Ethnic and socioeconomic disparities in initiation of second‐line antidiabetic treatment for people with type 2 diabetes in England: A cross‐sectional study

Abstract Aims To assess any disparities in the initiation of second‐line antidiabetic treatments prescribed among people with type 2 diabetes mellitus (T2DM) in England according to ethnicity and social deprivation level. Materials and methods This cross‐sectional study used linked primary (Clinical Practice Research Datalink) and secondary care data (Hospital Episode Statistics), and the Index of Multiple Deprivation (IMD). We included people aged 18 years or older with T2DM who intensified to second‐line oral antidiabetic medication between 2014 and 2020 to investigate disparities in second‐line antidiabetic treatment prescribing (one of sulphonylureas [SUs], dipeptidyl peptidase‐4 [DPP‐4] inhibitors, or sodium‐glucose cotransporter‐2 [SGLT2] inhibitors, in combination with metformin) by ethnicity (White, South Asian, Black, mixed/other) and deprivation level (IMD quintiles). We report prescriptions of the alternative treatments by ethnicity and deprivation level according to predicted percentages derived from multivariable, multinomial logistic regression. Results Among 36 023 people, 85% were White, 10% South Asian, 4% Black and 1% mixed/other. After adjustment, the predicted percentages for SGLT2 inhibitor prescribing by ethnicity were 21% (95% confidence interval [CI] 19–23%), 20% (95% CI 18–22%), 19% (95% CI 16–22%) and 17% (95% CI 14–21%) among people with White, South Asian, Black, and mixed/other ethnicity, respectively. After adjustment, the predicted percentages for SGLT2 inhibitor prescribing by deprivation were 22% (95% CI 20–25%) and 19% (95% CI 17–21%) for the least deprived and the most deprived quintile, respectively. When stratifying by prevalent cardiovascular disease (CVD) status, we found lower predicted percentages of people with prevalent CVD prescribed SGLT2 inhibitors compared with people without prevalent CVD across all ethnicity groups and all levels of social deprivation. Conclusions Among people with T2DM, there were no substantial differences by ethnicity or deprivation level in the percentage prescribed either SGLT2 inhibitors, DPP‐4 inhibitors or SUs as second‐line antidiabetic treatment.

prescribed SGLT2 inhibitors compared with people without prevalent CVD across all ethnicity groups and all levels of social deprivation.
Conclusions: Among people with T2DM, there were no substantial differences by ethnicity or deprivation level in the percentage prescribed either SGLT2 inhibitors, DPP-4 inhibitors or SUs as second-line antidiabetic treatment.

| INTRODUCTION
Most healthcare systems report inequities in disease incidence, healthcare delivery and outcomes according to people's socioeconomic status and ethnicity. 1,2 For countries with single-payer systems such as England, national recommendations from agencies like the National Institute for Health and Care Excellence (NICE) encourage access to effective and cost-effective interventions to maximize clinical benefit while also reducing health inequalities. [3][4][5] Nonetheless in countries such as England, inequities in using healthcare interventions according to people's socioeconomic characteristics persist for diseases such as cardiovascular disease (CVD), 6 chronic kidney disease (CKD), 7 and type 2 diabetes mellitus (T2DM). 8 There are inequities in T2DM prevalence and outcomes according to ethnicity 8 and deprivation. 9 People of Black and South Asian ethnicity, and people with lower income or lower educational attainment have a higher prevalence of T2DM, worse blood glucose control and earlier onset of macro-and microvascular complications compared with people of ethnicities other than Black and South Asian, higher incomes or higher educational attainment. [8][9][10][11][12][13] Ethnic minorities also tend to experience delays in T2DM treatment intensification when clinically indicated (therapeutic inertia), 12 which may contribute to worse outcomes compared with White people. [14][15][16][17][18] Other ethnic and socioeconomic inequities in T2DM treatment that could impact clinical outcomes, such as the type of second-line antidiabetic treatment prescribed at treatment intensification from metformin monotherapy, are less well understood. Hence, we chose to examine the potential disparities in second-line antidiabetic treatment prescribing by ethnicity and deprivation status.
For people with T2DM whose glycated haemoglobin (HbA1c) levels are poorly controlled, an important choice is which second-line oral antidiabetic therapy to prescribe in addition to metformin. 19 Between 2015 and 2021, NICE technology appraisals and clinical guidelines recommended that, for most people with T2DM, several second-line oral treatment options should be available, including sodium-glucose cotransporter 2 (SGLT2) inhibitors, dipeptidyl peptidase-4 (DPP-4) inhibitors or the lower-cost option of sulphonylureas (SUs). 19,20 Updated NICE guidelines (2022) recommend SGLT2 inhibitors for individuals at high risk of or with prevalent CVD but that, for other eligible patients, any of these three treatments may be suitable. 21 The decision to allow local discretion in the choice of second-line treatment may reflect the uncertainty over comparative effectiveness and cost-effectiveness of these three treatment choices, which is partly related to the lack of randomized controlled trials (RCTs) providing head-to-head comparisons of these three oral antidiabetic drugs. In contrast, international diabetes guidance and consensus reports recommend SGLT2 inhibitors for people with established atherosclerotic CVD, heart failure and CKD, 22  The CPRD is a large, population-based dataset covering approximately 20% of the UK population and is representative in terms of age, sex and ethnicity. 26,27 These data include clinical diagnoses, laboratory test results, and prescribing information recorded in primary care. Linkage of CPRD data to HES data is available for approximately 80% of people in the CPRD registered at general practices in England.
HES data include diagnoses and demographic information related to all NHS-funded hospitalizations. 28 The IMD is commonly used in epidemiological research as a proxy for socioeconomic status in England.
It ranks individuals according to deprivation status based on their postcode, and is usually reported in quintiles (1 being the least deprived, 5 being most deprived). 29 ONS mortality data include information on all deaths registered in England and Wales. 30

| Patient and public involvement
One patient and public (PP) representative (P.C.) was involved in this study's design, provided feedback on this manuscript, and is a coauthor. The PERMIT study protocol describes PP contributions to the study design. 25 PP representatives will assist with drafting lay summaries, which we will share on the study website (https://www.lshtm.ac. uk/research/centres-projects-groups/permit) and at study workshops with a wider group of multi-ethnic PP representatives. We will work with the Centre for Ethnic Health Research, led by co-author K.K., to make culturally adapted lay summaries.

| Study population
We included people aged 18 years or older with a T2DM diagnosis, in whom incident second-line oral antidiabetic treatment was prescribed for the first time between January 1, 2014 and March 31, 2020 after first-line antidiabetic treatment with metformin monotherapy. We used the complete historical general practice (GP) electronic health record to ensure this was the first time each person had a record of being prescribed an SU, a DPP-4 inhibitor or an SGLT2 inhibitor. The second-line therapy, an SU, a DPP-4 inhibitor or an SGLT2 inhibitor, had to have been added on to metformin, and had not replaced it. These three treatments constituted approximately 99% of the second-line treatments prescribed, therefore, other second-line antidiabetic treatments were excluded from this study. 23,31 Eligible people had to have had a prescription for metformin monotherapy within 60 days prior to the first prescription for second-line treatment to ensure they were continuous users of metformin monotherapy prior to intensification. Also, to ensure the second-line treatments were an addition to, rather than a switch from, metformin, the individuals were required to have been prescribed metformin on the same day or within 60 days after the first prescription for the second-line antidiabetic treatment.
We excluded women with a record of pregnancy within 12 months prior to second-line treatment initiation since antidiabetic prescribing guidelines are different for this population. 19 We also excluded people whose last recorded estimated glomerular filtration rate (eGFR) was less than 30 ml/min/1.73 m 2 since metformin is contraindicated in this group, and SGLT2 inhibitors are not recommended for this group in the United Kingdom for the purpose of lowering blood glucose. 19,32

| Definitions of ethnicity and deprivation
We defined ethnicity according to clinical and demographic codes recorded within the CPRD or linked HES data prior to or on the same day as the first-ever prescription date for one of the three second-line antidiabetic treatments of interest, that is, the index date. Ethnicity was grouped into 16 categories in primary care and 11 categories in secondary care, which we further re-grouped as the following: (1) White, (2) South Asian, (3) Black, and (4) Mixed/other (Table S1). We considered this re-grouping necessary to ensure sufficient sample sizes within each ethnic group, and to follow precedent studies using the same data sources, 12,33,34 as well as the ethnic groupings used in the 2011 England and Wales census. 35 If the two sources for ethnicity provided different categorizations, then we used ethnicity as defined in the CPRD since these data have been shown to be more reliable than HES inpatient data. 33 If no ethnicity data were available within the CRPD, we categorized ethnicity using HES data. If ethnicity was not recorded in either source, we considered ethnicity as missing and the individual was excluded from the complete case analyses.
We used the small area IMD to define deprivation. The IMD combines seven indices which capture dimensions of deprivation at the Lower-Layer Super Output Area or neighbourhood level, and ranks each neighbourhood from 1 to 32 844. 29 Neighbourhood rankings were divided into quintiles and used to compare relative levels of deprivation among people in this study living in different neighbourhoods in England.
We also considered how the proportion of patients receiving the alternative second-line treatments may differ according to calendar time, recognizing that the dissemination and awareness of the safety and efficacy of SGLT2 inhibitors for patients with T2DM increased over the time period, with the publication of important RCT results. [36][37][38] We considered this hypothesis in grouping calendar time into years 2014, 2015 to 2016, 2017 to 2018, and 2019 to 2020.

| Covariates
We adjusted for several additional variables, derived from data captured before or on the same day as the index date. These were sex, age, duration of time on metformin monotherapy, number of patients registered at the individual's general practice, geographic region, co-prescriptions for renin-angiotensin system inhibitors and/or statins, history of proteinuria, history of hypoglycaemia, clinical measures (body mass index [BMI] and HbA1c), smoking status, alcohol intake, and comorbidities at the time of second-line antidiabetic treatment initiation. Comorbidities included CKD stage (no known CKD, stages 1, 2, 3a, and 3b, assigned using the latest recorded eGFR), cancer (any), blindness, congestive heart failure, previous myocardial infarction (MI), unstable angina, previous stroke, other ischaemic heart disease, and uncontrolled hypertension based on the most recent blood pressure measures recorded in primary care. We defined prevalent CVD as a composite of heart failure, ischaemic heart disease, unstable angina, previous myocardial infarction, or previous stroke.

| Treatment prescribed
Our dependent variable of interest was incident second-line oral antidiabetic treatment prescribed (SUs, DPP-4 inhibitors, or SGLT2 inhibitors, in addition to metformin), defined using CPRD prescribing data.

| Analysis
We described baseline characteristics of the study population stratified by ethnicity and IMD. We then built mixed-effect multivariable, multinomial logistic regression models which compared the odds of initiating SGLT2 inhibitors and DPP-4 inhibitors versus SUs (reference outcome), as well as, in a separate model, SGLT2 inhibitors versus DPP-4 inhibitors (reference outcome), first adjusting for just age and sex. In the final adjusted model, we adjusted for all covariates, as well as mutual adjustment for ethnicity and deprivation (fixed effects) and clustering at the Clinical Commissioning Group (CCG) level (random effect). Because odds ratios can be misleading, particularly when the outcome is common, 39 we calculated and plotted predicted percentages from the adjusted model using recycled predictions. 40 These percentages refer to people prescribed each second-line antidiabetic treatment stratified by ethnicity, and separately by deprivation, while still adjusting for all measured covariates, and accounting for clustering at the CCG level. We obtained P values from Wald tests comparing the predicted percentage of being prescribed one of SUs, DPP-4 inhibitors or SGLT2 inhibitors by each non-White ethnic group versus White ethnic group and by deprivation Quintiles 2 to 5 versus deprivation Quintile 1. We also performed joint tests to test whether predicted percentages for each ethnic group or for each deprivation quintile were equal for each second-line antidiabetic treatment. These percentages and P values were used to support our final conclusions on disparities in second-line antidiabetic treatment prescribing by ethnicity or by deprivation. 39 We then stratified the adjusted predicted percentages by prevalent CVD status at baseline to determine if there were differences in prescribing by ethnicity or by deprivation quintile according to prevalent CVD status.
In the secondary analyses, we compared the change in odds ratios between the fixed-effect model (model including ethnicity, deprivation, and all covariates) and the mixed-effect model (the Data management and analyses were performed using Stata 17.

| Baseline characteristics
The study population included 36 023 people with complete data on all variables of interest who initiated second-line oral antidiabetic treatment during the study period with linked secondary care data ( Figure 1). Eighty-four percent of the cohort were White, 10% were South Asian, 4% were Black, and 1% were Mixed/other ethnicity. We excluded 6150 people with missing data for at least one variable, including 348 with missing ethnicity data and 20 with missing IMD data (Table S4).
Overall, 41% of the cohort was female, with a mean age of 59 years ( After stratifying by IMD quintile, we found that people in the most deprived quintile were over-represented (25%) and people in the least deprived quintile were under-represented (16%; There was no evidence of differences in adjusted predicted percentages for being prescribed DPP-4 inhibitors or SUs according to ethnicity ( Figure 2, Table S5). The results from the multinomial, multivariable logistic regression model used to calculate these adjusted predicted percentages are described in Table S6.

| Social deprivation and second-line antidiabetic treatment choice
The crude proportion of people prescribed each second-line antidiabetic treatment option across deprivation quintiles are presented in Table 2.
There was some evidence of a small difference in adjusted predicted percentages of people prescribed SGLT2 inhibitors according to deprivation: 19% (95% CI 17-21%) were prescribed SGLT2 inhibitors in the most deprived quintile, and 22% (95% CI 20-24%) were prescribed SGLT2 inhibitors in the least deprived quintile (P < 0.001; Figure 2, Table S5). Conversely, there was some evidence that people in the most deprived quintile had a small increase in the adjusted predicted percent of being prescribed DPP-4 inhibitors 44% (95% CI 42-47%) versus 42% (95% CI 39-45%) of people in the least deprived quintile (P = 0.04). There was no evidence of any differences in the adjusted predicted percentages of people prescribed SUs according to deprivation (P = 0.26).

| Second-line antidiabetic treatment prescribed among people with prevalent CVD
When stratifying by prevalent CVD status (n = 8466 with prevalent CVD, n = 27 557 without prevalent CVD), adjusted predicted percentages showed no substantial differences in SU prescribing across ethnicities and across deprivation quintiles. However, adjusted predicted percentages showed evidence of a slightly higher proportion of people prescribed DPP-4 inhibitors with prevalent CVD versus no CVD across all ethnicities. Conversely, adjusted predicted percentages showed evidence of slightly less SGLT2 inhibitor prescribing among people with prevalent CVD versus no CVD across all ethnicities (Table S7).

| Secondary analyses
Results were similar in the fixed-effect and mixed-effect models (Table S6), and we did not observe any substantial mediation by deprivation level on the association between ethnicity and second-line treatment prescribed (Table S8).  Second-line antidiabetic treatments prescribed changed over time (Tables S9-S11, Figure S1) prescribed SGLT2 inhibitors compared with those without prevalent CVD across time periods (Table S10).

| DISCUSSION
We found statistically significant, but small absolute differences in SU, DPP-4 inhibitor and SGLT2 inhibitor prescribing as second-line antidiabetic treatment, in combination with metformin, according to ethnicity and deprivation level, after accounting for several covariates and clustering at the CCG level in England. There was some evidence that across ethnic groups and levels of deprivation, people with prevalent CVD had a lower probability of being prescribed SGLT2 inhibitors compared with those without prevalent CVD.
It is reassuring that we did not observe substantial ethnic differences in second-line antidiabetic treatment prescribing, as previous research has described many other ethnic disparities related to T2DM. In the UK, ethnic minorities with T2DM had longer delays in intensification to second-line treatment than White people with T2DM, and experienced greater treatment inertia following identification of uncontrolled HbA1c. 12