Dapagliflozin, in combination with insulin, for treating type 1 diabetes Lead team presentation 1st appraisal committee B meeting Chair: Amanda Adler Lead team: Nicholas Latimer, Nigel Westwood, Sarah Wild ERG: Warwick Evidence NICE technical team: Sharlene Ting, Ross Dent, Nicole Elliott Company: AstraZeneca 26th March 2019 © NICE 2019. All rights reserved. Subject to notice of rights. The content in this publication is owned by multiple parties and may not be re-used without the permission of the relevant copyright owner.
Key issues
• The company did not present any clinical data to demonstrate that
dapagliflozin lengthens life in type 1 diabetes. Evidence for
dapagliflozin shows a very small improvement in quality of life
relative to placebo, but company model generates results that
suggest dapagliflozin improves quality of life and extends life. Given
the clinical evidence, do the model results have face validity?
• What is a clinically significant reduction in glycated haemoglobin
(HbA1c)?
• Is dapagliflozin clinically effective?
– At 52 weeks, dapagliflozin was associated with small reduction in
HbA1c; weight loss; very small relative improvement in quality of
life compared to placebo; increased risk of diabetic ketoacidosis
– No data on mortality
• How should treatment waning be modelled?
• How should stopping treatment be modelled? 2History of appraisal
Company submission: no subgroup, licensed dose
9th October 2018
unknown (5mg or 10mg)
24th January to 21st Technical engagement: draft technical report including
February 2019 questions (based on company submission and ERG report)
CHMP positive opinion: “type 1 diabetes mellitus, when
insulin alone does not provide adequate control of blood
1st February 2019
glucose levels despite optimal insulin therapy. Patients …
should not have a body mass index below 27 kg/m2”
Stakeholder feedback to technical engagement
• Company: new evidence and analyses on indicated
21st February 2019
population
• 2 clinical experts nominated by company
Final technical report: updated based on stakeholder
feedback
3Type 1 diabetes mellitus
• Autoimmune, metabolic disease → destroys insulin-producing pancreatic cells
• Haemoglobin A1c (HbA1c) measures ‘average’ blood glucose over time
• Blood glucose and pressure, but not body weight drive risk of complications
• Complications include:
– Vascular disease: coronary, cerebrovascular, peripheral: ‘macrovascular’
– Neuropathy: autonomic, sensory
– Retinopathy, cataracts, maculopathy
– Nephropathy
– Other: diabetic ketoacidosis (DKA), skin, psychological, etc.
– Treatment-related: low blood glucose (hypoglycaemia)
• UK prevalence: about 0.5% (400,000 people)
• Current treatment: insulin therapy
– In England, structured education, for example, ‘DAFNE’, is the norm
• In England, 70% of people with type 1 diabetes have HbA1c levels above 7.5%
(recommended targetPatient and clinical perspectives
• Management of condition can be demanding
• Constant risks and insulin dose adjustment can have considerable
psychosocial impact on wellbeing
• Affects quality of life
• No other therapy other than insulin, but difficult to achieve
consistently in-range glucose levels
• Good control of diabetes remains an unmet need
• Better management of HbA1c, and more time in range can lead to
fewer long-term complications
• Increased risk of diabetic ketoacidosis with dapagliflozin
5Relationship between high blood glucose
levels and complications – clinical evidence
• Hyperglycaemia associated with increased risk of complications
• 1 main randomised trial and follow-on epidemiological study from 1980s
Diabetes Control and Complications Trial (DCCT)
In type 1 diabetes, does improving metabolic control lower incidence of
diabetes-related complications over 5 to 10 years?
P: No retinopathy or retinopathy (n=1,441, aged 13-39, USA/Canada)
I: Insulin (multiple daily injections or pump) and target HbA1c ~6% (someone
without diabetes)
C: No more than 2 injections of insulin daily
O: Complications
Results: over mean follow-up of 6.5 years, reduced risk of microvascular
complications by over half
Epidemiology of Diabetes Interventions and Complications (EDIC)
P: willing participants from DCCT (>90%)
E: previously randomised to intensive glycaemic control
C: previously randomised to less tight glycaemic control (conventional)
O: complications 6Results of DCCT/EDIC: median HbA1c
pDapagliflozin (ForxigaTM) in type 1 diabetes
• Inhibits sodium-glucose cotransporter-2 (SGLT-2) → prevents ~90% glucose
reabsorbed in kidneys → increases urinary glucose excretion
• First oral medicine with an European licence (metformin has a French licence only)
• Marketing authorisation: “type 1 diabetes mellitus as an adjunct to insulin in
patients with body mass index ≥ 27 kg/m2, when insulin alone does not provide
adequate glycaemic control despite optimal insulin therapy”
– BMI restriction reflects safety concerns about DKA and is a subgroup
– Not recommended in patients with low insulin requirement
– “During treatment, insulin therapy should be continuously optimised to prevent
ketosis and diabetic ketoacidosis and insulin dose should only be reduced to
avoid hypoglycaemia”
– Patients should be able and committed to control ketone levels. They should be
educated about risk factors for diabetic ketoacidosis and how to recognise its
signs and symptoms
– Administration: 5mg, once daily, any time, with or without food
Does optimal insulin therapy include insulin pump, continuous glucose
monitoring and flash glucose monitoring?
8
How is ‘low insulin requirement’ defined?9
Decision problem
Population is a subgroup of evidence presented to regulators
NICE scope Company
Population Adults with type 1 diabetes on insulin Subgroup: patients with inadequate
therapy that does not adequately control of blood glucose and BMI
control blood glucose levels ≥ 27 kg/m2
Comparator Insulin with or without metformin Metformin not associated with
improvement in glycaemic control
(recent REMOVAL trial; clinical experts
advise off-label use in UK isDEPICT-1 and DEPICT-2 trials
Double-blind, randomised, placebo-controlled, international (UK sites)
Company defines ‘full analysis’ (FAS) and intention-to-treat’ (ITT) sets
8-week lead-in → 24-week double-blind treatment → 28-week extension (HbA1c unblinded)
Adults with Dapagliflozin (5mg or
8 weeks Primary endpoint at
inadequately 10mg) and insulin
to 24 weeks
controlled type 1 therapy
optimise • change from
diabetes despite (n=272) [N=1,059]
diet / baseline in HbA1c
optimised insulin
exercise
therapy (HbA1c n, indicated population;
/ insulin Placebo and insulin
7.5% to 10.5%) N, overall trial population
therapy
“Indicated” (n=289) [N=532]
population:
BMI ≥27kg/m2 Functional unblinding: insulin dosages cut by ≤20% → blood
glucose rise in placebo; on dapagliflozin → more urination
Key secondary endpoints:
• % of patients with fall in HbA1c by ≥0.5% without severe hypoglycaemia
• % change in bodyweight
• change in mean 24-hour blood glucose No mortality data
• change in % of blood glucose readings outside range
• % change in total daily insulin 10
Exploratory endpoints: EQ-5D-3L, Diabetes Treatment Satisfaction QuestionnaireIssue 1: Use of dapagliflozin in clinical practice
Background Stakeholder responses Technical team consideration
• 8-week, lead-in period • Optimised • 8-week lead-in period in
→ too short to stabilise management is part DEPICT adequate
HbA1c levels → may of standard care;
• Any carry-over effects from
affect results individualised
optimised management
• Unclear if people with a • Glycaemia in 8 likely to affect dapagliflozin
large reduction in HbA1c weeks before HbA1c and placebo arms similarly
from optimised measurement is
• Unclear if someone with
management would main driver of value
significant improvements in
have a different
• Improvements in HbA1c during optimisation
response to
lead-in period would period would respond
dapagliflozin than
affect all trial arms differently to dapagliflozin
people who do not
equally than someone who did not
What are the committee’s views?
11‘Indicated’ population: baseline
characteristics
DEPICT-1 DEPICT-2
Mean (standard deviation), unless
Dapagliflozin Placebo Dapagliflozin Placebo
specified
(n=145) (n=154) (n=127) (n=135)
Age (years) 46 (13) 45 (13) 43 (13) 45 (13)
Body mass index (kg/m2) 32 (5) 32 (4) 32 (4) 32 (4)
Duration of diabetes (years) 21 (12) 23 (12) 20 (10) 21 (12)
HbA1c (%) 8.5 (0.7) 8.4 (0.6) 8.4 (0.6) 8.4 (0.6)
Total insulin dose (rounded)
- Dose (IU) 72 (54) 73 (32) 72 (31) 69 (27)
- Dose/weight (IU/kg) 0.79 (0.64) 0.77 (0.28) 0.77 (0.28) 0.75 (0.25)
Method of administering insulin, %
- Injections 57% 58% 54% 59%
- Pump 43% 42% 46% 42%
Continuous glucose monitoring, % 29% 30% 36% 26%
HbA1c range at randomisation, %
- ≥7.5% toCONFIDENTIAL
Issue 3: Generalisability of DEPICT population
Background Stakeholder responses Technical team
consideration
• More people likely to take • Patients in DEPICT likely • Any guidance
drugs that affect the renin- reflect patients to be treated recommending
angiotensin-aldosterone with dapagliflozin in NHS dapagliflozin
system (RAAS) in NHS than for use in NHS
• Greater use of RAAS
in DEPICT (XXX) should exclude
blocking agents in UK than
people starting
• People on corticosteroids in DEPICT would not affect
on systemic
excluded from DEPICT HbA1c and weight lowering
corticosteroid
efficacy
therapy
• People on corticosteroids
excluded from DEPICT so
not to affect results
What are the committee’s views?
Would greater use of ACE inhibitors (and other factors) change absolute
baseline risk and therefore absolute difference in events?
Should dapagliflozin use exclude people starting on systemic corticosteroid
13
therapy?ERG comments on ‘indicated’ population
• Some results have not been provided by company so cannot
comment about generalisability
• Randomisation in DEPICT did not account for BMI stratification →
post-hoc subgroup breaks randomisation
• Compared to indicated population, UK patients have a:
– lower BMI (25.4 – 27.2 kg/m2 vs 31.5 – 31.9 kg/m2 )
– higher male population (56% – 60% vs 40% – 53%)
– lower use of insulin pump therapy (15.3% in England and 5.8% in
Wales vs 42% – 46%)
– similar mean HbA1c levels (8.6% in England vs 8.4% – 8.5%)
14
BMI, body mass index‘Indicated’ population: pooled trial results
on HbA1c and weight
At 52 weeks, small changes in HbA1c and weight results differ by analysis set
Company base case used data at 52 weeks from full analysis set
Adjusted mean change from baseline (standard error)
24 weeks 52 weeks
Dapagliflozin Placebo Dapagliflozin Placebo
Full analysis set (FAS): all randomly assigned patients who received at least one dose
of study drug during 24-week treatment period
HbA1c change -0.44 (0.05) -0.01 (0.05) -0.26 (0.06) 0.08 (0.06)
Weight change (%) -3.11 (0.29) -0.01 (0.30) -3.42 (0.29) 0.49 (0.30)
Weight change (kg) -2.86 (0.27) -0.01 (0.27) -3.15 (0.26) 0.45 (0.28)
Regardless of stopping randomised treatment (equivalent to Intention-to-treat; ITT)
HbA1c change -0.43 (0.05) -0.01 (0.05) -0.24 (0.06) 0.09 (0.06)
Weight change (%) -3.12 (0.29) -0.02 (0.30) -3.35 (0.27) 0.25 (0.28)
Weight change (kg) -2.87 (0.26) -0.02 (0.27) -3.08 (0.25) 0.23 (0.26)
Which analysis set is preferred? 15Issue 2: Minimum clinically significant reduction in HbA1c
Background Stakeholder comments
• Minimum clinically • Benefits of HbA1c reductions on complications are
significant change in continuous and not discrete
HbA1c levels → • Changes in glucose variability, hypoglycaemia and
consider measurement weight are important outcomes
error, natural variability • Absolute reduction of 0.3% is meaningful
in readings over time • A 10% reduction in risk is clinically meaningful
and baseline HbA1c • Dependent on baseline: achieving a 0.5%
levels reduction is more difficult starting from 7.5% than
9.5%
– Suggest 0.4% for a baseline HbA1c of‘Indicated’ population: insulin dose at 24
weeks using full analysis set
Adjusted mean change from
Difference from placebo
baseline (95% CI)
(95% CI; nominal p value)
Dapagliflozin Placebo
-10.1 1.5 -11.5
DEPICT-1
(-13.1, -7.0) (-2.1, 5.2) (-15.2, -7.5; pDEPICT: results on quality of life
Adjusted mean change from baseline (95% CI),
dapagliflozin vs placebo
Full analysis set
24 weeks 52 weeks
DEPICT-1 DEPICT-2 DEPICT-1 DEPICT-2
‘Indicated’ population
Overall treatment 1.7 1.7 1.0 1.7
satisfaction (DTSQ) (0.67, 2.82) (0.56, 2.92) (-0.27, 2.20) (0.48, 2.85)
Perceived frequency -0.6 -0.2 -0.3 -0.1
of hyperglycaemia (-0.88, -0.27) (-0.55, 0.10) (-0.58, 0.08) (-0.50, 0.25)
Perceived frequency -0.1 -0.1 -0.1 -0.2
of hypoglycaemia (-0.41, 0.23) (-0.45, 0.25) (-0.40, 0.29) (-0.52, 0.19)
Health status 1.4 7.8 2.6 6.6
(EQ VAS) (-1.96, 4.66) (2.25, 13.36) (-0.33, 5.51) (0.70, 12.59)
Overall trial population
Health status 1.5 2.0 1.6 1.6
(EQ VAS) (-0.80, 3.89) (-1.88, 5.77) (-0.50, 3.76) (-2.30, 5.49)
Health status 0.01 -0.02 0.01 -0.02
(EQ-5D-3L) (-0.02, 0.03) (-0.04, 0.01) (-0.01, 0.04) (-0.04, 0.01)
Is dapagliflozin effective at improving quality of life? 18
CI, confidence intervalAdjudicating DKA in DEPICT
19‘Indicated’ population: 52-week safety
Dapagliflozin is associated with ‘ketone-related’ adverse events in ‘indicated’ population
Company confirmed that there were no cases of Fournier’s gangrene in DEPICT
DEPICT-1 DEPICT-2
Full analysis set Dapagliflozin Placebo Dapagliflozin Placebo
(n=159) (n=154) (n=127) (n=135)
≥1 AE related to drug 37% 16% 32% 17%
AE leading to stopping 3.8% 3.9% 8.7% 5.2%
≥1 SAE related to drug 2.5% 0.6% 4.7% 2.2%
SAE leading to stopping 1.9% 0.6% 4.7% 1.5%
Death 0 0 0 0
AE of special interest
- Genital infection 18% 3.9% 12% 4.4%
- Urinary tract infection 10% 6.5% 13% 7.4%
- Renal function events 2.5% 0.6% 1.6% 0.7%
- Fractures 1.9% 3.9% 2.4% 0.7%
- Volume depletion events 0 1.9% 3.9% 2.2%
- Hypersensitivity 6.3% 3.2% 7.9% 8.1%
- Cardiovascular event 0.6% 0.6% 0.8% 0.7%
≥1 ketone-related SAE* 1.3% 0.6% 3.9% 0.7%
Ketone SAE* leading to stopping 0 0 3.1% 0
Definite diabetic ketoacidosis event 1.3% 1.3% 2.4% 0.7%
*Includes diabetic ketoacidosis, ketoacidosis and ketosis; (S)AE, (serious) adverse event
20
Are patients/clinicians likely to accept the increased risk of DKA?‘Indicated’ population: hypoglycaemia over
52 weeks (FAS)
Dapagliflozin is associated with an increased risk of hypoglycaemia
DEPICT-1 DEPICT-2
Full analysis set Dapagliflozin Placebo Dapagliflozin Placebo
(n=159) (n=154) (n=127) (n=135)
Events, n 4038 4158 3868 3730
Patients with ≥1 event, % 83% 79% 90% 85%
Severe – requires 3rd party help
- Events, n 27 17 16 45
- Patients with ≥1 event, % 13% 8% 10% 10%
- Exposure-adjusted incidence rate per 17.8 12.3 13.6 38.3
100 patient years
Documented symptomatic glucose 70
mg/dL (10.5 mmol/l)
- Events, n 3295 3453 3203 2967
- Patients with ≥1 event, % 79% 73% 87% 81%
- Exposure-adjusted incidence rate per 2177.7 2495.1 2719.2 2522.6
100 patient years
Are patients/clinicians likely to accept the increased risk of
21
hypoglycaemia?Summary of clinical evidence
• At 52 weeks, data from pooled DEPICT trials showed that
dapagliflozin was associated with:
– small reduction in HbA1c (0.26%) compared with baseline
– weight loss (3.15kg) compared with baseline
– very small relative improvement in quality of life compared to
placebo
– increased risk of diabetic ketoacidosis compared to placebo
• No data on mortality
Is dapagliflozin clinically effective? 22Cost effectiveness
23Where do QALY gains come from in
company’s model?
Treating
type 1 diabetes
Company assumes Company assumes
QALY gains here QALY gains here
Length of life Quality of life
Increase in QALYs comes from improving quality of life
and increasing length of life as a result of:
• reduction in HbA1c that lowers the risk of diabetes-
related complications 24
• weight loss that is associated with an increase in utility
QALY, quality-adjusted life yearCompany’s Cardiff type 1 diabetes model
• Patient-level, fixed-time-increment, Monte Carlo microsimulation → simulates
disease progression using risk equations over life-time horizon
• 6 month cycle; no half-cycle correction
• Risk equations to fit data from DCCT/EDIC for microvascular complications
and Swedish National Diabetes Registry for macrovascular complications
• Cohorts of 1,000 individual patients in each ‘run’ of model
• Company models each patient with same starting conditions: identical set of event
probabilities, unit costs and utility values are applied to their simulated progression.
Model captures random variability in outcomes between identical patients in each
cohort
• Patient cohort enters model with baseline characteristics and modifiable risk
factors. Variables’ values may change as simulation progresses, affecting risk of
complications
• Company assumed no progressive increase in risk factors (for example, HbA1c
and weight) based on clinical advice
In the company’s publication of model (McEwan et al. 2016), a progressive
increase in HbA1c of 0.045% was included compared to 0% in this appraisal.
Which is an appropriate assumption?
25
DCCT, Diabetes Control and Complications Trial; EDIC, Epidemiology of Diabetes Interventions and ComplicationsStudies in type 1 diabetes
Study Description Modelled?
DCCT/EDIC See slides 6 and 7. Assessed incidence and predictors Yes
of macrovascular and microvascular events
Wisconsin Epidemiologic Population-based study of 955 patients with T1DM in No
Study of Diabetic South Wisconsin, USA. Examined cumulative incidence
Retinopathy (WESDR) of macular oedema and relation to risk factors
Pittsburgh Epidemiology Prospective cohort of 1,124 patients with T1DM in or No
of Diabetes Complications near Pittsburgh, USA. Investigated risk of microvascular
Study (EDC) complications over time
Finnish Diabetic Prospective cohort of 29,906 patients with T1DM aged No
NephropathyValidation of model: company feedback (1)
• Model similar to established T1DM models: modelling approach and use of
DCCT/EDIC data to model disease progression (Company submission,
page 117)
• Cardiff model: internal and external validation, 2 peer-reviewed articles,
Mount Hood Diabetes Challenge (Company submission, page 177)
• Mount Hood Challenge: involve simulation of outcomes for hypothetical
patient cohorts and validation of model predictions against real-world data.
Ability of models to predict outcomes of clinical trials and observational
studies is assessed and compared
– No differences in prediction of events between model used in Mount
Hood and in this submission
– Company is unable to provide a comparison of Cardiff model results
against other modelling groups for T1DM analysis (Company’s
clarification response #1, B5)
27
T1DM, type 1 diabetes mellitusValidation of model: company feedback (2)
• Internal and external model validation (McEwan et al. 2016)
• Internal validation of: CT1DM Model’s equations to source data, and results of model’s
clinical endpoint predictions
• Available external clinical validation studies suitable for assessing model’s predictive
performance are limited in T1DM → DCCT/EDIC is basis of model’s progression rates
• External consistency of model’s predictions: compared with 5 other T1DM models (Sheffield
model; CRC; McQueen et al.; CORE model; Wolowacz et al.)
– CT1DM model started with baseline cohort, cost and health utility profiles consistent with
other models, and outputs compared over relevant time horizons
– Validation coefficient of determination for: clinical endpoints, R2 = 0.863 (internal R2 =
0.999; external R2 = 0.823); total costs R2 = 0.979; total QALYs R2 = 0.951
• External consistency of model’s predictions: compared with outputs from 3 economic
evaluations that used CORE model in NG17 (long-acting insulin and insulin regimens,
HbA1c thresholds, glucose monitoring strategies)
– CT1DM model started with baseline characteristics, costs and treatment profiles
consistent with NG17, and predicted outputs compared over relevant time horizons
– High degree of linear correlation between predicted endpoints in CT1DM model and
NG17; overall validation coefficient of determination R2 = 0.988
28
In what ways has the model been validated? In what ways has it not?Validation of model: DECLARE-TIMI 58 trial
Background Stakeholder comments
Company: • It is not appropriate to model outcomes for a
• No long-term data on population with type 2 diabetes using
dapagliflozin use in type 1 epidemiological evidence from type 1
diabetes diabetes → Cardiff T1DM model would not
• Data from type 2 diabetes be expected to accurately predict outcomes
supports continuation of treatment observed in DECLARE-TIMI 58 trial
differences between dapagliflozin
and placebo over 4 years • Benefits detected in DECLARE-TIMI study
(DECLARE-TIMI 58) are likely to apply to people with type 1
diabetes at same dose of drug. But patients
with type 1 diabetes may not be at such risk
of cardiovascular events because they will
be younger, less insulin resistant, and less
obese.
If DECLARE-TIMI, the large cardiovascular placebo-controlled safety
trial, is appropriate for cardiovascular safety, then how well does the
Cardiff T1DM model predict the cardiovascular results?
29Areas of uncertainty
Issue Why issue is important Impact on ICER
Company used data from DCCT/EDIC Unclear if lower If effectiveness of
to develop risk equations → predict magnitude of HbA1c dapagliflozin on
relationship between changes in HbA1c changes seen in DEPICT reducing risk of some
levels (among other risk factors) and than in DCCT would long-term
some long-term complications translate to reduced risk complications are over-
• Over 10 years, DCCT: intensive vs of long-term estimated in model →
conventional → 10 mmol/mol (2%) complications observed in likely worsen cost-
reduction in HbA1c (Slide 7) DCCT/EDIC effectiveness estimates
• DCCT relative changes are larger
than in DEPICT (0.26% at 1 year)
ERG not able to validate all parameter These are important Unknown
inputs for 3 of the 4 sub-models: components of the
• Diabetic retinopathy and macular simulation model
oedema progression
• Diabetic nephropathy
• Diabetic neuropathy
Do small reductions in HbA1c over a much shorter time period have a
proportional effect?
30
DCCT, Diabetes Control and Complications Trial; EDIC, Epidemiology of Diabetes Interventions and ComplicationsCardiff type 1 diabetes model: model run
• Retinopathy and macular oedema:
background diabetic retinopathy,
peripheral diabetic retinopathy,
severe vision loss, macular oedema
• Nephropathy: micro-albuminuria,
macro-albuminuria, end-stage renal
disease, dialysis, renal transplant
• Neuropathy: diabetic peripheral
neuropathy, ulcer and amputation
events (uncomplicated ulcer, deep
foot infection, foot ulcer and critical
ischaemia, minor amputation, major
amputation and fatal amputation)
• Cardiovascular disease: fatal and
non-fatal events
• Hypoglycaemia: symptomatic,
nocturnal and severe hypoglycaemia
• Depression not captured
31Modelled baseline patient characteristics
based on DEPICT
Baseline characteristic Mean (rounded) Standard error
Current age (years) 45 0.56
Proportion female 0.54 0.02
Proportion smokers 0.06 0.01
Duration of diabetes (years) 21.6 0.50
HbA1c 8.4% [10.8 mmol/mol] 0.03 [0.03]
Total cholesterol (mmol/l)* 4.8 0.04
HDL cholesterol (mmol/l)* 1.6 0.02
Systolic blood pressure (mmHg) 126 0.62
Diastolic blood pressure (mmHg) 77 0.40
Weight (kg) 92.1 0.69
eGFR (mL/min/1.73m2) ‡ 87 0.73
*Converted from mg/dL
32Modelled clinical history at baseline
Proportion of cohort
Standard
Mean
error
Cardiovascular disease 0.23 0.02
Background retinopathy 0.33 0.02
Microalbuminuria 0.11 0.01
Neuropathy 0.26 0.02
Peripheral vascular disease 0.03 0.01
Lower extremity amputation minor (assumed) 0.01 0
Hyperlipidaemia 0.55 0.02
Hypertension 0.46 0.02
Renin-angiotensin-aldosterone system inhibitor therapy 0.49 0.03
Proliferative retinopathy, severe vision loss, macular oedema,
macroalbuminuria, macroalbuminuria with impaired
glomerular filtration rate, dialysis, transplant, uncomplicated 0 0
foot ulcer, deep foot infection, foot ulcer and critical
ischaemia, major amputation
Does the modelled population represent the type of patient who cannot
otherwise achieve good glycaemic control in England?
What is the appropriate population to model – DEPICT population or people in
England with type 1 diabetes? 33Issue 4: Data in economic model
Background Stakeholder responses Technical team
consideration
Model should • Model cannot accommodate both 24 Not possible to
incorporate all available and 52 week data at same time include both 24-
trial data at 24 weeks • In company base case, 52-week week and 52-
and 52 weeks treatment effects applied week data in
• Sensitivity analysis using 24-week current model
treatment effects: improved ICER
estimate (£7,106 versus £9,175 in
base case)
ERG comments: disagree with
company’s response around inability to
implement suggested changes
Is the model fit for purpose if it cannot accommodate all trial data? 34Modelled treatment effects
Based on pooled DEPICT data Mean (standard error) Company base
at 52 weeks only case
Dapagliflozin Standard of Life years Utility
care gained gained
HbA1c change (%) -0.26 (0.06) 0.08 (0.06) 0.17 0.23
Total cholesterol change (mmol/l) No change No change - -
HDL change (mmol/l) No change No change - -
Systolic blood pressure change No change No change - -
(mmHg)
Diastolic blood pressure change No change No change - -
(mmHg)
Weight change (kg) -3.15 (0.26) 0.45 (0.28) 0 0.15
In which risk equation does lower HbA1c increase length of life?
How does the company model HbA1c over the long term?
What complications does weight change affect, if any?
Do the changes on life-years and utility have face validity? 35Issue 5: Extrapolating treatment effects after DEPICT (1)
Background Stakeholder comments
If treatment effect is not maintained • DEPICT showed no evidence of waning of
over time → health gains related to treatment effect on weight loss at 52 weeks
dapagliflozin would be lower → → continued and undiminished efficacy
worsen cost-effectiveness estimate
• For HbA1c, clinical experts suggest a
gradual increase after initial decrease, but
ERG comments HbA1c would not return to baseline
• Treatment effects on HbA1c wane • Sensible to account for any potential
from 24 to 52 weeks waning effect of treatment; unlikely all
• Benefit at 52 weeks is small and benefits will return to baseline for all
may add little benefit in preventing patients. Biological efficacy of drug does not
long-term complications except in seem to change with time, suggesting any
patients with a high risk of treatment waning may reflect clinical issues
cardiovascular disease
How are treatment effects of dapagliflozin expected to change over
time while on treatment?
36Issue 5: Extrapolating treatment effects after DEPICT (2)
Company base case:
• 52-week pooled DEPICT effects on HbA1c and weight applied to 1st cycle →
effects maintained while patients remain on treatment
• Stopping treatment: annual probability because of adverse events only in year 1
• Following stopping, risk factor levels return to baseline
Effect
Scenario Loss of effect Treatment stopping
data
Base 1-year probability because of
52-week Maintained while on treatment
case adverse events only
I 1-year probability because of
24-week Maintained while on treatment
adverse events only
II 52-week HbA1c and weight effects lost over
III 24-week second year of dapagliflozin treatment 1-year probability because of
IV 52-week HbA1c effect lost over second year of adverse events + all remaining
V dapagliflozin treatment; weight effect patients stop at 2 years
24-week
maintained
Which scenario is most clinically plausible? 37Changes in HbA1c and weight
Description of scenarios IV and V suggests that weight effect of dapagliflozin is
maintained, different to graphs. What happens to weight effect?
38
How do HbA1c and weight change in standard of care arm for each scenario?Issue 6: Stopping treatment Background • Some people may stop treatment for any reason in year 1 and beyond • Treatment should stop if there is no improvement in glycaemic control, based on a combination of change in HbA1c and hypoglycaemic events Stakeholder comments • Reasons to stop treatment: adverse events (diabetic ketoacidosis), renal decline, not effective (HbA1c, weight, hypoglycaemia, glycaemic variability), risk factors for adverse events (not compliant with ketone testing) • No explicit stopping rule; decision left to physician and individual patient • Stopping rules: – lack of response (HbA1c
Incidence of stopping treatment, adverse
events, DKA and hypoglycaemia – 52 weeks
Mean (Standard error)
‘Indicated’ population
Dapagliflozin Standard of care
Annual probability
- Stopping due to adverse events 0.06 (0.01) 0.05 (0.01)
- Urinary tract infection 0.11 (0.02) 0.07 (0.02)
- Genital tract infection 0.15 (0.02) 0.04 (0.01)
- Diabetic ketoacidosis 0.02 (0.01) 0.01 (0.01)
Annual number of events
- Non-severe, symptomatic
24.15 (0.30) 25.08 (0.31)
hypoglycaemia
- Severe hypoglycaemia 0.16 (0.02) 0.24 (0.03)
Overall trial population Dapagliflozin Placebo
5mg (n=271) (n=272)
% stopping for any reason 14.4% 18%
40Stopping treatment in standard of care arm
Company base case:
• Stopping treatment due to adverse events in 1st year. After stopping,
simulated risk factors (HbA1c and weight) revert to baseline levels
Rationale: to model both arms in a consistent manner based on
DEPICT data
Given that all patients are having background insulin therapy, is it
appropriate to model treatment stopping in patients in standard of care
arm?
41Issue 7: Modelling adverse events
Background Stakeholder responses Technical team
consideration
• DKA and severe • Rarity of Fournier’s Diabetic ketoacidosis
hypoglycaemia carry an gangrene precludes and severe
important risk of death and including it in model in hypoglycaemia carry
this should be accounted for any meaningful way an important risk of
in the model death and should be
• Literature suggests that accounted for in model
• Emerging serious adverse associated deaths
events should be included related to severe
in model hypoglycaemia and
diabetic ketoacidosis
are rare in UK
The company did not include the possibility of death from DKA,
hypoglycaemia or Fournier’s gangrene in its base case. Should these
be included? 42Issue 8: Utility approach
Background Stakeholder responses Technical team
consideration
Utility drives model • DEPICT not powered to detect While it may be
as most of difference differences in quality of life, and appropriate to use
in quality-adjusted longer-term trials required to capture utility values
life years (QALYs) beneficial effect of HbA1c and sourced from
are from differences reducing weight on complications literature for
in quality of life modelling, technical
rather than length of • Substantial evidence base linking team would still
life diabetes-related complications with have preferred to
quality of life is more robust than see scenario
using short-term trial data analysis including
trial EQ-5D data
ERG comments: satisfied with the
company’s overall approach
43Issue 9: Disutilities
Background Stakeholder responses Technical team
consideration
• Same source • Some utility decrements were not • DKA impacts
(Peasgood et al. 2016) sourced from Peasgood et al. because quality of life
should be used for as they lacked face validity and should be
many of utility changes • Sensitivity analysis demonstrates that recognised in
as possible use of non-significant event disutilities modelling
• Company used Lee et and/or disutility related with BMI increase • Preferred if
al. 2005 for utility from Peasgood et al. do not alter cost- company
change per change in effectiveness conclusions of base case provided
BMI → higher than analysis scenario
Peasgood et al. • Most appropriate application of utility analysis using
• DKA have an important decrements is additive approach in
impact on quality of life NG17
and should be in model ERG comments: • Unclear if
• Whether an additive or • agrees with company’s approach to ‘additive’
multiplicative approach estimating baseline utility approach is
should be used for • DKA disutility should be in base case most
disutilities depends on • prefers to see 3rd approach to modelling appropriate
source of data utilities from NG17 (minimum utility)
What are the committee’s views? 44Source of utilities
Parameter (Disutilities assumed equal in all Source
years)
Baseline utility T1DM and disutilities for Peasgood et al. 2016 (UK study reporting
background diabetic retinopathy, uncomplicated utility and disutility of T1DM complications
foot ulcer, minor and major amputation from DAFNE)
Cardiovascular disease, proliferative diabetic
retinopathy, severe visual loss, macular oedema,
NG17; Beaudet et al. 2014 (type 2 diabetes)
dialysis, transplant, neuropathy, deep foot
infection, foot ulcer and critical ischaemia
Microalbuminuria NG17
Thokala et al. 2014 (Sheffield Type 1
Macroalbuminuria
Diabetes Policy Model)
Lee et al. 2005 (UK study with mean BMI
Body mass index, per unit change
similar to DEPICT-1 [27.3 vs 28.5 kg/m2])
Currie et al. 2006 (multivariate model →
severity/frequency of hypoglycaemia related
Hypoglycaemia to fear of hypoglycaemia and changes in
utility (EQ-5D) using UK population of 1,305
patients with T1DM and T2DM)
Diabetic ketoacidosis NG17
Urinary or genital tract infection Barry et al. 1998
45Utilities used in company model
Company base case: cumulative
Parameter Mean events
Dapagliflozin Standard of care
Baseline utility T1DM 0.878 - -
Cardiovascular disease (fatal/non-fatal) 0.075 723 723
Background diabetic retinopathy 0.027 443 466
Proliferative diabetic retinopathy 0.070 229 275
Severe vision loss 0.074 57 64
Macular oedema 0.040 362 410
Microalbuminuria 0.000 319 359
Macroalbuminuria 0.017 211 241
Dialysis 0.169 - -
Transplant 0.023 11 12
Neuropathy 0.084 441 480
Uncomplicated foot ulcer 0.125 973 1022
Deep foot infection 0.170 496 524
Foot ulcer and critical ischaemia 0.170 204 214
Minor amputation 0.117 197 207
Major amputation 0.117 97 103
Body mass index, per unit change ±0.008 - -
Urinary or genital tract infection 0.003 - -
46Issue 10: Costs
Background Stakeholder responses Technical team
consideration
• Average cost of insulin • Average cost of insulin included Effect of additional
should include human human insulin and insulin analogues ketone monitoring
insulin and not only • To align with SmPC, reduction in should be explored in
insulin analogues insulin dose related to dapagliflozin model. Company’s
• Reduction in baseline from DEPICT has not been included scenario analysis in
insulin dose at both 24 in model. Differences in total insulin which ketone
weeks and 52 weeks dose are modelled between arms as monitoring for people
should be used to be a result of different weight profiles having dapagliflozin is
consistent with efficacy • Suggest daily ketone monitoring for 3 times more than
data 1st week and then ≥ weekly for 1st 3 that of people having
• Effect of additional months, then ‘sick day’ rules standard of care most
ketone monitoring closely reflects
should be explored in ERG comments: agrees with clinical experts’
model company’s approach to calculate insulin comments
treatment costs
What are the committee’s views? 47Costs of ketone monitoring (1)
Company base case:
• Patients monitor ketones during periods at risk → independent of treatment choice
→ cost of ketone monitoring balanced across treatment arms → no additional cost
of ketone monitoring
Company modelled 3 scenarios:
• On starting dapagliflozin, 4 weeks of daily ketone monitoring (period corresponds to
drop in total daily insulin dose observed in DEPICT trials) → additional one-off cost
of £49.11 in dapagliflozin arm (3 packs)
• 1 pack of ketone testing strips in standard of care arm vs 2 packs for dapagliflozin
arm
• 1 pack of ketone testing strips in standard of care arm vs 3 packs for dapagliflozin
arm (based on proportion experiencing ketosis in dapagliflozin (3/286=1.0%) vs
placebo (1/289=0.3%) arms in DEPICT trials)
48Costs of ketone monitoring (2) • Ketone monitoring checked during acute illness, stress or when glucose is elevated. DAFNE ‘sick day’ rules: – minor illness: no ketones (
Summary of company base case for
indicated population
• Company models effect of 52-week changes to HbA1c and weight
• Modelled treatment effects applied to risk factors in 1st cycle; after, risk
factors assumed to remain constant while patients on treatment.
• Stopping dapagliflozin: due to adverse events in 1st year. After stopping,
HbA1c and weight revert to baseline levels and company assumes
hypoglycaemia, diabetic ketoacidosis and adverse events rates same as
placebo
• Stopping standard of care: due to adverse events in 1st year. After
stopping, HbA1c and weight revert to baseline levels
• Include treatment-related adverse events (urinary and genital tract
infections), DKA and hypoglycaemia
• Baseline utility: estimated from Peasgood et al. and value reflected
baseline characteristics of indicated population (0.865)
• Insulin cost: daily insulin cost per kg (£0.019/kg)
50Company cost-effectiveness results
∆ cost (£) ∆ QALY ICER
Company base case £3,575 0.39 £9,175
A. 24-week effects £3,002 0.42 £7,106
B. 52-week effects wane and treatment stops at 2 years £480 0.04 £11,011
C. Apply annual stopping rate for any reason to year 1 onwards £1,348 0.18 £7,604
D. Apply disutility of 0.0091 to DKA £3,575 0.39 £9,198
E. 4% DKA fatal £3,406 0.35 £9,618
F. 4.45% severe hypoglycaemia fatal £5,709 0.71 £8,037
G. Utility estimates from Peasgood et al. (inc. BMI) £3,575 0.28 £12,620
H. Ketone monitoring I (one-off 3 packs dapagliflozin) £3,625 0.39 £9,301
I. Ketone monitoring II (1 pack placebo vs 2 packs dapagliflozin) £3,824 0.39 £9,813
J. Ketone monitoring III (1 pack placebo vs 3 packs dapagliflozin) £4,070 0.39 £10,444
K. Use ITT results £3,627 0.37 £9,850
L. Annual probability of stopping in placebo arm = 0 £3,564 0.40 £8,964
M. Multiplicative utility decrements £3,575 0.27 £13,038
Cumulative application of multiple alterations listed above:
- C, D, E, F, G, J, K & L £2,026 0.31 £6,618
- C, D, E, F, G, J, K, L & A £2,171 0.29 £7,487
- C, D, E, F, G, J, K, L & B £836 0.09 £9,465
- C, D, E, F, G, J, K, L & M £2,026 0.29 £7,018
51Other issues
Innovation
• Dapagliflozin is first adjunct to insulin licensed in UK, using a
different mechanism of action to insulin
• It may not represent a step-change in management of type 1
diabetes
Equalities
• No equalities issues identified in submissions or academic report
52End of Part 1
53You can also read