Study population

We identified 73,895 new users of GLP-1 receptor agonists and 113,207 new users of SGLT2 inhibitors (Table 1). Among GLP-1 receptor agonist users, the mean age (± SD) was 62.9 ± 12.2 years, 58.8% were men, 67.1% were obese (BMI ≥30 kg/m2) and 23.0% had a history of ASCVD. Among SGLT2 inhibitors users, the mean age (± SD) was 65.1 ± 11.5 years, 64.8% were men, 10.7% had a history of heart failure, 15.3% had a history of chronic kidney disease, and 29.2% had a history of ASCVD.

Table 1 Characteristics of new users of SGLT2 inhibitors and GLP-1 receptor agonists Treatment discontinuation

The cumulative incidences of treatment discontinuation and reinitiation among GLP-1 receptor agonist users and SGLT2 inhibitor users using various length of grace periods are shown in Figs 1 and 2, respectively.

Fig. 1

Treatment discontinuation (a), treatment reinitiation (b) and the proportion of patients covered by treatment (c) across various lengths of grace period used to define treatment discontinuation among GLP-1 receptor agonist users

Fig. 2

Treatment discontinuation (a), treatment reinitiation (b) and the proportion of patients covered by treatment (c) across various lengths of grace period used to define treatment discontinuation among SGLT2 inhibitor users

For GLP-1 receptor agonist users, the cumulative incidence of treatment discontinuation was 23.6% (95% CI 23.2, 23.9) at 1 year and 38.5% (95% CI 38.2, 38.9) at 3 years, when using a grace period of 90 days. Discontinuation rates were substantially reduced to 16.2% (95% CI 16.0, 16.5) at 1 year and 31.0% (95% CI 30.6, 31.4) at 3 years when using a grace period of 180 days, and to 23.3% (95% CI 23.0, 23.7) at 3 years when using a grace period of 365 days. Conversely, discontinuation rates were increased to 28.1% (95% CI 27.7, 28.4) at 1 year and 43.6% (95% CI 43.2, 44.0) at 3 years when using a grace period of 60 days (ESM Table 4).

For SGLT2 inhibitor users, the cumulative incidence of treatment discontinuation was 27.9% (95% CI 27.6, 28.1) at 1 year and 45.9% (95% CI 45.5, 46.2) at 3 years, when using a grace period of 90 days. Discontinuation rates were substantially reduced to 18.8% (95% CI 18.6, 19.0) at 1 year and 36.8% (95% CI 36.5, 37.1) at 3 years when using a grace period of 180 days, and to 28.8% (95% CI 28.5, 29.1) at 3 years when using a grace period of 365 days. Conversely, discontinuation rates were increased to 33.7% (95% CI 33.4, 34.0) at 1 year and 50.6% (95% CI 50.3, 50.9) at 3 years when using a grace period of 60 days (ESM Table 4).

Figure 3 shows that, while there were no or small differences in discontinuation rates between subgroups according to status of ASCVD, chronic kidney disease and heart failure among users of GLP-1 receptor agonists, discontinuation rates were higher among those with normal weight vs other weight categories. For example, the HR for treatment discontinuation was 0.65 (95% CI 0.61, 0.68) for obesity class II/III vs normal weight (Table 2). Figure 4 shows that discontinuation rates for SGLT2 inhibitors differed little between subgroups of patients by status of ASCVD, chronic kidney disease and heart failure, although some of the differences were statistically significant. The largest difference was seen among patients with heart failure vs those without (HR for discontinuation 0.81, 95% CI 0.79, 0.84) (Table 2).

Fig. 3

Treatment discontinuation across subgroups of ASCVD (a), chronic kidney disease (b), heart failure (c) and BMI (d) among GLP-1 receptor agonist users. Cut-offs for BMI classes are given in ESM Table 2. sdHR, sub-distributional HR

Table 2 Sex- and age adjusted sub-distributional HRs for discontinuation of treatment in new users of GLP-1 receptor agonists and SGLT2 inhibitorsFig. 4

Treatment discontinuation across subgroups of ASCVD (a), chronic kidney disease (b), heart failure (c) and BMI (d) among SGLT2 inhibitor users. Cut-offs for BMI classes are given in ESM Table 2. sdHR, sub-distributional HR

ESM Figs 1 and 2 shows that calendar year of drug initiation had little effect on the cumulative incidence of treatment discontinuation for both study drugs.

Table 2 shows that several variables were associated with a higher or lower risk for treatment discontinuation, with the associations being of similar direction and magnitude for both GLP-1 receptor agonist users and SGLT2 inhibitors users. Examples of variables associated with a higher or lower risk for treatment discontinuation for both drugs were age, place of birth, education, income and number of diabetes drugs used.

Treatment reinitiation

Among those who discontinued a GLP-1 receptor agonist, the cumulative incidence of reinitiating treatment was 41.1% (95% CI 40.5, 41.7) at 1 year and 57.4% (95% CI 56.7, 58.1) at 3 years when using a grace period of 90 days in the definition of treatment discontinuation (Fig. 1 and ESM Table 5). Similarly, among those who discontinued an SGLT2 inhibitor, the cumulative incidence of reinitiating treatment was 40.4% (95% CI 39.9, 40.8) at 1 year and 55.7% (95% CI 55.2, 56.2) at 3 years when using a grace period of 90 days (Fig. 2 and ESM Table 5). In line with the analyses of treatment discontinuation, the proportion of patients reinitiating treatment were substantially reduced or increased, respectively, when applying longer and shorter grace periods. When using a grace period of 60, 90 or 180 days to define treatment discontinuation, most treatment reinitiations occurred within the first year after discontinuation, but this pattern was not seen with a grace period of 365 days.

Proportion of patients covered

Figures 1 and 2 show the proportion of patients who were covered among GLP-1 receptor agonist users and SGLT2 inhibitor users, respectively. For GLP-1 receptor agonists, the proportion of patients covered was 81.7% at 1 year, 76.7% at 3 years and 75.1% at 5 years when using a grace period of 90 days. For SGLT2 inhibitors, the proportion of patients covered was 78.5% at 1 year, 69.8% at 3 years and 68.2% at 5 years when using a grace period of 90 days. The results were largely similar across the various lengths of grace period for both drugs (ESM Table 6).

Adherence

Among 72,733 GLP-receptor agonist users who were alive 1 year after the date of filling their first prescription, 68% had ≥80% of proportion of days covered, 60% had ≥90% proportion of days covered, and 22% had 100% proportion of days covered. The 1-year mean proportion of days covered was 78% (SD 39) and the 1-year median proportion of days covered was 95% (IQR 63–100).

Among 111,104 SGLT2 inhibitor users who were alive 1 year after the date of filling their first prescription, 64% had ≥80% of proportion of days covered, 58% had ≥90% proportion of days covered, and 33% had 100% proportion of days covered. The 1-year mean proportion of days covered was 77% (SD 30) and the 1-year median proportion of days covered was 97% (IQR 55–100).

ESM Table 7 shows the mean and median proportions of days covered across subgroups of ASCVD status, chronic kidney disease status, heart failure status, BMI category and year of initiation.

Treatment trajectories and switching of drug within the same drug class

Between 1 January 2017 and 31 December 2019, 35,105 patients initiated treatment with a GLP-1 receptor agonist (liraglutide: 22,008 [62.7%]; semaglutide: 6714 [19.1%]; other GLP-1 receptor agonist: 6383 [18.2%]) and 55,988 patients initiated treatment with an SGLT2 inhibitor (empagliflozin: 48,499 [86.6%]; dapagliflozin: 6929 [12.4%]; other SGLT2 inhibitors: 560 [1.0%]).

At 3 years after drug initiation, of the 35,105 GLP-1 receptor agonist users, 14,910 (42.5%) had discontinued treatment, 18,481 (52.6%) had continued treatment and 1714 (4.9%) had died or emigrated (Fig. 5). A total of 8032 (22.9%) patients had switched GLP-1 receptor agonist, with the most common drug switch being from liraglutide to semaglutide (n= 5839; 26.5% of all liraglutide initiators). Of those who had discontinued treatment, 8632 (57.9%) later reinitiated treatment but 6278 (42.1%) did not.

Fig. 5

Treatment trajectories and switching between drugs within the same drug class for GLP-1 receptor agonists. ‘Other GLP-1 receptor agonists’ comprise exenatide, lixisenatide and dulaglutide

At 3 years after drug initiation, of the 55,988 SGLT2 inhibitor users, 28,961 (51.7%) had discontinued treatment, 24,750 (44.2%) had continued treatment and 2277 (4.1%) had died or emigrated (Fig. 6). A total of 1156 (2.1%) patients had switched to another SGLT2 inhibitor. Of those who had discontinued treatment, 15,851 (54.7%) later reinitiated treatment but 13,110 (45.3%) did not.

Fig. 6

Treatment trajectories and switching between drugs within the same drug class for SGLT2 inhibitors. ‘Other SGLT2 inhibitors’ comprise canagliflozin and ertugliflozin