FormalPara Key Summary Points

Why carry out this study?

We previously demonstrated that a premorbid prescription of either glucagon-like peptide-1 receptor agonists (GLP-1RA) or sodium-glucose cotransporter-2 inhibitors (SGLT-2i), compared to dipeptidyl peptidase 4 inhibitors (DPP-4i), is associated with reduced severity of COVID-19 through analyses of observational data from the National COVID Cohort Collaborative (N3C).

With access to a sixfold larger N3C cohort gathered over approximately 3 years of the pandemic (1 January 2020–15 September 2022), we reassessed the association of GLP-1RA or SGLT-2i prescriptions alone and in combination with COVID-19 severity.

What was learned from this study?

Prescriptions for GLP-1RA and SGLT-2i continue to be associated with lower COVID-19 mortality compared to prescriptions for DPP-4i.

When compared to the use of SGLT-2i alone, concomitant prescription of GLP-1RA/SGLT-2i is associated with lower odds of secondary outcomes, including emergency room visits, hospitalizations, and mechanical ventilation, suggesting that there may be additive, protective effects from the concomitant use of GLP-1RA/SGLT-2i in the context of COVID-19.

Digital Features

This article is published with digital features, including graphical abstract, to facilitate understanding of the article. To view digital features for this article, go to https://doi.org/10.6084/m9.figshare.25257022.

Introduction

Chronic comorbid conditions such as diabetes are a risk factor for severe adverse coronavirus disease 2019 (COVID-19) outcomes, including hospitalization, invasive mechanical ventilation, and death [1,2,3]. Early efforts aimed to identify modifiable risk factors to minimize COVID-19 severity in this population. Glycemic control is thought to be one high-risk factor associated with severity of COVID-19 infection in people living with diabetes [4, 5].

Using observational data from the National COVID Cohort Collaborative (N3C), we previously demonstrated that premorbid prescription of two antihyperglycemic medication classes, glucagon-like peptide-1 receptor agonists (GLP-1RA) and sodium-glucose-cotransporter 2 inhibitors (SGLT-2i), compared to dipeptidyl peptidase 4 inhibitors (DPP-4i) prescription, associate with lower odds of multiple adverse outcomes among people with diabetes diagnosed with COVID-19 prior to 25 February 2021 (n = 12,446) [6]. It is unknown whether this association remains robust to the development of new variants, natural immunity, and effective vaccines. We thus reevaluated the association of GLP-1RA and SGLT-2i prescriptions on severe COVID-19 outcomes in an approximately sixfold larger cohort (n = 78,806) that covered a longer period of the pandemic. Additionally, given increasing prescriptions of SGLT-2i and GLP-1RA in combination, two agents which operate through distinct mechanisms that may provide unique benefits in the setting of COVID-19, we examined the impact of concomitant GLP-1RA/SGLT-2i prescription on COVID-19 severity.

Methods

Study Design

In this study, we analyzed real-world observational data of 78,806 adults from the N3C cohort [7], which includes individuals with at least one positive PCR test result for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) after 1 January 2020 [6, 8, 9]. We gained permission to use the deidentified electronic medical health medical record data via the data-use request process through the National Covid Cohort Collaborative (N3C) enclave. Our general study design and methods, including statistical analyses, have been previously described [6].

Briefly, we analyzed data through 15 September 2022 and included adults aged ≥ 18 years who had any prescription of GLP-1RA, SGLT-2i or DPP-4i within 24 months prior to a COVID-19 diagnosis. A diagnosis of type 2 diabetes was not required for inclusion in the study. Prescription information reflects prescriptions written during ambulatory visits and does not reflect dispensing or adherence. In analyses where DPP4i were used as the comparator, we excluded those persons with concomitant prescription of DPP-4i and GLP-1RA/SGLT-2i (Electronic Supplementary Material [ESM] Fig. S1). We did not exclude people who were included in our prior analysis (n = 12,446).

In this article, cohorts with a prescription for a particular drug will be referred to as arms (e.g., “GLP-1RA arm”). To ensure consistency with our prior analysis, individuals with prescriptions for both GLP-1RA and SGLT-2i (n = 11,594) contributed to both exposure arms in the comparison with individuals with prescriptions for DPP-4i. Additional analyses compared concomitant GLP-1RA/SGLT-2i prescription (“GLP-1RA/SGLT-2i arm”) to prescription with GLP-1RA or SGLT-2i alone.

We defined the first positive SARS-CoV-2 PCR as the index date and the primary outcome as 60-day mortality following a positive PCR. Secondary outcomes included emergency room (ER) visits, hospitalization, and mechanical ventilation (intubation or ventilation) within 14 days of a positive PCR test. We used data up to 24 months before the index date to identify drug exposure, continuous variables, medical history, and demographics.

Statistical Analysis

Standardized mean differences (SMD) were used to compare baseline characteristics before and after propensity score weighting (PSW) [10, 11]. In the primary analysis, we used targeted maximum likelihood estimation (TMLE) to estimate odds ratios (ORs) and 95% confidence interval (CI) [10, 11]. For sensitivity analyses, we used inverse probability treatment-weighted (IPTW) logistic regression. We performed post-hoc analyses on two restricted cohorts (individuals aged 45–80 years and individuals with an estimated glomerular filtration rate (eGFR) ≥ 45 mL/min) and a sensitivity analysis using only sex and age as covariates to evaluate the impact of imputing missing data in covariates. Analyses were performed using Palantir Foundry hosted within the N3C enclave, a cloud-based FedRAMP moderate secure enclave [7], and statistical programs Python and R.

Ethics Compliance

The protocol of this study was registered with the European Network of Centres for Pharmacoepidemiology and Pharmacovigilance (ENCePP) on 5 October 2020 (Number 37860). The University of North Carolina at Chapel Hill Office of Human Research Ethics determined that the study protocol did not constitute research on human subjects.

The analyses described in this publication were conducted with data or tools accessed through the NCATS N3C Data Enclave (https://covid.cd2h.org) and N3C Attribution & Publication Policy v 1.2-2020-08-25b supported by NCATS U24 TR002306 and Axle Informatics Subcontract NCATS-P00438-B. The N3C data transfer to NCATS is performed under a Johns Hopkins University Reliance Protocol IRB00249128 or individual site agreements with NIH.

This study research was possible because of the patients whose information is included within the data and the organizations (https://ncats.nih.gov/n3c/resources/data-contribution/data-transfer-agreement-signatories), and scientists who have contributed to the ongoing development of this community resource [7]. The study was performed in accordance with the Declaration of Helsinki (1964) and its later amendments [12].

Results

By 15 September 2022, 75 sites across the USA had contributed data on 15,540,911 individuals to the N3C database, of whom 4,671,046 had a positive SARS-CoV-2 PCR test. The present study included 78,806 individuals across 63 sites. While the first analysis evaluated 13 months (January 2020 to February 2021), this subsequent analysis extends to 33 months during the evolving pandemic (January 2020 to September 2022). Table 1 presents crude and weighted characteristics of the study sample. The total study population had a mean (± standard deviation [SD]) age of 58.7 (± 13.3) years. Those individuals in the DPP-4i arm were older and had a lower mean body mass index (BMI) than those in the GLP-1RA and SGLT-2i arms, respectively. The prevalence of most comorbid conditions was higher in the DPP-4i arm, but the prevalence of comorbid cardiovascular-related diseases was highest in the SGLT-2i arm. The subpopulations of interest were similar following PSW (Table 1). Where PSW exposure arms remained imbalanced, TMLE analysis improved the chance of correct model specification.

Table 1 Demographics and clinical characteristics before and after propensity score weighting, according to premorbid prescription for total study population and for glucagon-like peptide-1 receptor agonist, sodium-glucose cotransporter-2 inhibitor, and dipeptidyl peptidase-4 inhibitor arms
Full size table

Crude primary and secondary outcomes are summarized in ESM Table S1. The GLP-1RA and SGLT-2i arms associated with a lower 60-day mortality, with proportions of 2.68% and 2.97%, respectively, compared to 7.00% in the DPP-4i arm. Figure 1 provides ORs (95% CI) for all outcomes estimated by TMLE comparing the GLP-1RA or SGLT-2i arms to the DPP-4i arm. ORs for the primary outcome 60-day mortality were lower for the GLP-1RA (OR 0.64, 95% CI 0.56–0.72) and SGLT-2i (OR 0.62, 95% CI 0.57–0.68) arms compared to the DPP-4i arm. ORs were also significantly lower for all secondary outcomes with GLP-1RA and SGLT-2i prescription, with the exception of mechanical ventilation. IPTW analyses are presented in ESM Fig. S2. ORs for 60-day mortality were lower for the GLP1-RA (OR 0.64, 95% CI 0.56–0.72) and SGLT2i (OR 0.62, 95% CI 0.57–0.68) arms compared to the DPP4i arm. GLP1-RA and SGLT2i use was also associated with lower ORs for all secondary outcomes, including ER visits, hospitalization, and mechanical ventilation. Two post-hoc cohort analyses (age restricted: 45–80 years and eGFR restricted: ≥ 45 mL/min/1.73 m2; ESM Tables S2, S3) and a sensitivity analysis for age and sex adjustment (ESM Table S4) yielded results similar to the primary analysis with lower odds for all outcomes, except for mechanical ventilation.

Fig. 1
figure 1

Forest plot depicting TMLE-estimated ORs for primary and secondary outcomes for people with a COVID-19 diagnosis and prescription for GLP-1RA (A) and SGLT-2i (B) use compared with DPP-4i use, respectively. Single asterisk (*) indicates within 60 days after positive SARS-CoV-2 PCR test; double dagger sign (‡) indicates within 14 days after positive SARS-CoV-2 test; hash sign (#) indicates mechanical ventilation (intubation or ventilation). CI Confidence interval, DPP-4i dipeptidyl peptidase-4 inhibitor, GLP-1RA glucagon-like peptide 1 receptor agonist, SARS-CoV-2 severe acute respiratory syndrome coronavirus 2, OR odds ratio, SGLT-2i sodium glucose co-transporter 2 inhibitor, TMLE targeted maximum likelihood estimation

Full size image

Crude and weighted baseline information for individuals prescribed GLP-1RA and SGLT-2i alone and in combination are presented in Table 2. The percentage of individuals with comorbid conditions, including renal, hepatic and cardiovascular-related disease, was slightly lower in the concomitant GLP-1RA/SGLT-2i arm compared to the monotherapy arms. Conversely, use of other antihyperglycemic agents was higher in the GLP-1RA/SGLT-2i arm. Exposure arms were similar following PSW.

Table 2 Demographics and clinical characteristics before and after propensity score weighting, according to premorbid prescription for total, glucagon-like peptide-1 receptor agonist (GLP-1RA) monotherapy, sodium-glucose cotransporter-2 inhibitor (SGLT-2i) monotherapy, and concomitant GLP-1RA/SGLT-2i arms
Full size table

Comparison of crude primary and secondary outcomes for the GLP-1RA/SGLT-2i arm and the GLP-1RA and SGLT-2i arms (ESM Table S5) indicated that concomitant GLP-1RA/SGLT-2i prescription was associated with lower 60-day mortality (2.58%) compared to monotherapy (2.83% for GLP-1RA and 3.24% for SGLT-2i). The concomitant GLP-1RA/SGLT-2i prescription arm showed lower rates for secondary outcomes than the SGLT-2i arm, but similar rates when compared to the GLP-1RA arm.

TMLE-estimated ORs comparing the GLP-1RA and SGLT-2i arms, respectively, with the concomitant GLP-1RA/SGLT-2i arm (Fig. 2) resulted in similar odds for 60-day mortality for the GLP-1RA/SGLT-2i co-prescription arm compared to the GLP-1RA (OR 0.92, 95% CI 0.81–1.05) and SGLT-2i (OR 0.88, 95% CI 0.76–1.01) monotherapy arms. Lower odds were observed for all secondary outcomes, including ER visits, hospitalization, and mechanical ventilation in the concomitant GLP-1RA/SGLT-2i arm compared to the SGLT-2i monotherapy arm, whereas similar odds were observed when concomitant GLP-1RA/SGLT-2i use was compared to GLP-1RA use alone. IPTW-estimated ORs comparing the GLP-1RA and SGLT-2i arms with the concomitant GLP-1RA/SGLT-2i arm (ESM Fig. S3) demonstrated lower rates for 60-day mortality for both comparisons. Concomitant GLP-1RA/SGLT-2i prescription was also associated with lower odds for all secondary outcomes, although mechanical ventilation only trended toward lower odds in the comparison to GLP-1RA alone.

Fig. 2
figure 2

Forest plot depicting TMLE-estimated ORs for primary and secondary outcomes for people with a COVID-19 diagnosis and prescription for GLP-1RA and SGLT-2i combined use compared with GLP-1RA monotherapy (A) and SGLT-2i (B) monotherapy, respectively. Single asterisk (*) indicates within 60 days after positive SARS-CoV-2 PCR test; double dagger sign (‡) indicates within 14 days after positive SARS-CoV-2 test; hash sign (#) indicates mechanical ventilation (intubation or ventilation). CI confidence interval, DPP-4i dipeptidyl peptidase-4 inhibitor, GLP-1RA glucagon-like peptide 1 receptor agonist, OR odds ratio, SARS-CoV-2 severe acute respiratory syndrome coronavirus 2, SGLT-2i sodium glucose co-transporter 2 inhibitor, TMLE targeted maximum likelihood estimation

Full size image

Discussion

Since the COVID-19 pandemic began, diabetes has emerged as a risk factor for severe COVID-19, with results from meta-analyses suggesting a nearly twofold increased mortality risk [1]. Given that COVID-19 was the fourth leading cause of death in the USA in 2022 [6, 13], effective strategies to improve COVID-19 outcomes among people with diabetes are needed. To this end, antihyperglycemic medication use presents an attractive target with plausible biological mechanisms. GLP-1RA and SGLT2i inhibitors have garnered particular attention due to their anti-inflammatory effects and well-established cardiovascular risk reduction in high-risk individuals [14, 15]. We and others have demonstrated an association between the use of GLP-1RA and SGLT-2i and reduced adverse outcomes of COVID-19 [6, 16,17,18,19,20,21,22]. Whether this association remained as the pandemic progressed, with novel variants and increasing natural and vaccine-induced immunity, has not been established. Using a sixfold larger cohort than our original analysis [6] and data from a timepoint (15 September 2022) further into the pandemic, the present study provides further evidence supporting the association of GLP-1RA and SGLT-2i with improved COVID-19 outcomes compared to premorbid DPP-4i prescription.

In contrast, the DARE-19 study examined acute prescription of SGLT-2i in the setting of COVID-19. This double-blind randomized controlled trial investigated whether the SGLT-2i dapagliflozin provided organ protection in non-critically ill hospitalized people with COVID-19 and at least one cardiometabolic risk factor when initiated within 4 days of SARS-CoV-2 infection. A trend toward benefit was observed in the composite outcome of organ dysfunction or death but was not statistically significant [23]. It is plausible that premorbid SGLT-2i use, as examined in our study, provides more protection than initiation after SARS-CoV-2 infection. Consistently, results from other studies suggest that SGLT2i and GLP-1RA monotherapy confer lower risk of outcomes compared to DPP4i monotherapy when prescribed prior to hospitalization for COVID-19 [19].

Additionally, we found that concomitant GLP-1RA/SGLT-2i prescription trended toward improved 60-day mortality when compared to GLP-1RA or SGLT2i monotherapy but did not reach statistical significance. Dual therapy was associated with similar odds of secondary outcomes as GLP-1RA monotherapy but was associated with statistically significantly lower odds of all secondary outcomes when compared with SGLT-2i monotherapy. These findings are consistent with those from randomized trials suggesting that the cardiorenal benefits of GLP-1RA and SGLT-2i are independent of each other [24]. The impact of dual therapy is encouraging given that users of GLP-1RA/SGLT-2i combination therapy were more likely to be treated with additional antihyperglycemic agents, particularly insulin, as many studies have suggested that insulin use is associated with a higher risk of adverse outcomes and may indicate more advanced diabetes [19,20,21].

DPP-4i, which was chosen as a comparator, also has hypothesized immunomodulatory qualities that solicited attention as a potential COVID-19 therapeutic [25]. Yet, the results of observational studies of DPP-4i-related impact on COVID-19 have been inconclusive. The findings of a recent small randomized-controlled trial suggest improvement in COVID-19 severity in hospitalized people with hyperglycemia treated with DPP-4i compared to those receiving insulin alone [26]. The results of meta-analyses also suggest improved outcomes with DPP-4i compared to non-users [27]. Our data suggest that GLP-1RA and SGLT-2i outperform DPP-4i, although prospective data are limited.

The mechanism by which GLP-1RA and SGLT-2i protect against severe COVID-19 outcomes is unknown but may relate to established anti-inflammatory, immunomodulatory, cardiorenal, and metabolic effects [28, 29]. While their effects are likely multifactorial, future studies should examine whether these agents modulate innate or vaccine-induced immunity in people living with diabetes. The results from several studies indicate an association between lower effectiveness of COVID-19 vaccines for severe COVID-19-related outcomes in people with diabetes [30, 31]. Consistently, low anti-SARS-CoV-2 antibody levels on hospital admission associate with severe COVID-19-related outcomes in people with type 2 diabetes [32]. Whether the use of GLP-1RA, SGLT2i, or GLP-1RA/SGLT2i in combination modulate innate or vaccine-induced antibody response should be explored as a potential mechanism for their benefit in the setting of COVID-19.

Our observational study is limited by the potential of residual confounding, with particular attention to the socioeconomic demographic receiving these agents. Nevertheless, our findings are consistent with randomized-controlled trials that have repeatedly demonstrated the cardiorenal and mortality benefits of these two classes of medications.

Conclusion

Our study supports earlier findings that premorbid GLP-1RA or SGLT-2i prescribing was associated with lower mortality and other secondary outcomes in the setting of COVID-19 compared to DPP-4i prescribing. Furthermore, we provide the first evidence of potential synergistic effects from concomitant GLP-1RA/SGLT-2i use on COVID-19 severity.