Use of Oral Anti-Diabetic Drugs and Risk of Hospital and Intensive Care Unit Admissions for Infections

Abstract

Background:

Sepsis is one of the leading causes of hospital mortality, and diabetes is a risk factor for the development of infections. Although strong evidence has shown an association between metformin and reduced risk of infections, the risk of developing infections with newer classes of oral anti-diabetic drugs (OADs) has been less certain. Our study aims to examine the association between outpatient OAD use and hospital admissions for infections.

Methods:

The study cohort included 1.39 million adults with diabetes utilizing the Veterans Health Affairs Corporate Data Warehouse. Multivariate logistic regression was used to estimate the effect of each drug class on hospital admission for infection while adjusting for covariates.

Results:

After adjusting for covariates, those who took metformin during the study period had 3.3% lower odds of hospital admission for infection compared to those who were never on metformin (OR 0.97, 95% CI 0.95–0.98). OADs that were associated with a statistically significant increased odds of being admitted included meglitinides (OR 1.22, 95% CI 1.07–1.38), SGLT2 inhibitors (OR 1.16, 95% CI 1.08–1.24), alpha-glucosidase inhibitors (OR 1.09, 95% CI 1.04–1.15), and DPP4 inhibitors (OR 1.04, 95% CI 1.01–1.06).

Conclusions:

Metformin was associated with lower odds of hospital admission for infection while meglitinides, SGLT2 inhibitors, alpha-glucosidase inhibitors, and DPP4 inhibitors were associated with higher odds of admission for infection.

Introduction

Diabetes is one of the most common chronic diseases with an estimated 34.2 million diagnosed and undiagnosed cases in the United States, accounting for 10.5% of the US population.¹ Morbidity and mortality continue to be high due to a variety of causes including complications of diabetes, atherosclerosis, cardiovascular diseases, stroke, and severe infections including pneumonia and sepsis. Hyperglycemia has been associated with altered immune function, inflammation, and increased severity of infections,^{2, 3} and patients with type 2 diabetes have an increased risk of death associated with hospitalizations for infections.⁴ During the recent pandemic, diabetes was found to be one of the most frequent comorbidities in people with COVID-19 with a prevalence rate reported between 7 and 30%.^{5, 6} Furthermore, patients with diabetes who developed COVID had severe complications with increased morbidity and mortality.⁷

Patients with type II diabetes are often treated with oral antidiabetic medications, and over the last few years, several new classes of oral antidiabetic medications have been developed. Clinicians now have many options for oral therapies including, metformin, sulfonylureas, alpha-glucosidase inhibitors, meglitinides, thiazolidinediones (TZDs), dipeptidyl peptidase-4 (DPP-4) inhibitors, and sodium-glucose cotransporter 2 (SGLT2) inhibitors. Interestingly, many of these agents have effects beyond control of hyperglycemia, which include anti-inflammatory, antiaging and immunomodulatory properties.⁸ In preclinical studies, we and others have shown that metformin and pioglitazone can enhance host response to infections in vitro and in vivo models of infections.^9–12 However, repurposing of these drugs for immunomodulatory effects will need further investigations.

Recent clinical studies suggest an association between pre-admission metformin use and decreased rates of sepsis and inpatient mortality.^{13, 14} However, the development of infections associated with hospitalizations, among patients with type 2 diabetes taking newer classes of OADs has not been as well studied and have provided conflicting results.^{13, 15–17} Given the immunomodulatory effects of oral antidiabetics, we questioned if patients with diabetes who receive these medications are less prone to infections. The objective of our study is to examine the association between outpatient OAD use and hospital admissions for infections.

Methods:

Study design, setting, and population

This is a nationwide retrospective cohort study utilizing the Veterans Health Affairs Corporate Data Warehouse. A HIPAA waiver was obtained, as this research provided minimal risk to the subjects as it did not involve contact with patients or any intervention. The only risk possible is a loss of confidentiality which was minimized as all data was only accessed behind VINCI firewall and only accessible by essential personnel. In addition, an informed consent waiver was obtained as the patients were not in contact with the research staff and obtaining such consent for each subject would have made this study impractical to carry out.

We identified patients age 18–100 with diabetes who filled at least one OAD prescription from January 1^st, 2013 to December 31^st, 2017. Diabetes was defined as one use of the diagnostic codes ICD-9 code 250.xx or ICD-10 code E09–14.x associated with an outpatient visit with a primary care physician, or any two uses of the diagnostic codes, and outpatient prescription of a medication with Veterans Health Administration National Drug File Code for antihyperglycemic medications (HS500, HS501, HS5009).

Exposure Assessment

We assessed seven OAD drug classes and their individual impact on hospital admission for infection: Sulfonylureas, Meglitinides, Metformin, Alpha-glucosidase inhibitors, TZD’s, DPP-4 inhibitors, and SGLT2 inhibitors. Patients were classified as yes/no for each drug class. A patient who filled a prescription for a drug within any of the seven drug classes of interest at any point during the study period was classified as taking the drug class; otherwise the patient was classified as not taking the drug class.

Outcome Assessment

The primary endpoint was admission to a hospital for an infection yes/no. Admission for infection was defined based on the presence of an ICD9 or ICD10 code associated with an infectious condition on the admission or discharge problem list. A list of the ICD9 and ICD10 codes can be found on supplemental table 1 and 2. Patients with one or more admissions with an infection over the study period were counted as yes, patients with zero admissions for infection over the study period were counted as no. Patients who reached the primary endpoint within the first three months of the study period were excluded from analysis since for these patients we did not have sufficient clinical data prior to admission.

Covariates

Covariates included patient age, smoking status, birth sex, Elixhauser comorbidity index (to adjust for general health), and race. Patient age was calculated on 2/7/2020 for all subjects. Smoking status was derived from ICD codes and defined as yes/no. Subjects who had at least two diagnoses of current smoker or current tobacco user during the study period were classified as smokers; subjects who had one or zero diagnoses were considered non-smokers. Elixhauser score during the study period was calculated using ICD codes for each patient using the readmission index developed by Moore, et al.¹⁸

Statistical Analysis

Descriptive statistics were calculated for exposures and covariates. To analyze the association of each drug class with admission for infection we used seven multivariate logistic regression models, one for each drug class of interest. Odds ratios and 95% confidence intervals were calculated for each drug class to determine their association with admission for infection. For each model the outcome was hospital admission for infection (yes/no) and the main exposure was the drug class (yes/no). All models were adjusted for the same five covariates. Analyses were done in SAS 9.4 (SAS Institute Cary, NC) and R Studio v4.0.2.¹⁹

Results:

This cohort included 1,393,825 patients with diabetes. Their baseline characteristics are shown on Table 1. As expected, patients with a hospital admission for an infectious condition had a higher Elixhauser Comorbidity Index as compared to those who were never admitted for an infectious condition (14.2 vs 1.87).

Table 1.

Baseline Characteristics of not admitted vs admitted patients

Demographics
Age (SD)	70.52 (11.00)	70.65 (10.38)
Sex
Male	1,252,658 (95.7%)	81,945 (96.3%)
Female	56,064 (4.3%)	3,158 (3.7%)
Race
White	942,714 (72.0%)	63,519 (74.6%)
Black	244,305 (18.7%)	16,376 (19.2%)
Asian	11,848 (0.9%)	400 (0.5%)
American Indian or Alaskan Native	12,969 (1.0%)	923 (1.1%)
Native Hawaiian or Other Pacific Islander	14,835 (1.1%)	809 (1.0%)
Clinical Data
Smoking Status
Yes	262,588 (20.1%)	21,374 (25.1%)
No	1,046,134 (79.9%)	63,729 (74.9%)
Elixhauser Comorbidity Index (SD)	1.07(11.37)	14.2 (15.99
Requiring ICU Stay
Yes	N/A	6,400 (7.5%)
No	N/A	78,703 (92.5%)
Drug Class
Sulfonylureas
Yes	601,542 (46.0%)	38,679 (45.4%)
No	707,180 (54.0%)	46,424 (54.6%)
Alpha-glucosidase Inhibitors
Yes	27,618 (2.1%)	1,931 (2.3%)
No	1,281,104 (97.9%)	83,172 (97.7%)
Meglitinides
Yes	3,299 (0.3%)	298 (0.4%)
No	1,305,423 (99.7%)	84,805 (99.6%)
TZD’s
Yes	53,720 (4.1%)	2,858 (3.4%)
No	1,255,002 (95.9%)	82,245 (96.6%)
DPP-4 Inhibitors
Yes	99,186 (7.6%)	5,987 (7.0%)
No	1,209,536 (92.4%)	79,116 (93.0%)
SGLT2 Inhibitors
Yes	14,585 (1.1%)	870 (1.0%)
No	1,294,137 (98.9%)	84,233 (99.0%)
Metformin
Yes	976,034 (74.6%)	55,513 (65.2%)
No	332,688 (25.4%)	29,590 (34.8%)

Table 2 presents the odds ratios (ORs) for hospital admission for an infectious condition in association with each class of OAD compared to patients who were not on that class of OAD. After adjusting for age, sex, race, smoking status, and Elixhauser Comorbidity Index, those who took metformin during the study period had 3.3% lower odds of hospital admission for infection compared to those who were never on metformin (OR 0.97, 95% CI 0.95–0.98). OADs observed to have higher odds of admissions for infections as compared to non-users of that class of medication included meglitinides (OR 1.22, 95% CI 1.07–1.38), SGLT2 inhibitors (OR 1.16, 95% CI 1.08–1.24), alpha-glucosidase inhibitors (OR 1.09, 95% CI 1.04–1.15), and DPP4 inhibitors (OR 1.04, 95% CI 1.01–1.06). There was no statistically significant difference in admission for infections observed with sulfonylurea or TZD use.

Table 2.

ORs for hospital admission for an infectious condition in association with each class of OAD compared to patients who were not on that class of OAD

Meglitinides	1.22	1.07–1.38
SGLT2 Inhibitors	1.16	1.08–1.24
Alpha-glucosidase inhibitors	1.09	1.04–1.15
DPP-4 Inhibitors	1.04	1.01–1.07
Sulfonuryeas	0.99	0.97–1.00
TZDs	0.98	0.94–1.02
Metformin	0.97	0.95–0.98

Discussion:

To our knowledge, this is the largest national retrospective cohort study using a Veterans Affairs Administration database to examine the relationship between outpatient OAD use and the risk of hospitalization for infections. We found that metformin was associated with lower odds of admission for infection while meglitinides, SGLT2 inhibitors, alpha-glucosidase inhibitors, and DPP4 inhibitors were associated with higher odds of admission for infection. The effect sizes for most OADs were small with likely limited clinical relevance to individual patients. However, as OADs are widely prescribed, there could be implications at a population level.

Our results corroborate findings from a nationwide cohort of diabetic patients treated with OADs in Taiwan. This study observed lower odds for sepsis in metformin users vs never users (OR 0.80, 95% CI 0.77– 0.83) and increased odds for sepsis in meglitinide users vs never users (OR 1.32, 95% CI 1.25–1.40).¹⁷ In addition, similar observations were reported in a Swedish study where metformin therapy showed reduced risk of acidosis and serious infection in patients with diabetes (adjusted HR 0.85, 95% CI 0.74–0.97).²⁰

Emerging evidence indicates that metformin has favorable effects on health beyond those associated with improvement in glycemia. Observational studies suggest that diabetic individuals treated with metformin manifest a survival benefit even when compared to non-diabetic controls. ^{21, 22} Our findings add to the growing amount of evidence that supports metformin’s beneficial role in the prevention of infections.^{13, 17, 20, 23} In vitro studies have demonstrated that hyperglycemia disrupts host immunity via several mechanisms including the suppression of polymorphonuclear neutrophil function and intracellular opsonic and bactericidal activity.^{24, 25} However, several studies have demonstrated metformin’s effect on the integrity of host immunity independent of its effects on blood glucose levels^26–30. Metformin in itself has been shown to have some anti-microbial properties against several bacterial and viral pathogens.^{26, 27, 31} In addition, metformin has been demonstrated to counteract the inhibition of chemotaxis induced by exposure of neutrophils to lipopolysaccharide through its role in the inhibition of mitochondrial complex I and activation of 5’ adenosine monophosphate-activated protein kinase (AMPK).³² Metformin continues to be recommended as first line therapy for patients with diabetes, and our findings are consistent with this recommendation from the perspective of infection prevention.

In recent years, the addition of several newer classes of OADs has created a paradigm shift in the treatment of diabetes. As more options are available, clinicians now consider far more than hemoglobin A1c and blood glucose targets in choosing 2^nd and 3^rd step therapies after metformin. While outcome studies have clearly demonstrated the cardioprotective and renoprotective effects of SGLT-2 inhibitors in type 2 DM,^33–35 the early inconsistent findings on the risk of urinary tract infections (UTIs) generated confusion among clinicians. In 2015, after 19 cases of urosepsis and pyelonephritis was reported to the US Food and Drug Administration (FDA), the FDA revised all labels for SGLT2 inhibitor to include warnings about the risk of severe UTIs.³⁶ However, despite this early concern, 3 major outcome studies^37–39 and a recent meta-analysis including data from 86 randomized clinical trials including a total of 50,880 patients failed to report a difference in risk of UTIs when SGLT2 inhibitors were compared with placebo.⁴⁰

The findings of our study report a modest increase odds of hospital admissions for all infections in patients treated with SGLT2 inhibitors as compared to those not treated with SGLT2 inhibitors. While the initial concern for UTIs is not supported by recent data, our study suggests that there may be other infections that could contribute to the observed increased admissions for infections. Of note, the same major trials that failed to show increased risk of UTIs in SGLT2 inhibitors did report an observed 4 to 9 fold increase of mild genital mycotic infections generally responsive to topical therapy in SGLT2 inhibitor groups as compared to placebo.^37–39

Although our study found a statistically significant increase in risk of infections with DPP-4 inhibitor use, the small effect size is likely of limited clinical relevance to individual patients. Aside from its role in regulating plasma glucose, DPP-4 is known to have functions in immune regulation that give rise to concerns regarding infection risk with its inhibition.⁴¹ CD26 is the membrane bound form of DPP-4, and it is found on the surface of many cell lines including B,T, NK lymphocytes and macrophages.⁴² CD26/ DPP-4 is thought to have many potential roles in immune regulation by inducing cleavage of immunoregulating substrates,⁴³ direct lymphocyte regulation⁴⁴, and co-stimulatory effects on T-cell activation.⁴⁵ The existing clinical data on risk of infections related to DPP-4 inhibitors is conflicting. Data from clinical trials suggested an increase in upper respiratory infections (URIs) and UTIs with DDP-4 inhibitors.^{46, 47} Furthermore, an analysis of the international pharmacovigilance database VigiBase reported higher rate of URIs associated with DPP-4 inhibitors as compared to metformin.¹⁶ However several large population based observational studies have failed to show an increased risk for infections in DPP-4 inhibitor use,^{13, 17, 48, 49} and a meta-analysis published in 2018 including 74 clinical trials did not demonstrate an increased overall risk of infection with DDP-4 inhibitor use.⁴²

The main strengths of our study are the population-based design and large cohort size that included 1.39 million patients across the United States. However, observational studies of this nature have several challenges. Firstly, our VA based cohort skewed heavily towards male (95.8%) patients as compared to the general US population which is estimated to be 49.1% male.⁵⁰ Second, in our analysis, exposure to each OAD was independent to when the patient was admitted to the hospital for an infection. Although retrospective cohort studies in general cannot infer causality, this limitation in our analysis further emphasizes this point. Third although we utilized the Elixhauser Co-morbidity Index to adjust for severity of illness, our data did not factor in duration since diabetes diagnosis or severity of diabetes in general that could be a significant confounder to our findings. Furthermore, prescription of different kinds of OAD is also related to the underlying conditions, such as cardiac and renal disease, insulin usage was not considered because of the retrospective nature of the study. Finally, an additional limitation is the lack of comparability between odds ratios in this study due to variation in reference level groups. For each drug, the reference group is the patients in the study who were not on that drug at any point during the study period. As a result, the reference populations differ from drug to drug. Use of metformin was not found to be associated with use of other drugs, so we think it unlikely that its effect was confounded by the other drugs classes in this paper. Sulfonylureas were also found to be unrelated to use of other drugs. However, the five remaining drug classes were found to be associated with one another, resulting in similarity of the odds ratios due to confounding. Conclusions based on effect sizes of these five drugs are limited in this sense and will need further studies to fully account for confounding by use of other drugs. Additionally, it would also have been interesting to consider the types of infections, however, was beyond the scope of our current study. Our ongoing prospective studies will incorporate many of these details.

In conclusion, our findings suggest an association between outpatient use of certain OADs and the development of infections requiring hospital admission. Metformin was associated with lower odds of hospital admission for infection while meglitinides, SGLT2 inhibitors, alpha-glucosidase inhibitors, and DPP4 inhibitors were associated with higher odds of admission for infection. However, additional studies with time dependent exposure need to be carried out to further evaluate this association and infer causality.