Real-world use and side effects of semaglutide compared to other treatments

This was a retrospective, population-based, multicentre service evaluation. Data were collected from 42 general practices across the studied borough, West Midlands/England, for patients prescribed semaglutide between 2019 and 2024. Ethical approval was granted by the University of Birmingham Ethics Committee (Ref: MP2425 047, December 2024) and governance approval was obtained from the GP surgeries integrated healthcare ethics board (August 2024). Data were stored and analysed using Microsoft Excel; analysis included descriptive statistics and frequencies.

Anonymised patient-level data were extracted and compiled into a structured Excel database using a data share agreement across the respective practices. Variables included age, gender, ethnicity, semaglutide dose and prescription type, dates of first and last issue, HbA1c and weight measurements, GP practice, and reported adverse effects. Data cleaning and curation ensured completeness for inclusion in analyses.

For both HbA1c measurements and weight measurements, the starting point used for each patient was the latest measurement taken within the 6-month period before the first semaglutide issue. The end point used was the measurement taken within 6 months (before or after) of the most recent issue. Patient data were excluded if there was less than 1 month between their first and most recent issue; their HbA1c or weight were not measured within the 6-month period before their first issue; or their most recent measurement was more than 6 months before or after their most recent issue, as this was not deemed an up-to-date measurement. The mean and standard deviation were calculated for the start and end HbA1c and weight measurements for the whole patient population.

National semaglutide prescribing data were collected from the OpenPrescribing [23] database from January 2020 to September 2024, with data from 2019 being omitted due to only being available from the month of September, therefore data for the year 2019 was incomplete. National prescribing trends were compared with the population in the area studied to identify similarities and deviations.

ADR data were sourced from the Medicines and Healthcare products Regulatory Agency (MHRA) Yellow Card Scheme's interactive drug analysis profile (iDAP) [24] for the years 2020–2024. Data were analysed by system organ class. To calculate incidence, national prescription numbers (from OpenPrescribing) and local prescribing figures were used to derive ADR reporting rates.

HbA1c and weight were evaluated for treatment response. For each patient, the baseline was defined as the most recent value within six months prior to semaglutide initiation. The follow-up value was the closest measurement within six months of the final semaglutide prescription. Patients were excluded if data were missing, or if the interval between start and end measurements was <1 month or >6 months. The mean and standard deviation were calculated using Microsoft Excel.

GP practices were assigned a deprivation ranking using the Socioeconomic Index for Small Areas (SEISA) decile from the Higher Education Statistics Agency [25]. Practice postcodes were mapped to SEISA deciles (1 = most deprived and 10 = least deprived). Prescription volumes were analysed across deciles to assess socioeconomic patterns.

Between 2019 and 2024, semaglutide prescribing in the area studied (GP practices) increased initially, peaking in 2022 with 446 prescriptions, then declined sharply to 63 prescriptions by 2024. National prescribing followed a similar trend but showed a more gradual decline after March 2023 (Figure 1).

Among the 1403 patients prescribed semaglutide in the area studied, 82.5% identified as white, with White British patients accounting for 41%. The low number of White British patients prescribed semaglutide vs. 82% of the region population warrants further investigation. Other ethnicities, including South Asian patients, comprised 7.2% of those prescribed semaglutide and aligned with the local demographic (7.2%). Minoritised ethnic groups represented a small proportion of prescriptions, with some categories (e.g., South and Central American) accounting for a single patient. A total of 52% of patients were female and the age range was 20–91 years of age (Table 1).

The mean baseline HbA1c was 74.7 mmol/mol, falling to 63.9 mmol/mol, a reduction of 10.8 mmol/mol (~1.0% absolute HbA1c drop) (Figure 2).

Mean body weight decreased from 91.4 kg to 87.0 kg, representing a 4.8% reduction (Figure 3).

From 2020 to 2024, 23 patients (1.85%) in the area studied reported semaglutide-related ADRs, compared to a national reporting rate of 0.21%. Nationally, gastrointestinal symptoms were the most common (54%), followed by general disorders (13.3%) and nervous system disorders (10.7%), (Table 2 and Table 3).

Fifty-eight percent of patients transitioned from an acute to a repeat prescription, indicating ongoing use. However, 61% of patients had discontinued semaglutide by 2024. Among these, 24% received only acute prescriptions, with some receiving a single dose (Figure 4).

Prescribing was highest in practices located in the most deprived areas (SEISA deciles 1–2), with the single highest-prescribing practice situated in a decile 1 area. Prescribing levels declined as deprivation decreased (Figure 5).

Semaglutide prescribing rose rapidly following its national approval, peaking in 2022, before declining in 2023–2024. This temporal pattern mirrors national trends but was more pronounced locally. The timing of this decline aligns with national GLP-1RA shortages and the 2023 NHS Patient Safety Alert [26]. This suggests that access to semaglutide is especially vulnerable to supply chain pressures in areas with high treatment demand and constrained healthcare resources. The findings also reflect broader challenges in balancing demand driven by clinical indications versus off-label and cosmetic use [9,11].

Data showed that 83% of semaglutide recipients in the area studied identified as white (incl. British and other) and 1.3% identified as mixed ethnicity. The data shows that both white British (82.4% of region but only 41.3% of Rx) and mixed ethnicity (2.8% of region but only 1.3% of Rx) experience lower prescribing compared to South Asian and Black patients. This may be due to the elevated prevalence of type 2 diabetes and cardiometabolic risk in Black and South Asian populations [27]. Other white and European patients received 11.8% of all prescriptions but represent only 2.4% of the ethnicity of the region. While this could reflect differences in prescribing or potential differences in tolerability or efficacy, it also reflects clinical decision making. Previous analyses have highlighted how trial underrepresentation and limited subgroup data contribute to uncertainty about treatment effectiveness in certain populations [27,28].

The observed reduction in HbA1c (~11 mmol/mol) and mean weight loss (4.8%) are consistent with findings from the SUSTAIN and STEP trial programmes [7,8]. These results affirm the translational validity of semaglutide's efficacy in routine UK primary care, even within a population facing elevated comorbidity and deprivation. However, the lack of dose-specific or duration-adjusted data in our cohort limits firm attribution. Moreover, no clear association was found between socioeconomic status and treatment response, suggesting that semaglutide is likely effective across deprivation gradients when access is ensured.

The reported ADR incidence in the area studied (1.85%) was ninefold higher than the national rate (0.20%), yet both figures fall far below rates seen in clinical trials, where gastrointestinal symptoms affect up to 74% of patients [8]. This disparity highlights widespread underreporting, a known issue within passive pharmacovigilance systems [26]. Underreporting may result from low perceived severity, patient reluctance, lack of awareness, or limited guidance from healthcare professionals [26]. It is likely exacerbated in contexts where semaglutide is prescribed online or without adequate counselling, bypassing opportunities for ADR education [10,11,29]. Future pharmacovigilance efforts should explore the integration of automated prompts in electronic health records and proactive surveillance in community pharmacies and primary care.

Only 39% of the patients studied continued semaglutide treatment through the study period, with 61% discontinuing treatment. This may be partially explained by tolerability issues or clinical inertia following early discontinuation. However, this finding also coincides with the 2023 UK approval of tirzepatide, a dual GLP-1/GIP agonist, which demonstrated superior glycaemic and weight outcomes compared to semaglutide in the SURPASS programme [30]. The extent to which semaglutide discontinuation reflects planned treatment switching, treatment fatigue, or issues with access remains unclear. Future studies should investigate prescribing trajectories, switching behaviour, and patient-reported reasons for discontinuation.

There was a clear association between deprivation and prescribing rates, with the highest levels of semaglutide use in practices located in SEISA deciles 1 and 2. This reflects known associations between socioeconomic deprivation, higher diabetes prevalence, and earlier onset of complications [31,32]. It also raises important questions about resource allocation, treatment prioritisation, and the capacity of deprived areas to maintain continuity of care during supply disruptions. Notably, our findings also challenge assumptions that newer therapies are disproportionately accessed by affluent groups, demonstrating that effective primary care networks in deprived areas can deliver equitable access when supported.

These findings have several implications. First, clinicians should be supported to routinely counsel patients on common and serious adverse effects, including how and when to report these using systems like the Yellow Card Scheme. Structured follow-up, particularly after initiating GLP-1RAs, could improve early identification of ADRs and reduce unnecessary discontinuation. Second, efforts to curb inappropriate prescribing—particularly the off-label weight-loss use of Ozempic—should be accompanied by public health messaging to reduce demand driven by non-clinical motivations and unregulated online supply. Third, prescribing should be monitored at the population level to identify disparities by ethnicity or deprivation, using routinely collected NHS data.

This multicentre study enlisted a large sample size of 1403 patients from 42 GP practices, conducted within an area of known deprivation in the UK where there are evident health inequalities [22] and where, to our knowledge, research of this nature has not previously been conducted. However, key limitations include patient data requiring exclusion due to incomplete data available, and no access to the nature of ADRs experienced by the studied patients, co-morbidities, or concomitant medicines prescribed. Therefore, it cannot be known with certainty if HbA1c and weight reductions were solely due to semaglutide treatment. Data extracted from the Yellow Card and OpenPrescribing databases were for semaglutide in general, including uses in other indications, e.g., weight management where typically higher doses are prescribed than for T2DM; therefore, the ADRs may not reflect the doses of semaglutide used. As with all pharmacovigilance studies interpreting data from national reporting systems, this study is affected by reporting bias, whereby data included from the MHRA Yellow Card system may have included self- reported false positive ADR reports and underreporting; therefore, data may not be reflective of real-life incidences of adverse reactions experienced with semaglutide.

Future research should build on these findings by exploring ADR reporting behaviours among patients and clinicians, potentially through qualitative methods or surveys. Further analysis is needed to understand switching patterns to newer agents like tirzepatide and to examine long-term glycaemic durability. A broader study incorporating other localities in England would enable regional comparisons and generate more generalisable data. Expanding this analysis to include other GLP-1RAs would also help to identify class-wide or drug-specific trends in tolerability and adherence.