DPP4 inhibitors and COVID19 outcomes in patients with type II diabetes: a multicenter retrospective cohort study in Saudi Arabia

Abstract

Dipeptidyl peptidase-4 inhibitors (DPP-4is) have been hypothesized to impact COVID-19 outcomes by modulating immune and inflammatory responses. This study aimed to assess the association between DPP-4i use and COVID-19 severity in patients with type 2 diabetes. Adult (≥ 18 years) patients with T2DM hospitalized for COVID-19 across seven hospitals in Saudi Arabia between March 2020 and November 2021 were included. Outcomes were in-hospital mortality, progression to mechanical ventilation (MV), intensive care unit (ICU) admission, and length of stay. We performed time-to-event analyses using Cox proportional hazards models with propensity score weighting to account for confounders. Ordinal proportional odds model were for the length of stay outcomes. A total of 166 patients were included in the DPP-4i group, and 351 were included in the non-DPP-4i group. Propensity score weighting achieved a well-balanced comparison between the groups. The DPP-4i group showed a nonsignificant decrease in mortality (adjusted hazard ratio [HRadj] = 0.73; 95% confidence interval [CI]: 0.40–1.34; p = 0.319), a significant reduction in progression to MV (HRadj = 0.40; 95% CI: 0.21–0.77; p = 0.006), and a nonsignificant reduction in ICU admissions (HRadj = 0.83; 95% CI: 0.57–1.21; p = 0.338). Length of stay did not differ significantly between groups. This study revealed that prior usage of DPP-4is in T2DM patients with COVID-19 is linked to a significant reduction in progression to MV in this high-risk group. However, there is a need for further investigation through well-powered prospective studies and randomized controlled trials to confirm these findings.

Introduction

Diabetes mellitus (DM) is a known risk factor for infections, likely due to changes in the immune system [1]. Dysregulation of the innate immune system, such as disruption of cytokine signaling, is observed in individuals with both type 1 diabetes mellitus (T1DM) and type 2 diabetes mellitus (T2DM) [1,2,3]. Individuals with both types of DM have a higher infection rate than the general population and often worse outcomes [4, 5].

Mortality during the COVID-19 pandemic was a significant concern, especially among immunocompromised patients [6]. Moreover, patients with comorbid conditions, such as diabetes, had an increased risk of severe disease and mortality during the COVID-19 pandemic [6, 7]. Estimates suggest that a significant number of critically ill COVID-19 patients had diabetes, although the exact number varied by country [6, 8, 9]. In Saudi Arabia, the prevalence of diabetes among COVID-19 patients varies significantly (0.44–68.3%), with the largest study (7,260 patients) reporting a prevalence of 12.6% [10, 11].

In addition to the impaired immunity observed in patients with diabetes, the increased risk of severe COVID-19 is believed to be due to the overexpression of angiotensin converting enzyme type-2 (ACE-2) receptors [6]. ACE-2 receptors were found to be the point of entry for the virus (SARS-CoV-2) into the cell, and the overexpression of these receptors secondary to some antidiabetic and antihypertensive medications might in part explain the increased risk [6]. Some reports have also suggested that membrane-bound dipeptidyl peptidase-4 (DPP-4) might regulate the entry of SARS-CoV-2 into the respiratory system [12, 13]. This has led to investigations of the impact of the antidiabetic class of DPP-4is on COVID-19 infection, especially since this class has previously been found to exhibit beneficial anti-inflammatory and immunoregulatory effects [12, 14].

Several studies have investigated the use of DPP-4is and their associations with the severity of COVID-19, with conflicting results [12, 15]. A multicenter, retrospective case‒control study in northern Italy reported that, compared with standard of care, treatment with sitagliptin at the time of hospitalization was associated with reduced mortality, improved clinical outcomes, and a greater number of hospital discharges [16]. On the other hand, a retrospective cohort study revealed that the need for ICU admission was more likely to be observed in patients receiving DPP-4is [17]. A systematic review and meta-analysis published in 2022 that included 11 studies revealed that the use of DPP-4is was linked to lower mortality rates in COVID-19 patients [15].

To our knowledge, no data have been reported regarding the Saudi population. Thus, our study aimed to evaluate the associations between the use of DPP-4is in patients with T2DM who are infected with COVID-19 and the severity of COVID-19 in these patients.

Methods

Study design and settings

This retrospective cohort study included adult patients with T2DM who were admitted to multiple hospitals in Saudi Arabia for COVID-19. The involved sites were Johns Hopkins Aramco Healthcare (JHAH) (Dhahran/Al-Ahsa), National Guard Health Affairs (Riyadh/Al-Ahsa/Jeddah), King Fahad University Hospital (Dammam) and King Fahad Medical City (Riyadh). The study was conducted in accordance with the World Medical Association Declaration of Helsinki—Ethical Principles for Medical Research Involving Human Subjects (adopted 1964; updated 2013. The study was approved by the following institutional review boards (IRBs) (King Faisal University: KFU-REC/2021-02-23; King Abdullah International Medical Research Center: SP22R/014/02; King Fahad Medical City: H-01-R-012; JHAH: H-05-DH-044). The study adhered to the Observational Studies in Epidemiology (STROBE) Statement checklist [18].

Participant selection and data collection

IT staff at participating sites used their electronic health record systems to generate lists of patients who were hospitalized between March 2020 and November 2021 and who fulfilled the initial criteria of being admitted during the COVID-19 pandemic and having a diabetes diagnosis. Next, using the following inclusion and exclusion criteria, we screened each patient on these lists to determine their eligibility. Patients were included if they were adults (≥ 18 years), had a documented diagnosis of T2DM, and were admitted to the hospital due to SARS-CoV-2 infection, as determined by reverse transcription polymerase chain reaction (RT‒PCR). The index date for every patient was defined as the date of admission to the hospital with a positive PCR test for SARS-CoV-2.

We excluded patients with unconfirmed cases of SARS-CoV-2 infection, incomplete patient files, or other types of diabetes.

Demographic data and relevant baseline characteristics, such as comorbidities, home medication use, laboratory values, inpatient regimens, and outcomes, were manually extracted and encoded into the Research Electronic Data Capture (REDCap) system in a deidentified manner [19]. The exposure group (DPP-4i group) included patients with T2DM and documented use of DPP-4is prior to the current COVID-19 admission. Patients with T2DM who were not receiving DPP-4i therapy prior to the current COVID-19 admission were pooled into the nonexposure group (control group).

The classification of COVID-19 severity followed the WHO criteria, which were also incorporated into the Saudi Ministry of Health (MOH) guidelines. A moderate case was defined as nonsevere pneumonia without the need for additional oxygen. A severe case was defined as clinical signs of pneumonia (a fever, cough, fast breathing) and one or more symptoms, such as a respiratory rate of ≥ 30 breaths per minute, difficulty breathing, signs of respiratory distress, and an oxygen saturation (SpO2) of 93% or lower in ambient air. For critical cases, at least one of the following criteria had to be met: acute respiratory distress syndrome (ARDS), septic shock, an altered mental state, or failure of multiple organs [20].

Study outcomes

The main outcome of the study was patient mortality during the index hospitalization. The other outcomes included progression to mechanical ventilation (MV), admission to the intensive care unit (ICU), length of hospital stay, and length of ICU stay.

Statistical analysis

Continuous variables are reported as medians with interquartile ranges (IQRs), whereas categorical variables are presented as frequencies with percentages (%). Statistical significance was determined via the Mann‒Whitney test for continuous data and the chi‒square test or Fisher’s exact test for categorical data. Survival analysis of the categorical outcomes was conducted. To address potential confounders thoroughly, propensity score weighting was employed. This approach involves calculating the probability of treatment assignment (i.e., DPP-4is) via a logistic regression model that includes variables associated with treatment allocation [21]. Propensity scores are then used to estimate overlap weights, where each unit’s weight is determined by the probability of being assigned to the opposite group, as described by Li et al. [22]. Overlap weights are constrained and designed to minimize the asymptotic variance of the weighted average treatment effect within the class of balancing weights. In this study, these calculations were performed via the ‘weightit’ package in R [23].

To assess postweight balance, diagnostic tools such as propensity score distributions and Love plots were utilized. Balance was evaluated on the basis of the absolute standardized mean difference (SMD), where values < 0.01 indicate better balance. These diagnostic tools were applied via the ‘cobalt’ package in R [24].

A proportional Cox hazard model was applied after weighting to estimate the relative effect, and robust standard error estimation was used to account for the survey design [25]. The results are summarized with hazard ratios (HRs) and 95% confidence intervals (CIs). A weighted Kaplan‒Meier model was fitted to visualize absolute survival probabilities, and for the progression to mechanical ventilation outcome, we used the cumulative incidence function to depict the event occurrence. The length of stay outcomes were treated as ordinal data; therefore, ordinal proportional odds models were used to analyze these outcomes. The proportional odds model is an extension of the Wilcoxon‒Mann‒Whitney‒Kruskal‒Wallis‒Spearman test [26]. Under the ordinality assumption of the response, the model can account for covariates. A weighted proportional odds model was performed via the svyolr function from the survey package in R [25]. To account for multiplicity and to control type 1 error., we applied Benjamini-Hochberg (BH) Procedure to adjust p-values for multiple hypothesis testing. The method ranks the observed p-values in ascending order and adjusts them using an increasing threshold to control the expected proportion of false positives. All statistical analyses were performed using R Core Team (2023). R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria and SAS software version 9.4 (SAS Institute Inc., Cary, NC).

Results

Patient demographics, comorbidities, and baseline laboratory data

During the study period, we screened 3,053 patients with a DM diagnosis who were admitted to the hospital during the COVID-19 pandemic. A total of 2,536 patients were excluded because they had an unconfirmed diagnosis of COVID-19, had T1DM, were younger than 18 years, or had incomplete patient data (Fig. 1). A total of 517 patients were included in the final analysis: 166 in the DPP-4i group and 351 in the control group. The mean age of the patients in the study was comparable between the groups (69.0 years in the DPP-4i group vs. 67.0 years in the control group; p = 0.465), and no statistically significant difference in the severity of COVID-19 at admission was detected between the two groups. However, patients in the control group had a significantly higher BMI(28.9 kg/m² in the DPP-4i group vs. 31.5 kg/m² in the control group; p = 0.011), whereas a greater proportion of patients with coronary artery disease was observed in the DPP-4i group than in the control group (14.5% vs. 8.0%; p = 0.022). In addition, there were no statistically significant differences between the two groups in laboratory values at admission except for HbA1c and C-reactive protein levels, Since the DPP-4i group had slightly higher median HbA1c (8.5% [7.3–10.4%]) than the non-DPP-4i group (7.9% [6.8–9.8%]) (p = 0.032). In contrast, the median CRP in the DPP-4i group was significantly lower (10.2 mg/L [3.1–43.4 mg/L]) compared to the non-DPP-4i group (30.2 mg/L [8.4–101.0 mg/L]) (p < 0.001).

Fig. 1

Flow chart of study inclusion

Abbreviations: DPP-4i: Dipeptidyl peptidase-4 inhibitors

The complete baseline characteristics are presented in Table 1.

Table 1 Patients’ baseline characteristics (n = 517)

Patients’ medication use at admission

With respect to the other antidiabetic agents used, a greater proportion of patients in the DPP-4i group were on metformin, SGLT-2i, or thiazolidinedione, whereas more patients in the control group were on insulin. Other home medications were largely nondifferent between the groups except for the greater use of statins and aspirin in the DPP-4i group than in the control group (statin: 63.3% vs. 51.0%, p = 0.009; aspirin: 36.1% vs. 25.6%, p = 0.014) (Table 1). Among patients in the DPP-4i group, the majority were on sitagliptin (83.1%), 72.3% had been on DPP-4is for at least 3 months prior to admission, and 46.4% continued DPP-4i use during their current hospitalization (Table 1 footnotes).

Patient management during hospitalization

With respect to the COVID-19 treatment regimen used during admission, a greater proportion of patients in the control group were on IV steroids, remdesivir and lopinavir/ritonavir (Table 2). In addition, the oxygen requirements upon admission were greater in the control group than in the DPP-4i group (Table 2).

Table 2 COVID-19 regimen used during admission

Patient outcomes

For the main endpoint of mortality during hospitalization, the incidence was lower in the DPP-4i group than in the control group (9.6% vs. 18.8%, unadjusted HR = 0.55, 95% CI: 0.31 to 0.97, p = 0.040). In addition, a lower proportion of patients in the DPP-4i group needed MV and ICU admission than did patients in the control group (MV: 7.2% vs. 20.8%, unadjusted HR = 0.30, 95% CI: 0.16 to 0.56, p < 0.001; ICU admission: 26.5% vs. 37.0%, unadjusted HR = 0.66, 95% CI: 0.47 to 0.93, p < 0.021). The unadjusted Kaplan‒Meier curves for the mortality model are displayed in Fig. 2, Panel A. The unadjusted CIF plot is displayed in Fig. 2, Panel C. In addition, there was no difference between the two groups in terms of length of hospital stay or length of ICU stay. The details of the clinical outcomes are shown in Table 3.

Fig. 2

(Panel A) Unadjusted Kaplan Meir curve for the mortality outcome; (Panel B) Weighted Kaplan Meir curve for the mortality outcome. (Panel C) Unadjusted Cumulative Incidence Function (CIF) for the progression to mechanical ventilation outcome. (Panel D) Adjusted Cumulative Incidence Function (CIF) for the progression to mechanical ventilation outcome. Abbreviations: DPP4i: Dipeptidyl peptidase-4 inhibitors; HR: Hazard Ratios; CI: Confidence intervals

Table 3 Clinical outcomes results

Weighting and survival analysis

Table S1 presents the covariate balance after weighting, showing that all absolute SMDs were close to 0. The distributions of the propensity scores before and after weighting, along with the Love plots, are shown in Figure S1 and Fig. 3, respectively. The results of the survival analysis are summarized in Table 3. The Kaplan‒Meier curves for the mortality model are displayed in Fig. 3, Panel B. For progression to mechanical ventilation, the adjusted HR was 0.40 (95% CI: 0.21 to 0.77, p = 0.006). The adjusted CIF for this outcome is displayed in Fig. 2, Panel D. In Fig. 4, the forest plot provides a visualized summary that illustrates the estimates from the unadjusted and adjusted Cox proportional hazard models for mortality and progression to mechanical ventilation outcomes. The number of ICU admissions was not significantly different between the groups, with an adjusted HR of 0.83 (95% CI: 0.57 to 1.21, p = 0.338). There was no difference in the weighted analysis for length of stay between the two groups. See Table 3.

Fig. 3

Love’s plot of the covariate balance

Covariates balance are achieved by Covariate balance was achieved via the overlap weights methodology according to Li (2017) by weighting each unit proportional to its probability of assignment to the opposite group. The estimand is referred to “ATO” (average treatment effect in the overlap). SMD closer to zero are indicative of good balance. These computations were performed using the WeightIt package. Covariates balance diagnostics are assessed using the cobalt package in R

Abbreviations: WHO: World Health Organization; COVID-19: coronavirus disease 2019; HbA1c: hemoglobin A1c; SGLT-2i: Sodium-glucose transport protein-2 inhibitor; GLP-1RA: Glucagon-like peptide-1 receptor agonist; ACEI: Angiotensin-converting enzyme inhibitors; ARBs: Angiotensin II receptor blockers

Fig. 4

Forest plot of the survival analysis outcomes. Note: Weighted Cox proportional hazard model was used for the adjusted analysis

Discussion

This observational cohort study revealed a reduction in hospital mortality among patients with DM who were hospitalized with COVID-19 and treated with DPP-4is compared with those who were not. When examining the results for this outcome, the mortality reduction was significant in the unadjusted model but not in the adjusted survival analysis. Notably, the survival analysis included weighting for many potential confounders to achieve a balance between the two arms, and the 95% CI surrounding the hazard ratio was wide, capturing the uncertainty. While the adjusted model did not show a statistically significant reduction in mortality, the observed direction of effect aligns with previous findings [15, 16], warranting further investigation.

We observed a statistically significant reduction in the progression to mechanical ventilation in favor of the DPP-4i group in both the unadjusted and adjusted analyses. The reduction in this outcome may have influenced the mortality outcomes. These clinical benefits align with the results from a multicenter case‒control study conducted in Italy by Solare et al., in which sitagliptin was associated with a significant reduction in mortality (HR = 0.44; 95% CI: 0.29 to 0.66) and the need for mechanical ventilation (HR = 0.27; 95% CI: 0.11 to 0.62) [16]. In our study, sitagliptin was used by 83.1% of patients who were exposed to DPP-4is prior to admission. Moreover, a recent meta-analysis by Zein et al., which included 11 primarily retrospective studies, revealed a significant reduction in mortality in the DPP-4i group compared with the non-DPP-4i group, with a pooled odds ratio (OR) of 0.75 (95% CI: 0.56 to 0.99) and nonsignificant heterogeneity among the studies in the meta-analysis (I [2]: 42.9, p = 0.064) [15]. The meta-regression in this meta-analysis revealed that this mortality benefit was not affected by age, sex, hypertension status, study design, or geographical location. Similarly, the effects of prior treatment with DPP-4is versus other glucose-lowering drugs on COVID-19-related outcomes in patients with type 2 diabetes were also examined by Emral and colleagues in a multicenter, retrospective cohort study. Their data suggest that among diabetes patients diagnosed with COVID-19, prior DPP-4i therapy is associated with lower mortality (9.5 vs. 11.8%; p < 0.001) as well as lower rates of ICU hospitalization and/or mechanical ventilation (21.7 vs. 25.2%; p = 0.001). In addition, the rates of ICU admission and/or the need for MV were lower in the DPP-4i group (21.7% vs. 25.2%; p = 0.001) [27].

In contrast to our findings, a retrospective study conducted in Singapore reported that patients on DPP-4i had higher odds of ICU admission compared to those on other glucose-lowering medications, with an adjusted relative risk (aRR) of 4.07 (95% CI: 1.42–11.66, p = 0.009) and for mechanical ventilation (aRR 2.54, 95% CI: 0.43–14.99). Notably, the diabetes subgroup analysis in this study included only 76 patients, with just 27 patients on DPP-4i, limiting the robustness of these findings [28]. Furthermore, the study evaluated multiple endpoints and subgroups, introducing a potential risk of false discovery due to alpha inflation. The large nationwide observational study in England by Khunti et al. [29]. assessed diabetes prescriptions in patients with COVID-19 and found that DPP-4i use was associated with increased mortality (HR 1.07, 95% CI: 1.01–1.13). However, the authors noted that national guidelines for DPP-4i use primarily target older adults, particularly those with frailty, which may have influenced the observed association [30].

The proposed mechanisms for such benefits are the previously described anti-inflammatory and immunomodulatory properties of DPP-4is. Interestingly, the DPP-4 receptor (also known as CD26) has been identified as one of the functional receptors for the spike protein of Middle East respiratory syndrome coronavirus (MERS-CoV), which can mediate virus entryinto the cell. Therefore, inhibiting the DPP-4 pathway could disrupt the DPP-4-mediated virulence of SARS-CoV-2, thereby improving the clinical outcomes of patients with COVID-19. DPP-4 is expressed in abundance in several cells and tissues, including immune cells such as T cells, activated B cells, and activated natural killer cells, influences both innate and adaptive immune responses and significantly affects T-cell activation and cytokine production. This modulation includes the upregulation of proinflammatory cytokines and the cytotoxic activity of CD8 + T cells. DPP-4 is differentially expressed across T-cell subsets, impacting their cytokine secretion patterns and potentially the immune response balance. Additionally, inhibiting DPP-4 activity in B cells reduces their activation, which makes DPP-4 a potential target for managing immune dysregulation in patients with COVID-19 [31].

In addition to its transmembrane form, DPP-4 is also present in soluble form (called sDPP-4) in various body fluids, including serum, urine, saliva, and cerebrospinal fluid [32,33,34]. The levels of sDPP-4 were found to be significantly reduced in patients admitted with severe COVID-19 ¹³. With the lessons learned from the MERS-CoV epidemic, for which 84.5% of cases were reported in Saudi Arabia, studies have recognized the important role of DPP-4 in severe infections [35]. A study by Alkharsah et al. compared sDPP4 levels in healthy Saudi participants’ plasma samples with those of other Arab or non-Arab subjects, which could explain the increased risk for MERS endemicity among Saudis [36]. Their study also revealed various single-nucleotide polymorphisms (SNPs) that could be associated with reduced sDPP-4 levels. These SNPs were genotypes AG of the SNP rs35128070A > G, GT of the SNP rs1861978G > T, AA and AG of the SNP rs79700168G > A, and AA and AG of the SNP rs17574, all of which were associated with lower levels of sDPP-4 in Saudi subjects. Moreover, more human data are needed, as DPP-4 inhibition led to an increase in circulating sDPP-4 levels in animal models and could play a role in patient responses [37].

This study has several limitations that should be considered when interpreting the findings. The retrospective design and the potential presence of both measured and unmeasured confounding factors may limit the strength of our conclusions. To mitigate these limitations, we employed a weighting method in the Cox proportional hazards model, which may address some of the confounding. However, this approach cannot fully account for unmeasured confounding; therefore, the use of a randomized study design remains the gold standard. The study question remains unanswered by clinical trials, as many planned studies examining this association were terminated or withdrawn due to logistical challenges [38,39,40]. Treatment with DPP-4is can differ in terms of duration, agent of choice, and level of adherence. While there are data that suggest immuno-regulatory roles for DPP-4is with longer-term use (more than 3 months) [41], the effect of a shorter duration could not be fully investigated, as many patients were already on these agents for more than 3 months before their admission for COVID-19. Although we gathered information on the particular DPP-4i drugs, our study’s retrospective design made it difficult for us to reliably record the exact amount of time each patient received DPP-4i medication prior to admission and the dose as well. However, we were able to report which patients had used DPP-4i for less than three months or for three months or more before being admitted. Additionally, we acknowledge that while length of stay is a commonly used metric, it may be influenced by non-medical factors such as hospital capacity, bed availability, health insurance coverage, financial constraints, and isolation policies. Therefore, it may not reliably reflect the true clinical outcomes. Furthermore, while there are reports showing benefits of DPP-4i use, there is contrasting evidence that shows no such benefits in different populations, which may limit the generalizability of our findings to different populations or settings [42]. These findings can vary due to the heterogeneity of COVID-19 variants which emerged after our study. Future studies can study the impact of different variants on the study outcomes.

Conclusion

This study demonstrated that prior DPP-4i use among patients with T2DM who have COVID-19 is associated with better clinical outcomes in terms of a reduction in progression to mechanical ventilation. These findings suggest that DPP-4is may have a possible therapeutic role in this high-risk population. While promising, these findings should be interpreted with caution because of the observational nature of the study. Future randomized controlled trials are essential to confirm the efficacy of DPP-4is in patients with T2DM and SARS-CoV-2 infection. Finally, we still need to further explore the mechanism of this effect in animal studies to fully understand the relationship between the DPP-4 receptor and SARS-CoV-2 infection.