3 The company's submission

The Appraisal Committee (section 7) considered evidence submitted by the company and a review of this submission by the Evidence Review Group (ERG; section 8).

Clinical effectiveness

3.1 The company identified 11 studies that evaluated empagliflozin for treating type 2 diabetes. The company's submission included details of 8 of these 11 studies: 7 randomised controlled trials and 1 long‑term extension study (1245.31). The 3 studies excluded were 1 that evaluated empagliflozin as monotherapy (1245.20), EMPA‑REG OUTCOM (1245.25) and EMPA‑REG‑JAPAN (1245.52). The company did not explain why these studies were excluded. One of the included studies (1245.48) compared empagliflozin with placebo as monotherapy and therefore is not relevant to this appraisal.

3.2 The long‑term extension study (1245.31) recruited patients from 3 trials: 2 in which empagliflozin was evaluated as a combination therapy (1245.19 and 1245.23) and 1 monotherapy trial (1245.20). The results were presented separately for the patients from each trial.

3.3 Study 1245.23 comprised 2 separate sub‑studies: EMPA‑REG‑MET, evaluating empagliflozin plus metformin, and EMPA‑REG‑METSU, evaluating empagliflozin plus metformin and a sulfonylurea. Another study, 1254.19, evaluated empagliflozin plus pioglitazone or pioglitazone plus metformin. Two studies, 1245.33 and 1245.49, evaluated empagliflozin as an add‑on to basal insulin and multiple daily injections of insulin respectively, with or without other oral antidiabetic agents.

3.4 Study 1245.36 (in patients with renal impairment) included some patients with moderate to severe renal impairment. Because the summary of product characteristics states that empagliflozin 10 mg and 25 mg should not be initiated in patients with an estimated glomerular filtration rate (eGFR) less than 60 ml/min/1.73 m², only the subgroup of patients with mild renal impairment (eGFR of 60–90 ml/min/1.73 m²) is relevant for this appraisal.

3.5 All but 1 of the relevant studies had 3 treatment arms: empagliflozin 10 mg, empagliflozin 25 mg and placebo. Study 1245.28 instead compared empagliflozin 25 mg with glimepiride (a sulfonylurea) as a dual therapy on a background of metformin. Also, study 1245.23 included an open‑label treatment arm comprising patients with very poor glycaemic control whose baseline glycated haemoglobin (HbA_1c) was more than 10%. The patients in this arm had empagliflozin 25 mg.

3.6 The duration of the relevant studies varied from 24 weeks (1245.19 and 1245.23) to 2 years (1245.28). The efficacy and safety results at 76 weeks for patients enrolled in the 1245.19 and 1245.23 trials were available in study 1245.31. The studies evaluating empagliflozin as an add‑on to insulin therapy lasted for 78 weeks (1245.33) and 52 weeks (1245.49).

3.7 The primary outcome measure in the trials was change in the levels of HbA_1c from baseline. The results showed that empagliflozin at both doses (10 mg or 25 mg) was associated with a statistically significant reduction in mean HbA_1c compared with placebo in patients on different background therapies, including insulin. These reductions were maintained throughout the duration of treatment in the long‑term extension study (1245.31). The glycaemic control achieved with empagliflozin 25 mg in patients with metformin background therapy was statistically non‑inferior compared with glimepiride at week 104 in trial 1245.28. Empagliflozin also showed a statistically significantly better reduction in HbA_1c compared with placebo in patients with mild renal impairment (1245.36). The adjusted change from baseline in mean HbA_1c level from the relevant studies is summarised in table 1.

Table 1 Adjusted mean change from baseline in mean HbA_1c level (%) ±SE

Trial	Duration	Placebo/active comparator	Empagliflozin 10 mg	Empagliflozin 25 mg
Patients on baseline pioglitazone or pioglitazone plus metformin (dual or triple therapy)
1245.19	At week 24	−0.11±0.07	−0.59±0.07	−0.72±0.07
1245.19	P value	–	<0.0001	<0.0001
1245.31 (patients from study 1245.19)	At week 76	−0.01±0.07	−0.61±0.07	−0.70±0.07
1245.31 (patients from study 1245.19)	P value	–	<0.0001	<0.0001
Dual therapy: patients on baseline metformin
1245.23 (metformin only sub‑study)	At week 24	−0.13±0.05	−0.70±0.05	−0.77±0.05
1245.23 (metformin only sub‑study)	P value	–	<0.0001	<0.0001
1245.31 (patients from 1245.23 metformin only sub‑study)	At week 76	−0.01±0.05	−0.62±0.05	−0.74±0.05
1245.31 (patients from 1245.23 metformin only sub‑study)	P value	–	<0.0001	<0.0001
1245.28 (compared with glimepiride)	At week 104	−0.55±0.03	–	−0.66±0.03
1245.28 (compared with glimepiride)	P value (non‑inferiority)	–	–	<0.0001
Triple therapy: patients on baseline metformin plus SU
1245.23 (metformin plus SU sub‑study)	At week 24	−0.17±0.05	−0.82±0.05	−0.77±0.05
1245.23 (metformin plus SU sub‑study)	P value	–	<0.001	<0.001
1245.31 (patients from 1245.23 metformin plus SU sub‑study)	At week 76	−0.03±0.06	−0.74+0.06	−0.72+0.06
1245.31 (patients from 1245.23 metformin plus SU sub‑study)	P value	–	<0.0001	<0.0001
Add‑on to insulin: patients on baseline insulin ± other anti‑diabetics
1245.33	At week 18	−0.01±0.07	−0.57±0.07	−0.71±0.07
	P value	–	<0.0001	<0.0001
	At week 78	−0.02±0.09	−0.48±0.08	−0.64±0.09
	P value	–	<0.0001	<0.0001
1245.49	At week 18	−0.50±0.06	−0.94±0.06	−1.02±0.06
	P value	–	<0.0001	<0.0001
	At week 52	−0.81±0.08	−1.18±0.08	−1.27±0.08
	P value (non‑inferiority)	–	<0.0001	<0.0001
Empagliflozin in patients with mild renal impairment
1245.36 (subgroup with mild renal impairment)	At week 24	0.06±0.07	−0.46±0.07	−0.63±0.07
	P value	–	<0.0001	<0.0001
	At week 52	0.06±0.08	−0.57±0.08	−0.60±0.08
	P value	–	<0.0001	<0.0001
Abbreviations: HbA1c, glycated haemoglobin; SE, standard error; SU, sulfonylurea.

3.8 Important secondary outcomes included change in body weight and blood pressure from baseline. In a study of dual therapy (1245.23 EMPA‑REG MET), at week 24 compared with placebo, empagliflozin 10 mg resulted in mean weight loss of 1.6 kg and empagliflozin 25 mg resulted in a loss of 2.0 kg. Similarly, at week 24 in triple therapy (in study 1245.23 EMP‑REG METSU) compared with placebo empagliflozin 10 mg reduced weight by 1.8 kg and empagliflozin 25 mg reduced it by 2.0 kg. The long‑term extension study (1245.31) confirmed that weight loss from baseline achieved at week 24 was largely maintained at week 76. Both doses of empagliflozin with basal insulin regimens (in study 1245.33) were associated with much greater weight loss compared with placebo at week 78: 3.6 kg for empagliflozin 10 mg and 3.1 kg for empagliflozin 25 mg. In combination with multiple daily injections of insulin (1245.49), compared with placebo empagliflozin reduced mean body weight by 2.39 kg (10 mg) and 2.48 kg (25 mg). Reductions in systolic blood pressure ranged from 1.4 mm Hg in the 1245.49 trial to 4.8 mm Hg in the metformin‑only sub‑study of trial 1245.23.

3.9 Health‑related quality of life data were collected in 6 trials that compared empagliflozin with placebo (including a trial of empagliflozin as a monotherapy, 1245.20). The mean EQ‑5D utility index score at baseline was comparable across the 6 trials and ranged between 0.791 and 0.813. Across all trials the addition of empagliflozin did not result in a clinically meaningful change in quality of life, with baseline EQ‑5D utility index scores being maintained throughout the trials. The company's submission presents pooled data from the 6 trials at different time points (weeks 4, 6, 12, 18, 24, 40 and 52). The company also stated that no differences in EQ‑5D score were evident in any subgroups based on age, sex, BMI, country, blood pressure, HbA_1c level at baseline, eGFR at baseline, prior cardiovascular events, time since diagnosis, race or cardiovascular risk predictor. The trials also collected data using a visual analogue scale (EQ VAS), and the company reported that change from baseline EQ VAS was similar across all treatment groups at all time points.

3.10 The company's submission presented adverse events as reported in the individual studies. In general, the proportions of patients who experienced any adverse events, severe adverse events or adverse events leading to discontinuation of trial medication were similar between both empagliflozin groups and placebo across all trials. In most trials, adverse events leading to discontinuation were more frequent in the placebo group than in the empagliflozin groups. Adverse events in more than 5% of patients in any randomised group in the trials were: urinary tract infections, balanitis, upper respiratory tract infections, bronchitis, nasopharyngitis, influenza, cough, diarrhoea, hypoglycaemia, hyperglycaemia, dyslipidaemia, hypertension, arthralgia, back pain, pain in extremity, headache, dizziness and depression.

3.11 Hypoglycaemic events, urinary tract infection, genital infections, volume depletion and fractures were considered to be 'adverse events of special interest' and reported separately. The data showed that treatment with empagliflozin did not lead to an increase in hypoglycaemic events, except when empagliflozin was administered with a sulfonylurea (the 1245.23 EMPA‑REG‑METSU sub‑study and patients moving from the same sub‑study in extension study 1245.31) or with insulin as background therapy (1245.33 and 1245.49). Across all trials, genital infections (generally of mild to moderate intensity) were consistently more frequent in the empagliflozin groups than with placebo. The incidence of urinary tract infections was similar across both empagliflozin groups and placebo, although it was reported that empagliflozin was associated with a greater frequency in women compared with placebo. In addition, both genital and urinary tract infections were more common in women than men. The frequency of volume depletion was low across all clinical studies and comparable between all treatment groups. The rates of fracture were very low and similar for all treatment groups across all empagliflozin trials.

3.12 The company's submission considered dipeptidyl peptidase‑4 (DPP‑4) inhibitors and other sodium‑glucose cotransporter‑2 (SGLT‑2) inhibitors (dapagliflozin and canagliflozin) to be the comparators for empagliflozin. In the absence of any head‑to head trial, the company performed indirect comparisons by means of network meta‑analyses. The company conducted a systematic literature review to identify randomised controlled trials that evaluated the comparators. Five networks of randomised controlled trials were considered, each 1 including trials that compared the interventions for patients whose diabetes was no longer responding adequately to:

metformin (for dual therapy)
metformin plus sulfonylurea (for triple therapy)
thiazolidinediones (for dual therapy)
thiazolidinediones and metformin (for triple therapy)
insulin therapy plus other oral antidiabetic drugs (as an add‑on to insulin therapy).

3.13 The outcomes compared in the network meta‑analyses in the company's submission included change from baseline in HbA_1c, systolic blood pressure, and body weight. Safety outcomes were hypoglycaemia (severe and non‑severe), urinary tract infection and genital tract infection. For continuous outcomes, Bayesian network meta‑analysis was used to determine the mean differences in change from baseline and associated 95% credible intervals between all interventions. For binary outcomes, the proportions of events were modelled in a logistic regression framework, and relative risks and associated 95% credible intervals were estimated.

3.14 Several trials were available in which metformin was background therapy and so an uninformed random‑effects model was applied for that network. For the other 4 background therapies (metformin plus a sulfonylurea, thiazolidinedione, thiazolidinedione plus metformin and insulin), all comparisons were based on single trials (except empagliflozin compared with placebo for add‑on to insulin). The company used a conventional fixed‑effects model to account for heterogeneity in these networks.

3.15 Results of the original network meta‑analyses are not presented here because the company submitted a new set of network meta‑analyses in response to consultation on the appraisal consultation document.

Company's response to consultation

3.16 In response to consultation, the company provided new network meta‑analyses. The new networks focused on trials of empagliflozin, canagliflozin, dapagliflozin and sitagliptin compared with placebo. Trials with metformin and a sulfonylurea were also included to complete the networks where needed. The population, primary outcomes and adverse events in the new network meta‑analyses were the same as in the original network meta‑analyses.

3.17 The results of the new network meta‑analyses are presented in table 2.

Table 2 New network meta-analysis results for mean change from baseline in HbA_1clevel (%)

Treatment	Versus placebo (95% credible intervals)	Versus empagliflozin 10 mg (95% CI)	Versus empagliflozin 25 mg (95% CI)
Dual therapy with metformin (52‑week data)
Sulfonylurea	−0.52 (−0.64, −0.40)	0.06 (−0.08, 0.21)	0.10 (0.02, 0.17)
Empa 10 mg	−0.58 (−0.72, −0.45)	−	0.03 (−0.10, 0.16)
Empa 25 mg	−0.62 (−0.73, −0.50)	−0.03 (−0.16, 0.10)	−
Dapa 10 mg	−0.47 (−0.58, −0.36)	0.11 (−0.04, 0.27)	0.15 (0.03, 0.26)
Cana 100 mg	−0.52 (−0.68, −0.37)	0.06 (−0.11, 0.23)	0.09 (−0.03, 0.21)
Cana 300 mg	−0.65 (−0.80, −0.50)	−0.07 (−0.24, 0.10)	−0.03 (−0.15, 0.08)
Sita 100 mg	−0.52 (−0.67, −0.37)	0.07 (−0.10, 0.24)	0.10 (−0.02, 0.22)
Triple therapy with metformin and a sulfonylurea (52‑week data)
Empa 10 mg	−0.71 (−0.88, −0.54)	−	−0.02 (−0.19, 0.15)
Empa 25 mg	−0.69 (−0.86, −0.52)	0.02 (−0.15, 0.19)	−
Cana 100 mg	−0.75 (−0.96, −0.54)	−0.04 (−0.31, 0.23)	−0.06 (−0.33, 0.21)
Cana 300 mg	−0.97 (−1.18, −0.76)	−0.26 (−0.53, 0.01)	−0.28 (−0.55, −0.01)
Sita 100 mg	−0.60 (−0.85, −0.35)	0.11 (−0.19, 0.41)	0.09 (−0.21, 0.39)
Triple therapy with metformin and a thiazolidinedione (24‑week data)
Empa 10 mg	−0.44 (−0.66, −0.22)*	−	0.15 (0.06, 0.36)
Empa 25 mg	−0.59 (−0.80, −0.38)*	−0.15 (−0.36, 0.06)	−
Cana 100 mg	−0.63 (−0.82, −0.44)*	−0.19 (−0.49, 0.10)	−0.04 (−0.32, 0.24)
Cana 300 mg	−0.77 (−0.93, −0.61)*	−0.33 (−0.61, −0.06)	−0.18 (−0.44, 0.08)
Sita 100 mg	−0.70 (−0.84, −0.56)*	−0.26 (−0.52, 0.00)	−0.11 (−0.36, 0.14)
Add on to insulin (52‑week data)
Empa 10 mg	−0.48 (−0.64, −0.33)	−	0.16 (0.00, 0.32)
Empa 25 mg	−0.64 (−0.80, −0.49)	−0.16 (−0.32, 0.00)	−
Dapa 2.5 mg	−0.40 (−0.55, −0.25)	0.08 (−0.14, 0.30)	0.24 (0.02, 0.46)
Dapa 5 mg	−0.50 (−0.65, −0.35)	−0.01 (−0.23, 0.20)	0.14 (−0.07, 0.36)
Dapa 10 mg	−0.57 (−0.72, −0.42)	−0.08 (−0.30, 0.13)	0.07 (−0.14, 0.29)
Abbreviations: Cana, canagliflozin; CI, credible intervals; Dapa, dapagliflozin; Empa, empagliflozin; Sita, sitagliptin. * Compared with control (metformin plus thiazolidinedione) rather than placebo.

Table 3 New network meta‑analysis results for mean change from baseline in weight (kg)

Treatment	Versus placebo (95% credible intervals)	Versus empagliflozin 10 mg (95% CI)	Versus empagliflozin 25 mg (95% CI)
Dual therapy with metformin (52‑week data)
Sulfonylurea	2.47 (1.91, 3.03)	4.22 (3.63, 4.80)	4.80 (4.50, 5.10)
Empa 10 mg	−1.75 (−2.27, −1.22)	−	0.58 (0.06, 1.10)
Empa 25 mg	−2.33 (−2.83, −1.83)	−0.58 (−1.10, −0.06)	−
Dapa 10 mg	−2.21 (−2.87, −1.55)	−0.47 (−1.17, 0.24)	0.12 (−0.41, 0.65)
Cana 100 mg	−1.98 (−2.71, −1.25)	−0.24 (−0.99, 0.51)	0.35 (−0.21, 0.90)
Cana 300 mg	−2.33 (−3.06, −1.60)	−0.59 (−1.34, 0.17)	−0.00 (−0.56, 0.55)
Sita 100 mg	0.07 (−0.69, 0.83)	1.81 (1.03, 2.60)	2.40 (1.80, 2.99)
Triple therapy with metformin and a sulfonylurea (52‑week data)
Empa 10 mg	−2.00 (−2.49, −1.51)	−	0.05 (−0.49, 0.59)
Empa 25 mg	−2.05 (−2.56, −1.55)	−0.05 (−0.59, 0.49)	−
Cana 100 mg	−1.28 (−2.11, −0.44)	0.72 (−0.23, 1.68)	0.77 (−0.19, 1.75)
Cana 300 mg	−2.28 (−3.11, −1.46)	−0.28 (−1.24, 0.68)	−0.22 (−1.20, 0.74)
Sita 100 mg	0.37 (−0.53, 1.27)	2.37 (1.35, 3.40)	2.42 (1.38, 3.45)
Triple therapy with metformin and a thiazolidinedione (24‑week data)
Empa 10 mg	−2.14 (−2.82, −1.45)*	−	−0.15 (0.86, −0.54)
Empa 25 mg	−1.98 (−2.68, −1.28)*	0.15 (−0.54, 0.86)	−
Cana 100 mg	−2.55 (−3.36, −1.74)*	−0.41 (−1.48, 0.64)	−0.57 (−1.64, 0.49)
Cana 300 mg	−3.49 (−4.29, −2.69)*	−1.35 (−2.41, −0.30)	−1.51 (−2.56, −0.45)
Sita 100 mg	0.20 (−0.40, 0.80)*	2.33 (1.42, 3.24)	2.18 (1.26, 3.09)
Add on to insulin (52‑week data)
Empa 10 mg	−1.41 (−1.82, −1.00)	−	0.40 (−0.06, 0.86)
Empa 25 mg	−1.81 (−2.26, −1.36)	−0.40 (−0.86, 0.06)	−
Dapa 2.5 mg	−1.35 (−1.90, −0.81)	0.06 (−0.62, 0.74)	0.46 (−0.25, 1.17)
Dapa 5 mg	−1.43 (−1.97, −0.90)	−0.02 (−0.70, 0.65)	0.38 (−0.33, 1.08)
Dapa 10 mg	−2.04 (−2.58, −1.49)	−0.63 (−1.31, 0.05)	−0.23 (−0.93, 0.48)
Abbreviations: Cana, canagliflozin; CI, credible intervals; Dapa, dapagliflozin; Empa, empagliflozin; Sita, sitagliptin. * Compared with control (metformin plus thiazolidinedione) rather than placebo.

Table 4 New network meta‑analysis results for mean change from baseline in systolic blood pressure

Treatment	Versus placebo (95% credible intervals)	Versus empagliflozin 10 mg (95% CI)	Versus empagliflozin 25 mg (95% CI)
Dual therapy with metformin (52‑week data)
Sulfonylurea	1.31 (−1.06, 3.67)	4.20 (1.79, 6.58)	5.80 (4.41, 7.19)
Empa 10 mg	−2.89 (−4.84, −0.94)	−	1.60 (−0.34, 3.56)
Empa 25 mg	−4.49 (−6.44, −2.57)	−1.60 (−3.56, 0.34)	−
Dapa 10 mg	−3.80 (−6.65, −0.94)	−0.90 (−3.78, 1.91)	0.70 (−1.41, 2.78)
Cana 100 mg	−2.25 (−5.05, 0.51)	0.63 (−2.15, 3.42)	2.23 (0.23, 4.24)
Cana 300 mg	−3.42 (−6.20, −0.66)	−0.54 (−3.35, 2.26)	1.07 (−0.94, 3.07)
Sita 100 mg	0.60 (−2.55, 3.74)	3.49 (0.34, 6.62)	5.09 (2.63, 7.56)
Triple therapy with metformin and a sulfonylurea (52‑week data)
Empa 10 mg	−2.80 (−4.89, −0.72)	−	−0.09 (−2.19, 1.98)
Empa 25 mg	−2.72 (−4.94, −0.48)	0.09 (−1.98, 2.19)	−
Cana 100 mg	−3.82 (−6.75, −0.85)	−1.00 (−4.59, 2.59)	−1.10 (−4.77, 2.60)
Cana 300 mg	−3.00 (−5.81, −0.21)	−0.20 (−3.69, 3.28)	−0.28 (−3.88, 3.27)
Sita 100 mg	2.99 (−0.35, 6.33)	5.79 (1.86, 9.74)	5.70 (1.68, 9.69)
Triple therapy with metformin and a thiazolidinedione (24‑week data)
Empa 10 mg	−4.24 (−6.91, −1.54)*	−	−0.18 (−2.82, 2.54)
Empa 25 mg	−4.06 (−6.72, −1.36)*	0.18 (−2.54, 2.82)	−
Cana 100 mg	−4.11 (−6.98, −1.22)*	0.10 (−3.87, 4.13)	−0.05 (−4.02, 3.93)
Cana 300 mg	−3.54 (−6.46, −0.54)*	0.69 (−3.29, 4.68)	0.50 (−3.48, 4.53)
Add on to insulin (52‑week data)
Empa 10 mg	−2.45 (−4.03, −0.88)	−	0.05 (−1.52, 1.61)
Empa 25 mg	−2.51 (−4.07, −0.93)	−0.05 (−1.61, 1.52)	−
Dapa 2.5 mg	−0.66 (−3.32, 2.06)	1.79 (−1.28, 4.91)	1.84 (−1.22, 4.97)
Dapa 5 mg	−2.38 (−5.01, 0.27)	0.07 (−3.01, 3.16)	0.12 (−2.95, 3.21)
Dapa 10 mg	−3.11 (−5.76, −0.44)	−0.65 (−3.74, 2.42)	−0.60 (−3.70, 2.47)
Abbreviations: Cana, canagliflozin; CI, credible intervals; Dapa, dapagliflozin; Empa, empagliflozin; Sita, sitagliptin. * Compared with control (metformin plus thiazolidinedione) rather than placebo.

Evidence Review Group's comments on the company's clinical‑effectiveness evidence

3.18 The ERG considered the trials to be good quality but commented that the lack of head‑to‑head trials against the main comparators (DPP‑4 inhibitors or other SGLT‑2 inhibitors) was the main weakness of the evidence base. The demographic characteristics were well balanced across treatment groups except in study 1245.49, in which the proportion of men was much lower in the placebo arm than the empagliflozin arms (39.9% placebo, 52.2% empagliflozin 10 mg, 44.2% empagliflozin 25 mg).

3.19 The ERG noted that the company's submission did not report outcome data on change in lipid levels for any trial, but that change in lipid profiles for studies 1245.19 and 1245.23 had already been published. The results of study 1245.23 showed that in comparison with placebo, both doses of empagliflozin were associated with a statistically significant reduction in most of the components of serum lipids. Study 1245.19 also showed that both doses of empagliflozin reduced high‑density lipoprotein cholesterol statistically significantly compared with placebo. The changes in other fractions of lipids were not statistically significant.

3.20 The ERG identified many errors in the company's original network meta‑analyses. The ERG commented that the systematic review process was inadequately described, lacking details on inclusion criteria for studies, justification for excluded studies, quality assessment and data extraction process for included studies. The ERG was also concerned that the company had not done any sensitivity analyses or statistical tests.

Evidence Review Group's comments on the company's new clinical‑effectiveness evidence

3.21 The ERG stated that the new network meta‑analyses provided by the company were robust and successfully addressed the shortcomings of the original network meta‑analyses.

Cost effectiveness

3.22 The company originally submitted a patient‑level state transition model which had been developed for this appraisal. The ERG's critique of the model highlighted several errors which would invalidate any results and so the ERG concluded that the original model and its results were unreliable. As part of the consultation on the appraisal consultation document, the company were asked to provide further analyses, including revised estimations of the incremental cost‑effectiveness ratios of empagliflozin using a validated economic model.

3.23 In response to consultation, the company provided a new cost‑effectiveness model. The new model was an individual patient‑level microsimulation model using IMS CORE. It modelled individual patients' transitions between health states using a fixed cycle length of 1 year over a lifetime horizon. An NHS and personal social services perspective was taken and costs and benefits were discounted at 3.5%.

3.24 The model simulated the incidence of the complications of diabetes based on baseline characteristics of the patient and the treatment's initial impact on HbA_1c, systolic blood pressure and BMI. Complications included in the model were: fatal and non‑fatal myocardial infarction, fatal and non‑fatal stroke, angina, congestive heart failure, peripheral vascular disease, microalbuminuria, gross proteinuria, haemodialysis, diabetic retinopathy, cataract, macular oedema, severe vision loss, neuropathy, ulcer and amputation. Diabetes‑related deaths and general mortality were also modelled.

3.25 Health‑related quality of life values for the model were drawn from the UKPDS 62 and Sullivan et al. (2011). Quality of life decrements were applied to severe hypoglycaemic events, non‑severe hypoglycaemic events, genital tract infections, urinary tract infections and post‑urinary tract infection events. The IMS CORE model associated a quality of life change of 0.0038125 with each BMI point increase or decrease in people with a BMI of 25 kg/m² or higher.

3.26 Data from the company's new network meta‑analyses were used to model the clinical effectiveness of the drugs. If clinical effectiveness data were not available from the network meta‑analyses, the clinical effectiveness of a comparator was assumed to be the same as empagliflozin 10 mg or empagliflozin 25 mg. Treatment‑related adverse effects were hypoglycaemic events (severe and non‑severe), urinary tract infections and genital tract infections. If a patient's HbA_1c exceeded 7.5%, they were switched to insulin, which was associated with a change in costs, an increase in BMI and an adverse event rate of zero.

3.27 The new cost‑effectiveness model compared empagliflozin 10 mg and empagliflozin 25 mg with dapagliflozin 10 mg, sitagliptin 100 mg, canagliflozin 100 mg and canagliflozin 300 mg, each in dual therapy with metformin or as an add on to insulin therapy. It compared empagliflozin 10 mg and empagliflozin 25 mg with sitagliptin 100 mg, canagliflozin 100 mg and canagliflozin 300 mg, each in triple therapy with metformin and a sulfonylurea or metformin and a thiazolidinedione.

3.28 The company stated that only direct costs were included in its new model. The company sourced the relevant costs of managing complications from published studies, including UKPDS and previous NICE appraisals. Some costs were inflated to 2012 prices using the Personal Social Services Research Unit inflation rates. The cost of insulin was a weighted average annual cost based on prescribing data and NHS list prices. Drug costs and the cost of testing strips were included in the cost of insulin, and the annual costs of needle and test strips were included in the cost of intravenous insulin.

3.29 The base case results from the new cost‑effectiveness model are presented in table 5. Where clinical‑effectiveness data for particular parameters of the comparators were not available from the network meta‑analyses, the corresponding data for empagliflozin 10 mg or empagliflozin 25 mg were used. The results in table 5 are based on using the clinical effectiveness of empagliflozin 10 mg. There were slight differences in the results when the clinical effectiveness of empagliflozin 25 mg was used instead, but these did not have a substantial effect on the ICERs.

Table 5 Company's new base‑case results

Treatments	Costs	Net	QALYs	Net	ICER (£/QALY)
Dual therapy with metformin
Empagliflozin 25 mg	£61,535	–	7.995	−	−
Dapagliflozin 10 mg	£61,609	£74	7.964	−0.031	Dominated
Canagliflozin 100 mg	£61,719	£184	7.955	−0.040	Dominated
Empagliflozin 10 mg	£61,761	£226	7.963	−0.032	Dominated
Sitagliptin 100 mg	£61,778	£243	7.899	−0.096	Dominated
Canagliflozin 300 mg	£61,912	£377	7.990	−0.005	Dominated
Triple therapy with metformin plus a sulfonylurea
Empagliflozin 25 mg	£58,711	−	7.564	−	−
Empagliflozin 10 mg	£58,778	£67	7.571	0.007	£9571
Canagliflozin 100 mg	£58,794	£16	7.569	−0.002	Dominated
Canagliflozin 300 mg	£59,000	£222	7.616	0.045	£4933
Sitagliptin 100 mg	£59,390	£390	7.466	−0.150	Dominated
Triple therapy with metformin plus a thiazolidinedione
Sitagliptin 100 mg	£58,644	−	7.553	−	−
Canagliflozin 100 mg	£58,751	£107	7.579	0.026	£4115
Empagliflozin 25 mg	£58,854	£103	7.561	−0.018	Dominated
Canagliflozin 300 mg	£59,106	£355	7.614	0.035	£10,143
Empagliflozin 10 mg	£59,166	£60	7.542	−0.072	Dominated
Add‑on to insulin
Canagliflozin 100 mg	£60,235	−	7.545	−	−
Dapagliflozin 10 mg	£60,360	£125	7.545	0.000	Dominated
Empagliflozin 25 mg	£60,428	£193	7.534	−0.011	Dominated
Empagliflozin 10 mg	£60,539	£304	7.523	−0.022	Dominated
Sitagliptin 100 mg	£60,564	£329	7.511	−0.034	Dominated
Canagliflozin 300 mg	£60,599	£364	7.583	0.038	£9579
Abbreviations: ICER, incremental cost‑effectiveness ratio; QALY, quality‑adjusted life year. Note: comparator effectiveness was based on empagliflozin 10 mg where data were not available from the network meta‑analyses. A treatment is 'dominated' when it is both less effective and more costly than its comparator.

Company's new sensitivity analyses

3.30 The company did 2 sensitivity analyses. One analysis modelled BMI converging over time, and the other assumed no change in systolic blood pressure with sitagliptin for pairwise comparisons of empagliflozin and sitagliptin only. The results of the sensitivity analyses were not substantially different from the results of the base‑case analyses.

Evidence Review Group's critique of the company's new cost‑effectiveness model

3.31 The ERG highlighted that the IMS CORE model used by the company in its new cost‑effectiveness modelling has been validated and used in other NICE appraisals.

3.32 The ERG stated that rerunning the base case in IMS CORE gave different results from those reported in the company's response to consultation. However, the rerun results were qualitatively the same as the company's results. The ERG highlighted that the main cause of the differences in the ICERs was the very small difference in costs and QALYs when comparing treatments.

3.33 The ERG stated that it was not clear how the rates of non‑severe hypoglycaemia used in the model were calculated, but highlighted that the relative rates between comparators were the same as those in the new network meta‑analyses. They also highlighted that the changes to the IMS CORE default costs and utilities, and the source of the data for rates of genital tract infections, were not clear.

3.34 The ERG stated that second‑order sampling was used for the company's deterministic analyses. As a consequence of using second‑order sampling, the company's deterministic analyses provided probabilistic results. The ERG stated that the company did not present truly deterministic results.

Evidence Review Group's exploratory and sensitivity analyses using the new cost‑effectiveness model

3.35 The ERG's new exploratory analyses revised the cost of complications using UKPDS 65 and the quality of life values using UKPDS 62. The ERG also revised the rates of urinary tract infections for canagliflozin 300 mg in the analyses for empagliflozin as an add‑on to insulin. The ERG did not use second‑order sampling in its exploratory analyses so that the results of the analysis were truly deterministic.

3.36 The deterministic results of the ERG's new exploratory analysis are shown in table 6. Each treatment was compared with the next least costly treatment in each treatment regimen. If the previous treatment was dominated, the treatment was compared with the next least costly treatment that was not dominated. The results in the table are based on using the clinical effectiveness of empagliflozin 10 mg for any parameters of the comparators for which clinical‑effectiveness data were not available from the network meta‑analyses. There were slight differences in the results when the clinical effectiveness of empagliflozin 25 mg was used instead, but they did not have a substantial effect on the ICERs.

Table 6 Results of the ERG's new exploratory analyses

Treatment	Costs	Net	QALYs	Net	ICER
Dual therapy with metformin
Sitagliptin 100 mg	£41,554	−	8.136	−	−
Canagliflozin 100 mg	£41,626	£72	8.161	0.025	£1220
Empagliflozin 25 mg	£41,646	£20	8.203	0.042	£400
Dapagliflozin 10 mg	£41,675	£29	8.161	−0.042	Dominated
Empagliflozin 10 mg	£41,767	£121	8.178	−0.025	Dominated
Canagliflozin 300 mg	£42,192	£546	8.202	−0.001	£136,500
Triple therapy with metformin and a sulfonylurea
Empagliflozin 25 mg	£39,399	−	7.834	−	−
Canagliflozin 100 mg	£39,439	£40	7.864	0.030	£2105
Empagliflozin 10 mg	£39,479	£40	7.841	−0.023	Dominated
Canagliflozin 300 mg	£39,596	£157	7.894	0.030	£3,568
Sitagliptin 100 mg	£39,602	£6	7.782	−0.112	Dominated
Triple therapy with metformin and a thiazolidinedione
Sitagliptin 100 mg	£39,392	−	7.858	−	−
Canagliflozin 100 mg	£39,522	£130	7.850	−0.008	£4063
Empagliflozin 10 mg	£39,522	£130	7.814	−0.044	Dominated
Empagliflozin 25 mg	£39,633	£241	7.836	−0.022	Dominated
Canagliflozin 300 mg	£39,872	£480	7.865	0.007	£12,069
Add on to insulin
Empagliflozin 10 mg	£40,580	−	7.814	−	−
Sitagliptin 100 mg	£40,810	£230	7.817	0.003	Dominated
Canagliflozin 100 mg	£40,951	£371	7.832	0.018	£12,367
Dapagliflozin 10 mg	£41,008	£57	7.810	−0.022	Dominated
Empagliflozin 25 mg	£41,023	£72	7.804	−0.028	Dominated
Canagliflozin 300 mg	£41,292	£341	7.858	0.026	£11,367
Abbreviations: ICER, incremental cost‑effectiveness ratio; QALY, quality‑adjusted life year. Note: comparator effectiveness was based on empagliflozin 10 mg where data were not available from the network meta‑analyses. A BMI coefficient of −0.0038 was used. A treatment is 'dominated' when it is both less effective and more costly than its comparator.

ERG's sensitivity analyses

3.37 The ERG did a sensitivity analysis for the impact of different quality of life values associated with changes in BMI, because BMI has been an important parameter in similar NICE technology appraisals. For the sensitivity analyses, the ERG used no decrease and a decrease of 0.0061 in quality of life for each point change in BMI. The ERG used no decrease because it was easy to implement in the IMS CORE model. The decrease of 0.0061 was used because the ERG believed it was the coefficient that corresponded with the relevant interval of the EQ‑5D social tariff, and was the true coefficient applied by Bagust and Beale (2005).

3.38 In general, the ICERs decreased when the 0.0061 decrease in quality of life was used. For some comparisons, the comparators were no longer dominated and ICERs of above £70,000 per QALY gained were reported. The most notable change to the ICERs was for canagliflozin 300 mg compared with empagliflozin 25 mg in triple therapy with metformin and a thiazolidinedione. In this comparison, the ICER decreased from £68,571 per QALY gained to £9358 per QALY gained when a quality of life decrement of 0.0061 was used. For the analyses as an add‑on to insulin therapy, the ICER for sitagliptin 100 mg compared with empagliflozin 10 mg changed from £76,667 per QALY gained to being dominated by empagliflozin 10 mg.

3.39 Full details of all the evidence are available.

Empagliflozin in combination therapy for treating type 2 diabetes