An indirect treatment comparison of the efficacy of insulin degludec/liraglutide (IDegLira) and insulin glargine/lixisenatide (iGlarLixi) in patients with type 2 diabetes uncontrolled on basal insulin

Abstract

Aims: To obtain estimates of the relative treatment effects between insulin degludec/liraglutide (IDegLira) and insulin glargine U100/lixisenatide (iGlarLixi) in patients with type 2 diabetes mellitus (T2DM) uncontrolled on basal insulin therapy.

Materials and methods: Data from phase 3 trials providing evidence for estimating the relative efficacy and safety of IDegLira vs iGlarLixi in patients uncontrolled on basal insulin-only regimens were used in this analysis. Outcomes of interest were changes in HbA_1c, body weight and insulin dose, and rate ratio of hypoglycemia. The indirect comparison of the reported trial findings followed the principles of Bucher et al.

Results: IDegLira was estimated to provide a 0.44 [95% CI = 0.17–0.71] %-point reduction in HbA_1c compared with iGlarLixi. Body weight was reduced by 1.42 [95% CI = 0.35–2.50] kg with IDegLira compared with iGlarLixi. Insulin dose was comparable between the two interventions. The rate of severe or blood glucose-confirmed (self-measured plasma glucose [SMPG] ≤ 3.1 mmol/L) hypoglycemia with IDegLira was approximately half that of iGlarLixi (rate ratio = 0.51 [95% CI = 0.29–0.90]). However, using the American Diabetes Association definition of documented symptomatic hypoglycemia (SMPG ≤3.9 mmol/L) the rate was comparable between the two treatments (rate ratio = 1.07 [95% CI = 0.90–1.28]).

Limitations: The assumptions made in the indirect comparison and differences between the included trials in baseline HbA_1c levels, previous use of sulfonylureas, definitions of hypoglycemia, presence or absence of run-in period, the different duration of the trials, and the cross-over design of one of the trials.

Conclusions: The results of this indirect treatment comparison demonstrate that, among patients with T2DM uncontrolled on basal insulin, treatment with IDegLira results in a greater reduction of HbA_1c and a greater reduction in body weight compared with iGlarLixi at similar insulin doses.

Keywords:

• GLP-1
• basal insulin
• glycemic control network meta-analysis
• liraglutide
• insulin analogs

Introduction

Type 2 diabetes mellitus (T2DM) is a progressive disease primarily characterized by impaired insulin secretion, peripheral insulin resistance, and reduced post-prandial secretion and activity of glucagon-like peptide-1 (GLP-1) and other incretin hormones^1,2. Beta-cell function is typically half that of healthy individuals at the time of diagnosis, and, as the disease progresses, insulin output further declines, with many patients ultimately requiring treatment with exogenous insulin to reach target HbA_1c^1,3.

Although insulin is extremely efficacious, the presence of side effects, such as hypoglycemia and weight gain, are obstacles to both the initiation and optimization of insulin therapy^1,4. GLP-1 analogs stimulate the release of endogenous insulin in the presence of elevated blood glucose concentrations, reducing both fasting and postprandial blood glucose concentrations⁵. GLP-1 analogs can also reduce hunger and food intake, which promotes weight loss^6–8. They are generally well tolerated, but can cause transient gastrointestinal side-effects during the titration phase^6–8. Studies have shown that combining insulin therapy with GLP-1 analogs can yield improved glycemic control, alongside a lower risk of hypoglycemia and weight gain/weight neutrality, compared with insulin therapy alone in patients with T2DM^6,9–13.

Combination of insulin and a GLP-1 analog in a fixed-ratio, once-daily injection has several advantages over administration of the mono-components alone. These include ease of titration and a reduced number of daily injections, which could improve treatment adherence^14,15. In addition, fixed-ratio combinations of insulin and GLP-1 provide the opportunity to titrate GLP-1RA in small increments, thereby reducing gastrointestinal side-effects¹⁶. There are currently two approved fixed-ratio combinations of basal insulin and GLP-1 analog: IDegLira, a combination of insulin degludec (IDeg) and the GLP-1 analog liraglutide (ratio of 1 U IDeg/0.036 mg liraglutide)¹⁷; and iGlarLixi, a combination of insulin glargine (IGlar) U100 and the GLP-1 analog lixisenatide (ratio of 1 U iGlar/0.33 mg lixisenatide; a ratio of 1 U IGlar/0.50 mg lixisenatide is also available in the European Union)¹⁸. A recent systematic review and meta-analysis of publicly available studies with IDegLira and iGlarLixi estimated that, compared with control subjects, HbA_1c was reduced by 0.68% and 0.44%, respectively. The same review reported that end-of-trial HbA_1c was identical (∼ 6.5%) for both IDegLira and iGlarLixi, which could indicate that the differences in absolute HbA_1c reduction reflect different baseline HbA_1c levels¹⁹.

There are no head-to-head trials comparing IDegLira vs iGlarLixi, and, even if such a trial were conducted, it would be years before the results became available; thus, the only way to compare the clinical efficacy of these two combinations is via indirect comparisons. In order to support informed clinical and payer decision-making, the aim of the current study was to obtain indirect estimates of the relative treatment effects between these two interventions in patients with T2DM uncontrolled on basal insulin treatment. The analysis was based on published randomized controlled trial (RCT) evidence, and the limitations of indirect comparisons were accounted for whenever possible.

Methods

Ethics statement

This article is based on previously conducted studies, and does not involve any new studies of human or animal subjects performed by any of the authors.

Evidence base

Results from the following phase 3 trials of combination products in participants who had previously failed to achieve satisfactory glucose control using basal insulin-only regimens were publicly available at the time of this study (March 2017): for IDegLira, the DUAL II and DUAL V studies^9,12; and for iGlarLixi, the LixiLan-L trial¹³. Inadequate glycemic control was defined as HbA_1c of 7.5 − 10.0% [58 − 86 mmol/mol], inclusive, for DUAL II and LixiLan-L; and 7.0–10.0% [53–86 mmol/mol] inclusive for DUAL V^9,12,13. These three publications^9,12,13 were used as sources for the present indirect treatment comparison, together with data from Food and Drug Administration (FDA) briefing books^20,21 and an American Diabetes Association (ADA) 2016 poster²².

The trials used in the analysis are summarized in Tables 1 and 2. The principal differences between these trials included: differences in inclusion/exclusion criteria (and, therefore, in patient characteristics, as seen in Table 2), the presence of a 6-week run-in period for LixiLan-L (after which 28% of patients failed to meet a second set of inclusion criteria) vs no run-in for the DUAL trials, use of different basal insulins as comparators for LixiLan-L vs DUAL II, use of “capped” (a limit on the maximum dose of insulin that could be prescribed; DUAL II, LixiLan-L) vs uncapped (DUAL V) basal insulin dosing in reference arms, and lower self-measured plasma glucose (SMPG) titration targets for the DUAL trials than Lixilan-L. The DUAL II trial was double-blinded, whereas DUAL V and LixiLan-L were open label. Different thresholds and reporting criteria were used to define hypoglycemia in the DUAL trials and LixiLan-L. Study-specific relative treatment effects are presented in Table 3.

Table 1. Overview of studies in patients inadequately controlled on basal insulin-only regimens—study characteristics.

	LixiLan-L^13	DUAL II^9	DUAL V^12	SWITCH 2^25
Arms	IGlar U100	IDeg	IGlar U100	IDeg
	iGlarLixi	IDegLira	IDegLira	IGlar U100
Inclusion criteria	Basal insulin >6 months	Basal insulin ≥90 days	IGlar U100 ≥ 56 days	Basal insulin ≥26 weeks
	Basal insulin 15–40 U/day	Basal insulin 20–40 U/day	IGlar U100 20–50 U/day	Basal insulin ± OAD excluding SUs
	0–2 OADs (Metformin, SU, glinide SGLT2i, or DPP-4i)	Metformin ± SU/glinide	Metformin	HbA1c ≤ 9.5%
	HbA1c 7.5–10% (inclusive)	HbA1c 7.5–10% (inclusive)	HbA1c 7.0–10% (inclusive)	BMI ≤45
	SMPG ≤10–11.1 mmol/L	BMI ≥27	BMI ≤40	Age ≥18 years
	Age ≥18 years	Age ≥18 years	Age ≥18 years
Run-in phase	6 weeks Discontinue all OADs except metformin Titrate IGlar U100 and metformin	None	None	None
Post run-in inclusion criteria	HbA1c 7–10% (inclusive) SMPG ≤7.8 mmol/L IGlar U100 dose ≤50 U/day	N/A	N/A	N/A
Blinding	Open-label	Double-blind	Open-label	Double-blind
Concomitant	Metformin if used at screening	Metformin for all	Metformin for all	Continued pre-study OAD
Insulin caps	IGlar U100: 60 U	IDeg: 50 U	IGlar U100: No cap	IDeg: No cap
	iGlarLixi: 60 U	IDegLira: 50 U	IDegLira: 50 U	IGlar U100: No cap
Titration target	4.4–5.6 mmol/L	4.0–5.0 mmol/L	4.0–5.0 mmol/L	3.9–5.0 mmol/L
Duration	30 weeks	26 weeks	26 weeks	32 × 2 weeks (64 weeks) in crossover design

BMI, body mass index; DPP-4i, dipeptidyl peptidase-4 inhibitor; IDeg, insulin degludec; IDegLira, insulin degludec/liraglutide; IGlar U100, insulin glargine 100 U/mL; iGlarLixi, insulin glargine/lixisenatide; N/A, not applicable; OAD, oral anti-diabetic drug; SGLT2i, sodium-glucose co-transporter-2 inhibitor; SMPG, self-measured fasting plasma glucose; SU, sulfonylurea; U, units.

Table 2. Overview of studies in patients inadequately controlled on basal insulin-only regimens—baseline characteristics.

	LixiLan-L^Table Footnotea^13		DUAL II^9		DUAL V^12		SWITCH 2^25
	iGlarLixi	IGlar U100	IDegLira	IDeg	IDegLira	IGlar U100	Total
Sample size (n)	367	369	199	199	278	279	720
Age (years)	60 ± 9	60 ± 9	57 ± 9	58 ± 11	58 ± 10	59 ± 9	61 ± 11
White (%)	92	92	79	76	94	95	80
Duration diabetes (years)	12 ± 7	12 ± 7	10 ± 6	11 ± 7	12 ± 7	11 ± 7	14 ± 8
Insulin dose (units)	35 ± 9	35 ± 9	29 ± 8	29 ± 8	31 ± 10	32 ± 10	40^b (IDeg/IGlar U100)	43^b (IGlar U100/IDeg)
Weight (kg)	87.7 ± 14.5	87.1 ± 14.8	95.4 ± 19	93.5 ± 20	88.3 ± 17.5	87.3 ± 15.8	91.7 ± 19.5
BMI (kg/m^2)	31.3 ± 4.3	31.0 ± 4.2	33.6 ± 6	33.8 ± 6	31.7 ± 4.4	31.7 ± 4.5	32.2 ± 5.6
HbA1c (%)	8.1 ± 0.7	8.1 ± 0.7	8.7 ± 0.7	8.8 ± 0.7	8.4 ± 0.9	8.2 ± 0.9	7.6 ± 1.1
FPG (mmol/L)	7.3 ± 2.0	7.4 ± 2.1	9.7 ± 2.9	9.6 ± 3.1	8.9 ± 2.6	8.9 ± 2.9	7.6 ± 2.9
Previous SU (%, yes)	40	40	50	50	0	0	0

^a Baseline values are post run-in period.

^b Standard deviation not published.

Data are means ± standard deviation unless otherwise stated.

BMI, body mass index; FPG, fasting plasma glucose; IDeg, insulin degludec; IDegLira, insulin degludec/liraglutide; IGlar U100, insulin glargine 100 U/mL; iGlarLixi, insulin glargine/lixisenatide; SU, sulfonylurea.

Table 3. Study-specific treatment effects.

	LixiLan-L^a ^13^,^21	DUAL II^9^,^20	DUAL V^12^,^20^,^41	SWITCH 2^25^,^26
	iGlarLixi vs IGlar U100	IDegLira vs IDeg	IDegLira vs IGlar U100	IDeg vs IGlar U100
ETD [95% CI]
Change in HbA1c (%)	−0.52 [−0.63; −0.40]	−1.05 [−1.25; −0.84]	−0.59 [−0.74; −0.45]	0.09 [−0.04; 0.23]
Change in body weight (kg)	−1.37 [−1.81; −0.93]	−2.51 [−3.21; −1.82]	−3.20 [−3.77; −2.64]	−0.30 [−0.98; 0.38]^b
Daily actual insulin dose (U)	−0.3 [−1.8; 1.3]	−0.0 [−1.9; 1.8]	−25.5 [−28.9; −22.1]	−4.0 [−10.4; 2.4]^b
EOR [95% CI]
HbA1c < 7.0%	2.90 [2.14; 3.93]^d	5.44 [3.42; 8.66]	2.84 [2.00; 4.04]^d	NR
HbA1c < 7.0% w/o weight gain	3.34 [2.31; 4.84]^d	7.65 [4.50; 13.02]	4.08 [2.80; 5.94]^d	NR
HbA1c < 7.0% w/o hypoglycemia	2.03 [1.44; 2.86]^d	5.57 [3.36; 9.21]	2.85 [2.01; 4.05]^d	NR
HbA1c < 7.0% w/o weight gain w/o hypoglycemia	2.51 [1.61; 3.89]^d	7.44 [4.08; 13.57]	4.56 [2.96; 7.03]^d	NR
ERR [95% CI]
Severe or BG-confirmed hypoglycemia^e	1.00 [0.83; 1.20]	0.66 [0.39; 1.13]	0.43 [0.30; 0.61]	0.77 [0.70; 0.85]
ADA (SMPG ≤3.9 mmol/L) documented symptomatic hypoglycemia	0.72 [0.65; 0.80]^c	0.84 [0.73; 0.96]^c	0.46 [0.34; 0.62]	0.92 [0.87; 0.96]

^a Change values are for baseline/post run-in period.

^b CI calculated based on published SD and n in the two treatment arms.

^c ERR calculated using published event rates for each treatment arm and CI is calculated based on event rates and number of events for each treatment arm.

^d EOR calculated using the proportion of patients reaching target in each treatment arm and CI is calculated based on proportion of patients reaching target in each treatment arm and n in each treatment arm.

^e Defined as SMPG ≤3.1 mmol/L with or without symptoms in DUAL II and DUAL V, SMPG ≤3.1 mmol/L with symptoms in SWITCH 2, and SMPG ≤3.3 mmol/L with symptoms in LixiLan-L.

ADA, American Diabetes Association; BG, blood glucose; CI, confidence interval; EOR, estimated odds ratio; ERR, estimated rate ratio; ETD, estimated treatment difference; IDeg, insulin degludec; IDegLira, insulin degludec/liraglutide; IGlar U100, insulin glargine 100 U/mL; iGlarLixi, insulin glargine/lixisenatide; NR, not reported; SD, standard deviation; SMPG, self-measured plasma glucose; U, units; w/o, without.

In order to minimize bias in estimating relative treatment effects between competing interventions based on between-trial comparisons, it was important to, first, compare relative treatment effect measures vs the same reference intervention. Second, it was necessary to ensure that there are no differences in study, patient, or contextual factors between the trials that are associated with the relative treatment effects (i.e. effect modifiers)^23,24; or, if such differences exist, consider and discuss how they may affect the results. This analysis was conducted with these requirements in mind.

The first indirect comparison (“Network 1”) was based on the DUAL II and LixiLan-L trials and relied on the assumption that IDeg and IGlar U100 have similar expected outcomes, and that the effect of IDeg capped at 50 U maximum dose is similar to that of IGlar U100 capped at 60 U maximum dose (see Figure 1).

Figure 1. Indirect comparison of IDegLira and iGlarLixi among patients with type 2 diabetes uncontrolled on basal insulin therapy. IDeg, insulin degludec; IDegLira, insulin degludec/liraglutide; iGlarLixi, insulin glargine/lixisenatide; IGlar U100, insulin glargine U100.

In the second indirect comparison (“Network 2”, Figure 1), the assumption that IDeg and IGlar U100 have similar outcomes was avoided by connecting DUAL II and LixiLan-L through SWITCH 2^25,26, the only phase 3 study comparing IDeg with IGlar U100 in a population solely comprised of insulin-experienced patients with T2DM. The “DUAL II + SWITCH 2 path” provided an estimate of IDegLira relative to IGlar U100. Furthermore, DUAL V was included in the network connected directly with LixiLan-L through their common comparator IGlar U100. The difference in treatment effects for IDegLira relative to IGlar U100 based on DUAL V vs the estimate based on the “DUAL II + SWITCH 2 path” was assumed to be caused by the insulin cap in DUAL II. As such, Network 2 allowed adjustment for the impact of insulin cap on the indirect comparison of IDegLira and iGlarLixi, assuming its impact is the same for IDegLira relative to IGlar U100 and iGlarLixi relative to IGlar U100. This assumption is based on the similar mean daily insulin doses for the IDeg and IGlar treatment arms at end-of-trial for DUAL II (45 U) and LixiLan-L (47 U)^9,13. The analysis based on Network 2 was considered the primary indirect comparison.

An indirect comparison based on a network including only DUAL V and LixiLan-L was not considered because the common comparator, IGlar U100, was capped at 60 U in LixiLan-L vs no cap in DUAL V, and no adjustment would be possible.

Analysis

Outcomes of interest at 6 months of follow-up were: change from baseline in glycated hemoglobin (HbA_1c); change from baseline in body weight; insulin dose at end-of-trial; rate of ADA-documented symptomatic hypoglycemic events (SMPG ≤3.9 mmol/L); rate of severe or blood glucose [BG]-confirmed hypoglycemic events (SMPG ≤3.1 mmol/L with or without symptoms in the DUAL trials; SMPG ≤3.1 mmol/L with symptoms in SWITCH 2; SMPG ≤3.3 mmol/L with symptoms in LixiLan-L); proportion of patients achieving HbA_1c < 7%, HbA_1c < 7% without weight gain, HbA_1c < 7% without hypoglycemic events (severe or BG-confirmed in the DUAL trials; ADA-documented symptomatic in LixiLan-L), HbA_1c < 7% without weight gain or hypoglycemic events (severe or BG-confirmed in the DUAL trials; ADA-documented symptomatic in LixiLan-L).

Estimation of the treatment effects of IDegLira relative to iGlarLixi based on Network 1 were performed according to Bucher et al.²⁷, in which an indirect comparison of trial 1 (A vs C) and trial 2 (B vs C) can be represented as A–B = (A–C) – (B–C). The indirect comparison estimates based on Network 2 were obtained with fixed effects generalized linear models with parameters representing the treatment effects of each intervention in the network relative to IGlar U100 and one parameter representing the impact of the insulin cap on the relative treatment effects of IDegLira and iGlarLixi relative to IGlar U100. The parameter estimate for the effect of insulin cap corresponded to the difference in relative treatment effect between the “DUAL II + SWITCH 2 path” and DUAL V (Figure 1). Network 2 thereby adjusts for the insulin cap on IGlar U100 in LixiLan-L, and the estimate represents a situation where IGlar U100 had not been capped in LixiLan-L. Study-specific relative treatment effects regarding change from baseline in HbA_1c, weight, and end of trial insulin dose were assumed to follow a normal likelihood function. The relative treatment effects for response outcomes (odds ratios [ORs]) and hypoglycemic events (rate ratios) were log transformed and represented with a normal likelihood for the analysis as well. The indirect comparison analyses based on Network 2 were performed in a Bayesian framework with non-informative prior distributions for the model parameters using a Markov Chain Monte Carlo method, as implemented in the OpenBUGS software package^28,29.

This article is based on previously conducted studies, and does not involve any new studies of human or animal subjects performed by any of the authors.

Results

The estimated treatment effects of IDegLira relative to iGlarLixi regarding HbA_1c, body weight, and daily insulin dose are presented in Table 4. The estimates from Network 1 and Network 2 were similar. In the primary network (Network 2), IDegLira was estimated to provide a 0.44 [95% CI = 0.17–0.71] %-point reduction in HbA_1c compared with iGlarLixi. Body weight was reduced by 1.42 [95% CI = 0.35–2.50] kg with IDegLira compared with iGlarLixi. Insulin dose was comparable between the two interventions: estimated treatment difference IDegLira vs iGlarLixi, –3.6 [95% CI = –10.3–3.3] U. In the sensitivity analysis (Network 1), results were in the same direction, except for insulin dose, which was 0.3 [95% CI = –2.2–2.7] U higher with IDegLira.

Table 4. Mean difference of change in HbA_1c, change in body weight and daily insulin dose determined for Network 2 (the primary network) and Network 1.

Outcomes	Mean difference (IDegLira vs iGlarLixi) (95% CI)
	Network 1	Network 2 (Primary network)
Change in HbA1c (%)	−0.53 (−0.77, −0.29)	−0.44 (−0.71, −0.17)
Change in body weight (kg)	−1.13 (−1.96, −0.30)	−1.42 (−2.50, −0.35)
Daily insulin dose (U)	0.3 (−2.2, 2.7)	−3.6 (−10.3, 3.3)

The rate for severe or BG-confirmed hypoglycemia with IDegLira was approximately half the rate with iGlarLixi (rate ratio = 0.51 [95% CI = 0.29–0.90]) based on Network 2 (Table 5); however, it should be noted that BG-confirmed hypoglycemia was defined as SMPG ≤3.3 mmol/L in LixiLan-L, as opposed to SMPG ≤3.1 mmol/L in the other trials. Based on the ADA definition of documented symptomatic hypoglycemia (SMPG ≤3.9 mmol/L), the rate was comparable between the two treatments (rate ratio = 1.07 [95% CI = 0.90–1.28]). Rate ratios based on Network 1 showed a similar trend, but the difference between IDegLira and iGlarLixi for severe or BG-confirmed hypoglycemia was not statistically significant. In Network 1, the estimated proportion of patients with HbA_1c < 7.0% was higher (OR = 1.88 [95% CI = 1.08–3.27]) with IDegLira compared to iGlarLixi, as was the proportion of patients with HbA_1c < 7.0% without weight gain (OR = 2.29 [95% CI = 1.20–4.37]) (Table 6). ORs were also higher for the proportion of patients with HbA_1c < 7.0% without hypoglycemia (OR = 2.75 [95% CI = 1.49–5.06]), and the proportion of patients with HbA_1c < 7.0% without weight gain or hypoglycemia (OR = 2.97 [95% CI = 1.41–6.25]).

Table 5. Rate ratio of hypoglycemia determined for Network 2 (the primary network) and Network 1.

Outcome	Rate ratio (IDegLira vs iGlarLixi) (95% CI)
	Network 1	Network 2 (Primary network)
Severe or BG-confirmed hypoglycemia^a	0.66 (0.38, 1.16)	0.51 (0.29, 0.90)
ADA (SMPG ≤3.9 mmol/L) documented symptomatic hypoglycemia	1.17 (0.99, 1.38)	1.07 (0.90, 1.28)

^a Defined as SMPG ≤3.1 mmol/L with or without symptoms in DUAL II and DUAL V, SMPG ≤3.1 mmol/L with symptoms in SWITCH 2, and SMPG ≤3.3 mmol/L with symptoms in LixiLan-L.

ADA, American Diabetes Association; BG, blood glucose; CI, confidence interval; IDegLira, insulin degludec/liraglutide; iGlarLixi, insulin glargine/lixisenatide; SMPG, self-measured plasma glucose; U, units.

Table 6. Odds ratio for patients reaching HbA_1c < 7.0% and HbA_1c < 7.0% without weight gain, hypoglycemia, or either, based on Network 1.^a

Outcome	Odds ratio for IDegLira vs iGlarLixi (95% CI)
HbA1c < 7.0%	1.88 (1.08, 3.27)
HbA1c < 7.0% without weight gain	2.29 (1.20, 4.37)
HbA1c < 7.0% without hypoglycemia^b	2.75 (1.49, 5.06)
HbA1c < 7.0% without weight gain or hypoglycemia^b	2.97 (1.41, 6.25)

^a Responder analysis not carried out in SWITCH 2, therefore these estimates are not available for Network 2.

^b Analyses based on different definitions of hypoglycemia (IDegLira: SMPG ≤3.1 mmol/L; iGlarLixi: SMPG ≤3.9 mmol/L.

CI, confidence interval; IDegLira, insulin degludec/liraglutide; iGlarLixi, insulin glargine/lixisenatide; SMPG, self-measured blood glucose.

It was not possible to conduct the indirect comparison of the responder end-points using Network 2, since the necessary responder end-point results are not available for SWITCH 2.

Discussion

In this indirect treatment comparison, the clinical efficacy of IDegLira was compared indirectly with that of iGlarLixi. Two networks were used, testing different assumptions on the comparability of the pharmacodynamic profiles for IDeg and IGlar U100, and the impact of insulin dose capping within trials. The comparisons made in this analysis suggest that IDegLira would be associated with reduced HbA_1c and body weight, compared with iGlarLixi, and that these reductions were obtained using a lower or similar insulin dose. IDegLira may also be associated with a lower risk of severe or BG-confirmed hypoglycemia vs iGlarLixi.

The mean reduction in HbA_1c was 0.44% greater with IDegLira compared to iGlarLixi, and a greater proportion of patients reached HbA_1c < 7.0%. This improvement in glycemic control is considered clinically relevant in terms of the reduction in risk of diabetes-related micro- and macro-vascular complications, and could be augmented by the reduction in body weight^30,31. In numerous studies, it has been shown that HbA_1c and BMI are the main drivers of long-term clinical outcomes in patients with diabetes (including the risk of heart failure, myocardial infarction, stroke, blindness amputation, ulceration, and mortality)^30,32. These improvements may provide substantial cost savings for healthcare payers through reduced expenditure on the treatment of diabetes-related complications³³. Evidence from available cost-effectiveness analyses further support the role of HbA_1c and BMI in determining long-term outcomes; improvements in which have been shown to offset higher initial treatment costs^34,35. The estimated relative treatment effects for IDegLira vs iGlarLixi, derived from this indirect treatment comparison, could be used as a basis for future cost-effectiveness analyses.

Several factors need to be considered when interpreting the results of this indirect treatment comparison. These relate to assumptions made and differences between the trials in baseline HbA_1c levels, previous use of sulfonylureas, definitions of hypoglycemia, presence or absence of run-in period, the different duration of the trials, and the cross-over design of SWITCH 2. In general, the direction and magnitude of biases resulting from these cross-trial differences are difficult to predict and present a considerable limitation for all indirect comparisons, which rely on the assumption that all important prognostic factors have been balanced between trials. Each of these potential assumptions, and relevant considerations that should minimize any resulting bias, are discussed below.

First, the assumption made in Network 1 that IDeg and IGlar U100 have a similar effect on outcomes may be challenged. Data from clamp studies comparing these two insulins have shown that IDeg has a considerably longer half-life and much lower variability in BG-lowering activity compared with IGlar U100, and this difference is supported by the clinical outcomes in phase 3 RCTs^36–38 and in SWITCH 2³⁹. In particular, the assumption of similar clinical outcomes for IDeg and IGlar U100 is conservative with respect to the rate of hypoglycemia in the context of results from phase 3 RCTs with IDeg vs IGlar U100, in patients with T2D using insulin. In these trials, rates of hypoglycemia with IDeg were lower or similar to those with IGlar U100, and end-of-trial insulin doses were similar with IDeg and IGlar U100^40,41.

In Network 1 it was assumed that the impact of the insulin cap is the same for all treatments. There is a possibility that other study-level factors are correlated with the presence of an insulin cap and also contribute to the difference between the estimates for Network 1 and Network 2. As a result, the adjustment may have been confounded, leading to a biased indirect comparison of IDegLira relative to iGlarLixi. One possibility is that a higher insulin cap might make a treatment appear more effective in terms of glycemic control, but would also have effects on body weight and the rate of hypoglycemia. However, a recent analysis estimated that the effect of uncapping the IGlar U100 treatment arm in LixiLan-L would only have reduced HbA_1c by a further 0.01%, at maximum⁴². It should be noted that, in DUAL II, the mean end-of-trial insulin dose was 45 U in both treatment arms, and in LixiLan-L the mean end-of-trial insulin dose was similar at 47 U in both treatment arms. In DUAL II, 65% and 67% of patients treated with IDegLira and IDeg, respectively, reached the maximum insulin dose of 50 units; while in LixiLan-L 27% and 31% of patients treated with iGlarLixi and IGlar U100 reached the maximum dose of 60 units, respectively. This indicates that the insulin cap affected both arms similarly in each trial and, therefore, cancels out the relative comparison, justifying the assumption that the impact of the insulin cap is the same for all treatments.

Sub-group analyses of the LixiLan-L study indicate that reductions in HbA_1c with iGlarLixi compared with IGlar U100 are similar for different baseline HbA_1c sub-groups (<8% and ≥8%)²². We, therefore, believe that differences in baseline HbA_1c levels between trials are unlikely to have introduced bias (Table 2).

Non-severe hypoglycemia was defined differently between the DUAL and LixiLan-L studies. Severe hypoglycemia is widely considered to be any event in which the assistance of another person is required, but definitions of non-severe hypoglycemia vary widely and may or may not include symptomatic episodes of hypoglycemia, regardless of the BG concentration. The threshold of ≤3.9 mmol/L that is suggested by the ADA and European Association for the Study of Diabetes (EASD) for defining hypoglycemia is very close to the fasting plasma glucose target of 4.0–5.0 mmol/L often used in treat-to-target studies that compare new therapies. This means that many patients may record hypoglycemia during routine testing of BG concentrations, which could increase the number of episodes in each arm and make it more difficult to detect genuine differences. For this reason, some studies have used a lower target to provide a buffer between effective titration of insulin therapy and incipient hypoglycemia. The rate ratios for severe or BG-confirmed hypoglycemia in Table 5 are based on a target of ≤3.1 mmol/L in DUAL II, DUAL V, and SWITCH 2, and ≤3.3 mmol/L in LixiLan-L. However, we expect that the rate ratio between arms would have been similar in LixiLan-L if a target of ≤3.1 mmol/L had been used, and, thereby, that the comparison between IDegLira and iGlarLixi would have been similar.

It is notable that the ADA and EASD recently issued a joint position statement suggesting that BG levels of <3.0 mmol/L, which these bodies consider to be a clinically significant biochemical hypoglycemia, should be reported in clinical trials in preference to the <3.9 mmol/L threshold (depending on the purpose of the study)⁴³.

Comparing responder rates (proportion of patients achieving HbA_1c < 7% without hypoglycemic events, and the proportion of patients achieving HbA_1c < 7% without weight gain or hypoglycemic events) was complicated by the different definitions of hypoglycemia used in the DUAL studies and LixiLan-L. Notwithstanding, the estimates in the present analysis are the best possible estimates that can be made using current data, and could be considered as a conservative comparison of the composite end-points using a low hypoglycemia threshold. The hypoglycemia benefit for iGlarLixi compared to IGlar U100 is only observed in LixiLan-L at 3.9 mmol/L and not at 3.3 mmol/L. Hence, the composite end-points using a low threshold (3.1 or 3.3 mmol/L) instead of 3.9 mmol/L would likely be less favorable for iGlarLixi compared to IGlar U100 and, thereby, more favorable for IDegLira compared to iGlarLixi.

Whereas the LixiLan-L trial outcome period was preceded by a run-in period with IGlar U100 dose optimization, the DUAL II, DUAL V, and SWITCH 2 trials did not have a run-in period. It is difficult to determine what proportion of the change in HbA_1c in the LixiLan-L trial between the start date and the initiation of iGlarLixi is due to regression to the mean, and what is due to IGlar U100 dose optimization. However, the run-in period would only introduce bias to the results if it did not affect both treatment arms equally. Similarly, the DUAL studies had a lower titration target (4.0–5.0 mmol/L) compared with LixiLan-L (4.5–5.5 mmol/L), and, whilst this may have led to a greater reduction in HbA_1c, the target applied to both treatment arms within studies. The higher odds of reaching HbA_1c < 7% with IDegLira provide further evidence for the improved BG-lowering effect vs iGlarLixi, because the presence of a run in period would not affect this outcome as it would change in HbA_1c from baseline.

This indirect comparison provides an estimate of the relative benefits of IDegLira vs iGlarLixi in the absence of direct evidence, based on high-quality data gleaned from individual studies with large sample sizes. As such, the results of this study provide insights into the relative clinical utility of the only two currently available compounds within this novel class of glucose-lowering agents and may, thus, be used to inform guideline development and reimbursement discussions with respect to this new class of therapies. The robustness of the primary analysis based on Network 2 is supported by the sensitivity analysis based on Network 1, which demonstrated changes in end-points that were similar. The study has a high degree of transparency due to the inclusion of only publicly available data. Limitations that could introduce biases have been addressed, and possible mitigating factors discussed.

Conclusions

Fixed-ratio combinations of insulin and GLP-1 analogs have the potential for improving clinical outcomes in patients with T2DM uncontrolled on basal insulin. The results of this indirect treatment comparison demonstrate that, among patients with T2DM uncontrolled on basal insulin, treatment with IDegLira resulted in a greater reduction of HbA_1c, higher odds of reaching HbA_1c < 7%, and a greater reduction in body weight, compared with iGlarLixi, at similar insulin doses.

Transparency

Declaration of funding

Sponsorship for this study and article publication fees were funded by Novo Nordisk A/S, Søborg, Copenhagen, Denmark. Precision Health Economics received funding from Novo Nordisk to perform the analysis in this study.

Declaration of financial/other interests

ME has received honoraria and research awards from Novo Nordisk, Sanofi Aventis, MSD, and Novartis. LKB has participated in advisory panels and is on the speakers’ bureau for Novo Nordisk. JHB, US, and AA are employees of Novo Nordisk A/S. TJA was an employee of Novo Nordisk A/S during the time of the analysis and review of the manuscript. JPJ is an employee of Precision Health Economics. Peer reviewers on this manuscript have received an honorarium from JME for their review work, but have no other relevant financial relationships to disclose.

Acknowledgments

Writing, editing, and submission assistance in the preparation of this manuscript was provided by Paul Tisdale and Beverly La Ferla of Watermeadow Medical, an Ashfield company, part of UDG Healthcare plc. Support for this assistance was funded by Novo Nordisk.