Abstract
Aims –
To determine the association with cardiovascular (CV) outcomes of sodium glucose cotransporter-2 (SGLT-2) inhibitors compared with dipeptidyl peptidase-4 (DPP-4) inhibitors in patients with type 2 diabetes mellitus (T2DM) and chronic kidney disease (CKD).
Materials and Methods –
We conducted a population-based cohort study of new users of SGLT-2 inhibitors and DPP-4 inhibitors with T2DM and CKD using data from Optum Clinformatics DataMart. We assembled three cohorts: T2DM/no CKD, T2DM/CKD 1–2, and T2DM/ CKD 3a. Study outcomes were 1) time to first heart failure (HF) hospitalization; and 2) time to a composite CV endpoint comprised of non-fatal myocardial infarction (MI) or stroke. After inverse probability of treatment weighting, we used proportional hazards regression to estimate hazard ratios (HR) and 95% confidence intervals (CI).
Results –
New users of SGLT-2 inhibitors versus of DPP-4 inhibitors had lower risks of HF hospitalization in the T2DM/no CKD (HR, 0.76; 95% CI, 0.70, 0.82) and T2DM/CKD 1–2 (HR, 0.63; 95% CI, 0.48, 0.84), but no significant association was present in the T2DM/CKD 3a cohort. Compared with prescription of DPP-4 inhibitors, SGLT-2 inhibitors were associated with lower risks of non-fatal MI or stroke of 23% (HR, 0.77; 95% CI, 0.70, 0.85) in the T2DM/no CKD cohort, but no significant associations were present in the T2DM/CKD 1–2 and T2DM/CKD 3a cohorts.
Conclusions –
Incident prescription of SGLT-2 inhibitors was associated with lower risks of HF hospitalization but not with non-fatal MI or stroke despite suggesting benefit, relative to prescription of DPP-4 inhibitor across different stages of CKD.
INTRODUCTION
Diabetic kidney disease is a leading cause of kidney failure and a major risk factor for cardiovascular (CV) morbidity and mortality. Over the past decade, sodium-glucose cotransporter-2 (SGLT-2) inhibitors have garnered much attention for their CV and kidney benefits as shown in multiple CV outcome trials (CVOTs). The EMPA-REG-Outcome trial reported that empagliflozin, compared with placebo, reduced the risk of primary composite CV outcome consisting of death from CV causes, nonfatal myocardial infarction (MI), or nonfatal stroke as well as the risk of hospitalization for heart failure (HF).1 Patients randomized to the empagliflozin group also had significantly lower rates of death from CV causes and death from any cause compared with those in the placebo group.1 Subsequently, the CANVAS trial showed that canagliflozin, compared with placebo, reduced the risk of primary composite CV outcome and hospitalization for HF by 14% and 33%, respectively2 while dapagliflozin in the DECLARE-TIMI 58 trial demonstrated a 17% reduced risk of hospitalization for HF or CV death versus placebo, which was explained by a lower rate of hospitalization for HF.3 Results from these earlier CVOTs suggested that SGLT-2 inhibitors could be used for secondary prevention of major CV events and heart failure (HF) in patients with T2DM and either a previous CV event or risk factors for CV disease (CVD). However, these studies enrolled only a small number of patients with chronic kidney disease (CKD).
In contrast to these earlier CVOTs, the CREDENCE trial was conducted in patients with substantial albuminuria and/or impaired kidney function and had a primary cardiorenal composite endpoint. The CREDENCE trial showed that canagliflozin, when used in conjunction with angiotensin-converting enzyme (ACE) inhibitors or angiotensin II receptor blockers (ARBs) in adult patients with T2DM and albuminuric CKD (stages 2, 3a, and 3b), had clinically meaningful incremental CV and kidney benefits and could be safely initiated in patients with estimated glomerular filtration rate (eGFR) as low as 30 mL/min/1.73m.2,4,5
Some observational studies have used real-world data to examine the association of SGLT-2 inhibitor use with CV outcomes in patients with T2DM;6–11 however, these studies have not focused on the high-risk population of patients with both T2DM and CKD. Moreover, clinical trials comparing the effects of different oral glucose-lowering drugs (GLDs) on CV outcomes have been sparse, as most studies of newer GLDs have been placebo-controlled. Dipeptidyl peptidase-4 (DPP-4) inhibitors constitute another widely used class of oral GLDs with proven CV safety and noninferiority versus placebo although concerns about the potential risk of HF have been raised.12–15 DPP-4 inhibitors are recommended as second- or higher line therapy for patients with T2DM and are, therefore, a reasonable comparator for examining the effectiveness of another oral GLD class.
The aim of this study was to use real-world data to conduct a comparative effectiveness analysis of new users of SGLT-2 inhibitors versus DPP-4 inhibitors focusing on the risk of CV outcomes in patients with T2DM and CKD.
MATERIALS AND METHODS
Data Source
We collected data from Clinformatics™ Data Mart Database (OptumInsight, Eden, Prairie, MN), a de-identified database from a large national insurance provider. Detailed description of the Optum database is given in the Supplemental Material (Appendix 1). This study was approved by the Institutional Review Board of Stanford University and conducted in accordance with the Declaration of Helsinki guidelines.
Study population
We included all patients diagnosed with T2DM based on ICD-9 (250.x0 and 250.x2) and ICD-10 (E11) codes from January 1, 2013 to June 30, 2017 (N = 2,399,612) and retained only those who initiated SGLT-2 inhibitors or DPP-4 inhibitors (Figure 1). Patients were classified into CKD stages using the National Kidney Foundation Kidney Disease Outcomes Quality Initiative (NKF KDOQI) classification guidelines (Table S1).16 We used sex-specific data on urinary albumin-to-creatinine ratio (UACR) to assess albuminuria (>17 mg/g for men or >25 mg/g for women)17 and CKD stage was only determined if eGFR and UACR data were both not missing unless eGFR alone clearly indicated CKD stage 3a. Calculations were done using a single serum creatinine lab value closest to medication initiation date. Patients with CKD stages 3b-5 were excluded due to small number of patients in these groups (SGLT-2 inhibitors were not approved for patients with eGFR below 45 mL/min/1.73m2 until late 2019 when canagliflozin received approval for use with eGFR ≥30 mL/min/1.73m2). For patients with unknown race, we computed eGFR by assuming both black and non-black race then computing a population-weighted average eGFR. We derived three final cohorts: patients with diagnosed T2DM and no CKD (T2DM/no CKD), those with T2DM and CKD stage 1 or 2 (T2DM/CKD 1–2), and those with T2DM and CKD stage 3a (T2DM/CKD 3a). From each of the three assembled cohorts, we excluded patients who received SGLT-2 inhibitors or DPP-4 inhibitors in the previous six months to ensure that the patient entered the cohort on the day of his or her first use of these drugs. We also excluded those who were <18 years of age, did not have ≥6 months of continuous enrollment before drug initiation, and had a history of secondary or gestational diabetes. Patients could contribute person-time to more than one of these three cohorts but could only contribute once to each cohort (Figure 2). The three final cohorts—T2DM/no CKD, T2DM/CKD 1–2, and T2DM/CKD 3a—consisted of 132,979, 7793, and 10,646 patients, respectively.
Figure 1.
Flowchart of cohort creation in The Clinformatics Data Mart, OptumInsight Life sciences database (OptumInsight, Eden, Prairie, MN). Patients can contribute to more than one cohort over time but only once to each cohort.
Figure 2.
Hypothetical examples of how a patient could contribute person-time to one or more of the three study cohorts but could only contribute once to each cohort. For example, patient 1 initiated an SGLT-2 inhibitor or a DPP-4 inhibitor while having T2DM but no CKD, so this patient contributed person-time to the T2DM/no CKD cohort despite subsequently developing CKD stage 1 or 2. Likewise, patient 2 initiated an SGLT-2 inhibitor or a DPP-4 inhibitor while having T2DM and CKD stage 1or 2, so this patient contributed person-time to the T2DM/CKD stages 1–2 cohort despite subsequently developing CKD stage 3a. Patient 3 initially initiated the drug exposure of interest while having T2DM but no CKD and, therefore, contributed person-time to the T2DM/no CKD cohort. Then there was discontinuation of drug use but the patient initiated the drug again after developing CKD stage 1 or 2, and could therefore contribute additional person-time to the T2DM/CKD stages 1–2 cohort.
Primary exposure
The primary exposure of interest was initiation of an SGLT-2 inhibitor or a DPP-4 inhibitor at the time of cohort entry or after a change in CKD stage. The main analysis used a per-protocol approach where patients were followed until medication discontinuation, defined as no new prescription filled within 45 days of last prescription filled + supply date or switching to a new medication class.
Study outcomes
The outcomes of interest were 1) time to first HF hospitalization, and 2) time to first composite CV endpoint consisting of non-fatal MI or stroke in the two and a half years since medication initiation. These outcomes were ascertained using ICD-9/ICD-10 diagnosis codes described in Table S2. In previous studies, the positive predictive value of such claims-based algorithms were ≥80%.18–21 For outcome ascertainment, we required either a single inpatient claim or an outpatient claim followed by either inpatient or second outpatient claim within 30 days. Patients were censored for end of healthcare enrollment or end of the study period (December 31, 2017), whichever came first.
Patient characteristics and covariates
We ascertained data on sociodemographic variables using the Optum SES Member file (SES data file version 7.0). In addition, we considered the following as potential covariates: comorbidities defined using ICD-9/ICD-10 codes, use of GLDs six months prior to cohort entry as well as number of diabetes drugs at the time of cohort entry, and use of other drugs.
Statistical analysis
We performed all analyses separately for each of the three cohorts and compared SGLT-2 inhibitor versus DPP-4 inhibitor users using standardized differences in each cohort.22 We used inverse probability of treatment weighting (IPTW) to mitigate selection bias by imbalances of observed characteristics between SGLT-2 and DPP-4 users.23 Weights were computed from multivariable logistic regression models and included the variables listed in Table 1 to estimate the propensity for SGLT-2 inhibitor versus DPP-4 inhibitor use. We computed stabilized weights defined as the inverse of the estimated propensity multiplied by a constant equal to the observed proportion of patients treated with a SGLT-2 inhibitor. Stabilization does not affect the point estimate but reduces the variability of the IPTW weights. For the small number of weights that were still too large, they were truncated and reset to the boundary level (10 or 0.1) if they were larger than 10 or smaller than 0.1.24,25
Table 1.
Baseline characteristics of patients with type 2 diabetes mellitus initiating SGLT-2 inhibitor or DPP-4 inhibitor by CKD stage†
| Patient Characteristics | T2DM/no CKD (N = 132,979) | T2DM/CKD Stages 1–2 (N = 7793) | T2DM/CKD Stage 3a (N=10,646) | |||
|---|---|---|---|---|---|---|
| SGLT-2i (N = 46,696) | DPP-4i (N = 86,283) | SGLT-2i (N = 3464) | DPP-4i (N = 4329) | SGLT-2i (N = 2594) | DPP-4i (N = 8052) | |
| Demographics | ||||||
| Age (years) | 56.7 ± 10.8 | 64.5 ± 12.6 | 57.2 ± 11.0 | 63.3 ± 12.2 | 66.0 ± 9.1 | 72.1 ± 9.2 |
| Male sex, N (%) | 25,673 (55.0) | 44,770 (51.9) | 2159 (62.3) | 2575 (59.5) | 1293 (49.9) | 3885 (48.3) |
| Race, N (%) | ||||||
| White | 25,472 (54.6) | 43,341 (50.2) | 1469 (42.4) | 1508 (34.8) | 1422 (54.8) | 3921 (48.7) |
| Black | 4414 (9.5) | 9840 (11.4) | 349 (10.1) | 452 (10.4) | 210 (8.1) | 983 (12.2) |
| Asian | 1293 (2.8) | 4175 (4.8) | 122 (3.5) | 255 (5.9) | 79 (3.1) | 365 (4.5) |
| Missing | 9683 (20.7) | 16,260 (18.8) | 812 (23.4) | 930 (21.5) | 563 (21.7) | 1504 (18.7) |
| Hispanic ethnicity, N (%) | 5834 (12.5) | 12,667 (14.7) | 712 (20.6) | 1184 (27.4) | 320 (12.3) | 1279 (15.9) |
| Income, N (%) | ||||||
| <$40,000 | 9075 (19.4) | 22,069 (25.6) | 1079 (24.9) | 662 (19.1) | 611 (23.6) | 2305 (28.6) |
| $40,000 - $49,999 | 3007 (6.4) | 6241 (7.2) | 301 (7.0) | 234 (6.8) | 161 (6.2) | 621 (7.7) |
| $50,000 - $59,999 | 3208 (6.9) | 6534 (7.6) | 337 (7.8) | 217 (6.3) | 180 (6.9) | 701 (8.7) |
| $60,000 - $74,999 | 4772 (10.2) | 8474 (9.8) | 402 (9.3) | 335 (9.7) | 270 (10.4) | 840 (10.4) |
| $75,000 - $99,999 | 6994 (15.0) | 11,040 (12.8) | 517 (11.9) | 474 (13.7) | 331 (12.8) | 908 (11.3) |
| ≥$100,000 | 11,641 (24.9) | 15,861 (18.4) | 805 (18.6) | 876 (25.3) | 572 (22.1) | 1262 (15.7) |
| Missing, N (%) | 9683 (20.7) | 16,260 (18.8) | 666 (19.2) | 888 (20.5) | 469 (18.1) | 1415 (17.6) |
| Reported comorbidities, N (%) | ||||||
| Arrhythmia | 1140 (2.4) | 5595 (6.5) | 94 (2.7) | 184 (4.3) | 152 (5.9) | 843 (10.5) |
| Cardiovascular disease | 618 (1.3) | 3588 (4.2) | 51 (1.5) | 137 (3.2) | 75 (2.9) | 457 (5.7) |
| Congestive heart failure | 764 (1.6) | 5364 (6.2) | 65 (1.9) | 168 (3.9) | 139 (5.4) | 824 (10.2) |
| Coronary artery disease | 3096 (6.6) | 10,644 (12.3) | 278 (8.0) | 423 (9.8) | 356 (13.7) | 1541 (19.1) |
| Hypertension | 21,474 (46.0) | 44,981 (52.1) | 1735 (50.1) | 2192 (50.6) | 1558 (60.1) | 4907 (60.9) |
| Peripheral vascular disease | 797 (1.7) | 4244 (4.9) | 100 (2.9) | 206 (4.8) | 106 (4.1) | 600 (7.5) |
| Cancer | 852 (1.8) | 2972 (3.4) | 71 (2.1) | 144 (3.3) | 83 (3.2) | 462 (5.7) |
| Liver disease | 718 (1.5) | 1387 (1.6) | 49 (1.4) | 71 (1.6) | 43 (1.7) | 160 (2.0) |
| Use of medications, N (%) | ||||||
| No. of diabetes drugs on the day of cohort Entry | ||||||
| 0 | 3895 (8.3) | 14,423 (16.7) | 182 (5.3) | 466 (10.8) | 170 (6.6) | 1311 (16.3) |
| 1 | 15,124 (32.4) | 35,356 (41.0) | 930 (26.9) | 1605 (37.1) | 777 (30.0) | 3383 (42.0) |
| 2 | 16,398 (35.1) | 27,725 (32.1) | 1268 (36.6) | 1647 (38.1) | 1002 (38.6) | 2516 (31.3) |
| >3 | 11,279 (24.2) | 8779 (10.2) | 1084 (31.3) | 611 (14.1) | 645 (24.9) | 842 (10.5) |
| Past use of GLP-a1a in prior 6 months | 17,933 (38.4) | 11,968 (13.9) | 1483 (42.8) | 708 (16.4) | 977 (37.7) | 983 (12.2) |
| Past use of insulin in prior 6 months | 14,672 (31.4) | 13,715 (15.9) | 1421 (41.0) | 926 (21.4) | 1059 (40.8) | 1636 (20.3) |
| Past use of metformin in prior 6 months | 30,759 (65.9) | 50,926 (59.0) | 2384 (68.8) | 2973 (68.7) | 1468 (56.6) | 4164 (51.7) |
| Past use of sulfonylurea in prior 6 months | 16,266 (34.8) | 34,980 (40.5) | 1334 (38.5) | 1894 (43.8) | 1054 (40.6) | 3568 (44.3) |
| Past use of TZD in prior 6 months | 3991 (8.6) | 5419 (6.28) | 308 (8.89) | 263 (6.08) | 247 (9.52) | 541 (6.72) |
| Past use of medications in prior 1 year | ||||||
| Angiotensin-converting enzyme inhibitors | 20,505 (43.9) | 38,778 (44.9) | 1716 (49.5) | 2227 (51.4) | 1230 (47.4) | 3875 (48.1) |
| Angiotensin II receptor blockers | 11,951 (25.6) | 21,943 (25.4) | 1108 (32.0) | 1291 (29.8) | 975 (37.6) | 2761 (34.3) |
| Anticoagulants | 1883 (4.0) | 6113 (7.1) | 129 (3.7) | 244 (5.6) | 210 (8.1) | 896 (11.1) |
| Antidepressants | 2178 (4.7) | 3535 (4.1) | 128 (3.7) | 161 (3.7) | 142 (5.47) | 142 (5.47) |
| Anxiolytics, hypnotics, or sedatives | 8052 (17.2) | 14,904 (17.3) | 539 (15.6) | 638 (14.7) | 526 (20.3) | 1568 (19.5) |
| Antipsychotics | 1073 (2.3) | 2688 (3.1) | 86 (2.5) | 90 (2.1) | 89 (3.4) | 248 (3.1) |
| Beta blockers | 13,088 (28.0) | 31,549 (36.6) | 1092 (31.5) | 1493 (34.5) | 1181 (45.5) | 4079 (50.7) |
| Beta-2 agonist inhalants | 5215 (11.2) | 10,279 (11.9) | 351 (10.1) | 479 (11.1) | 347 (13.4) | 1041 (12.9) |
| Calcium blockers | 1380 (3.0) | 3699 (4.3) | 124 (3.6) | 154 (3.6) | 155 (6.0) | 481 (6.0) |
| Loop diuretics | 4019 (8.6) | 12,413 (14.4) | 310 (9.0) | 408 (9.4) | 533 (20.6) | 1916 (23.8) |
| Other diuretics | 954 (2.0) | 2117 (2.5) | 56 (1.6) | 73 (1.7) | 109 (4.2) | 324 (4.0) |
| Glucocorticoid inhalants | 2328 (5.0) | 5633 (6.5) | 141 (4.1) | 220 (5.1) | 204 (7.9) | 621 (7.7) |
| Nitrates | 1725 (3.7) | 5041 (5.8) | 126 (3.6) | 195 (4.5) | 178 (6.9) | 695 (8.6) |
| NSAIDs | 11,626 (24.9) | 20,369 (23.6) | 905 (26.1) | 1045 (24.1) | 680 (26.2) | 1887 (23.4) |
| Opioids | 15,939 (34.1) | 30,080 (34.9) | 1204 (34.8) | 1474 (34.1) | 1098 (42.3) | 3213 (39.9) |
| Oral glucocorticoids | 6603 (14.1) | 12,893 (14.9) | 481 (13.9) | 676 (15.6) | 467 (18.0) | 1442 (17.9) |
| Platelet inhibitors | 2717 (5.8) | 7726 (9.0) | 246 (7.1) | 338 (7.8) | 314 (12.1) | 1112 (13.8) |
| Statins | 29,864 (64.0) | 56,141 (65.1) | 2441 (70.5) | 3070 (70.9) | 1960 (75.6) | 5975 (74.2) |
| Thiazides | 4979 (10.7) | 11,014 (12.8) | 394 (11.4) | 565 (13.1) | 461 (17.8) | 1363 (16.9) |
| Laboratory measurements | ||||||
| Serum albumin (mg/dL) | 4.3 ± 0.3 | 4.3 ± 0.3 | 4.3 ± 0.3 | 4.3 ± 0.3 | 4.3 ± 0.3 | 4.2 ± 0.3 |
| Missing, N (%) | 31,485 (67.4) | 66,375 (76.9) | 122 (3.5) | 194 (4.5) | 137 (5.3) | 555 (6.9) |
| Hemoglobin (g/dL) | 14.0 ± 1.5 | 13.6 ± 1.7 | 14.1 ± 1.6 | 13.8 ± 1.6 | 13.4 ± 1.6 | 13.0 ± 1.7 |
| Missing, N (%) | 29,216 (62.6) | 59,059 (68.5) | 601 (17.4) | 639 (14.8) | 403 (15.5) | 1275 (15.8) |
| HbA1c (%) | 8.5 ± 1.9 | 8.4 ± 1.8 | 9.1 ± 1.9 | 8.8 ± 1.9 | 8.5 ± 1.7 | 8.0 ± 1.6 |
| Missing, N (%) | 22,330 (47.8) | 46,886 (54.3) | 128 (3.7) | 122 (2.8) | 183 (7.1) | 624 (7.8) |
Abbreviations: CKD, chronic kidney disease; DPP-4i, DPP-4 inhibitor; GLP-1a, GLP-1 agonist; HbA1c, hemoglobin A1c; NSAIDs, nonsteroidal anti-inflammatory drugs; T2DM, type 2 diabetes mellitus; TZD, thiazolidinedione
Reported as means and standard deviations unless noted otherwise.
We used Cox proportional hazards regression to model time to the outcome of study. Inspection of the scaled Schoenfeld residuals after application of the Cox proportional hazards model showed effects that varied over time for some of the cohorts for both of the outcomes analyzed (Figure S1). More specifically, there was a slight protective association with SGLT2-inhibitors during the first three months of follow-up period that subsequently flattened out and the P-values for the test of proportional hazards assumption for these cohorts were <0.05, most probably due to a large sample size. Therefore, the hazard ratios (HR) computed from the final analysis can be interpreted as average HR over follow-up time.
We conducted several sensitivity analyses to test the robustness of our primary findings. First, we used a cutoff of UACR >30 mg/g to define albuminuria in classifying patients as having CKD stage 1 or 2. Second, we re-estimated the propensity score for each of the CKD cohorts by including laboratory measurements such as HbA1c to further account for underlying glucose control and performed a separate imputation analysis in the CKD cohorts. Third, we carried forward the exposure to the initiated drug without considering drug discontinuation or switching to mimic an intention to treat approach and address potential informative censoring.
Methods on handling of missing data are described in the Supplemental Material (Appendix 2).
RESULTS
Baseline characteristics of patients in the three different cohorts prior to conducting IPTW are summarized in Table 1. Across all three cohorts, patients prescribed SGLT-2 inhibitors, relative to those prescribed DPP-4 inhibitors, were younger, more frequently male, and had lower general burden of reported comorbidities and higher frequency of past prescriptions of other GLDs in the prior 6 months with the exception of sulfonylureas. Figure S2 shows that all patient characteristics were well balanced between initiators of SGLT-2 inhibitors and those of DPP-4 inhibitors after applying propensity score-based IPTW adjustment and imputing missing data, as observed by standardized mean differences that fell below the pre-defined limit of 0.1 (10%) threshold. All covariates were balanced between users of SGLT-2 inhibitors and DPP-4 inhibitors across all three different cohorts (Table S3).
Among initiators of SGLT-2 inhibitors, the unadjusted incidence rates for HF hospitalization were 2.9 per 100 person-years for patients with T2DM/no CKD, 3.1 per 100 person-years for patients with T2DM/CKD 1–2, and 8.5 per 100 person-years for patients with T2DM/CKD 3a (Table 2). The unadjusted incidence rates for non-fatal MI or stroke in cohorts of patients with T2DM/no CKD, T2DM/CKD 1–2, and T2DM/CKD 3a were 1.8, 2.6, and 3.6 per 100 patient-years, respectively. In patients initiating DPP-4 inhibitors, the unadjusted incidence rates for HF hospitalization were 8.8 per 100 person-years for patients with T2DM/no CKD, 6.6 per 100 person-years for patients with T2DM/CKD 1–2, and 15.1 per 100 person-years for patients with T2DM/CKD 3a. The crude incidence rates for non-fatal MI or stroke in cohorts of patients with T2DM/no CKD, T2DM/CKD 1–2, and T2DM/CKD 3a were 4.3, 4.1, and 6.0 per 100 person-years, respectively.
Table 2.
Incidence rates (per 100 person-years) of cardiovascular outcomes in patients with type 2 diabetes mellitus by treatment type and diabetes and CKD stage
| T2DM/no CKD | ||||||
|---|---|---|---|---|---|---|
| SGLT-2i | DPP-4i | |||||
| # events | Total person-time† | Incidence rate | # events | Total person-time | Incidence rate | |
| Heart failure hospitalization | 1476 | 51,757 | 2.9 | 8133 | 92,402 | 8.8 |
| Composite CV endpoint‡ | 942 | 52,352 | 1.8 | 4200 | 96,811 | 4.3 |
| T2DM/CKD 1–2 | ||||||
| SGLT-2i | DPP-4i | |||||
| # events | Total person-time | Incidence rate | # events | Total person-time | Incidence rate | |
| Heart failure hospitalization | 119 | 3541 | 3.1 | 286 | 4338 | 6.6 |
| Composite CV endpoint‡ | 93 | 3571 | 2.6 | 182 | 4436 | 4.1 |
| T2DM/CKD 3a | ||||||
| SGLT-2i | DPP-4i | |||||
| # events | Total person-time | Incidence rate | # events | Total person-time | Incidence rate | |
| Heart failure hospitalization | 235 | 2779 | 8.5 | 1250 | 8297 | 15.1 |
| Composite CV endpoint‡ | 104 | 2914 | 3.6 | 547 | 9077 | 6.0 |
Abbreviations: CKD, chronic kidney disease; CV, cardiovascular; DPP-4i, DPP-4 inhibitor; IR, incidence rate; PT, person-time; SGLT-2i, SGLT-2 inhibitor; T2DM, type 2 diabetes mellitus
In years.
Defined as non-fatal myocardial infarction or stroke.
In the Cox proportional hazards regression models from per-protocol analysis, SGLT-2 inhibitor use was associated with a 24% lower risk of developing HF hospitalization compared with use of DPP-4 inhibitors (HR, 0.76; 95% CI, 0.70, 0.82) in patients with T2DM/no CKD (Figure 3). In patients with T2DM/CKD 1–2, SGLT-2 inhibitor use was associated with a 37% lower risk of developing HF hospitalization compared with use of DPP-4 inhibitors (HR, 0.63; 95% CI, 0.48, 0.84). The relative hazard for HF hospitalization was numerically lower but did not reach statistical significance in patients with T2DM/CKD 3a (HR, 0.89; 95% CI, 0.74, 1.07).
Figure 3.
Risk of heart failure admission to hospital and composite cardiovascular endpoint (non-fatal myocardial infarction or non-fatal stroke) associated with SGLT-2 inhibitors versus DPP-4 inhibitors in IPTW per-protocol analyses of patients with type 2 diabetes mellitus by CKD stage.
Prescription of SGLT-2 inhibitors was associated with 23% (HR, 0.77; 95% CI, 0.70, 0.85) lower risk of non-fatal MI or stroke, compared with prescription of DPP-4 inhibitors, in patients with T2DM/no CKD. Prescription of SGLT-2 inhibitors was not statistically associated with the risk of non-fatal MI or stroke, compared with prescription of DPP-4 inhibitors, in patients with T2DM/CKD 1–2 (HR, 0.85; 95% CI, 0.63, 1.14) or in those with T2DM/CKD 3a (HR, 0.79; 95% CI, 0.60, 1.14). In the ITT analyses, all of these associations of interest were slightly attenuated but remained significant (Table S4). Furthermore, sensitivity analyses in which a cutoff of UACR >30 mg/g was used to define albuminuria in classifying patients as having CKD stage 1 or 2 (Table S5) and underlying glucose control was accounted for by including additional laboratory measurements (Table S6) did not lead to any substantial changes in the main findings.
DISCUSSION
In this population-based cohort study, we found that the prescription of SGLT-2 inhibitors was associated with a 24% and 23% reduction in the risk of hospitalization for HF and composite CV outcome of MI or stroke, respectively, in patients with T2DM and no evidence of kidney disease, compared with the use of DPP-4 inhibitors. The use of SGLT-2 inhibitors was associated with a 37% reduced risk of hospitalization for HF in patients with T2DM and CKD stages 1–2 compared with the prescription of DPP-4 inhibitors. The associations between prescription of SGLT-2 inhibitors versus DPP-4 inhibitors and HF hospitalization in patients with T2DM and CKD stage 3a as well non-fatal MI or stroke in patients with T2DM and CKD stages 1–3a suggested benefit but did not reach statistical significance.
These findings complement and are in line with those reported in the EMPA-REG-OUTCOME, CANVAS, DECLARE-TIMI 53, CREDENCE, and DAPA-CKD trials, all of which used placebo control.1–4,26 In general, these trials found a significant reduction in the risk of hospitalization for HF by 27–39% with the use of SGLT-2 inhibitors versus placebo and the effects across the drug class were consistently substantial. Our findings when comparing SGLT-2 inhibitors relative to an active control, namely DPP-4 inhibitors, suggest a beneficial effect of SGLT-2 inhibitors on HF hospitalization in patients with T2DM and, to a lesser extent, on non-fatal MI or stroke, which are supported by findings of CREDENCE and DAPA-CKD. In our study, we found numerically lower risks of non-fatal MI or stroke that did not reach statistical significance in the T2DM/CKD 1–2 and T2DM/CKD 3a cohorts. The CREDENCE trial saw a 20% reduction in risk of major adverse cardiovascular events in patients among whom about half had established CVD or multiple risk factors with known CKD.4 It is possible that the protective effect against major adverse cardiovascular events is more pronounced in patients with advanced renal impairment, which we could not capture in our study due to limited number of patients with CKD stage 3a.
Findings from RCTs are useful for guiding clinical practice, but they may be of limited use in a real-world setting due to modest sample sizes, relatively short follow-up times, and biased selection of participants under idealized and controlled conditions that do not necessarily reflect a real-world clinical setting. CVOTs also evaluate drug efficacy in high-risk patients with established CVD or CV risk and thereby render less evidence on drug efficacy or anticipated effectiveness in the setting of primary prevention or in lower-risk patient groups.
Population-based observational studies can substantively complement data derived from clinical trials. Larger studies conducted in more representative populations than those usually selected for RCTs may better estimate the long-term comparative effectiveness and safety of drugs. Recent findings from the RCT DUPLICATE program in which comparisons were made between the findings of RCTs and those of noninterventional real-world evidence that emulated trials as closely as possible showed that overall, the agreement between RCT and real-world evidence estimates was good for all antidiabetic drug trials, and suggests the use of active comparators that are used in similar indicated populations in real-world evidence studies.27 Unlike all of the large SGLT-2 inhibitor RCTs, our study used DPP-4 inhibitors as the active comparator. In routine practice, providers and patients often face a choice between two or more treatment options. Therefore, comparative data on the benefits of SGLT-2 inhibitors versus another GLD class could be informative for clinical decision-making.
Our study focused on patients with CKD and found sizeable benefits consistent with results from the CREDENCE and DAPA-CKD trials.4,26 Furthermore, our findings are in line with those of other recent observational studies and our stratified analysis by CKD stage adds novelty to the existing literature. A retrospective cohort study using real-world Optum data found that during a 30-month period, the HR for HF hospitalization associated with the SGLT-2 inhibitor canagliflozin were 30% less likely versus a DPP-4 inhibitor in adult patients with T2DM, and HR for composite CV endpoint that was comprised of acute MI and stroke was 11% less likely versus a DPP-4 inhibitor.10 These findings are comparable to ours, which were based on a similar follow-up time and used the same data source. Other observational studies have reported protective associations of SGLT-2 inhibitors and the risk of adverse CV events in comparison with other GLDs. The CVD-REAL-Study was a multinational study in which data were collected from the United States, Norway, Denmark, Sweden, Germany, and the United Kingdom and estimates were pooled to determine the weighted effect size. The study reported a 39% reduced risk of hospitalization for HF in patients with T2DM who were new initiators of SGLT-2 inhibitors versus those of other GLDs. Subsequently, the CVD-REAL 2 Study found reduced risks of MI and stroke in addition to reduced risk of HF hospitalization in patients with T2DM initiating SGLT-2 inhibitors versus other GLDs.6,28 More recent data from the CVD-REAL 2 study showed lower risk of hospitalization for HF and total mortality regardless of left ventricular ejection fraction status.29
The ongoing EMPRISE study has generated promising real-world evidence to support the benefit and impact of empagliflozin, complementing the findings of the EMPA-REG-OUTCOME trial. An interim analysis confirmed a substantial reduction in hospitalization for HF associated with the use of empagliflozin versus sitagliptin across a diverse patient population in routine clinical practice, including in patients with and without history of CVD.30 However, the study is not yet powered for subgroup analyses and only 7% of the patients in the propensity score-matched cohort had CKD. In light of the fact that many real-world observational studies were not able to examine the comparative effectiveness of SGLT-2 inhibitor versus other GLDs in patients with CKD, our study findings add clinically important and relevant data to the literature. Similar associations were found in a population-based cohort study among patients with T2DM and established CVD within the US Department of Defense Military Health System who newly initiated SGLT-2 inhibitors versus non-SGLT-2 inhibitors.31,32
Strengths of our study include large sample size, a new user design, clinically relevant endpoints and prescription information based on real-world data, adjustment for measured confounders using IPTW to minimize the impact of confounding by indication, handling of missing data using multiple imputation methods, stratified analysis across different stages of CKD, and analysis of nationally representative data for patients enrolled in employer-based insurance programs that may better reflect the real-world effectiveness of SGLT-2 inhibitors in routine practice.
Our study has some limitations. First, due to the observational nature of the study, we cannot exclude the possibility of residual confounding. While we adjusted for the likelihood of treatment, we could not capture all relevant covariates, and it is possible that one or more unmeasured covariates might have influenced the adjusted relative hazards calculated. Second, we had to truncate the study at 2.5 years, which may not be sufficient time for CV events to manifest themselves in a patient population with lower prevalence of established CVD or CV risk factors. Third, mortality data were not available in the claims database to examine associations with CV mortality or all-cause mortality, or to evaluate associations with traditional MACE which is usually defined as CV death or non-fatal MI or stroke. Fourth, in order to conserve sample size, we accounted for large amount of missing data on race in calculating eGFR through a sensitivity analysis where we assumed the patient was black then assuming he or she was non-black and taking the population-weighted average of the two. By doing this, we were able to reduce bias related to removing information available from patients with missing race, but we may have misclassified some patients whose eGFR was around the 60 mL/min/1.73m2 cutoff between stage 2 and 3a CKD. Similarly, we may have misclassified some patients with eGFR 60 or above with missing data on proteinuria or albuminuria. The sex-specific UACR cutoff we used in this study may be less relevant in routine practice; however, sensitivity analysis using a UACR cutoff of >30mg/g showed no material differences in the main findings. Lastly, our findings may not be generalizable to patients with other types of insurance or no insurance coverage with different underlying socioeconomic or other clinical and patient characteristics.
In this population-based cohort of patients with T2DM, initiation of SGLT-2 inhibitor was associated with lower risks of HF hospitalization, but not with non-fatal MI or stroke despite suggesting benefit, compared with initiation of DPP-4 inhibitor across different stages of CKD. Our findings suggest the real-world effectiveness of SGLT-2 inhibitors and complement data from RCTs as well as those of other observational studies of SGLT-2 inhibitors that did not specifically investigate strata defined by CKD. This study suggests that the SGLT-2 inhibitors have potential CV benefits compared with the DPP-4 inhibitor class when used in routine care of patients with T2DM with and without CKD.
Supplementary Material
ACKNOWLEDGMENTS
This work was supported by the National Institutes of Health grant numbers 5K01DK110221 (Dr. Rhee) and K24DK085446 (Dr. Chertow).
REFERENCES
- 1.Zinman B, Wanner C, Lachin JM, et al. Empagliflozin, Cardiovascular Outcomes, and Mortality in Type 2 Diabetes. N Engl J Med. 2015;373(22):2117–2128. [DOI] [PubMed] [Google Scholar]
- 2.Neal B, Perkovic V, Mahaffey KW, et al. Canagliflozin and Cardiovascular and Renal Events in Type 2 Diabetes. New England Journal of Medicine. 2017;377(7):644–657. [DOI] [PubMed] [Google Scholar]
- 3.Wiviott SD, Raz I, Bonaca MP, et al. Dapagliflozin and Cardiovascular Outcomes in Type 2 Diabetes. New England Journal of Medicine. 2018;380(4):347–357. [DOI] [PubMed] [Google Scholar]
- 4.Perkovic V, Jardine MJ, Neal B, et al. Canagliflozin and Renal Outcomes in Type 2 Diabetes and Nephropathy. N Engl J Med. 2019;380(24):2295–2306. [DOI] [PubMed] [Google Scholar]
- 5.Rhee JJ, Jardine MJ, Chertow GM, Mahaffey KW. Dedicated kidney disease-focused outcome trials with sodium-glucose cotransporter-2 inhibitors: Lessons from CREDENCE and expectations from DAPA-HF, DAPA-CKD, and EMPA-KIDNEY. Diabetes Obes Metab. 2020;22 Suppl 1:46–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Kosiborod M, Cavender MA, Fu AZ, et al. Lower Risk of Heart Failure and Death in Patients Initiated on Sodium-Glucose Cotransporter-2 Inhibitors Versus Other Glucose-Lowering Drugs: The CVD-REAL Study (Comparative Effectiveness of Cardiovascular Outcomes in New Users of Sodium-Glucose Cotransporter-2 Inhibitors). Circulation. 2017;136(3):249–259. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Kosiborod M, Lam CSP, Kohsaka S, et al. Cardiovascular events associated with SGLT-2 inhibitors versus other glucose-lowering drugs: the CVD-REAL 2 study. J Am Coll Cardiol. 2018;71(23):2628–2639. [DOI] [PubMed] [Google Scholar]
- 8.Birkeland KI, Jørgensen ME, Carstensen B, et al. Cardiovascular mortality and morbidity in patients with type 2 diabetes following initiation of sodium-glucose co-transporter-2 inhibitors versus other glucose-lowering drugs (CVD-REAL Nordic): a multinational observational analysis. Lancet Diabetes Endocrinol. 2017;5(9):709–717. [DOI] [PubMed] [Google Scholar]
- 9.Toulis KA, Willis BH, Marshall T, et al. All-Cause Mortality in Patients With Diabetes Under Treatment With Dapagliflozin: A Population-Based, Open-Cohort Study in The Health Improvement Network Database. J Clin Endocrinol Metab. 2017;102(5):1719–1725. [DOI] [PubMed] [Google Scholar]
- 10.Patorno E, Goldfine AB, Schneeweiss S, et al. Cardiovascular outcomes associated with canagliflozin versus other non-gliflozin antidiabetic drugs: population based cohort study. BMJ. 2018;360:k119. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Dawwas GK, Smith SM, Park H. Cardiovascular outcomes of sodium glucose cotransporter-2 inhibitors in patients with type 2 diabetes. Diabetes, Obesity and Metabolism. 2019;21(1):28–36. [DOI] [PubMed] [Google Scholar]
- 12.Rosenstock J, Perkovic V, Johansen OE, et al. Effect of Linagliptin vs Placebo on Major Cardiovascular Events in Adults With Type 2 Diabetes and High Cardiovascular and Renal Risk: The CARMELINA Randomized Clinical Trial. JAMA. 2019;321(1):69–79. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Green JB, Bethel MA, Armstrong PW, et al. Effect of Sitagliptin on Cardiovascular Outcomes in Type 2 Diabetes. N Engl J Med. 2015;373(3):232–242. [DOI] [PubMed] [Google Scholar]
- 14.Scirica BM, Bhatt DL, Braunwald E, et al. Saxagliptin and cardiovascular outcomes in patients with type 2 diabetes mellitus. N Engl J Med. 2013;369(14):1317–1326. [DOI] [PubMed] [Google Scholar]
- 15.White WB, Cannon CP, Heller SR, et al. Alogliptin after acute coronary syndrome in patients with type 2 diabetes. N Engl J Med. 2013;369(14):1327–1335. [DOI] [PubMed] [Google Scholar]
- 16.KDOQI Clinical Practice Guidelines for Chronic Kidney Disease: Evaluation C, and Stratification. http://www.kidney.org/professionals/kdoqi/guidlines_ckd/p4_class_g1.htm
- 17.Warram JH, Gearin G, Laffel L, et al. Effect of duration of type I diabetes on the prevalence of stages of diabetic nephropathy defined by urinary albumin/creatinine ratio. J Am Soc Nephrol. 1996;7(6):930–937. [DOI] [PubMed] [Google Scholar]
- 18.Kiyota Y, Schneeweiss S, Glynn RJ, et al. Accuracy of Medicare claims-based diagnosis of acute myocardial infarction: estimating positive predictive value on the basis of review of hospital records. Am Heart J. 2004;148(1):99–104. [DOI] [PubMed] [Google Scholar]
- 19.Wahl PM, Rodgers K, Schneeweiss S, et al. Validation of claims-based diagnostic and procedure codes for cardiovascular and gastrointestinal serious adverse events in a commercially-insured population. Pharmacoepidemiol Drug Saf. 2010;19(6):596–603. [DOI] [PubMed] [Google Scholar]
- 20.Tirschwell DL, Longstreth WT Jr., Validating administrative data in stroke research. Stroke. 2002;33(10):2465–2470. [DOI] [PubMed] [Google Scholar]
- 21.Saczynski JS, Andrade SE, Harrold LR, et al. A systematic review of validated methods for identifying heart failure using administrative data. Pharmacoepidemiol Drug Saf. 2012;21 Suppl 1(0 1):129–140. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med. 2009;28(25):3083–3107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Robins JM, Hernán MA, Brumback B. Marginal structural models and causal inference in epidemiology. Epidemiology. 2000;11(5):550–560. [DOI] [PubMed] [Google Scholar]
- 24.Cole SR, Hernán MA. Constructing inverse probability weights for marginal structural models. Am J Epidemiol. 2008;168(6):656–664. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Harder VS, Stuart EA, Anthony JC. Propensity score techniques and the assessment of measured covariate balance to test causal associations in psychological research. Psychol Methods. 2010;15(3):234–249. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Heerspink HJ, Stefánsson BV, Correa-Rotter R, et al. Dapagliflozin in patients with chronic kidney disease. N Engl J Med. 2020;383(15):1436–1446. [DOI] [PubMed] [Google Scholar]
- 27.Franklin JM, Patorno E, Desai RJ, et al. Emulating randomized clinical trials with nonrandomized real-world evidence studies: first results from the RCT DUPLICATE Initiative. Circulation. 2021;143(10):1002–1013. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Kosiborod M, Lam CSP, Kohsaka S, et al. Cardiovascular events associated with SGLT-2 Inhibitors versus other glucose-lowering drugs: The CVD-REAL 2 Study. J Am Coll Cardiol. 2018;71(23):2628–2639. [DOI] [PubMed] [Google Scholar]
- 29.Lam CSP, Karasik A, Melzer-Cohen C, et al. Association of sodium-glucose cotransporter-2 inhibitors with outcomes in type 2 diabetes with reduced and preserved left ventricular ejection fraction: Analysis from the CVD-REAL 2 study. Diabetes Obes Metab. 2021;23(6):1431–1435. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Patorno E, Pawar A, Franklin JM, et al. Empagliflozin and the risk of heart failure hospitalization in routine clinical care. Circulation. 2019;139(25):2822–2830. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Udell JA, Yuan Z, Rush T, et al. Cardiovascular outcomes and risks after initiation of a sodium glucose cotransporter 2 inhibitor: results from the EASEL population-based cohort study (Evidence for Cardiovascular Outcomes With Sodium Glucose Cotransporter 2 Inhibitors in the Real World). Circulation. 2018;137(14):1450–1459. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Udell JA, Yuan Z, Ryan P, et al. Cardiovascular outcomes and mortality after initiation of canagliflozin: Analyses from the EASEL Study. Endocrinol Diabetes Metab. 2020;3(1):e00096. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.