Results Summary
What was the research about?
Type 2 diabetes is a long-term health problem that causes blood sugar levels to rise. Some people with diabetes also have chronic kidney disease, or CKD. In CKD, the kidneys don’t work well to remove waste from the blood.
Metformin is a type of medicine used to treat diabetes. People once thought metformin could cause serious problems for people with CKD. But studies have shown that metformin is safe for patients with mild CKD. In this study, the research team wanted to learn if using metformin is safe and effective for patients with diabetes and moderate to severe CKD.
The research team first compared health records of patients starting metformin with patients starting four other types of diabetes medicines:
- Sulfonylureas
- DPP-4 inhibitors
- GLP-1 receptor agonists
- SGLT2 inhibitors
Because the study didn’t have enough complete data for patients with diabetes and moderate to severe CKD starting these medicines, the research team looked at health records of a larger group of patients with mild to moderate CKD currently using these medicines.
What were the results?
Because of the limited data for patients starting the medicines, the research team can’t say for sure if starting metformin was as safe as, or worked better than, the other types of medicines. Metformin appeared to be similar to sulfonylureas and better than DPP-4 inhibitors in reducing average blood sugar levels.
For the larger group of patients currently using these medicines, compared with metformin, sulfonylureas had an increased risk of severe low blood sugar levels. The other medicines didn’t have the increased risk.
While using metformin, patients didn’t have an increased risk of a hospital stay for acidosis. Acidosis is a serious problem that can happen when blood sugar is too high for too long.
What did the research team do?
The research team looked at health record and insurance claims data from 2012 to 2017 for patients with diabetes and CKD. All were ages 65 and older and had Medicare. Data were collected from New York, North Carolina, and Tennessee.
For the larger group of patients currently using the medicines, the research team compared:
- 5,370 patients using metformin with 4,192 patients using sulfonylureas
- 5,471 patients using metformin with 2,524 patients using DPP-4 inhibitors
- 6,958 patients using metformin with 802 patients using GLP-1 receptor agonists
- 6,602 patients using metformin with 255 patients using SGLT2 inhibitors
Patients with diabetes gave input on the study.
What were the limits of the study?
Fewer patients who used newer medicines such as GLP-1 receptor agonists and SGLT2 inhibitors had data available than the research team expected. As a result, the team had limited ability to compare metformin with these medicines. The team also couldn’t assign patients by chance to the medicines; factors other than the type of medicine, like patient traits such as age or health, may have affected the results.
Future research could use a study design that assigns patients by chance to the medicines.
How can people use the results?
Patients with diabetes and CKD and their doctors can use the results when considering metformin.
Professional Abstract
Objective
To compare the safety and effectiveness of metformin versus other non-insulin medicines in preventing severe hypoglycemia and other adverse outcomes, and in improving glycated hemoglobin (HbA1c) levels and other clinical outcomes among patients with type 2 diabetes and chronic kidney disease (CKD)
Study Design
| Design Element | Description |
|---|---|
| Design | Observational: cohort study |
| Population | Adults ages 65 and older diagnosed with type 2 diabetes and moderate to severe CKD with EHR and Medicare insurance data in PCORI-sponsored clinical research networks in New York, North Carolina, and Tennessee; final sample size varied depending on the specific intervention, examination of the comparator pair, and the outcome |
| Interventions/ Comparators |
4 medicine pairwise comparisons:
|
| Outcomes |
Primary: severe hypoglycemia, change in HbA1c Secondary:
|
| Timeframe | Up to 4-year follow-up from time of initial prescription fill |
This retrospective cohort study compared metformin versus other non-insulin diabetes medicines in health outcomes among patients with type 2 diabetes and moderate to severe CKD.
Researchers used electronic health record (EHR) data from 2012 to 2017, linked to administrative Medicare insurance claims data from three clinical research networks. Cohorts were created independently in New York, North Carolina, and Tennessee.
The original sample included new users of one of the non-insulin diabetes medicines based on prescription claims. However, the number of patients with complete data for pairwise comparisons was too small to conduct the planned analyses. Researchers then included a large cohort of existing users of metformin with type 2 diabetes and moderate to severe CKD. The larger cohort analysis used propensity score weighting to match patients and compared:
- 5,370 patients who used metformin versus 4,192 patients who used sulfonylureas
- 5,471 patients who used metformin versus 2,524 who used dipeptidyl peptidase-4 (DPP-4) inhibitors
- 6,958 patients who used metformin versus 802 patients who used glucagon-like peptide 1 (GLP-1) receptor agonists
- 6,602 patients who used metformin versus 255 patients who used sodium-glucose cotransporter 2 (SGLT2) inhibitors
To assess safety and clinical outcomes, researchers reviewed data from EHRs and insurance claims.
Patients with diabetes provided input throughout the study.
Results
Because of the smaller than expected number of new medicine users with complete data, the study had inadequate power to detect small differences between the four medicines in pairwise comparisons but was able to detect large, statistically significant differences. The results of this analysis indicated that metformin was comparable to sulfonylureas (adjusted hazard ratio [aHR]=0.12; 95% confidence interval [CI]: -0.18, -0.42) and superior to DPP-4 inhibitors in reducing HbA1c (aHR=0.54; 95% CI: 0.16, 0.93).
In the larger cohort analysis, compared with patients using metformin, patients using sulfonylureas had an increased risk of hypoglycemia (aHR=1.73; 95% CI: 1.03, 2.93); patients using DPP-4 inhibitors, GLP-1 receptor agonists, and SGLT2 inhibitors did not. Use of metformin was not associated with increased rates of hospitalization for acidosis at any level of patients’ estimated glomerular filtration rate (eGFR).
Limitations
The small number of patients using newer medicines, such as SGLT2 inhibitors and GLP-1 receptor agonists, with complete data on study outcomes limited researchers’ ability to compare metformin with these medicines. Without randomization, this observational study cannot account for unmeasured confounding variables that may have affected the outcomes.
Conclusions and Relevance
In this study, among patients with type 2 diabetes and mild to moderate CKD, metformin appeared to be effective and did not appear to be associated with an increased risk of acidosis or reduction in kidney function.
Future Research Needs
Future research could compare metformin with newer diabetes medicines using randomized controlled study designs or observational studies with greater numbers of patient records.
COVID-19-Related Study
Examining the Effects of High Blood Sugar Levels and Non-Insulin Diabetes Medicines on Risk of Hospitalization for COVID-19 among Patients with Diabetes
Results Summary
In response to the COVID-19 public health crisis in 2020, PCORI launched an initiative to enhance existing research projects so that they could offer findings related to COVID-19. The initiative funded this study and others.
What was this COVID-19 study about?
Type 2 diabetes is a long-term illness that causes blood sugar levels to rise. Patients with diabetes may have a higher risk of severe COVID-19. Questions remain about how to reduce these risks.
The research team did two studies using health records from patients with type 2 diabetes:
- Study 1 looked at how higher HbA1c levels affected the risk of a hospital stay or death from COVID-19. HbA1c is a blood test that measures patients’ average blood sugar levels over three months.
- Study 2 looked at whether taking different diabetes medicines affected the risk of a hospital stay for COVID-19. These medicines were metformin, sulfonylureas, DPP-4 inhibitors, GLP-1 receptor agonists, and SGLT2 inhibitors.
What were the results?
Study 1. As patients’ HbA1c levels increased, the risk of a hospital stay for COVID-19 increased. The risk of dying from COVID-19 while in the hospital also increased.
Study 2. Taking different diabetes medicines didn’t change the risk of a hospital stay for COVID-19.
What did the research team do?
Study 1. The research team reviewed health records for 4,192 patients ages 65 and older with type 2 diabetes who didn’t take insulin. Patients had an HbA1c level of 5.7 percent or higher between December 2019 and March 2020. All had Medicare and lived in New York City. The team reviewed records through June 2020.
Among patients, 26 percent were White, 24 percent were Black, 4 percent were Asian, and 46 percent didn’t report a race or reported their race as other. The average age was 76, and 60 percent were women.
Study 2. The research team reviewed health records for 2,982 patients from the first study. All patients had received at least one prescription within the last year for one of the five diabetes medicines. The team looked to see if the use of each of these medicines affected patients’ risk of a hospital stay for COVID-19.
Among patients, 22 percent were White, 25 percent were Black, 4 percent were Asian, and 50 percent didn’t report a race or reported their race as other. The average age was 76, and 62 percent were women.
Patients with diabetes, caregivers of people with diabetes, and clinicians provided input during both studies.
What were the limits of the study?
These studies looked at health records. Patient traits or other factors not included in these records, such as how often patients took their medicine, may have affected the results. The studies took place in one large urban city. Results may differ for patients in other places. In study 2, the number of patients taking two of the medicines was too small to be sure of the results.
How can people use the results?
People with type 2 diabetes and their doctors can use these results when considering the risk for severe COVID-19.
Professional Abstract
In response to the COVID-19 public health crisis in 2020, PCORI launched an initiative to enhance existing research projects so that they could offer findings related to COVID-19. The initiative funded this study and others.
Background
Patients with type 2 diabetes have a higher risk of severe COVID-19 than patients without diabetes. Some studies also suggest that having elevated glycated hemoglobin (HbA1c) levels may increase the risk of hospitalization for COVID-19. Questions remain about whether different diabetes medicines affect COVID-19 outcomes, such as risk of hospitalization.
Objective
To examine the association between (1) elevated HbA1c levels and (2) use of non-insulin diabetes medicines and the risk of hospitalization among adult patients with type 2 diabetes
Study Design
| Design Element | Description |
|---|---|
| Design | Observational: cohort study |
| Population |
Electronic health records and Medicare insurance claims data from a PCORI-sponsored clinical research network in New York City
|
| Outcomes |
Study 1: hospitalization for COVID-19 (primary), in-hospital death from COVID-19 (secondary) Study 2: hospitalization for COVID-19 |
| Timeframe |
Study 1: December 15, 2019–June 15, 2020 Study 2: March 15, 2019–June 15, 2020 |
Two retrospective cohort studies examined the association between (1) baseline HbA1c levels and (2) use of non-insulin diabetes medicines and the risk of hospitalization for COVID-19 among patients with type 2 diabetes. For both studies, researchers used electronic health records and Medicare insurance claims data from a clinical research network in New York City.
Study 1 examined HbA1c levels and included 4,192 adult patients with type 2 diabetes who had at least one HbA1c measurement in the three months prior to March 15, 2020. Among patients, 26% were White, 24% were Black, 4% were Asian, and 46% did not report a race or reported their race as other. The average age was 76, and 60% were female.
Study 2 examined the use of non-insulin diabetes medicines among a subset of 2,982 patients from study 1. Patients received at least one prescription in the three months prior to March 15, 2020, for either metformin, a sulfonylurea, a DPP-4 inhibitor, a GLP-1 receptor agonist, or an SGLT2 inhibitor. Among patients, 22% were White, 25% were Black, 4% were Asian, and 50% did not report a race or reported their race as other. The average age was 76, and 62% were female. Patients had not used insulin to control their diabetes.
Patients with diabetes, caregivers, and clinicians provided input during both studies.
Results
Study 1. As patients’ HbA1c levels increased by 1%, the risk of hospitalization for COVID-19 increased by 14% (adjusted hazard ratio [aHR]: 1.14; 95% confidence interval [CI]: 1.05, 1.23) and the risk of in-hospital death increased by 18% (aHR: 1.18; CI: 1.05, 1.33).
Study 2. The use of individual, specific non-insulin diabetes medicines was not significantly associated with hospitalization for COVID-19.
Limitations
As an observational study, unknown variables like medication adherence may have affected the outcomes. This study used data from one large urban setting. Results may differ in other locations. In study 2, the sample sizes were too small for GLP-1 receptor agonists and SGLT2 inhibitors to draw reliable conclusions.
Conclusions and Relevance
In this study, among people with type 2 diabetes, higher HbA1c levels were associated with an increased risk of hospitalization for COVID-19. Use of different non-insulin diabetes medicines did not affect the risk of hospitalization.
Peer Review Summary
Peer review of PCORI-funded research helps make sure the report presents complete, balanced, and useful information about the research. It also assesses how the project addressed PCORI’s Methodology Standards. During peer review, experts read a draft report of the research and provide comments about the report. These experts may include a scientist focused on the research topic, a specialist in research methods, a patient or caregiver, and a healthcare professional. These reviewers cannot have conflicts of interest with the study.
The peer reviewers point out where the draft report may need revision. For example, they may suggest ways to improve descriptions of the conduct of the study or to clarify the connection between results and conclusions. Sometimes, awardees revise their draft reports twice or more to address all of the reviewers’ comments.
Peer reviewers commented and the researchers made changes or provided responses. Those comments and responses included the following:
- The reviewers requested justification for the researchers’ use of single imputation methods to account for missing data in their study, as well as their use of population means for the imputation rather than other patient characteristics. The researchers justified their choice by demonstrating that several publications supported the use of single imputation plus indicator variables in propensity score analyses and that multiple imputation is not as worthwhile when using point estimates rather than variance estimates. The researchers justified using the sample mean rather than a more complex estimating model using multiple patient characteristics by stating there is little evidence that the more complex analysis would better improve balance or reduce the potential for bias in the imputed data.
- The reviewers asked why the researchers did not include insulin among the medications they analyzed, especially since the researchers noted that early studies had shown that insulin was related to poorer outcomes in patients with diabetes mellitus (DM) and a COVID-19 infection. The researchers explained that they made the decision to exclude insulin early in project development because patients with DM taking insulin tended to be those who were more difficult to manage. The researchers added language in the discussion to explain that exclusion of these patients from the study means that the study results are not generalizable to insulin users.
- The reviewers commented on the inclusion of patients taking multiple antidiabetic medications in analyses along with patients on monotherapy. They advised the researchers to examine whether outcomes were different if the researchers completed sensitivity analyses using only patients on monotherapy. The researchers added this analysis to their report, but also noted that the head-to-head analysis goes through propensity score weighting which can balance the possible bias from concomitant medication use.
Final Enhancement Report
This COVID-19 study's final enhancement report is expected to be available by January 2024.
Final Research Report
This project's final research report is expected to be available by January 2024.
Journal Citations
Related Journal Citations
Peer-Review Summary
Peer review of PCORI-funded research helps make sure the report presents complete, balanced, and useful information about the research. It also assesses how the project addressed PCORI’s Methodology Standards. During peer review, experts read a draft report of the research and provide comments about the report. These experts may include a scientist focused on the research topic, a specialist in research methods, a patient or caregiver, and a healthcare professional. These reviewers cannot have conflicts of interest with the study.
The peer reviewers point out where the draft report may need revision. For example, they may suggest ways to improve descriptions of the conduct of the study or to clarify the connection between results and conclusions. Sometimes, awardees revise their draft reports twice or more to address all of the reviewers’ comments.
Peer reviewers commented and the researchers made changes or provided responses. Those comments and responses included the following:
- The reviewers questioned how the researchers modified the study design to improve study power, particularly regarding prevalent user design and intent-to-treat (ITT) analysis. The researchers explained that the ITT analysis was always planned for aim 1 sensitivity analysis, and that ITT was always planned as the primary analysis for aim 2. The prevalent user design as well as time-varying exposure designs were added early in the project not to improve study power but because the researchers were interested the long-term impact of hypoglycemia from sulfonylureas and in the effect of taking concomitant diabetes medication. They also articulated the specific limitations of prevalent user designs and time-varying approaches in the discussion section.
- The reviewers suggested further that the researchers address their rationale for the study changes given the differences these changes caused for study interpretation. The researchers added a discussion about the rationale and potential for bias that changes to the study created, especially expansion of the study sample from only new users of the medication to including prevalent users. They noted that the two groups addressed very different questions.
- The reviewers asked about the use of ‘time-varying covariates’ because it was not clear what these were or how they were analyzed. The researchers explained that this referred to exposures to antidiabetic drugs during the follow-up period so that these changes could be incorporated into the statistical model. To reduce confusion, the researchers replaced the term time-varying covariates with more descriptive language in the text.