Study participants and compliance

Participants were recruited between 9 June 2020 and 7 June 2021; the last participant completed the trial on 5 November 2021. In total, 669 people were screened, 413 were randomised and 411 were treated (DGs 1, 2 and 4, n=50 each; DG3, n=52; DG5, n=51; DG6, n=49; semaglutide up to 1.0 mg qw, n=50; placebo, n=59; ESM Fig. 2). Of the 411 participants treated, 80 (19.5%) prematurely discontinued treatment, 53 (12.9%) owing to an AE. Important protocol deviations were reported for 62 of all randomised participants (15.0%), with two-thirds (n=41) due to restricted medication use. All 411 participants treated were analysed for efficacy (full analysis set: all randomised participants who received one or more dose of the study drug and had analysable data for one or more efficacy endpoint) and safety (treated set: all randomised participants who received one or more dose of the study drug). Baseline characteristics and demographics were similar between DGs (N=411); 83.7% of participants were White and mean ± SD age was 57.3±9.8 years, BMI 33.9±6.0 kg/m2 and HbA1c 64.7±9.2 mmol/mol (8.1±0.8%) (Table 1). The population included in this study was representative of a large study population of people with type 2 diabetes with respect to age and HbA1c levels [17]; however, most participants were White and had a higher bodyweight, due to the inclusion criteria of this trial.

Table 1 Participant baseline characteristics and demographics Primary endpoint

Absolute HbA1c (mixed model for repeated measures [MMRM] estimates; primary endpoint) decreased from baseline after 16 weeks’ treatment with survodutide, with a markedly weaker treatment effect observed in DG1 (0.3 mg qw) than in the other DGs; no obvious dose-dependent effects were observed between DG2–6 (Fig. 1). The results of the multiple contrast test showed that the contrasts of all predefined dose–response models were significant in terms of non-flat dose–response for the absolute change from baseline in HbA1c after 16 weeks of treatment at a one-sided α=0.025. According to the final multiple comparisons procedure with modelling (MCPMod) averaging model, the predicted dose–response reached a plateau at 1.8 mg qw survodutide, with no increase in treatment effect seen at doses higher than this (ESM Fig. 3a). After 16 weeks’ treatment with survodutide, adjusted mean (95% CI) HbA1c levels decreased from a baseline (mean ± SD) of 64.7±9.2 mmol/mol (8.07±0.84%; N=411) as follows: DG1 (n=41), −9.92 mmol/mol (−12.27, −7.56; −0.91% [−1.12, −0.69]); DG2 (n=46), −15.95 mmol/mol (−18.27, −13.63; −1.46% [−1.67, −1.25]); DG3 (n=36), −18.72 mmol/mol (−21.15, −16.29; −1.71% [−1.94, −1.49]); DG4 (n=33), −17.01 mmol/mol (−19.59, −14.43; −1.56% [−1.79, −1.32]); DG5 (n=44), −17.84 mmol/mol (−20.18, −15.51; −1.63% [−1.85, −1.42]); DG6 (n=36), −18.38 mmol/mol (−20.90, −15.87; −1.68% [−1.91, −1.45]). The decrease from baseline was significantly greater for all survodutide groups compared with placebo (−1.62 mmol/mol [−3.83, 0.59]; −0.15% [−0.35, 0.05]; n=49) at all tested time points (p<0.0001 for all DGs and time points except DG1 week 5 [p=0.0004]). After 16 weeks’ treatment, low-dose survodutide treatment (0.9 mg qw [DG2]) reduced HbA1c to approximately the same extent as semaglutide (n=45) up to 1.0 mg qw ( −15.95 mmol/mol [−1.46%] vs −16.07 mmol/mol [−1.47%], respectively). Descriptive statistics of the primary endpoint revealed similar results to the MMRM analysis (ESM Fig. 3b); survodutide reduced mean ± SD HbA1c by up to 19.5 mmol/mol (1.88%) in both DG3 (n=36) and DG6 (n=36) after 16 weeks, with a low dose (DG2, n=46) again showing similar results to the reductions seen with semaglutide (n=45) up to 1.0 mg qw (−14.9±10.2 mmol/mol [−1.37±0.93%] vs −16.4±9.2 mmol/mol [−1.50±0.84%], respectively).

Fig. 1

MMRM estimates for the absolute change in HbA1c from baseline to EoT. aSemaglutide arm was open label

Secondary endpoints

The relative and absolute reduction from baseline in bodyweight was greater with increasing survodutide dose, with bodyweight loss seen in all survodutide DGs in a dose-dependent manner (Fig. 2, ESM Fig. 4). After 16 weeks of treatment, the relative decrease in bodyweight from baseline (key secondary endpoint) for DG2–6 was significantly greater than for placebo (p<0.001), with a maximum adjusted mean (95% CI) MMRM estimate for relative bodyweight reduction of −8.7% (−10.1, −7.3; DG6, n=37; Fig. 2). Survodutide doses of ≥1.8 mg qw produced greater adjusted mean (95% CI) bodyweight reductions than semaglutide up to 1.0 mg qw (DG3 [n=36] vs semaglutide [n=45]: −6.6% [−7.9, −5.3] vs –5.3% [−6.6, −4.1]). Results of the multiple contrast test showed that all predefined dose–response models were significant in terms of non-flat dose–response for the relative change from baseline in bodyweight at a one-sided α=0.025. A significant dose–response was seen in the final MCPMod averaging model and did not reach a plateau (ESM Fig. 4a). Descriptive statistics of the relative change from baseline in bodyweight were similar to the MMRM analysis (ESM Fig. 4b). The adjusted mean (95% CI) MMRM estimates for absolute changes from baseline in bodyweight after 16 weeks’ treatment were consistent with the relative changes, with favourable results seen for DG3–6 (up to −8.4 kg [−9.7, −7.1]; DG6, n=37) compared with semaglutide (n=45) up to 1.0 mg qw (−5.2 kg [−6.4, −4.0]) (ESM Fig. 4c).

Fig. 2

MMRM estimates for the relative change in bodyweight from baseline to EoT. aSemaglutide arm was open label

Analysis of additional secondary endpoints showed that the proportion of participants with ≥5% and ≥10% reductions in bodyweight after 16 weeks’ treatment increased dose-dependently with survodutide (Table 2). In DG6, 57.1% of participants (n=28) had ≥5% and 34.7% (n=17) had ≥10% bodyweight reductions; this compares with 6.8% (n=4) and 0%, respectively, for placebo and 38.0% (n=19) and 16.0% (n=8), respectively, for semaglutide up to 1.0 mg qw. The probability of achieving ≥5% or ≥10% bodyweight loss was significantly greater in DG2–6 for ≥5% loss (OR [95% CI] DG2, 7.92 [2.43, 25.74]; DG3, 17.68 [5.21, 60.03]; DG4, 25.87 [7.31, 91.55]; DG5, 21.75 [6.57, 72.04]; DG6, 35.00 [9.84, 124.47]) and DG3–6 for ≥10% loss (DG3, 25.17 [1.35, 470.86]; DG4, 33.00 [1.78, 613.20]; DG5, 42.43 [2.37, 761.05]; DG6, 84.51 [4.71, >999]) compared with placebo (Table 2).

Table 2 Proportion of participants achieving ≥5% and ≥10% bodyweight reductions

Waist circumference decreased from baseline with both survodutide and semaglutide treatment, but data were highly variable and associated with wide 95% CIs (ESM Fig. 5). The maximum adjusted mean ± SEM MMRM estimate for decrease in waist circumference from baseline was observed after 16 weeks’ treatment in DG6 (−10.5±1.7 cm; n=36). Adjusted mean ± SEM MMRM estimates for the placebo-corrected absolute change from baseline after 16 weeks’ treatment were significant for DG4 (−4.6±2.25 cm; n=35; p=0.041) and DG6 (−8.4±2.24 cm; n=36; p=0.0002).

Further efficacy endpoints

Treatment with survodutide or semaglutide reduced the mean 7-point self-monitoring of blood glucose (SMBG) level to a greater extent than placebo at week 16; the maximum mean ± SD decrease from baseline was observed in DG3 (−3.03±2.47 mmol/l; n=36) (Fig. 3), and decreases were more pronounced at post-mealtime time points than at pre-mealtime time points.

Fig. 3

Change from baseline in 7-point SMBG. Blood glucose measurements were collected before each meal (assuming three meals a day), approximately 2 h after each meal and at bedtime on a single day during screening (baseline) and a single day between the last dose of study drug and EoT (week 16). Mean overall time points represent the mean per participant per visit of the seven blood glucose measurements at baseline and week 16. Data are presented per DG. The centre line denotes the median value, with the symbols within the boxes denoting the mean. The box boundaries mark the upper and lower quartile of the dataset. The whiskers indicate the variability of the data; whiskers are drawn to the nearest value within 1.5× the IQR of the upper and lower quartiles. Any observations outside of these values are plotted with symbols. aSemaglutide arm was open label

Minor treatment effects were observed for all domains of the TFEQ-R18 V2, hunger VAS and PGI-S; a full description of these results is given in the ESM for the 16 week time frame (ESM Tables 1 and 2).

Safety

A total of 77.8% of participants (n=235) treated with survodutide reported at least one treatment-emergent adverse event (TEAE); 52.5% of those receiving placebo (n=31) and 52.0% receiving semaglutide (n=26) also reported TEAEs (Table 3). Of these, severe AEs were reported by 16 participants treated with survodutide (5.3%), four receiving placebo (6.8%) and none receiving semaglutide; these were mostly GI disorders (n=8/16 [50%] for survodutide; n=1/4 [25%] for placebo). Serious AEs were reported by 3.6% of participants receiving survodutide (n=11: DG1, n=1; DG2, n=4; DG3, n=3; DG4, n=2; DG5, n=1) and 5.1% receiving placebo (n=3) (Table 3, ESM Table 3). AEs led to treatment discontinuation in 15.9% of participants receiving survodutide (n=48), 5.1% receiving placebo (n=3) and 4.0% receiving semaglutide (n=2). Most of these discontinuations occurred within the first 6 weeks of the study (n=38/53, 71.7%), coinciding with the dose-escalation period, and were mainly due to GI disorders (survodutide: n=36/48, 75.0%; placebo n=1/3, 33.3%; semaglutide: 0) such as nausea and vomiting. More discontinuations due to AEs were observed in those with baseline bodyweight <100 kg than those with bodyweight ≥100 kg. GI disorders were the most frequently reported AEs across all DGs, occurring in 55.3% of participants receiving survodutide, 22.0% receiving placebo and 28.0% receiving semaglutide.

Of participants treated with survodutide, 58.6% (n=177) reported drug-related AEs, compared with 22.0% receiving placebo (n=13) and 38.0% of those receiving semaglutide (n=19) (Table 3). The majority of AEs were GI disorders (nausea, vomiting, diarrhoea and dyspepsia), which were reported for 50.0% (n=151) receiving survodutide. Drug-related AEs were classed as serious for 1.3% of participants receiving survodutide (n=4); DG1 (abdominal pain and vomiting), DG2 (mouth ulceration, autoimmune disorder and pharyngeal ulceration), DG3 (dehydration) and DG4 (diarrhoea) (n=1 each; Table 3, ESM Table 3).

Mean heart rate was slightly increased from baseline in all treatment groups; the mean increase after 16 weeks was 2.3–7.3 beats per min (bpm) across the survodutide DGs, 5.9 bpm for semaglutide and 1.67 bpm for placebo. The mean increases in heart rate from baseline were below 10 bpm at all time points except for two in DG4 (10.3 bpm at week 7; 10.1 bpm at week 16). There were no new onsets reported in the QTcF (QT interval corrected for heart rate using the method of Fridericia) interval categories >480–500 msec or >500 msec; one participant each from DG3 (2.0%) and the placebo group (1.7%) reported an increase in QTcF interval of >60 msec. The changes from baseline in QTcF interval were considered minor and no increased risk of cardiovascular events was identified.

Pharmacodynamic endpoints

Treatment with survodutide did not lead to any clear dose-dependent reductions in the NASH-related Fib-4 score, APRI and NAFLD fibrosis score relative to placebo (Table 4). Mean ± SD Pro-C3 levels were substantially lowered from baseline over time in all survodutide DGs (up to −7.3±9.4 µg/l in DG5; n=43) and the semaglutide group (−5.2±7.3 µg/l; n=44) compared with placebo (−0.1±10.2 µg/l; n=48) (Table 4). Changes from baseline in ELF score were detected in all treatment groups, with decreases observed in DG2–6, up to a mean ± SD change of −0.2±0.5 in DG6 (n=36) compared with an increase of 0.2±0.4 in the placebo group (n=49) (Table 4). The changes from baseline in other exploratory biomarkers are presented in Table 4.

Table 4 Absolute change from baseline in exploratory variables at EoT

Plasma exposure to survodutide increased with escalating weekly or biweekly doses, with trough concentrations increasing in an approximately dose-proportional manner (ESM Fig. 6). Visually, steady state for survodutide appeared to be achieved at week 8 for DG1–5 and at week 12 for DG6 (ESM Fig. 6).

Glucagon levels decreased dose-dependently from baseline over 16 weeks’ treatment with survodutide, suggesting target engagement of GCGRs and GLP-1Rs (ESM Fig. 7a). The mean ± SEM changes after 16 weeks were most pronounced in DG5 and DG6 (−11.0±2.0 pmol/l [n=32] and −15.1±3.4 pmol/l [n=22], respectively), and no relevant changes from baseline were observed in the semaglutide or placebo groups. Although small treatment effects on plasma amino acid levels were observed in DG6 compared with semaglutide up to 1.0 mg qw and placebo following 16 weeks’ treatment, these were only notable for alanine; the mean ± SEM decrease in alanine from baseline to EoT in DG6 (−58.4±24.2 µmol/l; n=36) was indicative of target engagement at GCGRs (ESM Fig. 7b).