Document resource
WHAT IS ALREADY KNOWN ON THIS TOPIC Incretin-based therapies such as dipeptidyl peptidase-4 inhibitors (DPP4is) or glucagon-like peptide-1 receptor agonists (GLP1RAs) have anti-inflammatory and anti-infectious properties and have been hypothesized to modify risk of COVID-19 infection and outcomes among persons with type 2 diabetes.No clinical studies have examined the risk of COVID-19 infection among DPP4i or GLP1RA users.DPP4i or GLP1RA use has not been associated with reduced risk of mortality from COVID-19 in previous cohort studies, but there is less evidence on DPP4i or GLP1RA use and risk of clinical outcomes, such as hospitalizations or cardiovascular events.WHAT THIS STUDY ADDS Our study suggests DPP4i or GLP1RA use is not associated with a reduced risk of COVID-19 infection compared with use of sodium-glucose cotransporter-2 inhibitors or sulfonylureas.Among persons with COVID-19, DPP4i or GLP1RA use was associated with a lower risk of hospitalization at 30 days.HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY In the absence of any randomized controlled trial data, our study adds to the existing body of evidence on how DPP4i or GLP1RAs associate with COVID-19 infection and outcomes.Introduction Persons with type 2 diabetes (T2DM) have an elevated risk of complications and mortality from COVID-19, while the presence of cardiovascular disease (CVD) has been associated with adverse outcomes and mortality related to COVID-19 among those with T2DM. 1 2 Treatment with glucose-lowering medications has been suggested to decrease both the risk of COVID-19 infection and of complications from COVID-19.3 Dipeptidyl peptidase-4 (DPP4) inhibitors (DPP4i) and glucagon-like peptide-1 receptor agonists (GLP1RAs), are both incretin-based therapies that are of particular interest given their possible anti-infectious and anti-inflammatory properties.4 DPP4 is a functional coronavirus receptor, thus it has been suggested that DPP4 inhibition might interfere with coronavirus binding.4 GLP1RAs have anti-inflammatory properties and in preclinical studies have been shown to reduce pulmonary inflammation and preserve lung function in lung injury.4 However, clinical evidence to date has been inconsistent. A 2021 systematic review of 11 observational studies noted conflicting results regarding the effect of DPP4i and COVID-19-related mortality, finding a neutral effect for DPP4i use versus non-use on the risk of COVID-19-related mortality.5 Evidence surrounding DPP4i use has not been clarified by studies conducted since this meta-analysis.6–8 A 2021 systematic review of nine observational studies found that GLP1RA use among people with COVID-19 was associated with reduced mortality,9 though a large UK cohort study found no benefit of GLP1RA for the outcome of COVID-19-related mortality.10Existing studies on DPP4i and GLP1RA use in COVID-19 focus primarily on mortality. Given the potential risk of increased cardiovascular (CV) events among those with T2DM infected with COVID-19, further evidence on CV outcomes would clarify the role of DPP4i and GLP1RA in the context of COVID-19 and may be useful to clinicians to guide therapy decisions in practice. The aim of this study was to examine the association between DPP4i or GLP1RA and COVID-19 infection, as well as subsequent adverse outcomes compared with sodium-glucose cotransporter-2 inhibitors (SGLT2is) or sulfonylureas (SUs) use among persons with T2DM.Methods We conducted a population-based cohort study in Ontario, Canada using linked healthcare administrative data among persons with T2DM ≥66 years of age.Data sources We conducted this study using healthcare administrative datasets at ICES, an independent, non-profit research institute whose legal status under Ontario’s health information privacy law allows it to collect and analyze healthcare and demographic data, without consent, for health system evaluation and improvement. Information on diabetes status was obtained from the Ontario Diabetes Database (ODD). 11 Sociodemographic data were obtained from the Ontario Registered Persons Database. Information on diagnoses was obtained from the Canadian Institute for Health Information (CIHI) Discharge Abstract Database (DAD), CIHI National Ambulatory Care Reporting System (NACRS), and Ontario Health Insurance Plan Physician Claims database. Laboratory data were from the Ontario Laboratory Information System (OLIS). COVID-19 data were obtained from the integrated COVID dataset (C19INTGR) which captures COVID-19 testing episodes in Ontario from OLIS. Frailty status was calculated using the Hospital Frailty Risk Score.12 Prescription drug use data were from the Ontario Drug Benefit Database, which provides dispensed prescription data for medications on the provincial formulary in Ontario for those ≥65 years of age. Outcome data were obtained from C19INTGR, DAD, and NACRS. A detailed description of datasets used, and definitions, is in online supplemental eTable 1. Datasets were linked using unique encoded identifiers and analyzed at ICES.SP110.1136/bmjdrc-2024-004677.supp1Supplementary dataPopulation The study population included all persons residing in Ontario ≥66 years of age with T2DM as of January 1, 2020, who had a COVID-19 PCR test between January 15, 2020 and July 26, 2021. We used this age cut-off because prescription drug data are only available for persons ≥65 years of age in Ontario (we allowed people a 1 year period to enroll in the program). The index date for our analysis was the date of the first COVID-19 test a person had during the study period. A person was considered to have T2DM if they were listed in the ODD 11 and were ≥30 years when they were diagnosed with diabetes. All individuals in the cohort were required to be users of metformin as of the index date (defined as filling ≥1 prescription for metformin in the 100-day period leading up to index date). We restricted the cohort to people on metformin to minimize selection bias and confounding by severity/prognosis (as patients already on metformin would be more likely to be at a similar course in their T2DM trajectory compared with those not on metformin). For individuals with multiple COVID-19 tests, we excluded all tests after the first test (ie, one record per patient; the first test a person received). We also excluded persons who received a combination of DPP4i and SGLT2i/SU or GLP1RA and SGLT2i/SU based on a 100-day lookback from the index date. We evaluated outcomes in the overall cohort. We also conducted a preplanned stratified analysis in two groups: a cohort with a history of either coronary artery disease (CAD), heart failure (HF), stroke or chronic kidney disease (CKD), and a cohort without any of those comorbidities.Exposures The exposure of interest was prevalent use of DPP4i or GLP1RA on the index date. These drugs were grouped together given that they are both incretin-based therapies that are alleged to exert anti-infective and anti-inflammatory effects and potentially modify the course of COVID-19 infection. 4 The comparator was prevalent use of SGLT2i or SU on the index date. We chose to group SGLT2i/SU together as a comparator to minimize selection bias and increase statistical power. SGLT2i and SU may be used at a similar point in the T2DM treatment pathway according to Canadian guidelines, and may each be used at a similar point to DPP4i or GLP1RA.13 We classified a person as using a medication if they filled ≥1 prescription for a medication in the 100 days leading up to the index date.Outcomes The primary outcome was COVID-19 infection, defined as a positive test in the C19INTGR dataset. Secondary outcomes were clinical outcomes among persons with a positive/presumptive positive test from the primary analysis. Persons with a positive test were followed for 30 days after the test to evaluate secondary outcomes, which included: all-cause hospitalization, intensive care unit admission, mechanical ventilation, CV event (composite of hospitalization for myocardial infarction, HF, stroke, myocarditis, pericarditis, arrhythmia, need for pacemaker or implantable cardioverter defibrillator insertion, resuscitated cardiac arrest), venous thromboembolism, all-cause mortality, and a composite of a CV event or death. Outcome definitions are in online supplemental eTable 1.Analysis We described baseline characteristics using descriptive statistics. We then compared primary and secondary outcomes between DPP4i/GLP1RA users and SGLT2i/SU users. We reported crude and weighted risk differences (RDs), relative risks (RRs), and corresponding 95% CI. For adjusted analyses, we first calculated a propensity score based on the following confounders: age, sex, rurality (small, medium, large community size), neighborhood income quintile, public health unit, immigration status (immigrated to Ontario since 1985), hypertension, mean years of hypertension among those with hypertension, mean years of diabetes, CAD, HF, stroke, atrial fibrillation, CKD, liver disease, cancer, any history of organ transplant, HIV, lung disease, frailty, serum creatinine (SCr), hemoglobin, white blood cell count, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, hemoglobin A1c, statin use (≥1 prescription in 100 days before index date), and insulin use (≥1 prescription in 100 days before index date). For laboratory tests, we used the most recent test since January 1, 2015. See online supplemental eTable 1 for all definitions. We then used inverse probability of treatment weighting (IPTW) with stabilized weights to create a pseudo-population in which drug treatment (ie, DPP4i/GLP1RA vs SGLT2i/SU) was independent of measured confounders. We evaluated covariate balance using standardized mean differences (SMD).14 After ensuring groups were balanced (SMD <0.1) in terms of measured confounders, we calculated the weighted RD and RR. We used this analytic approach for the overall cohort and in our analysis stratifying by history of CAD, HF, CKD, or stroke.In pre-planned sensitivity analyses, we evaluated the primary and secondary outcomes restricting the exposure group to users of DPP4i only or users of GLP1RA only. We also performed an analysis where we restricted the comparator group to SGLT2i use only. We conducted a pre-specified stratified analysis according to presence or absence of CAD, HF, stroke, or CKD. In a post-hoc analysis, we conducted a statistical comparison of clinical outcomes between those without CAD, HF, stroke, or CKD and those with these conditions using a Z test comparing the weighted RD of these two subgroups (at the 5% significance level). We re-estimated the propensity score and repeated the IPTW procedure to rebalance covariates for the main analysis, subgroup analyses, and sensitivity analyses. We also calculated the E value for all estimates.15Permissions This dataset was created at ICES under section 45 of Ontario’s Personal Health Information Protection Act, which did not require Research Ethics Board (REB) review (under section 45 ICES studies are legally exempt from having to apply to an REB). Patients or the public were not involved in the design, conduct, reporting, or dissemination of our research. Data used for this study are not publicly available.Results We identified 5,504,276 people with a COVID-19 test over the study period, of which 226,131 were ≥65 years old with T2DM. A total of 81,332 persons were taking metformin and thus were eligible to be included in our cohort. There were 26,485 users of DPP4i/GLP1RA (mean age 76, 47% female, 1,829 were GLP1RA users), 14,487 users of SGLT2i/SU (mean age 75, 39% female), and 40,360 persons who used neither DPP4i/GLP1RA nor SGLT2i/SU (mean age 77, 49% female). The rate of insulin use was 25% in the DPP4i/GLP1RA group compared with 20% in the SGLT2i/SU group. The rate of statin use was 80% in the DPP4i/GLP1A group compared with 83% in the SGLT2i/SU group. Detailed baseline characteristics are in table 1. Covariate balance results are in online supplemental eTable 2.Table 1Baseline characteristics before weightingCharacteristicTotalDPP4i/GLP1RASGLT2i/SUN=81,332*N=26,485N=14,487Age Median (Q1–Q3)75 (70–82)75 (70–82)73 (69–79)Sex Male, n (%)43,258 (53.2)13,931 (52.6)8,904 (61.5)Income quintile, n (%) 120,036 (24.6)6,724 (25.4)3,319 (22.9) 217,962 (22.1)5,852 (22.1)3,204 (22.1) 316,347 (20.1)5,338 (20.2)2,966 (20.5) 414,212 (17.5)4,493 (17.0)2,638 (18.2) 512,775 (15.7)4,078 (15.4)2,360 (16.3)Rurality, n (%) Large urban (>100K people)64,262 (79.0)21,388 (80.8)10,987 (75.8) Medium urban (10–100K)7,703 (9.5)2,402 (9.1)1,486 (10.3) Small town (<10K)9,367 (11.5)2,695 (10.2)2,014 (13.9)Recent immigrant†, n (%)10,852 (13.3)3,757 (14.2)1,890 (13.0)Medical conditions, n (%) CAD14,075 (17.3)4,314 (16.3)3,356 (23.2) HF4,827 (5.9)1,763 (6.7)831 (5.7) Stroke2,323 (2.9)884 (3.3)327 (2.3) CKD12,432 (15.3)5,160 (19.5)2,020 (13.9) Hypertension69,665 (85.7)22,794 (86.1)12,269 (84.7) Previous cancer5,293 (6.5)1,773 (6.7)878 (6.1) Atrial fibrillation10,106 (12.4)3,353 (12.7)1,709 (11.8) Lung disease26,839 (33.0)8,953 (33.8)4,347 (30.0) Liver disease, none–mild438 (0.5)165 (0.6)71 (0.5) Liver disease, moderate–severe250 (0.3)102 (0.4)43 (0.3) HIV91 (0.1)29 (0.1)18 (0.1) Transplant recipient174 (0.2)77 (0.3)22 (0.2)Frailty score, n (%) 0–523,262 (28.6)7,296 (27.5)4,499 (31.1) 6–1510,183 (12.5)3,878 (14.6)1,337 (9.2) >153,956 (4.9)1,755 (6.6)358 (2.5) Missing43,931 (54.0)13,556 (51.2)8,293 (57.2)Hypertension duration, mean (SD)18.9 (8.0)19.0 (8.1)18.6 (8.0)Diabetes duration, mean (SD)13.7 (7.6)15.0 (7.6)14.9 (7.6)Medication use, n (%) Statins63,980 (78.7)21,097 (79.7)11,969 (82.6) Insulin14,108 (17.3)6,619 (25.0)2,908 (20.1) DPP4i–24,214 (91.0)– GLP1RA–1,829 (6.9)– SGLT2i––9,393 (64.8) SU––5,094 (35.2)Laboratory values, mean (SD) LDL (mmol/L)1.70 (0.79)1.65 (0.78)1.65 (0.76) HDL (mmol/L)1.23 (0.36)1.20 (0.35)1.19 (0.34) SCr (mmol/L)90.1 (34.8)93.4 (38.2)90.0 (32.5) Hemoglobin (g/L)129 (18)127 (18)134 (18) HbA1c (%)6.99 (1.15)7.15 (1.24)7.27 (1.22)*Includes those not on DPP4i, GLP1RA, SGLT2i, or SU.†Immigrated to Canada 1985 or later.CAD, coronary artery disease; CKD, chronic kidney disease; DPP4i, dipeptidyl peptidase 4 inhibitor; GLP1RA, glucagon like peptide 1 receptor agonist; HbA1c, glycated hemoglobin; HDL, high-density lipoprotein; HF, heart failure; LDL, low-density lipoprotein; SCr, serum creatinine; SGLT2i, sodium glucose cotransporter 2 inhibitor; SU, sulfonylurea.COVID-19 infection COVID-19 infection results are shown in table 2. In the overall cohort, there was no difference in the risk of infection between groups. The crude rate of infection in the DPP4i/GLP1RA group was 10% (2,733/26,485) compared with 9.2% in the SGLT2i/SU group (1,332/14,487). The weighted RD was −0.06% (95% CI −0.79% to 0.66%) and the weighted RR was 0.99 (95% CI 0.93 to 1.07). In stratified analysis, DPP4i/GLP1RA use was associated with a higher risk of COVID-19 infection compared with SGLT2i/SU use among persons with a history of either CAD, HF, CKD, or stroke (weighted RD 1.33%, 95% CI 0.07% to 2.59% and weighted RR 1.15, 95% CI 1.00 to 1.31), while there was no association among persons without CAD, HF, CKD, or stroke. Sensitivity analyses were consistent with results from the main analysis.Table 2COVID-19 infection rates for DPP4i/GLP1RA users compared with SGLT2i/SU usersGroupPositive tests (n)Tested (N)Crude rate (95% CI)Weighted rate (95% CI)Weighted RD, % (95% CI)Crude RR (95% CI)Weighted RR (95% CI)DPP4i/GLP1RA2,73326,48510.3% (9.96% to 10.7%)10.3% (9.96% to 10.7%)−0.06 (−0.79 to 0.66)1.12 (1.05 to 1.19)0.99 (CI 0.93 to 1.07)SGLT2i/SU1,33214,4879.19% (8.74% to 9.68%)10.4% (9.77% to 11.0%)RefRefRefNo history of CAD, HF, CKD, or strokeDPP4i/GLP1RA1,78117,30110.3%10.3−0.86 (−1.75 to 0.04)1.04 (0.97 to 1.12)0.92 (0.85 to 1.00)SGLT2i/SU9209,3139.88%11.1RefRefRefHistory of CAD, HF, CKD, or strokeDPP4i/GLP1RA9529,18410.4% (9.76% to 11.0%)10.4% (9.76% to 11.0%)1.33 (0.07 to 2.59)1.30 (1.17 to 1.45)1.15 (1.00 to 1.31)SGLT2i/SU4125,1747.96% (7.26% to 8.74%)9.03% (8.00% to 10.2%)RefRefRefRR weighted using inverse probability treatment weighting, adjusting for: age, sex, rural/urban residency, public health unit, neighborhood income quintile, immigration status, number of hospitalizations in the prior year, hypertension, duration of hypertension, duration of diabetes, HIV, history of organ transplant, cancer, arrhythmia, coronary artery disease, chronic kidney disease, heart failure, stroke, liver disease, history of lung disease, insulin therapy, statin therapy, frailty, SCr, LDL, HDL, HbA1c, hemoglobin, WBC count (coronary artery disease, chronic kidney disease, heart failure, stroke not used in subgroup analyses).Bolded values indicate statistically significant estimates.CAD, coronary artery disease; CKD, chronic kidney disease; DPP4i, dipeptidyl peptidase 4 inhibitor; GLP1RA, glucagon like peptide 1 receptor agonist; HbA1c, glycated hemoglobin; HDL, high-density lipoprotein; HF, heart failure; LDL, low-density lipoprotein; RD, risk difference; RR, relative risk; SCr, serum creatinine; SGLT2i, sodium glucose cotransporter 2 inhibitor; SU, sulfonylurea; WBC, white blood cell.Clinical outcomes among those testing positive for COVID-19 The 30-day clinical outcome results are shown in table 3. DPP4i/GLP1RA use was associated with lower risk of hospitalization (weighted RD −6.72%, 95% CI −10.4% to −3.02% and weighted RR 0.79, 95% CI 0.70 to 0.89) compared with SGLT2i/SU use. There was no difference in the associated risk of any other outcomes but there was a trend towards a lower associated risk of CV events in the DPP4i/GLP1RA group (weighted RD −1.91%, 95% CI −4.00% to 0.18%, and RR 0.73, 95% CI 0.54 to 1.00).Table 3Clinical outcomes at 30 days for DPP4i/GLP1RA users compared with SGLT2i/SU users testing positive for COVID-19OutcomeCrude outcome rate, n (%)EstimatesDPP4i/GLP1RASGLT2i/SUWeighted RD, % (95% CI)Crude RR (95% CI)Weighted RR (95% CI)N=2,733N=1,332–––All-cause hospitalization695 (25.5)370 (27.8)−6.72 (−10.4 to –3.02)0.92 (0.82 to 1.02)0.79 (0.70 to 0.89)ICU admission148 (5.4)90 (6.8)−1.06 (−2.89 to 0.77)0.80 (0.62 to 1.03)0.84 (0.62 to 1.12)Mechanical ventilation87 (3.2)56 (4.2)−0.52 (−1.87 to 0.84)0.76 (0.54 to 1.05)0.86 (0.59 to 1.26)CV events*144 (5.3)78 (5.9)−1.91 (−4.00 to 0.18)0.90 (0.69 to 1.18)0.73 (0.54 to 1.00) MI hospitalization25 (0.9)21 (1.6)−0.54 (−1.34 to 0.26)0.58 (0.33 to 1.03)0.63 (0.33 to 1.18) HF hospitalization60 (2.2)33 (2.5)−1.15 (−2.56 to 0.25)0.89 (0.58 to 1.35)0.66 (0.41 to 1.04) Stroke hospitalization12 (0.4)9 (0.7)−0.38 (−1.03 to 0.26)0.65 (0.27 to 1.54)0.53 (0.21 to 1.34) Arrhythmia hospitalization65 (2.4)30 (2.3)−0.36 (−1.73 to 1.01)1.06 (0.69 to 1.62)0.87 (0.52 to 1.45) Cardiac arrest13 (0.5)8 (0.6)−0.09 (−0.61 to 0.43)0.79 (0.33 to 1.91)0.84 (0.32 to 2.20)VTE32 (1.2)15 (1.1)0.03 (−0.73 to 0.78)1.04 (0.57 to 1.91)1.02 (0.53 to 1.98)Death165 (6.0)72 (5.4)−0.03 (−1.93 to 1.86)1.12 (0.85 to 1.46)0.99 (0.73 to 1.36)CV event/death287 (10.5)137 (10.3)−1.50 (−4.09 to 1.09)1.02 (0.84 to 1.24)0.87 (0.70 to 1.09)RR weighted using inverse probability treatment weighting, adjusting for: age, sex, rural/urban residency, public health unit, neighborhood income quintile, immigration status, number of hospitalizations in the prior year, hypertension, duration of hypertension, duration of diabetes, HIV, history of organ transplant, cancer, arrhythmia, coronary artery disease, chronic kidney disease, heart failure, stroke, liver disease, history of lung disease, insulin therapy, statin therapy, frailty, SCr, LDL, HDL, HbA1c, hemoglobin, WBC count.Bolded values indicate statistically significant estimates.*Myocarditis, pacemaker insertion contributed to outcome but not listed in table since ≤5 events per outcome.CV, cardiovascular; DPP4i, dipeptidyl peptidase 4 inhibitor; GLP1RA, glucagon like peptide 1 receptor agonist; HbA1c, glycated hemoglobin; HDL, high-density lipoprotein; HF, heart failure; ICU, intensive care unit; LDL, low-density lipoprotein; MI, myocardial infarction; RD, risk difference; RR, relative risk; SCr, serum creatinine; SGLT2i, sodium glucose cotransporter 2 inhibitor; SU, sulfonylurea; VTE, venous thromboembolism; WBC, white blood cell.Stratified analysis results for clinical outcomes are reported in tables 4 and 5. Among persons without CAD, HF, CKD, or stroke, DPP4i/GLP1RA use was associated with lower risk of mechanical ventilation (weighted RD −1.77%, 95% CI −3.69% to 0.15% and weighted RR 0.63, 95% CI 0.40 to 0.99) and all-cause hospitalization (weighted RD −5.65, 95% CI −9.74 to 1.56 and weighted RR 0.80, 95% CI 0.69 to 0.93) compared with SGLT2i/SU use. There was no difference in the associated risk of any other outcomes. Among persons with a history of CAD, HF, CKD, or stroke, the associated risk of all-cause hospitalization was lower in the DPP4i/GLP1RA group (weighted RD −9.78%, 95% CI −18.1% to −1.48% and weighted RR 0.76, 95% CI 0.61 to 0.94) while the associated risk of mechanical ventilation was higher (weighted RD 2.05%, 95% CI 0.51% to 3.59% and weighted RR 2.44, 95% CI 1.11 to 5.38).Table 4Clinical outcomes at 30 days among COVID-19 positive patients without CAD, HF, stroke, or CKD, comparing DPP4i/GLP1RA users to SGLT2i/SU usersOutcomeCrude outcome rate, n(%)EstimatesDPP4i/GLP1RASGLT2i/SUWeighted RD, % (95% CI)Crude RR (95% CI)Weighted RR (95% CI)N=1,781N=920–––All-cause hospitalization407 (22.9)230 (25.0)−5.65 (−9.74 to 1.56)0.91 (0.70 to 1.05)0.80 (0.69 to 0.93)ICU admission90 (5.1)60 (6.5)−1.20 (−3.35 to 0.96)0.77 (0.56 to 1.06)0.81 (0.56 to 1.16)Mechanical ventilation54 (3.0)44 (4.8)−1.77 (−3.69 to 0.15)0.63 (0.43 to 0.94)0.63 (0.40 to 0.99)CV events*74 (4.2)41 (4.5)−0.77 (−2.87 to 1.33)0.93 (0.64 to 1.35)0.84 (0.54 to 1.31) MI hospitalization14 (0.8)9 (1.0)−0.03 (−0.74 to 0.67)0.80 (0.35 to 1.85)0.96 (0.40 to 2.29) HF hospitalization22 (1.2)16 (1.7)−0.90 (−2.18 to 0.38)0.71 (0.37 to 1.35)0.58 (0.29 to 1.15) Arrhythmia hospitalization42 (2.4)17 (1.8)0.25 (−1.35 to 1.84)1.28 (0.73 to 2.23)1.12 (0.53 to 2.54) Cardiac arrest6 (0.3)7 (0.8)−0.34 (−0.94 to 0.27)0.44 (0.15 to 1.31)0.50 (0.16 to 1.56)VTE21 (1.2)10 (1.1)0.03 (−0.89 to 0.95)1.08 (0.51 to 2.29)1.03 (0.46 to 2.28)Death87 (4.9)46 (5.0)−0.82 (−3.01 to 1.37)0.98 (0.69 to 1.38)0.86 (0.57 to 1.28)CV event/death151 (8.5)78 (8.5)−0.66 (−3.31 to 1.99)1.00 (0.77 to 1.30)0.93 (0.69 to 1.25)RR weighted using inverse probability treatment weighting, adjusting for: age, sex, rural/urban residency, public health unit, neighborhood income quintile, immigration status, number of hospitalizations in the prior year, hypertension, duration of hypertension, duration of diabetes, HIV, history of organ transplant, cancer, arrhythmia, liver disease, history of lung disease, insulin therapy, statin therapy, frailty, SCr, LDL, HDL, HbA1c, hemoglobin, WBC count.Bolded values indicate statistically significant estimates.*Myocarditis, pacemaker insertion, stroke hospitalization contributed to outcome but not listed in table since ≤5 events per outcome.CAD, coronary artery disease; CKD, chronic kidney disease; CV, cardiovascular; DPP4i, dipeptidyl peptidase 4 inhibitor; GLP1RA, glucagon like peptide 1 receptor agonist; HbA1c, glycated hemoglobin; HDL, high-density lipoprotein; HF, heart failure; ICU, intensive care unit; LDL, low-density lipoprotein; MI, myocardial infarction; RD, risk difference; RR, relative risk; SCr, serum creatinine; SGLT2i, sodium glucose cotransporter 2 inhibitor; SU, sulfonylurea; VTE, venous thromboembolism; WBC, white blood cell.Table 5Clinical outcomes at 30 days among COVID-19 positive patients with CAD, HF, stroke, or CKD, comparing DPP4i/GLP1RA users to SGLT2i/SU usersOutcomeCrude outcome rate, n(%)EstimatesDPP4i/GLP1RASGLT2i/SUWeighted RD, % (95% CI)Crude RR (95% CI)Weighted RR (95% CI)N=952N=412––All-cause hospitalization288 (30.3)140 (34.0)−9.78 (−18.1 to −1.48)0.89 (0.75 to 1.05)0.76 (0.61 to 0.94)ICU admission58 (6.1)30 (7.3)−1.86 (−6.16 to 2.43)0.84 (0.55 to 1.28)0.77 (0.44 to 1.35)Mechanical ventilation33 (3.5)12 (2.9)2.05 (0.51 to 3.59)1.19 (0.62 to 2.28)2.44 (1.11 to 5.38)CV events*70 (7.4)37 (9.0)−4.21 (−9.42 to 1.00)0.82 (0.56 to 1.20)0.64 (0.39 to 1.03) MI hospitalization11 (1.2)12 (2.9)−0.89 (−2.45 to 0.66)0.40 (0.18 to 0.89)0.56 (0.23 to 1.39) HF hospitalization38 (4.0)17 (4.1)−1.63 (−5.05 to 1.79)0.97 (0.55 to 1.69)0.71 (0.37 to 1.35) Arrhythmia hospitalization23 (2.4)13 (3.2)−1.31 (−4.38 to 1.76)0.77 (0.39 to 1.50)0.65 (0.27 to 1.56)VTE11 (1.2)≤5–––Death78 (8.2)26 (6.3)0.84 (−3.40 to 5.07)1.30 (0.85 to 1.99)1.11 (0.63 to 1.96)CV event/death136 (14.3)59 (14.3)−3.92 (−10.3 to 2.41)1.00 (0.75 to 1.32)0.78 (0.55 to 1.13)RR weighted using inverse probability treatment weighting, adjusting for: age, sex, rural/urban residency, public health unit, neighborhood income quintile, immigration status, number of hospitalizations in the prior year, hypertension, duration of hypertension, duration of diabetes, HIV, history of organ transplant, cancer, arrhythmia, liver disease, history of lung disease, insulin therapy, statin therapy, frailty, SCr, LDL, HDL, HbA1c, hemoglobin, WBC count.Bolded values indicate statistically significant estimates.*Myocarditis, pacemaker insertion, stroke hospitalization, cardiac arrest contributed to outcome but not listed in table since ≤5 events per outcome.CAD, coronary artery disease; CKD, chronic kidney disease; CV, cardiovascular; DPP4i, dipeptidyl peptidase 4 inhibitor; GLP1RA, glucagon like peptide 1 receptor agonist; HbA1c, glycated hemoglobin; HDL, high-density lipoprotein; HF, heart failure; ICU, intensive care unit; LDL, low-density lipoprotein; MI, myocardial infarction; RD, risk difference; RR, relative risk; SCr, serum creatinine; SGLT2i, sodium glucose cotransporter 2 inhibitor; SU, sulfonylurea; VTE, venous thromboembolism; WBC, white blood cell.Supplementary analyses In our post-hoc sensitivity analysis, there was no statistical difference in the associated risk of any clinical outcomes for those with CAD, HF, CKD or stroke, compared with those without, except for mechanical ventilation ( online supplemental eTable 3). E values for clinical outcomes are in online supplemental eTable 4 (the E value is the “minimum strength of association an unmeasured confounder needs to have with a treatment and outcome to fully explain away a treatment-outcome association”).15 Sensitivity analyses were generally consistent with the main analyses. Restricting the exposure to GLP1RA use only resulted in low event rates and imprecise estimates (online supplemental eTable 5); point estimates were lower than the main analysis for all outcomes, but the 95% CIs overlapped. Restricting the comparison group to SGLT2i users alone did not substantially change results (online supplemental eTable 6).Discussion Summary of findings In this population-based cohort study, GLP1RA/DPP4i use was not associated with COVID-19 infection, though in a stratified analysis among patients with a history of CAD, HF, stroke, or CKD, the associated risk of infection was higher in the GLP1RA/DPP4i group compared with SGLT2i/SU. Among those who tested positive for COVID-19, GLP1RA/DPP4i use was associated with a reduced risk of all-cause hospitalization, and there was a possible trend towards a lower associated risk of CV events. Point estimates for remaining outcomes were primarily in the direction of benefit for GLP1RA/DPP4i. Since our GLP1RA/DPP4i cohort was comprised primarily of DPP4i users, our findings are most applicable to persons using DPP4i.Comparison to existing literature Preclinical studies show GLP1RAs have anti-inflammatory properties, reduce pulmonary inflammation and preserve lung function in lung injury. 4 There is a lack of clinical evidence on the effect of DPP4i/GLP1RA on risk of COVID-19 infection. While there has been a suggestion that DPP4i or GLP1RA might prevent COVID-19 infection based primarily on animal studies, the clinical relevance has been debated.4 To our knowledge, no clinical studies have examined the role of DPP4i/GLP1RA on COVID-19 risk. Our findings do not support the hypothesis that DPP4i or GLP1RA reduce risk of COVID-19 infection. Evidence on the effect of DPP4i and GLP1RA on COVID-19 outcomes is inconsistent. Systematic review and meta-analyses of observational studies point to conflicting findings surrounding the association between DPP4i use and mortality among those with COVID-19, with two meta-analyses demonstrating no association between DPP4i use and mortality,5 16 and another finding reduced odds of mortality with DPP4i use, though with a wide 95% CI reaching 0.99.17 Observational studies on DPP4i are generally limited by small sample sizes (eg, three available meta-analyses of observational studies have a pooled totals of 1,933 patients,17 7,012 patients,16 and 6,022 patients)5 and differences in comparator groups, which make it challenging to compare results. A large cohort study (2,851,465 persons) conducted in the UK published since these systematic reviews found no association between DPP4i use and mortality. Our DPP4i/GLP1RA group consisted mostly of DPP4i users and our findings are consistent with existing evidence on the association between DPP4i and mortality.10There is a paucity of evidence on the association between DPP4i and outcomes such as CV events and hospitalizations in the context of COVID-19 infection. Our study adds to existing literature by demonstrating that DPP4i/GLP1RA use is associated with reduced risk of hospitalizations at 30 days among persons testing positive for COVID-19, while there was a possible trend towards a lower risk of CV events. Our findings should be contrasted against those of landmark DPP4i trials such as Saxagliptin Assessment of Vascular Outcomes Recorded in Patients-Thrombolysis in Myocardial Infarction (SAVOR-TIMI) 53, Trial Evaluating Cardiovascular Outcomes with Sitagliptin (TECOS), and Examination of Cardiovascular Outcomes with Algogliptin Versus Standard of Care (EXAMINE), which showed DPP4i do not lower risk of CV events among persons with T2DM.18–20 Our study was limited to persons with COVID-19 and only followed individuals for 30 days to examine CV outcomes. The role of DPP4i in mitigating CV events among COVID-19 patients has not been extensively examined in clinical studies to date, but has been hypothesized to be due to DPP4i potentially inhibiting myocardial inflammation caused by SARS-CoV-2.4 21 Thus, if there is any benefit of DPP4i in lowering risk of CV events, it may be confined to a short-term period around COVID-19 infection rather than a long-term reduction in risk (which would not be expected based on landmark DPP4i trials). In addition to a paucity of evidence on DPP4i/GLP1RA use and CV events, there has been little examination of whether there is any difference in outcomes among those with CAD, stroke, HF, and CKD in users of these medications. A large UK cohort study10 demonstrated no association between DPP4i and mortality among those without CVD (HR 0.93, 95% CI 0.84 to 1.03), but a small increased risk of mortality (HR 1.21, 95% CI 1.10 to 1.33) among those with CVD.Existing studies also suggest inconsistent associations between GLP1RA use and all-cause mortality in COVID-19 positive patients.9 10 One systematic review of nine studies found an association between GLP1RA use and reduced odds of mortality among patients with COVID-199 but a large UK cohort study published since that systematic review found no association.10 The relatively low GLP1RA use in our cohort and low event rates in GLP1RA users suggest more data are required before definitive conclusions can be made.Limitations There is potential for unmeasured confounding in our study. We did not have data on body mass index, smoking behavior, cardiac function, and did not account for specialist care. 5 Patients on SGLT2i may have had worse cardiac function as these medications are indicated in the setting of HF. While we did adjust for SCr, we did not adjust for estimated glomerular filtration (eGFR). While we did adjust for SCr, we did not adjust for eGFR. Since advanced CKD is associated with worse outcomes, disparity in eGFR between groups might have led to bias in estimation of the benefits of DPP4i/GLP1RA in terms of mortality, hospitalization, and CV events. We did not account for the sequence of T2DM medications, although we did adjust for duration of T2DM. Since the exposure and comparator have the same indications, we do not suspect this introduced bias. We did not account for the duration of T2DM medication use. It is possible that differences in duration of medication use could introduce bias due to confounding. Our study population was older adults who are at higher risk of adverse outcomes compared with younger adults. Our findings may not be generalizable to younger adults with T2DM. The mean glycated hemoglobin (HbA1c) in our cohort was 7.2% and our findings may not be generalizable to individuals with higher HbA1c levels. Further, given increasing uptake of vaccines since our follow-up, and viral mutagenicity over time, our findings may not be generalizable in the current context of COVID-19. The reason for the difference in the mechanical ventilation outcome is unclear. It may be a marker of persons with CAD, stroke, HF, or CKD having worse COVID-19 outcomes compared with those without22 23 or type I error. We examined all-cause hospitalization as an outcome. It is possible the difference in risk of hospitalization was not necessarily due to COVID-19. However, SGLT2i have been suggested to reduce risk of all-cause hospitalization compared with DPP4i24 and SU have been suggested to carry similar risk of all-cause hospitalization.25 Our study spanned the first two waves of the COVID-19 pandemic in Ontario. There were differences in practice throughout these waves (eg, testing policies, masking policies). However, we do not expect that such differences would be different between the two groups. We combined GLP1RA and DPP4i users into one category and planned to examine outcomes for each individual agent in a sensitivity analysis. However, given only one GLP1RA is covered by Ontario’s pharmacare program, the rate of use of GLP1RA in our study was low (in the sensitivity analysis examining GLP1RA use only, outcome estimates were imprecise while analyzing DPP4i alone did not change findings). It is also possible that people may have had private insurance that covered a GLP1RA which we could not capture. We examined a number of clinical outcomes and included analyses in different subgroups, thus there is a possibility of type I error. Finally, the statistical comparison of the difference in clinical outcomes between those with CAD, HF, stroke, or CKD and those without, was conducted post-hoc and therefore this result should be interpreted with caution.Implications and future directions Clinicians should ideally have evidence to guide whether (and when) initiation of a specific T2DM medication is warranted to modify COVID-19 infection risk and risk of adverse outcomes. Our study provides evidence on the role of T2DM therapy and COVID-19 infection risk, demonstrating no association, and also provides evidence on incretin-based therapies and CV event risk among COVID-19 positive patients (areas where there is a current lack of evidence). Given the observational nature of the study, and the important limitations, these findings remain hypothesis-generating. In the absence of available randomized controlled trial (RCT) data, our findings add to the evidence base on this topic. RCTs had been proposed to investigate the role of DPP4i more definitively; however, many trials have been withdrawn or terminated, 26 27 while another is still not recruiting.28 Future research should examine the timing of T2DM medication use with respect to COVID-19 infection risk and adverse outcomes, specifically when an agent could optimally be started (eg, could a DPP4i be started when a patient contracts COVID-19 or do they have to be a prevalent user and for how long?). Until more answers to these questions are available from RCTs, any possible impact of T2DM medications on COVID-19 infection or related outcomes should not be used to inform clinical decisions.Conclusion In this population-based cohort study of older adults with T2DM taking metformin, use of DPP4i/GLP1RA was not associated with COVID-19 infection but was associated with lower risk of all-cause hospitalization and a possible trend towards lower associated risk of CV events among those testing positive for COVID-19.