Trends in Prescribing Preferences for Antidiabetic Medications Among Patients With Type 2 Diabetes in the U.K. With and Without Chronic Kidney Disease, 2006–2020

Kidney disease affects more than one in three adults with type 2 diabetes mellitus (T2DM) (1), and T2DM is a key contributor to the global increase in the incidence of end-stage kidney disease (ESKD), dialysis, and kidney transplantations (2). Compared with T2DM alone, the co-occurrence of T2DM and kidney disease augurs a clinical course characterized by greater insulin resistance (3), accelerated progression of T2DM (1,4), and an increased risk of cardiovascular events and overall mortality (5,6). Additionally, the therapeutic management of T2DM itself is further complicated by the presence of chronic kidney disease (CKD), because decreased kidney function impairs the clearance of most glucose-lowering medications, predisposing patients to drug-related toxicities, particularly hypoglycemia (7).

In recent years, data from clinical trials have shown that sodium–glucose cotransporter inhibitors (SGLT2 is) reduce by ∼40% the incidence of adverse kidney outcomes such as progression of kidney disease and ESKD (8,9), and select glucagon-like peptide 1 receptor agonists (GLP-1RAs) reduce such events by 22% to 36%, primarily through their salutary effects on albuminuria (10,11). Consequently, package inserts from the U.S. Food and Drug Administration and summaries of product characteristics from the European Medicines Agency for select SGLT2is (namely, canagliflozin and dapagliflozin) now include new indications highlighting the reduced risk of ESKD and deleterious kidney outcomes among patients codiagnosed with T2DM and CKD. Furthermore, the joint American Diabetes Association and European Association for the Study of Diabetes consensus reports, as well as the U.K. National Institute for Health and Care Excellence guidelines recommend SGLT2is for patients with T2DM with an estimated glomerular filtration rate (eGFR) between 30 and 60 mL/min/1.73 m² to prevent progression of kidney disease (12–14).

Over the past two decades alone, >20 new medications, including entirely new drug classes like SGLT2is (available in U.K. since 2012), dipeptidyl peptidase 4 inhibitors (DPP4 is; available in U.K. since 2006), and GLP-1 RA (available in U.K. since 2005), have been approved for the treatment of T2DM. However, there is a paucity of data comparing the prescription use and uptake patterns among patients with T2DM with and without kidney disease. Moreover, it is also unclear whether the recent publication of landmark trials and changes in clinical guidelines and drug labels have resulted in corresponding changes in the use of select medications among these populations. Accordingly, using electronic health care records from general practices in the U.K. from 2006 to 2020, we identified a cohort of patients with T2DM and assessed the overall, first-line, and second line (i.e., after metformin) medication-initiation patterns among patients with and without CKD. Furthermore, among patients with CKD, we examined medication use across varying degrees of kidney function.

The study was approved by the Rutgers University Institutional Review Board and the Clinical Practice Research Datalink (CPRD) Independent Scientific Advisory Committee.

We used CPRD Gold data from 2006 to 2020. CPRD is an integrated, longitudinal, population-based database that contains electronic health care records for >15 million patients sourced from >700 practices throughout the U.K. (15). The data have been extensively validated, are of research quality, and representative of the U.K. population with respect to age, sex, and ethnicity (16–19). Medical information, including patient symptoms, diagnoses, and lifestyle measures (e.g., smoking status), is recorded using Read Codes (version 2) by the general practice staff during consultations. Prescriptions issued by the provider are captured in the database along with the product name and British National Formulary Codes. Information on laboratory results (e.g., serum creatinine, hemoglobin A_1c [HbA_1c]) are available through preexisting linkages with laboratories in the U.K. Other relevant data elements of interest include patient demographics (e.g., age, biological sex, practice sites) and anthropometric assessments (e.g., BMI).

Within the data, we identified instances of medication initiation (defined as more than one prescription) of the following glucose-lowering therapies: metformin, sulfonylureas (SUs), SGLT2i, DPP4i, GLP-1RA, thiazolidinediones, and insulin. New treatment initiation was defined as nonuse of any study medication from a given class in the baseline period (i.e., 365 days prior to treatment initiation). Cohort membership was restricted to initiation episodes with a recorded serum creatinine value; diagnosis of T2DM; and no evidence of type 1 or gestational diabetes, cancer, polycystic ovary syndrome, acute kidney injury, end-stage kidney disease, or HIV; and to practices that were up to standard for at least 1 year prior to the index date (15). Patients were allowed to contribute more than one episode of treatment initiation as long as the eligibility criteria were met at the time of initiation.

The most recent serum creatinine value within 1 year of the index date was used to estimate the glomerular filtration rate via the modified Chronic Kidney Disease Epidemiology Collaboration Equations, which does not consider race (20). These equations have been extensively validated and recommended by the National Kidney Foundation and the American Society of Nephrology (21). An eGFR threshold of <60 mL/min/1.73 m² was used to designate patients as having CKD. Using Kidney Disease Improving Global Outcomes guidelines, we further classified patients into the following CKD stages: 3a (eGFR, 45–59 mL/min/1.73 m²), 3b (30–44 mL/min/1.73 m²), and 4 (15–29 mL/min/1.73 m²) (22).

In addition to kidney function, we also assessed information on relevant clinical characteristics, including sociodemographic variables (e.g., age at treatment initiation, biological sex), medical conditions (e.g., complications of diabetes, cardiovascular conditions), and the most recently available HbA_1c. These characteristics were reported by CKD status over the study period (2006–2020); for the year 2020, the most recent calendar year for our study; and by individual medication classes.

All analyses were performed using SAS 9.4 (SAS Institute Inc, Cary, NC). The period from 2006 to 2020 was segmented into 15 calendar-year intervals, and each episode of treatment initiation was assigned to one of these intervals according to the date of initiation. We described the overall trends in medication-initiations patterns among those with and without CKD by estimating the proportion of individual drug classes initiated (numerator) over all antidiabetic medication initiations (denominator).

Furthermore, we examined changes in the first- and second-line (after metformin) initiations, defining first-line therapy by nonuse of any prior glucose-lowering therapies in the 365 days prior to the index date. Second-line therapy was defined as initiation of a glucose-lowering medication after having only used metformin monotherapy during the 365 days prior to the index date. Patients initiating disparate second-line therapies differed with respect to key characteristics (e.g., patients initiating DPP4i were older); therefore, we estimated a multinomial regression that modeled the probability of initiating the following four most frequently used second-line therapies as the outcome: SGLT2i, GLP-1RA, SU, or DPP4i (referent category). For this regression analysis, we restricted our cohort to patients receiving metformin monotherapy (i.e., naïve to any second-line antidiabetic therapy), and adjusted for the effects of age, biological sex, and HbA_1c.

We reported initiation patterns by stage of kidney disease (eGFR: 15–29, 30–44, 45–59 mL/min/1.73 m²) and described medication use among a cohort of patients with preserved eGFR (i.e., ≥60 mL/min/1.73 m²) and moderately to severely increased albuminuria, as indicated by a urine albumin to creatinine ratio >30 mg/g. In addition to describing medication-initiation events (i.e., new initiations of antidiabetic medications), we also reported the secular trends in the prevalence of use of antidiabetic medications by CKD status where the requirement for new use was waived. Finally, we reported the trends in initiation patterns in which we also included initiation episodes with missing or unknown T2DM status. For all analyses, we conducted Cochrane–Armitage tests to assess for presence of a trend for each drug class over time and reported the corresponding P values.

We identified 542,081 new episodes of medication initiations across the study period (see Supplementary Fig. 1 for the CONSORT flow diagram). After the inclusion and exclusion criteria were applied, there were 269,585 remaining eligible episodes, and of patients involved in these episodes, 38,622 initiating medication use had CKD. Notably, information on serum creatinine level (within 1 year of medication initiation) was available for >85% of patients, and 75,175 of 542,081 episodes (13.8%) were excluded due to missing serum creatinine data.

Compared with their counterparts without kidney disease, patients with CKD were older, more likely to be female, and had a higher prevalence of micro- and macrovascular complications of T2DM, but the two groups had similar levels of antidiabetic medication use during the baseline period (Table 1). These differences in patient characteristics were consistent for the most recent calendar year in our study (Supplementary Table 1). Compared with patients initiating other classes of antidiabetic medications, those using SGLT2i and GLP-1RA were more likely to be younger, have tried other antidiabetic therapies during the baseline period, and were least likely to have CKD (see Supplementary Table 2 for information on clinical characteristics by antidiabetic class).

Among patients with CKD and over the study period, metformin initiations as a proportion of overall medication initiations declined by 30.8 percentage points (59.2% in 2007 to 28.4% in 2020; P < 0.01), coinciding with a 27.9 percentage-point increase in use of DPP4i (from 0.4 to 28.3%; P < 0.01) (Fig. 1A and Supplementary Table 3A). Notably, although the use of other glucose-lowering therapies remained stable or decreased modestly, SGLT2i use increased by 8.9 percentage points, from 0.5 to 9.4% (P < 0.01).

Among patients without CKD (Fig. 1B and Supplementary Table 3B), metformin use declined similarly by 27.7 percentage points, from 61.8% in 2007 to 34.1% in 2020 (P < 0.01). By contrast, SGLT2i use increased markedly by 26.1 percentage points, from 2.4 to 28.5% (P < 0.01), and eventually becoming one of the most-initiated drug classes after metformin. By 2020, patients without CKD were three times more likely to initiate SGLT2i compared with those with CKD (28.5% vs. 9.4%; P < 0.01). The use of GLP-1RA remained low, regardless of presence of kidney disease.

Metformin was the predominant first-line medication initiated among patients with and without CKD (Fig. 2; Supplementary Table 4). However, its use was markedly lower among those with CKD. By the end of the study period (2020), metformin constituted 66.5% vs. 85.8% of medication initiations among patients with and without CKD, respectively. By 2020, DPP4i (15.5% of patients with CKD vs. 2.3% without CKD) and SUs (12.7% of patients with CKD vs. 6.2% without CKD) were more likely to be used as first-line therapy among patients with CKD compared with those who did not have CKD.

The choice of second-line therapy initiated after metformin also varied as a function of CKD status (Fig. 3; Supplementary Table 5). Among those with CKD, DPP4i became the most-initiated drug class, increasing by 56.8 percentage points, from 0.6% in 2007 to 57.4% in 2020 (P < 0.01), replacing SUs, which decreased from a peak of 77.3% in 2009 to 30.3% in 2020 (P < 0.01). Notably, the use of SGLT2i initiations as second-line therapy remained relatively low, only constituting 7.8% of second-line medication initiations in 2020 (P < 0.01). Comparatively, over the same period, the use of SGLT2i among patients without CKD increased by 44.9 percentage points, from 1.4% in 2013 to 46.3% in 2020, surpassing SU use in 2018 and DPP4i prescriptions in 2019 to become the most frequently initiated second-line drug class (P < 0.01). By 2020, patients without CKD were nearly six times more likely to initiate SGLT2i as second-line therapy compared with those with CKD (46.3% vs. 7.9%).

Figure 4 and Supplementary Table 6 describe the initiation trends for glucose-lowering therapies by stage of kidney disease. Among patients with stage 3a CKD, use of metformin declined, but metformin retained its position as the most-initiated medication class, followed by DPP4i and SUs. SGLT2i use remained low, increasing marginally from 0.8% in 2013 to 11.5% in 2020 (P < 0.01).

Among patients with more severe kidney disease, DPP4i initiations increased rapidly, especially after 2011 and 2012, eventually constituting 34.2% and 40.2% of all initiations among patients with stage 3b and 4 CKD and becoming the treatment of choice for this population. SGLT2i use among patients with stage 3b CKD increased from 0.1% in 2013 to 6.8% in 2020 and was negligible among those with stage 4 CKD.

Supplementary Table 7 describes the initiation patterns of patients with diabetes with preserved eGFR and moderately or severely increased albuminuria. The overall, first-, and second-line medication-initiation patterns for these patients largely resembled those of the non-CKD cohort.

Among patients using metformin monotherapy, and after adjusting for the effects of age, biological sex, and HbA_1c, those with CKD were less likely to initiate SGLT2i than DPP4i (adjusted odds ratio [OR] 0.20; 95% CI 0.17–0.24), equally likely to initiate GLP-1RA (OR 1.24; 95% CI 0.93–1.64), and more likely to initiate SUs (OR 1.21; 95% CI, 1.14–1.28) (Supplementary Table 8).

Findings for changes in the prevalence of use of antidiabetic medications were similar to our primary analysis, which examined medication initiations (Supplementary Table 9). The prevalence of SGLT2i use increased from 0.1 to 4.2% of all antidiabetic medications over the study period among those with CKD compared with an increase from 0.4 to 14.5% among those without CKD. Use of GLP-1RA remained low throughout the study, constituting <2% and <3% of prescribed medications among patients with and without CKD, respectively. Findings from the sensitivity analysis in which we included patients with missing or unknown T2DM status were similar to the primary findings; for instance, over the study period, initiation of SGLT2i increased by 9.3 percentage points, from 0.4 to 9.7%, among patients with CKD, and 23.1 percentage points, from 2.1 to 25.2%, among those without CKD (Supplementary Table 10).

We leveraged a large database of U.K. electronic medical records between 2006 to 2020 to compare antidiabetic medication-initiation patterns among patients with T2DM with and without CKD. With the emergence and increasing availability of newer therapeutics, we noted a decline in the initiation of metformin relative to other glucose-lowering agents over the study period. SGLT2i initiations increased modestly among patients with CKD, but this increase was more pronounced among those without CKD. Notably, by the end of the study period (2020), patients without CKD compared with patients with CKD were three and six times more likely to initiate SGLT2i overall and as second-line therapy, respectively; these differences in SGLT2i uptake among patients with CKD persisted after adjustments for age and biological sex. We found that DPP4i became the treatment of choice among patients with more severe kidney disease (i.e., stage 3b and 4 CKD) and as second-line therapy among patients with CKD. Meanwhile, GLP-1RA prescribing remained minimal throughout the study period, regardless of CKD status.

Despite evidence from landmark trials demonstrating cardiorenal benefits among patients with CKD, as well as multiple guidelines recommending their use, we found a slower uptake of SGLT2i among patients with CKD compared with those without CKD (12,13). The observed contrast in the rate of uptake of SGLT2i among CKD and non-CKD populations may be attributable to four factors. First, despite ongoing guidance recommending SGLT2i for patients with an eGFR of <60 mL/min/1.73 m² irrespective of HbA_1c targets (13), the diminished ability of SGLT2i to reduce serum glucose levels in the setting of moderate to severe kidney disease may limit its use in the real world (23). Second, there may be potential concerns about SGLT2i among patients with CKD relating to transient increases in serum creatinine levels and volume depletion (24). Third, SGLT2is are neither studied nor indicated in severe kidney disease (eGFR <30 mL/min/1.73 m²) and thus may be avoided in patients nearing this kidney function threshold (i.e., stage 3b CKD). Fourth, compared with patients without CKD, those with CKD were older and had a higher overall comorbidity burden, which may have led to the preferential use of medications with well-established safety profiles among older and frailer populations (i.e., DPP4i); however, we found that these differences in use persisted after adjustments for age in our regression analysis. Finally, because the U.K. has universal health care coverage, considerations relating to lack of insurance coverage or medication costs are less likely to be significant barriers for use of branded products such as SGLT2i and GLP-1RA, compared with the U.S., where such factors may play a more important role (25). However, there may still exist regional variation in the use of branded products such as SGLT2i, due to formulary restrictions imposed by local health boards or clinical commissioning groups (26).

Despite the publication of recent clinical guidelines promoting GLP-1RAs for patients with high-risk cardiovascular conditions and the demonstration of kidney benefits for some agents within this class (i.e., liraglutide, semaglutide, and dulaglutide), we found minimal use of GLP-1RA in our data throughout the study period, regardless of CKD status, which is consistent with prior data in CPRD (27,28). Notably, unlike SGLT2i, GLP-1RAs retain their glucose-lowering effect in patients with impaired kidney function. A possible explanation for the limited usen of GLP-1RAs could be their lack of inclusion in the National Institute for Health and Care Excellence (a branch of the U.K. Department of Health and Social Care that develops national evidence-based guidelines and quality standards for multiple disease states) T2DM guidelines as second-line therapy (29). Furthermore, until the recent approval of oral semaglutide, GLP-1RAs were exclusively available as injectable formulation, which may also have constrained their clinical use. Our finding of an increase in DPP4i use as second-line therapy for patients with CKD and as the treatment of choice among patients with more severe kidney disease (eGFR <45 mL/min/1.73 m²) can likely be ascribed to its well-established safety profile and tolerability among patients with severe kidney impairment (30). However, with the possible exception of saxagliptin, which was not commonly prescribed in our data and may increase cardiovascular risk, all agents within this class have a neutral cardiorenal profile.

Metformin remained the predominant first-line therapy prescribed regardless of CKD status, though its use was lower among patients with CKD. This likely was due to concerns about lactic acidosis; however, with the publication of the recent European Medicines Agency and U.S. Food and Drug Administration guidance recommending its use in patients with kidney disease and an eGFR >30 mL/min/1.73 m², we saw a corresponding, albeit marginal, increase in its use among patients with CKD (31). Finally, in spite of its propensity for hypoglycemia development in patients with CKD and multiple guidelines recommending against its use (32), SU initiations among patients with CKD, especially as second-line therapy, remained consistently high in our data throughout the study period.

Using a population-representative electronic health care database from general practices in the U.K. over 15 years, we evaluated in our study the trends in the use of glucose-lowering therapies in patients with T2DM and CKD, and we contextualized these findings by juxtaposing them against a non-CKD cohort. Laboratory data to estimate kidney function were available for >85% of initiation episodes, allowing for an unbiased appraisal of medication use. Moreover, the availability of albuminuria data allowed us to evaluate medication use among a population with diabetes and CKD with preserved eGFR (33). However, study findings should be viewed in light of certain limitations. First, CPRD data do not allow for the differentiation of prescriptions originating from primary versus secondary care (e.g., endocrinologists), nor does it include information on prescriptions originating from within the hospital or emergency department visits. Second, medication records in our data capture prescribing information, not prescription fills. Accordingly, a patient prescribed a medication without having it filled would still be considered as having initiated that medication in our study. Third, we used a 12 month washout window to define new medication initiation; however, it is possible for a patient to have used the medication prior to this 12-month washout period. In conclusion, although SGLT2i initiations among patients with CKD increased modestly over the study period, by the end of the study in 2020, compared with patients with CKD, those without CKD were three and six times more likely to initiate SGLT2i overall and as second-line therapy, respectively. Notably, the use of GLP-1RA remained low throughout the study years, regardless of CKD status, whereas DPP4i and SUs were frequently initiated in patients with CKD. Compared with DPP4is, which have a neutral cardiorenal profile, and SUs, which have an increased risk of hypoglycemia development in patients with CKD, both SGLT2i and GLP-1RA confer significant advantages with respect to the prevention of deleterious cardiorenal events. As the landscape of diabetes management continues to shift from a glucocentric approach that prioritized reductions in HbA_1c to one that considers long-term cardiorenal health, work is needed to identify and eliminate potential barriers associated with the uptake of SGLT2i and GLP-1RA among patients with CKD.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. All authors were involved in the conception and design of the study, analysis and interpretation of data, critical revision of the manuscript for important intellectual content, and final approval of the manuscript. J.L. and C.V.D. drafted the manuscript, and C.V.D. provided administrative and logistical support, statistical expertise, and overall study supervision. J.L. assembled and analyzed the data. J.L. and C.V.D. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.