THOMAS SJ CRABTREE,1,2,3 KAREN ADAMSON,4 ALEX BICKERTON,5 ALISON EVANS,6 SUZANNE PHILLIPS,6 ALISON GALLAGHER,7 NIELS LARSEN,2 DENNIS BARNES,8 KETAN DHATARIYA,9,10 BENJAMIN CT FIELD,11,12 ISKANDER IDRIS,2,3 ROBERT EJ RYDER,1 on behalf of all ABCD semaglutide audit contributors
1 City Hospital, Sandwell and West Birmingham Hospitals NHS Trust, Birmingham, UK
2 Royal Derby Hospital, University Hospitals of Derby and Burton NHS Trust, UK
3 Faculty of Medicine and Health Sciences, University of Nottingham, Nottingham, UK
4 St John’s Hospital, Livingston, UK
5 Yeovil District Hospital, Yeovil District Hospital NHS Foundation Trust, Yeovil, UK
6 Gloucestershire Royal Hospital, Gloucestershire Hospitals NHS Foundation Trust, Gloucester, UK
7 Leicester General Hospital, University Hospitals of Leicester NHS Trust, Leicester, UK
8 Tunbridge Wells Hospital, Maidstone and Tunbridge Wells NHS Trust, UK
9 The Norfolk and Norwich Hospital, Norfolk and Norwich University Hospitals NHS Trust, Norwich, UK
10 Norwich Medical School, University of East Anglia, Norwich, UK
11 Department of Clinical and Experimental Medicine, University of Surrey, Guildford, UK
12 Department of Diabetes and Endocrinology, Surrey and Sussex Healthcare NHS Trust, Redhill, UK
Address for correspondence: Dr Tom Crabtree
Royal Derby Hospital, Uttoxeter Road, Derby, DE22 3NE, UK.
E-mail: T.Crabtree@nhs.net
Background: Previous randomised controlled trials have observed individual differences in response to Glucagon-Like Peptide-1 Receptor Agonists (GLP1RA) according to baseline characteristics such as glycated haemoglobin (HbA1c) and weight. The Association of British Clinical Diabetologists (ABCD) launched a nationwide UK audit in January 2019 to assess the clinical utility, efficacy and safety of injectable semaglutide in routine practice. The aim of this analysis was to investigate associations between baseline characteristics and HbA1c and weight reductions with semaglutide in real-world use.
Methods: Data were extracted from the secure online tool and individuals who had baseline and follow-up data available within a defined 6 (3-9) month window were included. Variables were assessed as both continuous variables and categorical variables in a multivariate regression model. Missing data were multiply imputed.
Results: In total, 620 individuals were included. Baseline characteristics: (mean±SD) age was 58.7±10.7 years, HbA1c 81.6±18.5 mmol/mol (9.5±1.7%), weight 108.2±24.2 kg and body mass index (BMI) 37.6±7.6 kg/m2. Median diabetes duration was 11.2 years (IQR 6.6-16) and 50.5% (313/620) of subjects were male. The median follow-up time was 0.5 years. HbA1c reduced by 14.9 mmol/mol (95% CI 13.5, 16.1) [-1.4% (95% CI - 1.2, -1.5)]; p<0.001; and weight reduced by 4.2kg (95% CI 3.6, 4.8; p<0.001). Higher HbA1c, younger age and GLP1RA naïvety were associated with larger HbA1c reduction. Higher baseline weight/BMI and GP1RA naïvety were associated with larger weight reduction.
Conclusion: In this real-world study, baseline HbA1c and weightwere important predictors of HbA1c and weight reduction outcomes following initiation of semaglutide in routine clinical practice. Our data mirror existing randomised controlled trial data, but further evidence is being collected over a longer follow-up period.
Br J Diabetes 2023;23:94-100
https://doi.org/10.15277/bjd.2023.418
Key words: Semaglutide, audit, real-world
Type 2 diabetes (T2DM) is estimated to affect 3.4 million people in England and 537 million worldwide.1,2 It is increasingly recognised as a heterogenous condition.3-5 Individuals with different clinical phenotypes may have different responses to the range of therapies currently available and there is increasing interest in utilising precision medicine approaches to stratify individuals towards the best treatment option for them.6,7 Whilst work is ongoing to identify potential biomarkers and genetic factors, their cost may prohibit their wide use in public-funded healthcare systems.8 Routinely measured clinical data may be useful to stratify individuals, identifying those who are most likely to benefit from a given treatment, though evidence in this area is remains limited.5
Glucagon-like peptide 1 receptor agonists (GLP1RAs) are increasingly used in the management of T2DM, with greater emphasis on their use in recent national and international guidelines.9-11 Semaglutide is a once-weekly injectable GLP1RA licensed for the treatment of T2DM; it is also available as a once- daily oral preparation.12
The SUSTAIN series of randomised controlled trials consistently demonstrated efficacy of semaglutide, with mean weight loss of 4.11 kg and HbA1c reductions of 15 mmol/mol at the maximally titrated 1mg dose when results were pooled by meta-analysis.13 Reductions in major adverse cardiovascular events have also been reported across the trials.14,15 Although SUSTAIN-6 reported increased rates of retinopathy, these findings were not observed across the trials by meta-analysis.16,17 However, the trials included were of relatively short duration: work is ongoing to understand retinopathy risk with more prolonged follow-up.18
Several trials reported no difference in HbA1c and weight response to semaglutide across individuals with different baseline clinical characteristics,19 but two reported differential responses across baseline characteristics. First, SUSTAIN-8 demonstrated larger HbA1c reductions in individuals with higher baseline BMI and HbA1c levels.20 Second, real-world evidence from an Italian clinic found HbA1c reductions were greatest at six months in individuals with higher baseline HbA1c, greater weight loss and longer diabetes duration.21 This latter study also found that higher baseline BMI predicted greater weight loss and that GLP1RA-naïve individuals reached similar end-points to those switching from alternative GLP1Ras. Previous Association of British Clinical Diabetologists (ABCD) audit programme has also reported the additional benefits of switching to semaglutide from alternative GLP1RA preparations.22 Only minor differences have been observed in terms of race or ethnicity across the SUSTAIN trails.23
The nationwide ABCD injectable semaglutide audit was launched in January 2019 with the aim to collect routine clinical data on individuals using semaglutide in the UK and to assess outcomes in real-world clinical practice. The aim of this analysis was to assess the clinical characteristics which may predict HbA1c and weight responses associated with semaglutide in the real world, using data from the ABCD audit.
Participating centres were invited to collect anonymised, routinely collected, clinical data on individuals commenced on injectable semaglutide (0.25-1 mg, Ozempic) to a secure online tool. De-identified data were then extracted for analysis in this cohort study. All individuals with baseline data and at least one follow-up entry for HbA1c and/or weight were included. Individuals with large amounts of missing data in key covariates of imputation (e.g. age, gender) were excluded (n=14). The follow-up period was defined as 6 months, with a window of 3- 9 months. Key co-variates were identified as follows: baseline HbA1c, baseline weight, age, gender, duration of diabetes and previous GLP1RA use. Ethnicity could not be assessed due to missing data. For all other variables, missing data were multiply imputed with 10 imputations, and Rubin’s rules were applied for the analysis.24
HbA1c and both absolute and relative weight change from baseline were assessed using a multivariate linear regression analysis incorporating the above characteristics into the model as continuous covariates. The overall model co-efficient and significance levels were also calculated and reported. Subgroup analyses were then performed by classifying these characteristics into groups as follow:
Statistical significance was defined as p<0.05. When including multiple groups pairwise comparisons were adjusted for multiple comparisons using a Bonferroni correction. All analyses were performed in Stata 16.
The ABCD audit programme is approved by the Confidentiality Advisory Group and a Caldicott Guardian.25
In total 620 individuals were included in this analysis. The flow diagram for inclusion in this analysis is available in Appendix 1. The baseline characteristics are summarised in table 1.
Over a median follow-up time of 0.5 years (0.4-0.7), across the entire population HbA1c reduced by 14.9 mmol/mol (95% CI 13.5, 16.1; p<0.001) [-1.4% (95%CI -1.2, -1.5)]. Absolute weight reductions of 4.2kg (95% CI 3.6-4.8; p<0.001) and relative weight reductions of 3.9% (95% CI 3.4-4.5%; p<0.001) were observed.
The models and coefficients calculated for baseline characteristics when used in continuous data format are displayed in table 2. The coefficients represent the difference in HbA1c reduction per-one unit change in the covariate assessed (for continuous values) or compared to reference (for categorical i.e. GLP1RA-naïve vs previous GLP1RA use). Baseline HbA1c, baseline weight and previous GLP1RA use were significant predictors of HbA1c response to injectable semaglutide. Those with high baseline HbA1c levels had greater HbA1c reductions, and those switching from an alternative GLP1RA to semaglutide had attenuated HbA1c reductions.
Only baseline weight significantly predicted weight loss response to semaglutide, although previous GLP1RA use was trending towards statistical significance. Those with higher weight at baseline lost more weight in association with injectable semaglutide use. For relative weight change, no factors predicted weight loss in response to semaglutide.
The change in HbA1c from baseline in the subgroup analysis is demonstrated in figure 1. Those with HbA1c ≥80 mmol/mol (9.5%) had larger HbA1c reductions than those in the lower HbA1c groups, and those switched from alternative GLP1RA drugs had a smaller HbA1c reduction than GLP1RA- naïve individuals. This persisted for relative HbA1c changes, although we have not reported these as we considered they would not be clinically meaningful. In addition, individuals aged <60 years had greater HbA1c reductions than those aged ≥60 years.
The change in absolute weight from baseline in the subgroup analysis is demonstrated in figure 2. Those with BMI >40 kg/m2 at baseline had larger weight reductions associated with semaglutide than those with BMI 30-40 kg/m2 at baseline but not those with BMI <30 kg/m2. Those who switched to semaglutide from an alternative GLP1RA were experienced less weight loss than GLP1RA-naïve individuals. The relative change in weight from baseline is demonstrated in figure 3. No statistically significant differences were noted between subgroups.
Our results from the ABCD audit programme demonstrate that HbA1c and both absolute and relative weight reductions with injectable semaglutide translate evidence from randomised controlled trials into real-world clinical practice.
The HbA1c and weight reductions observed were comparable to those in the SUSTAIN trials.13,20 Additionally, our results mirrored the findings from SUSTAIN-8 in that those with higher HbA1c and weight baseline achieved larger reductions by follow-up.20 There was no significant difference noted between individuals with BMI <30 kg/m2 and those with BMI >40 kg/m2 .This is likely due to relatively small numbers in the lowest BMI group (n=74).
We did not observe any difference in response across different diabetes durations, in contrast to reports from SUSTAIN-8 and the real-world Italian study highlighted in the introduction.20,21 We report similar observations to SUSTAIN-8, with a statistically significant impact of baseline weight on HbA1c change, with heavier individuals achieving statistically larger reductions.20 It should be noted, however, that the coefficient for this is small, and therefore unlikely to be clinically meaningful. The difference across weight did not persist within the subgroup analysis.
GLP1RA-naïve individuals had larger HbA1c and weight reductions than those switching from other GLP1RA medications, although reductions were observed in both groups. This is a finding we have reported previously and that was reported in SUSTAIN-8.20,22
Strengths and limitations
The pragmatic real-world study design allows us to observe the use of injectable semaglutide in clinical practice and compare findings in this population to existing RCT data. Our findings confirm that trial results can be generalised to UK clinical practice, despite clinical practice including individuals who are less closely monitored or who perhaps have more extreme baseline characteristics and would therefore have been excluded from trials. In our analysis, baseline HbA1c (81.6 vs 67.2 mmol/mol in SUSTAIN-8) and weight (108 kg vs 90.6 kg in SUSTAIN-8) were higher, and diabetes duration (11.2 years vs 7.5 years in SUSTAIN-8) was longer than in clinical trials. If drug shortages persist these data may help in guiding who may benefit the most from semaglutide.26
Real-world data have their limitations. Although the use of multiple imputations may mitigate problems caused by missing data, it does not resolve this completely. Nonetheless, we excluded individuals with high levels of missing data in key covariates. Notably, dose data were not available. Although it is assumed that individuals were maximally titrated, assessment or adjustment for doses was not possible. Additionally, unobserved confounders may impact the results but we are reassured that our findings complement and mirror the existing extensive data on semaglutide well. Ethnicity and deprivation data were not available and further work should focus on uptake and outcomes in these key groups to assess real-world equitability.
In this real-world analysis from the ABCD audit programme, injectable semaglutide was associated with greater HbA1c reduction in those who were younger, vs. older, and in those who had higher, vs. lower, baseline HbA1c. Weight reductions were largest in those who weighed most at baseline. Whilst reductions in weight and HbA1c were observed in individuals switching to semaglutide from other members of the GLP1RA class, these were smaller than in GLP1RA-naïve individuals. Data collection for the ABCD audit is ongoing, and further follow-up will allow us to assess the impact of semaglutide on key outcomes in the long term.
Conflict of interest TSJC has received speaker fees and/or support to attend conferences from NovoNordisk, Lilly, Sanofi, Abbott Diabetes Care, Insulet and Dexcom.
Funding The ABCD audit programme is an independent audit programme supported by an unrestricted grant from NovoNordisk.
Ethics statement This study utilises data from a clinic audit and was assessed against health regulatory authority criteria. No ethical approvals were required for this analysis.