CGM-derived postprandial glucose with IcoSema versus other insulin regimens: a post hoc analysis of COMBINE 1 and 3 Authors: Christophe De Block, Malik Benamar, Ariel Fu, Raluca Maltesen, Francesco Giorgino Published: 2026 ABSTRACT SummaryBackgroundImpaired insulin and incretin responses, alongside elevated postprandial glucose (PPG) levels after meals, are hallmarks of type 2 diabetes (T2D) and can lead to postprandial hyperglycaemia. This post hoc analysis used PPG levels derived from continuous glucose monitoring (CGM) to investigate the effect of IcoSema versus other insulin-based regimens on PPG control in T2D.MethodsIn this post hoc analysis, a modified Glucose Rate Increase Detector (GRID) algorithm was used to assess the effect of IcoSema compared with once-weekly basal insulin icodec (icodec) or daily basal-bolus therapy (BBT) on PPG endpoints in adults with inadequately controlled T2D, using data from the COMBINE 1 and 3 trials. In the randomised, phase 3a COMBINE 1 and 3 trials, IcoSema (a once-weekly combination therapy of basal insulin icodec [icodec] and semaglutide) was investigated versus icodec (COMBINE 1) or daily BBT (insulin glargine U100 + insulin aspart; COMBINE 3). CGM data collected during weeks 48–52 were used to estimate mealtimes using the GRID algorithm and to derive PPG endpoints. The primary aim of this analysis was to use CGM data collected during the final 4 weeks of treatment in COMBINE 1 and 3 to determine the effect of IcoSema versus other insulin-based regimens on PPG control in individuals with T2D. Data were analysed using the full analysis set (all randomly assigned participants) and the on-treatment period. Participants were screened between June 1, 2022–March 13, 2023, in COMBINE 1 and November 30, 2021–September 28, 2022, in COMBINE 3. COMBINE 1 and 3 are registered on ClinicalTrials.gov (NCT05352815 and NCT05013229, respectively) and are complete.FindingsData from 1650 participants were analysed. In both trials, an average of 2.3 meals/day/participant were detected. IcoSema treatment resulted in statistically significantly lower mean peak PPG increment, 90-min PPG increment, and 120-min PPG increment values and a faster time to return to normal glucose levels versus once-weekly basal insulin (all p < 0.0001). The 120-min PPG increment was also statistically significantly lower with IcoSema versus daily BBT (p = 0.0009); however, there were no statistically significant differences in mean peak PPG increment, 90-min PPG increment, or time to return to normal glucose levels.InterpretationIcoSema provided statistically significantly better PPG control than once-weekly icodec and similar PPG control versus daily BBT with only one weekly injection. Future studies that prospectively evaluate CGM-derived PPG outcomes and their clinical relevance across different meal compositions in real-world treatment settings would be beneficial.FundingNovo Nordisk A/S.Keywords: Continuous glucose monitoring, Postprandial glucose, IcoSema, Type 2 diabetes Research in context Evidence before this study IcoSema is a once-weekly combination therapy of basal insulin icodec (icodec) and the glucagon-like peptide-1 (GLP-1) analogue semaglutide. To identify relevant phase 3 studies of combination therapies of basal insulin and GLP-1 receptor agonists, and their effects on postprandial glucose (PPG), we searched PubMed for relevant articles published in English between 1 January, 2000, and 5 February, 2026, using the terms (“fixed-ratio combination” AND “basal insulin” AND “glucagon-like peptide-1” AND “type 2 diabetes”). Currently available once-daily combination therapies include insulin glargine with lixisenatide (iGlarLixi) and insulin degludec with liraglutide (iDegLira). Previous studies have shown that, compared with basal insulin alone, both iGlarLixi and iDegLira improved glycaemic control, reduced PPG levels, and reduced residual hyperglycaemia. Compared with basal-bolus therapy (BBT), these fixed-ratio combination therapies had comparable glycaemic control, reduced hypoglycaemia, resulted in weight loss (versus weight gain with BBT), and reduced the number of injections required. In the COMBINE 1 phase 3a trial, which included participants with inadequately controlled type 2 diabetes (T2D), IcoSema (a once-weekly combination therapy) demonstrated superiority in the following outcomes versus icodec alone: change in glycated haemoglobin (HbA1c) levels; change in body weight; and overall lower hypoglycaemia rates. COMBINE 2 was a phase 3a trial in which IcoSema demonstrated superior reductions in HbA1c, and similar hypoglycaemia rates compared with once-weekly semaglutide. In the COMBINE 3 phase 3a trial, when compared with BBT (insulin glargine U100 + insulin aspart), IcoSema achieved non-inferior reductions in HbA1c, superiority in change in body weight, and superiority in hypoglycaemia rates. Added value of this study In this post hoc analysis, a modified version of the Glucose Rate Increase Detector (GRID) algorithm was used to assess the effect of IcoSema compared with icodec or daily BBT on PPG levels in participants with inadequately controlled T2D using data from the COMBINE 1 and 3 trials. CGM data were not collected in COMBINE 2 and therefore the trial was not included in this post hoc analysis. In COMBINE 1, IcoSema treatment resulted in statistically significantly lower mean peak PPG increment, 90-min PPG increment, and 120-min PPG increment values and a statistically significantly faster time to return to normal glucose levels compared with icodec (all p < 0.0001). In COMBINE 3, the 120-min PPG increment was statistically significantly lower with IcoSema versus daily BBT (p = 0.0009); however, there were no statistically significant differences in mean peak PPG increment, 90-min PPG increment, or time to return to normal glucose levels (all p > 0.05). Implications of all the available evidence In COMBINE 1, there was no statistically significant difference in fasting glucose levels between IcoSema and icodec; however, IcoSema treatment resulted in significantly greater improvements in HbA1c levels. These results suggest that the statistically superior HbA1c control with IcoSema versus icodec is probably due to the enhanced effects of IcoSema on PPG levels compared with those of icodec. Compared with daily BBT, IcoSema had comparable PPG control but reduced the treatment burden on individuals owing to the requirement for fewer injections (52 injections/year with IcoSema versus up to 1460 injections/year with daily BBT [assuming up to three bolus injections/day]). Future studies that prospectively evaluate CGM-derived PPG outcomes and their clinical relevance across different meal compositions in real-world treatment settings would be beneficial. Introduction In individuals without diabetes, a transient increase in postprandial glucose (PPG) occurs following a carbohydrate-containing meal, and this triggers an increase in insulin secretion from pancreatic β-cells and in incretin secretion from the epithelium of the gastrointestinal tract.1,2 An impaired insulin response and postprandial hyperglycaemia (glucose levels of ≥200 mg/dL [≥11.1 mmol/L]) are hallmarks of type 2 diabetes (T2D).3 If left uncontrolled, postprandial hyperglycaemia can increase the risk for, and contribute to the progression of, additional diabetes-associated complications.4, 5, 6, 7 Long-term studies in individuals with T2D have demonstrated that postprandial hyperglycaemia is also associated with increased risk of all-cause mortality and of cardiovascular disease and events.4, 5, 6, 7 PPG excursions and the time taken to return to normal preprandial glucose levels are dependent on many factors, including the size and type of food intake; an individual’s β-cell function, incretin response, and preprandial glycaemic control; and treatment regimen. Measurement of PPG and fasting plasma glucose (FPG) levels using continuous glucose monitoring (CGM) can provide detailed insights on an individual’s glycaemic profile in response to meals and throughout the day, as well on the relative contribution of PPG levels to overall glycaemic control.8 In individuals with T2D, improvements in both FPG and PPG are major contributors to achieving overall glycated haemoglobin (HbA1c) targets, in addition to aiding improvements in CGM metrics such as time in range (TIR; 70–180 mg/dL [3.9–10.0 mmol/L]) through the reduction of time above range (>180 mg/dL [>10.0 mmol/L]).1,8 There are various therapeutic options for the management of T2D (including basal insulin, metformin, and thiazolidinediones) that predominantly affect FPG levels and have relatively modest effects on PPG levels.9 If HbA1c targets are not met despite achieving preprandial glucose targets, clinical guidelines recommend targeting PPG levels with the addition of either a glucagon-like peptide-1 receptor agonist (GLP-1 RA), such as semaglutide, or the addition of bolus insulin.10,11 Semaglutide has been shown to lower both FPG and PPG levels significantly in people with T2D and/or obesity versus comparators (sitagliptin, insulin glargine, or exenatide).12,13 IcoSema (a once-weekly combination therapy of basal insulin icodec [icodec] and semaglutide) has been evaluated in the phase 3a COMBINE clinical trial programme (COMBINE 1–3).14, 15, 16 In COMBINE 1 and 3, the efficacy and safety profiles of IcoSema were investigated versus once-weekly basal insulin (icodec) or daily basal-bolus therapy (BBT; insulin glargine U100 [glargine U100] + insulin aspart [aspart]) in adults with inadequately controlled T2D receiving daily basal insulin.14,15 IcoSema showed superiority to once-weekly basal insulin (COMBINE 1) and non-inferiority to daily BBT (COMBINE 3) for the change in HbA1c from baseline to week 52.14,15 No statistically significant differences in FPG levels were seen between treatment groups in either trial.14,15 During weeks 48–52, TIR (70–180 mg/dL [3.9–10.0 mmol/L]) and time in tight range (TITR; 70–140 mg/dL [3.9–7.8 mmol/L]) were statistically significantly greater with IcoSema than with once-weekly basal insulin in COMBINE 1; no statistically significant differences were observed between the IcoSema and daily BBT groups in COMBINE 3 for either parameter.14,15,17 A Glucose Rate Increase Detector (GRID) algorithm was previously developed to use CGM data to estimate the rate of change of glucose levels and detect meals in individuals with diabetes.18 This post hoc analysis used a modified version of the GRID algorithm to analyse CGM data collected in the COMBINE 1 and 3 trials to estimate mealtimes and assess PPG parameters, such as the mean peak PPG increment. The aim of this analysis was to determine the effect of IcoSema versus other insulin-based regimens on PPG control in individuals with T2D. Methods Study design and participants COMBINE 1 and 3 were randomised, multinational, open-label, treat-to-target, phase 3a trials, each comprising a 2-week screening period, a 52-week treatment period, and a 5-week follow-up period. COMBINE 1 was conducted at 192 outpatient clinics, and hospital departments across 20 countries and regions, while COMBINE 3 was conducted at 109 outpatient clinics and hospital departments across 14 countries. The trial designs, methods, and statistical analyses have been published previously.14,15 There was no participant involvement in the design, conduct, and reporting of the trials. Both trials were in adults (aged ≥ 18 years) with T2D (HbA1c 7.0–10.0% [53.0–85.8 mmol/mol]) inadequately controlled with daily basal insulin (20–80 U/day) with or without oral glucose-lowering medications. Ethics COMBINE 1 and 3 were conducted in accordance with the principles of the Declaration of Helsinki19 and the Good Clinical Practice Guidelines of the International Conference for Harmonisation.20 The protocol, consent forms, and other relevant documents were reviewed and approved by the appropriate institutional review boards or independent ethics committees. The approving ethics committees for COMBINE 1 and 3 can be found in Supplementary Table S1. Written informed consent was obtained from all participants prior to study participation. Both trials are registered with ClinicalTrials.gov (NCT05352815 [COMBINE 1]; NCT05013229 [COMBINE 3]) and have been completed. Randomisation and masking Eligible participants were centrally randomised (1:1) to once-weekly IcoSema or a comparator (COMBINE 1: once-weekly icodec; COMBINE 3: daily BBT [once-daily glargine U100 with 2–4 daily injections of aspart]). In both trials, participants were randomised to the next available treatment according to the randomisation schedule. CGM data were blinded to participants and investigators in both trials. Procedures In COMBINE 1 and 3, IcoSema was initiated at a dose of 40 dose steps (equivalent to 40 U of icodec and 0.114 mg of semaglutide). IcoSema doses were adjusted weekly (±10 dose steps) based on pre-breakfast self-measured blood glucose (SMBG) values measured on the 2 days before titration and on the day of titration. For individuals randomised to icodec in COMBINE 1, the weekly dose was calculated as the pre-trial basal insulin dose multiplied by seven. The first weekly dose was increased by a 50% one-time additional dose. For the second injection, the calculated weekly dose was administered. Thereafter, the icodec dose was adjusted weekly in increments or decrements of 20 U to achieve a pre-breakfast SMBG value of 80–130 mg/dL (4.4–7.2 mmol/L). For individuals randomised to daily BBT in COMBINE 3, glargine U100 was administered once daily using a prefilled pen. Basal insulin doses were adjusted weekly (±3 U) based on pre-breakfast SMBG values measured on the 2 days before and on the day of titration. Aspart was administered two to four times daily with main meals at a starting dose of 4 U per meal. Bolus insulin doses were adjusted twice a week at intervals of 3–4 days based on preprandial or bedtime SMBG values measured on the 3 days before the titration day. During the first 8 weeks, aspart doses were only adjusted for safety reasons. At randomisation, all participants were provided with a blood glucose meter and instructed to measure their pre-breakfast SMBG levels daily. The blood glucose meter was also used to confirm hypoglycaemic events during the studies. In both COMBINE 1 and 3, recording PPG levels with blood glucose meters was not a mandatory requirement. No dietary instructions were provided to participants per protocol, and participants were not required to record timing, size, or composition of their meals in either trial. Participants wore a CGM device (Dexcom G6 [Dexcom, USA]) during weeks 48–52 (COMBINE 1 and 3). Blinded CGM data collected during this period were included in this analysis. CGM data were not used for dose titration or reporting of hypoglycaemia. A modified version of the GRID algorithm18 was used to identify meals from the CGM data and estimate their start time. The CGM signal was smoothed and resampled at 15-min intervals. A meal was detected if the glucose level was above 80 mg/dL (>4.4 mmol/L) and at least one gradient (measuring the rate of change of the CGM signal) was greater than 1.7 mg/dL/min (>0.094 mmol/L/min), two consecutive gradients were greater than 1.6 mg/dL/min (>0.09 mmol/L/min), or three consecutive gradients were greater than 1.5 mg/dL/min (>0.08 mmol/L/min). Meal onset was estimated to be 45 min before the detection point to accommodate biological absorption of the meal. Meals that were detected less than 2 h apart by the GRID method were combined, and PPG parameters were evaluated starting at the first detected meal. Breakfast was defined as meals with an estimated start time between 06:00 and 10:59; lunch as meals with an estimated start time between 11:00 and 15:59; and dinner as meals with an estimated start time between 16:00 and 21:59. Outcomes This post hoc analysis investigated the following CGM-derived PPG parameters: peak PPG increment (change from baseline to the maximum glucose level during the 90 min after the estimated meal start time); 30-, 60-, 90-, and 120-min PPG increments (change from baseline to 30, 60, 90, and 120 min since estimated meal start time); area under the PPG curve from meal start until 90 min after meal start, minus the baseline glucose level (AUC90min) or until 120 min after meal start, minus the baseline glucose level (AUC120min); time to peak PPG increment; and time to return to normal glucose levels (≤140 mg/dL [≤7.8 mmol/L]) (Fig. 1a). In instances when glucose levels never returned to normal levels (≤140 mg/dL [≤7.8 mmol/L]), data were handled as follows: if the glucose peak was ≤140 mg/dL (≤7.8 mmol/L), the time to this peak was used; and if glucose levels remained above 140 mg/dL (>7.8 mmol/L) and did not return to normal before the next meal, the time to the next meal was used. Baseline values were determined from CGM data collected during the 15 min before the estimated meal start by selecting the glucose level closest to the time of the meal start. Overview of (a) PPG parameters and (b) participant disposition. Panel a: any additional gradients during the 2 h after the estimated meal start time did not trigger a new meal identification but were considered part of the already identified meal. In practice, this meant that PPG was only evaluated for meals detected by the GRID method that occurred at least 2 h after a previous meal. aAUC90min is the CGM-derived change from baseline to 90 min after the estimated meal start time, adjusted for baseline value. AUC90min, area under the PPG curve from meal start until 90 min after meal start; CGM, continuous glucose monitoring; GRID, Glucose Rate Increase Detector; PPG, postprandial glucose. Statistical analysis For both COMBINE 1 and 3, sample size calculations were performed for the primary endpoints and have been previously published.14,15 No additional sample size calculations were performed for this post hoc analysis. PPG parameters were analysed using a linear mixed model for repeated measures (MMRM) with treatment and region as fixed factors. The residuals were modelled as having a compound-symmetric covariance structure within individuals across meals, with an additional assumption of no correlation between individuals. Details of the linear mixed model prototype equation can be found in the Supplementary Methods. The MMRM was chosen because it models correlated repeated measures and accommodates missing data under the missing-at-random assumption. Two-tailed t-tests were used to test for treatment differences (two-sided α = 0.05). No correction for multiplicity was applied. Data were analysed using SAS version 8.4 and R version 4.3.1. Role of the funding source This study was funded by Novo Nordisk. Representatives of Novo Nordisk were involved in the design of the study; data collection; data analysis; data interpretation; and the preparation, review, and approval of the manuscript. Novo Nordisk did not have the right to veto the decision to publish or control the decision about which journal the manuscript was to be submitted to; these decisions were made by the authors. Results Participants In COMBINE 1 and 3, participants were screened between June 1, 2022–March 13, 2023, and November 30, 2021–September 28, 2022, respectively. Of the 1970 participants in COMBINE 1 (n = 1291) and 3 (n = 679), 1870 received a CGM device per protocol; of these, 1650 had CGM data and meals detected during weeks 48–52 and were included in this analysis (Fig. 1b). The CGM device used had not been approved in China at the time of the trial; therefore, Chinese participants in COMBINE 1 (n = 100) did not receive a CGM device per protocol and have not been included in this analysis. At baseline, HbA1c level, body mass index, and duration of T2D were broadly similar between treatment groups in each trial and between the two trials.14,15 Meal detection In COMBINE 1 and 3, the modified GRID algorithm detected an average of 2.3 meals per participant per day in each treatment group (Fig. 2). In both trials and in both treatment arms in each trial, most meals (>83%) were identified between 06:00 and 21:59 (Supplementary Table S2). The proportion of meals identified during each of the standard mealtimes (breakfast, lunch, and dinner) was broadly similar across mealtimes, between treatment arms in each trial, and between trials (Supplementary Table S2). Overview of (a) the number of detected meals per day per participant and (b) the estimated start time of meals detected from CGM data (weeks 48–52). Meals detected per day. A meal was detected if the glucose level was >80 mg/dL (>4.4 mmol/L) with either one gradient of >1.7 mg/dL/min (0.095 mmol/L/min), two consecutive gradients of >1.6 mg/dL/min (>0.09 mmol/L/min), or three consecutive gradients of >1.5 mg/dL/min (>0.08 mmol/L/min). Days included are all the days from the first CGM time stamp to the last CGM time stamp at each visit. Aspart, insulin aspart; CGM, continuous glucose monitoring; glargine U100, insulin glargine U100; icodec, insulin icodec. PPG parameters in COMBINE 1: IcoSema versus once-weekly basal insulin Compared with icodec treatment, IcoSema treatment resulted in statistically significantly lower mean values of peak PPG increment (estimated treatment difference [ETD] [95% confidence interval (CI)], −4.2 [−5.51, −2.81] mg/dL; p < 0.0001); 90-min PPG increment (ETD [95% CI], −5.7 [−7.34, −4.07] mg/dL; p < 0.0001); and 120-min PPG increment (ETD [95% CI], −12.2 [−14.32, −10.09] mg/dL; p< 0.0001) (Table 1). Furthermore, the mean time to the peak PPG increment (ETD [95% CI], −1.4 [−1.90, −0.92] minutes; p < 0.0001) and the mean time to return to normal glucose levels (ETD [95% CI], −59.9 [−73.83, −46.03] minutes; p < 0.0001) were statistically significantly faster with IcoSema than with icodec (Table 1). There was no statistically significant difference in mean AUC90min (overall or by meal) between IcoSema and icodec (p > 0.05; Table 1); however, mean AUC120min was statistically significantly smaller with IcoSema than with icodec (ETD [95% CI], −273.2 [−368.91, −177.51] mg∗min/dL; p < 0.0001). By meal, mean AUC120min was statistically significantly smaller with IcoSema than with icodec for all standard mealtimes (breakfast, lunch, and dinner) (Supplementary Table S3). Overall, PPG levels were lower with IcoSema than with icodec when examined as mean values across all meals (Fig. 3a) or by breakfast, lunch, and dinner separately (Fig. 4a). Table 1PPG parameters based on estimated meal start from CGM data (weeks 48–52).Combine 1Combine 3IcoSema (n = 521)Icodec (n = 541)ETD (95% CI); p valueIcoSema (n = 301)Glargine U100 + aspart (n = 287)ETD (95% CI); p valueEstimated mean peak PPG increment, mg/dL61.665.7−4.2 (−5.51, −2.81); p < 0.000162.362.30.0 (−1.77, 1.84); p = 0.9708Estimated mean 30-min PPG increment, mg/dL−6.0−7.11.1 (0.61, 1.58); p < 0.0001−6.9−6.90.0 (−0.63, 0.73); p = 0.8857Estimated mean 60-min PPG increment, mg/dL35.936.3−0.3 (−1.08, 0.38); p = 0.347234.935.3−0.4 (−1.35, 0.52); p = 0.3835Estimated mean 90-min PPG increment, mg/dL53.459.1−5.7 (−7.34, −4.07); p < 0.000153.854.9−1.1 (−3.23, 1.07); p = 0.3234Estimated mean 120-min PPG increment, mg/dL50.963.1−12.2 (−14.32, −10.09); p < 0.000151.956.7−4.7 (−7.53, −1.95); p = 0.0009Estimated mean AUC90min, mg∗min/dL1431.61453.2−21.6 (−69.66, 26.51); p = 0.37891365.41390.0−24.6 (−90.88, 41.63); p = 0.4657Estimated mean AUC120min, mg∗min/dL3006.93280.1−273.2 (−368.91, −177.51); p < 0.00012958.43060.6−102.2 (−229.25, 24.86); p = 0.1147Estimated mean time to peak PPG increment, min73.875.2−1.4 (−1.90, −0.92); p < 0.000173.274.4−1.2 (−1.83, −0.48); p = 0.0008Estimated mean time to return to normal glucose levels (≤140 mg/dL), min256.6316.5a−59.9 (−73.83, −46.03); p < 0.0001287.5284.23.3 (−18.17, 24.81); p = 0.7619The peak PPG increment was defined as the change from baseline to the maximum glucose level during the 90 min after the estimated meal start time. PPG parameters were analysed using a linear mixed model for repeated measures with a compound symmetry covariance matrix. The model included treatment and region as fixed factors. p value: two-sided p value for test of no treatment difference. No correction for multiplicity was applied.Aspart, insulin aspart; AUC90min, area under the PPG curve from meal start until 90 min after meal start, adjusted for baseline value; AUC120min, area under the PPG curve from meal start until 120 min after meal start, adjusted for baseline value; CGM, continuous glucose monitoring; CI, confidence interval; ETD, estimated treatment difference (IcoSema – comparator); glargine U100, insulin glargine U100; icodec, insulin icodec; PPG, postprandial glucose.aEstimated time to return to normal glucose levels for 540 participants in the icodec arm. PPG parameters based on estimated meal start from CGM data (weeks 48–52). The peak PPG increment was defined as the change from baseline to the maximum glucose level during the 90 min after the estimated meal start time. PPG parameters were analysed using a linear mixed model for repeated measures with a compound symmetry covariance matrix. The model included treatment and region as fixed factors. p value: two-sided p value for test of no treatment difference. No correction for multiplicity was applied. Aspart, insulin aspart; AUC90min, area under the PPG curve from meal start until 90 min after meal start, adjusted for baseline value; AUC120min, area under the PPG curve from meal start until 120 min after meal start, adjusted for baseline value; CGM, continuous glucose monitoring; CI, confidence interval; ETD, estimated treatment difference (IcoSema – comparator); glargine U100, insulin glargine U100; icodec, insulin icodec; PPG, postprandial glucose. Estimated time to return to normal glucose levels for 540 participants in the icodec arm. Mean PPG curves after estimated meal start based on meal detection algorithms applied to the CGM signal (weeks 48–52) in (a) COMBINE 1 and (b) COMBINE 3. Curves show mean PPG after all meals. Observed data. The last available CGM measurement during the 15 min before meal start was used for time 0. Aspart, insulin aspart; CGM, continuous glucose monitoring; glargine U100, insulin glargine U100; icodec, insulin icodec; PPG, postprandial glucose. Mean PPG curves after estimated meal start by meal based on meal detection algorithms applied to the CGM signal (weeks 48–52) in (a) COMBINE 1 and (b) COMBINE 3. Curves show mean PPG after each meal. Observed data. The last available CGM measurement during the 15 min before meal start was used for time 0. Breakfast was defined as meals consumed between 06:00 and 10:59; lunch as meals consumed between 11:00 and 15:59; and dinner as meals consumed between 16:00 and 21:59. Aspart, insulin aspart; CGM, continuous glucose monitoring; glargine U100, insulin glargine U100; icodec, insulin icodec; PPG, postprandial glucose. PPG parameters in COMBINE 3: IcoSema versus daily BBT There were no statistically significant differences between IcoSema and daily BBT in mean peak PPG increment value (ETD [95% CI], 0.0 [−1.77, 1.84] mg/dL; p = 0.9708); 90-min PPG increment value (ETD [95% CI], −1.1 [−3.23, 1.07] mg/dL; p = 0.3234); AUC90min (ETD [95% CI], −24.6 [−90.88, 41.63] mg∗min/dL; p = 0.4657); or time to return to normal glucose levels (ETD [95% CI], 3.3 [−18.17, 24.81] minutes; p = 0.7619) (Table 1). However, the mean 120-min PPG increment value was statistically significantly lower with IcoSema than with daily BBT (ETD [95% CI], −4.7 (−7.53, −1.95) mg/dL; p = 0.0009; Table 1). Additionally, the mean time to the peak PPG increment was statistically significantly faster with IcoSema than with daily BBT (ETD [95% CI], −1.2 (−1.83, −0.48) minutes; p = 0.0008 Table 1). By meal, there were no statistically significant differences between IcoSema and daily BBT for AUC90min or AUC120min at breakfast, lunch, or dinner, except for a statistically significantly higher AUC90min with IcoSema at lunch (ETD [95% CI], 121.4 [18.25, 224.60] mg∗min/dL; p = 0.0212) (Supplementary Table S3). Mean CGM-derived PPG curves were similar between IcoSema and daily BBT when examined across all meals (Fig. 3b) or by each meal separately (Fig. 4b). Discussion This is the first detailed investigation of the impact of IcoSema on PPG parameters and characteristics. This post hoc analysis of data from COMBINE 1 and 3 showed that treatment with IcoSema resulted in statistically significantly better PPG control compared with once-weekly icodec and in similar PPG control to that achieved with daily BBT (glargine U100 + aspart). Compared with the once-weekly basal insulin, icodec, IcoSema resulted in a statistically significantly lower peak PPG increment and a statistically significantly faster time to return to normal glucose levels. There was no statistically significant difference in AUC90min for IcoSema compared with icodec; however, IcoSema treatment resulted in a significantly lower AUC120min than that obtained with icodec. These data suggest that, compared with icodec, IcoSema treatment resulted in lower PPG levels after meals, which in turn could also reduce the duration of postprandial hyperglycaemia. The effect of IcoSema was seen across all three daily meals, with lower PPG values observed for each meal type with IcoSema than with once-weekly basal insulin. In both COMBINE 1 and 3, in both treatment arms, estimated mean AUC90min and AUC120min values were highest for breakfast, followed by lunch, and then lowest for dinner. This may suggest a trend towards a higher carbohydrate content in meals eaten at breakfast than in those eaten at lunch or dinner. However, no formal supportive statistical analyses were completed to compare the magnitude of differences between mealtimes and information on meal composition was not collected to allow verification of this. Given that COMBINE 1 and 3 were multinational trials, the composition of the three daily meals probably varied between trial participants; however, despite this likely variability, there was still a greater benefit on PPG parameters with IcoSema than with icodec. Daily basal insulin is known to effectively control FPG levels, but with little effect on PPG.9,21,22 The improved PPG control with IcoSema versus once-weekly basal insulin is likely to be due to the semaglutide component. Semaglutide is known to reduce appetite and food intake, increase insulin sensitivity, increase prandial insulin, and suppress postprandial glucagon secretion in a glucose- dependent manner, reducing hepatic glucose output after meals and glucagon secretion. All of these may contribute to the lower PPG levels compared with icodec. These data from this post hoc analysis align with a previous study that showed that participants treated with the once-daily combination therapy IDegLira (a combination of once-daily basal insulin degludec and liraglutide [a GLP-1 RA]) had significantly better PPG control than those treated with degludec alone.23 There were no statistically significant differences in FPG levels between treatment arms in COMBINE 1, thus suggesting that the PPG control with IcoSema may contribute to the improvement in glycaemic control (HbA1c and TIR) obtained with IcoSema over that obtained with icodec alone. The improved PPG control with IcoSema versus icodec is also in line with the statistically significantly greater TITR reported with IcoSema,17 and this could translate into more time per day spent in euglycaemia with IcoSema than with icodec. In COMBINE 3, there were no statistically significant differences between IcoSema and daily BBT arms in peak PPG increment or in time to return to normal glucose levels. Additionally, there was only a small difference in 120-min PPG increment value compared with BBT. Similar PPG curves were observed after each meal between treatment arms. These data are in keeping with those previously reported for the primary COMBINE 3 trial and other post hoc analyses, which suggest that, compared with daily BBT, IcoSema could offer similar effects on glycaemic control parameters (FPG, PPG, HbA1c, TIR, and TITR), significantly lower rates of hypoglycaemia, and a superior reduction in body weight,15,17 with fewer injections required, which reduces the treatment burden on individuals (52 injections/year for IcoSema versus 1460 injections/year for daily BBT regimens [assuming up to three bolus injections per day]). It should also be noted that aspart was selected as the bolus insulin in the BBT arm of COMBINE 3, rather than faster aspart. Compared with aspart, faster aspart has a quicker absorption rate, which results in a greater initial reduction in glucose levels; therefore, if it was used, it could potentially have had a greater effect than aspart on PPG levels.24 Study strength and limitations This post hoc analysis had multiple strengths. These include the large number of participants from two multinational trials; in both trials, the trial populations had similar baseline characteristics between arms. Furthermore, the same criteria were used to detect each meal, which did not rely on accuracy of participant reporting. Moreover, blinding of the CGM data meant that participants were unlikely to adapt their behaviours based on these data. Finally, CGM data provide more data points than traditional SMBG measurements throughout the day and, unlike traditional SMBG measurements, provide a glucose curve for each day. This post hoc analysis also had some limitations, which should be considered. First, participants did not record postprandial SMBG during the CGM data collection periods. Second, data on the timing of meals, the composition of meals (carbohydrate/fat distribution), and the quantity of carbohydrates consumed were not collected during the trials. Third, in COMBINE 3, recording of the timing of bolus insulin dose administration was not mandatory, making it difficult to validate the data (i.e., to assess consistency between bolus injections and the detected meals in the CGM signal). Fourthly, the GRID algorithm was designed to detect large peaks in glucose levels; thus, small increments in glucose levels may not have been captured, and this could have skewed the mean glucose levels reported. Moreover, it can be assumed that meals containing a low amount of glucose (i.e., those rich in protein and fat) may not have been detected by the GRID algorithm, because they would result in a relatively smaller PPG increment than that resulting from a meal with a high glucose content (e.g., one rich in carbohydrates). This may have impacted the number of meals detected per day for each participant; indeed, an average of 2.3 meals/day/participant were detected, as opposed to the 3 meals/day/participant that might be expected (i.e., one for each of the three standard mealtimes [breakfast, lunch, and dinner]). Additionally, the GRID algorithm was unable to correct for PPG excursions that may have been triggered in response to hypoglycaemic episode corrections; however, reassuringly, the percentage of time spent with glucose levels below 54 mg/dL (<3.0 mmol/L) was very low (≤0.5%) across all treatment arms in COMBINE 1 and 3.14,15 Furthermore, the statistical models did not account for any differences between weekdays and weekends, temporal trends, or meal types. Lastly, the assessment of glycaemic variability would have provided additional context for interpretation. However, because this study focused specifically on PPG parameters to gain insights into meal-related glycaemic excursions, analyses of glycaemic variability derived from CGM data collected during weeks 48–52 of COMBINE 1 and 3 will be evaluated separately. In conclusion, among people with T2D, based on post hoc analyses of CGM data from COMBINE 1 and 3, once-weekly IcoSema showed statistically significantly better PPG control than once-weekly basal insulin and similar PPG control to that obtained with daily basal-bolus insulin therapy via treatment with only one weekly injection. Contributors A combination of academic authors and authors who are employees of Novo Nordisk (the trial sponsor) participated in developing the trial concepts and design, and in collecting the data. CDB, MB, AF, RM, and FG accessed and verified the underlying data. All authors participated in the critical review and drafting of the manuscript, read and approved the final version of the manuscript, and were responsible for the decision to submit for publication. Data sharing statement An authorised researcher can request access to clinical trial data by submitting a research proposal for review and approval by Novo Nordisk and an internal independent review panel. Requests are usually considered after the research is finished and the main results have been published. If the research supports a regulatory application, requests will be considered after the product and its intended use are approved in both the EU and the USA. Participants, clinical data will be anonymised, following an approved internal process, before data are shared to external third parties. For details on how to request access to clinical data, visit novonordisk-trials.com. Declaration of interests CDB reports consulting and speaking fees from Abbott, AstraZeneca, Boehringer Ingelheim, Eli Lilly, Insulet, Medtronic, and Novo Nordisk. MB, AF, and RM are employees of Novo Nordisk. FG has received grants from Eli Lilly and Roche Diabetes Care; has received consulting fees from Eli Lilly and Novo Nordisk; received support for attending meetings or for travel from AstraZeneca, Boehringer Ingelheim, Eli Lilly, LifeScan, Medtronic, Merck Sharp & Dohme, Novo Nordisk, Roche Diabetes Care, and Sanofi Aventis; has provided advisory board services for AstraZeneca, Boehringer Ingelheim, Eli Lilly, LifeScan, MedImmune, Medtronic, Merck Sharp & Dohme, Novo Nordisk, Roche Diabetes Care, and Sanofi Aventis; and declares unpaid leadership positions for the European Association for the Study of Diabetes and European Foundation for the Study of Diabetes. Acknowledgements This study was funded by Novo Nordisk A/S. COMBINE 1 and COMBINE 3 were sponsored by Novo Nordisk and are registered with ClinicalTrials.gov (NCT05352815, registered 24 April 2022; NCT05013229, registered 13 August 2021). Medical writing support was provided by Chloe Fletcher MSc of Oxford PharmaGenesis, Oxford, UK, with funding from Novo Nordisk. Footnotes Supplementary data related to this article can be found at https://doi.org/10.1016/j.eclinm.2026.103985. Supplementary data Supplementary Methods and Tables S1–S3 References Associated Data Supplementary Materials Supplementary Methods and Tables S1–S3