Comparative effectiveness of metformin versus sulfonylureas on kidney function decline or death among patients with reduced kidney function: a retrospective cohort study
=========================================================================================================================================================================

* Adriana M. Hung
* Amber J. Hackstadt
* Marie R. Griffin
* Carlos G. Grijalva
* Robert A. Greevy, Jr.
* Christianne L. Roumie

## Abstract

**Background:** Diabetes often causes kidney disease. In this study, we sought to evaluate if metformin use was associated with death or kidney events in patients with diabetes and concurrent reduced kidney function.

**Methods:** We used data from the Veterans Health Administration, Medicare and National Death Index databases to assemble a national retrospective cohort of veterans who were using metformin or sulfonylureas from 2001 through 2016 and who began follow-up at an estimated glomerular filtration rate (eGFR) of less than 60 mL/min/1.73 m2. The primary composite outcome was a kidney event (i.e., 40% decline in eGFR or end-stage renal disease) or death. The secondary outcome was a kidney event (eGFR decline or end-stage renal disease). We weighted the cohort using propensity scores and used Cox proportional models to estimate the cause-specific hazard of outcomes and of treatment nonpersistence as a competing risk. We stratified follow-up into 2 periods, namely the first 360 days from the start of follow-up, and 361 days and beyond.

**Results:** In the first 360 days, the propensity score–weighted cohort included 24 883 patients who used metformin and 24 998 who used sulfonylureas. There were 33.5 (95% confidence interval [CI] 30.9–36.3) and 43.0 (95% CI 40.1–46.0) deaths or kidney events per 1000 person-years for patients who used metformin or sulfonylureas, respectively (hazard ratio [HR] 0.78, 95% CI 0.71–0.85). For the secondary outcome of kidney events, the HR was 0.94 (95% CI 0.67–1.33). In the second period from 361 days onward, the primary outcome event rate was 26.5 (95% CI 24.7–28.5) per 1000 person-years for those who used metformin, compared with 36.3 (95% CI 34.2–38.6) per 1000 person-years for those who used sulfonylureas (HR 0.73, 95% CI 0.67–0.79). Results were consistent for kidney events alone (HR 0.73, 95% CI 0.59–0.91).

**Interpretation:** Metformin use for 361 days or longer after reaching an eGFR of less than 60 mL/min/1.73 m2 was associated with decreased likelihood of kidney events or death in patients with diabetes, compared with use of sulfonylureas. Metformin provided end-organ protection, in addition to glucose control.

Type 2 diabetes is the most common cause of chronic kidney disease and end-stage renal disease worldwide.1 Metformin is considered the first-line pharmacologic treatment for type 2 diabetes based on results from the United Kingdom Prospective Diabetes Study, which showed macrovascular benefits of metformin compared with sulfonylureas, but the study was underpowered to report on renal outcomes.2,3 Metformin reduces glycated hemoglobin (HbA1c), promotes weight loss and insulin sensitivity, and reduces the long-term risk of microvascular and macrovascular complications, compared with sulfonylureas or insulin.4 Sulfonylureas affect weight change and blood pressure, both known contributors to kidney dysfunction.5 However, the protective associations between metformin and kidney function appear to be independent of changes in body mass index, blood pressure and glucose control.6 Whether the beneficial association of metformin in patients with normal renal function extends to patients with mild-to-moderate kidney disease remains unknown.

We sought to test the hypothesis that persistent metformin use is associated with lower risk of kidney events and death among patients with diabetes and reduced kidney function, compared with use of sulfonylureas.

## Methods

### Study design

We conducted a retrospective cohort study of veterans who were cared for at the veterans health care system where the study was conducted, who were using metformin or sulfonylureas from 2001 through 2016 and who began follow-up at an estimated glomerular filtration rate (eGFR) of less than 60 mL/min/1.73 m2.

### Data sources

The data sources included the United States Veterans Health Administration (VHA), with linkage via a research identifier to data from the Medicare, Medicaid and National Death Index databases from 2001 through 2016. The clinical data warehouse of the VHA includes data derived from the electronic health record, designed for research to include identifiers to link multiple data sources. Pharmacy data included dispensed VHA prescriptions, date filled, days supplied and medication dose. Diagnostic and procedure information identified inpatient and outpatient encounters using structured data from the *International Classification of Diseases, 9th Revision* and *10th Revision*, depending on what data were available for a given year. We collected laboratory and vital signs data from the clinical data warehouse.7 The Veterans Affairs Information Resource Center assembles Medicare and Medicaid data for veteran enrollees. From these files, we obtained enrolment and prescription (Part D) data. Dates of death were included in the vital statistics and in the National Death Index.

### Study population

We assembled a retrospective cohort of patients with new onset diabetes. The underlying cohort included veterans (aged ≥ 18 yr) who received regular VHA care (i.e., a VHA encounter or prescription fill at least once every 180 d) in the 2 years before cohort entry, and who had a new prescription of metformin or a sulfonylurea (including glipizide, glyburide or glimepiride), with no fill of any diabetes medication within the previous 180 days.

We followed patients longitudinally; they were required to remain persistent to their incident diabetes regimen, with medication gaps no larger than 180 days, until they reached the index date. The index date and start of follow-up was when patients reached an eGFR of less than 60 mL/min/1.73 m2 (Appendix 1, Supplemental Figure 1, available at [www.cmajopen.ca/content/11/1/E77/suppl/DC1](http://www.cmajopen.ca/content/11/1/E77/suppl/DC1)). We excluded patients who added or switched medications at or before the index date and those who had dialysis, an organ transplant or hospice enrolment within 2 years before the index date. The index date was restricted to dates between Jan. 1, 2002, and Dec. 30, 2016, to allow sufficient collection of baseline data and follow-up.

![Figure 1:](http://www.cmajopen.ca/https://www.cmajopen.ca/content/cmajo/11/1/E77/F1.medium.gif)

[Figure 1:](http://www.cmajopen.ca/content/11/1/E77/F1)

Figure 1: 
Study flowchart. Note: VA = Veterans Affairs.

### Study variables

The study exposures were continued use of metformin or sulfonylureas on the index date. Follow-up began at the date of reduced kidney function (eGFR < 60 mL/min/1.73 m2) and continued until an outcome (as defined below), a competing risk (nonpersistence) or a censoring event.

We used creatinine measures to calculate eGFR using the Chronic Kidney Disease Epidemiology Collaboration equation. 8,9 The primary outcome was a composite of death or kidney event, an outcome used by clinical trials.10 Kidney events included either a sustained decline in eGFR of 40% for 3–12 months or end-stage renal disease, defined as renal replacement therapy (including dialysis), renal transplantation or an eGFR of less than 15 mL/min/1.73 m2 (Appendix 1, Supplemental Table 1).10–12 The secondary outcome was a kidney event. To avoid capturing a single reduced eGFR measure, which may represent acute kidney injury, we used the first confirmed event as the event date.

View this table:
[Table 1:](http://www.cmajopen.ca/content/11/1/E77/T1)

Table 1: 
Patient characteristics on index date of kidney function decline and at 361 days after the index date for persistent patients

We considered treatment nonpersistence — defined as 90 days without use of an antidiabetic agent, the addition of a new agent or switch to a different agent — as a competing risk event. All-cause death and treatment nonpersistence were competing risk events for the secondary outcome (kidney event). Censoring events were reaching day 181 of no contact with a VHA facility (inpatient, outpatient or pharmacy use) or study end (Dec. 31, 2016). We included study covariates (Appendix 1, Supplemental Table 2) measured closest to the index date and up to 720 days before the index date. Missing covariates were handled with multiple imputations using 20 iterations of chained imputations. The imputed values are found using predictive mean matching13 and canonical correlation analysis.9 We included indicators for missingness to account for potential informative missingness.

View this table:
[Table 2:](http://www.cmajopen.ca/content/11/1/E77/T2)

Table 2: 
Rates and hazard ratios for kidney composite outcomes among patients who persisted on metformin or sulfonylurea in matched weighted cohort in first 360 days and from day 361 onward of reaching reduced kidney function threshold*

### Statistical analysis

Propensity scores modelled the probability of metformin continuation, given covariates, Veterans Integrated Services Network and an indicator for imputed covariates. The weighted analysis balances the covariate distributions by assigning weights to patients in both exposure groups such that the weighted groups resemble each other group (average treatment effect in evenly matchable units). Both metformin and sulfonylurea groups were weighted so that their distribution of covariates resembled each other.14,15 We derived matching weights at the index date and at 361 days. We calculated standardized mean differences as the difference between groups in number of standard deviations. We estimated the cause-specific hazard of the primary composite outcome (kidney event or death) using Cox proportional hazards models between the metformin and sulfonylurea (referent) groups in a propensity score–weighted cohort. For the regression models, we used multiple imputation with predictive mean matching and 20 imputed data sets for the propensity score estimation and the covariate-adjusted Cox proportional hazard models to address missingness in the baseline covariates (Appendix 1, Supplemental Methods).

A plot of Schoenfeld residuals against time indicated that the proportional hazards assumption in the Cox model was not met.16,17 To meet these assumptions, we divided our study into 2 time periods (index date through the first 360 d, and day 361 onward for patients persistent on their medication). Multistate models for the propensity score–weighted cohort estimated cumulative incidence in the presence of competing risks of medication nonpersistence using the Aalen–Johansen estimator (Appendix 1, Supplemental Methods).18–21

We conducted sensitivity analyses adjusted for all covariates in the propensity score–weighted cohort for the evaluation of the primary outcome of death or kidney events. We examined prespecified subgroups in the time period from 361 days onward, including groups by age (≥ 65 yr, < 65 yr), race (Black and non-Black), baseline eGFR at the index date (eGFR ≥ 45 mL/min/1.73 m2, eGFR < 45 mL/min/1.73 m2) and the use of renin–angiotensin–aldosterone system (RAAS) inhibitors (yes, no). We conducted analyses using R ([http://www.r-project.org](http://www.r-project.org)).

### Ethics approval

The institutional review board of VHA Tennessee Valley Healthcare System approved this study.

## Results

We identified 74 096 new users of metformin and 28 967 new users of sulfonylurea who reached eGFR < 60 mL/min/1.73 m2 and began the first study period (Figure 1). Table 1 shows the weighted cohort characteristics on the index date and for those who persisted on their regimen on day 361 (Appendix 1, Supplementary Table 3 shows the unweighted cohort characteristics). The weighted cohort included 24 883 patients using metformin and 24 998 patients using sulfonylureas, including glipizide (*n* = 2750, 55.0%), glyburide (*n* = 10 999, 44.0%) and glimepiride (*n* = 250, 1.0%).

On day 361, we identified 12 571 patients using metformin and 12 637 patients using sulfonylureas in the recalculated propensity score–weighted cohort. Median follow-up in the weighted cohort for the study period beginning at day 361 was 1.5 (interquartile range [IQR] 0.6–3.2) years for patients taking metformin and 1.5 (IQR 0.6–3.2) years for those using sulfonylureas. All standardized mean differences in both weighted cohorts were less than 0.10 (Table 1).

### Estimated glomerular filtration rate values

The median historical eGFR before the index date was 69.6 (IQR 64.9–78.1) mL/min/1.73 m2 and the median difference between the historical and index date eGFR values was 14.6 (IQR 9.6–23.5) mL/min/1.73 m2 for patients taking metformin and 14.6 (IQR 9.6–23.2) mL/min/1.73 m2 for those taking sulfonylureas. The median time between eGFR measures was 4.6 (IQR 2.4–7.0) months for patients taking metformin and 5.0 (IQR 2.6–7.5) months for those taking sulfonylureas. The median number of days between the index and follow-up eGFR measures was 112 days, and the median follow-up eGFR was 54 mL/min/1.73 m2. The median eGFR closest to the 361-day time point was 63.5 (IQR 55.8–72.3) mL/min/1.73 m2 for patients using metformin and 63.7 (IQR 55.7–72.3) mL/min/1.73 m2 for those using sulfonylureas.

### Outcomes

#### First 360 days of follow-up

In the first 360 days, 10 951 (44.0%) of 24 883 patients taking metformin and 9822 (39.3%) of 24 998 patients taking sulfonylureas did not persist in treatment. Nonpersistent patients on metformin stopped the drug (*n* = 6358, 25.6%) or added other drugs (*n* = 4593, 14.2%). Nonpersistent patients on sulfonylureas stopped the drug (*n* = 6263, 25.1%) or added other drugs (*n* = 3559, 14.2%). Among those treated with metformin or sulfonylurea, 265 (1.1%) and 361 (1.5%), respectively, were censored for leaving the VA; 13 090 (52.6%) and 14 029 (56.1%), respectively, were censored for reaching the end of 360-day study period.

There were 576 events (527 deaths and 49 kidney events, including 41 instances of eGFR decline, 3 eGFR events and 5 instances of end-stage renal disease) for patients treated with metformin and 786 events (730 deaths and 56 kidney events, including 44 instances of eGFR decline, 2 eGFR events and 9 instances of end-stage renal disease) for those treated with sulfonylureas, yielding 33.5 (95% confidence interval [CI] 30.9–36.3) events and 43.0 (95% CI 40.1–46.0) events per 1000 person-years, respectively. The propensity score–weighted, cause-specific hazard ratio (HR) for death and kidney events among patients treated with metformin compared with those treated with sulfonylureas was 0.78 (95% CI 0.71–0.85). Covariate adjustment to the propensity score–weighted model yielded similar results (adjusted HR 0.79, 95% CI 0.72–0.87). When evaluating the secondary outcome of kidney events, with death as a competing risk, there were 2.9 (95% CI 2.2–3.8) events and 3.1 (95% CI 2.4–4.0) events per 1000 person-years for patients treated with metformin or sulfonylurea, respectively, yielding an HR of 0.94 (95% CI 0.67–1.33). When evaluating the secondary outcome of death, there were 30.6 (95% CI 28.2–33.3) events and 40 (95% CI 37.2–42.9) events per 1000 person-years for patients treated with metformin or sulfonylurea, respectively, yielding an HR of 0.76 (95% CI 0.69–0.84) (Table 2). The cumulative incidence plots showing the competing risks are shown in Figure 2A and Figure 2B.

![Figure 2:](http://www.cmajopen.ca/https://www.cmajopen.ca/content/cmajo/11/1/E77/F2.medium.gif)

[Figure 2:](http://www.cmajopen.ca/content/11/1/E77/F2)

Figure 2: 
Full Aalen–Johansen cumulative probability plot of a kidney event (i.e., 40% decline in estimated glomerular filtration rate or end-stage renal disease) or death (panel A) or of a kidney event (panel B) in the weighted cohort for the first 360 days after reaching an estimated glomerular filtration rate less than 60 mL/min/1.73 m2 by treatment group. Note: Met = metformin, Sul = sulfonylurea.

#### Day 361 of follow-up and onward

Among the 12 571 patients who persisted on metformin for at least 361 days after the index date, there were 747 primary composite events (637 deaths and 110 kidney events, including 107 instances of eGFR decline and 3 of end-stage renal disease). Among the 12 637 patients who persisted on sulfonylureas, there were 1033 events (884 deaths and 149 kidney events, including 148 with eGFR decline and 1 with end-stage renal disease) (Table 2). Incidence rates were 26.5 (95% CI 24.7–28.5) per 1000 person-years for patients on metformin versus 36.3 (95% CI 34.2–38.6) per 1000 person-years for those on sulfonylureas, yielding an HR of 0.73 (95% CI 0.67–0.79); we observed consistent results after adjusting for covariates (adjusted HR 0.76, 95% CI 0.70–0.83). For the secondary outcome, which evaluated kidney events and treated death as a competing risk, the event rate was 3.9 (95% CI 3.2–4.7) versus 5.2 (95% CI 4.5–6.1) events per 1000 person-years for patients treated with metformin or sulfonylureas, respectively (HR 0.73, 95% CI 0.59–0.91). For the secondary outcome of death, the event rate was 22.7 (95% CI 21.0–24.5) versus 31.5 (95% CI 29.5–33.5) events per 1000 person-years for patients treated with metformin or sulfonylureas, respectively (HR 0.72, 95% CI 0.66–0.79). Figure 3A and Figure 3B show the cumulative probabilities of death and kidney events or of kidney event alone, respectively.

![Figure 3:](http://www.cmajopen.ca/https://www.cmajopen.ca/content/cmajo/11/1/E77/F3.medium.gif)

[Figure 3:](http://www.cmajopen.ca/content/11/1/E77/F3)

Figure 3: 
Full Aalen–Johansen cumulative probability of a kidney event (i.e., 40% decline in estimated glomerular filtration rate or end-stage renal disease) or death (panel A) or of a kidney event (panel B) in the weighted cohort for those who persisted on their treatment for at least 361 days after reaching an estimated glomerular filtration rate less than 60 mL/min/1.73 m2 by treatment group. Note: Met = metformin, Sul = sulfonylurea.

For those patients who persisted on their drug therapy for at least 361 days after reaching reduced kidney function, the cumulative probability of death or kidney event at 3 years was 3.8% for patients on metformin and 6.4% for those on sulfonylureas (risk difference 2.6%). The number needed to treat with metformin is 38.5 patients to prevent 1 death or kidney event. For these same patients who remained on metformin or sulfonylureas, the cumulative probability of reaching a kidney event at 5 years was 0.76% (95% CI 0.64%–0.91%) versus 1.00% (95% CI 0.87%–1.20%), respectively, and was 1.10% (95% CI 0.95%–1.30%) versus 1.50% (95% CI 1.30%–1.70%) at 10 years, respectively.

### Subgroup analysis

Results stratified by age (≥ 65 yr v. < 65 yr), race (Black v. non-Black), eGFR (≥ 45 mL/min/1.73 m2 v. < 45 mL/min/1.73 m2) and use of RAAS inhibitors (yes v. no) were consistent with the main analysis, but CIs were wide for most subgroups (Figure 4).

![Figure 4:](http://www.cmajopen.ca/https://www.cmajopen.ca/content/cmajo/11/1/E77/F4.medium.gif)

[Figure 4:](http://www.cmajopen.ca/content/11/1/E77/F4)

Figure 4: 
Propensity score–weighted hazard ratios (HRs) for the primary and secondary outcomes by subgroup for patients persistent on therapy at 361 days after reaching an estimated glomerular filtration rate (eGFR) less than 60 mL/min/1.73 m2. Note: CI = confidence interval.

## Interpretation

Diabetes is the most common condition associated with kidney disease worldwide. In this national evaluation of patients with type 2 diabetes who developed an eGFR of less than 60 mL/min/1.73 m2, continued use of metformin was associated with a decreased risk of the composite outcome of death or kidney event, compared with use of sulfonylureas. In particular, use of metformin was associated with lower risk of death in the first 360 days. Continued use of metformin past 361 days was associated with lower risk of clinically important kidney outcomes. Accordingly, the risk difference for death or a kidney event for patients treated with metformin versus sulfonylureas was 2.6%; the number needed to treat was 38.5 patients with continued metformin for 3 years to prevent 1 death or kidney event.

Our results are consistent with work that showed metformin was associated with lower rates of kidney events compared with sulfonylureas among patients with preserved kidney function.6,22 Metformin has properties that may affect the kidney, including antioxidant, anti-inflammatory, antifibrotic and insulin-sensitizing properties.23–26 Many of these properties can potentially improve endothelial function in patients with kidney disease.27 Few studies have evaluated the association of metformin on kidney function decline among patients with moderately reduced kidney function, in whom metformin is now indicated as the first-line therapy for diabetes management.28 Patients with diabetes and reduced eGFR have higher mortality (all cause and cardiovascular) when compared with patients with diabetes and preserved kidney function.28,29 Our current study findings, taken in the context of previous reports, suggest that metformin should remain the first-line agent among those who develop kidney decline.6,22,30–33

### Limitations

We required persistence on diabetes incident medication at the index date (kidney threshold) and at 361 days beyond the index date for analyses. These criteria excluded patients who stopped, added or switched medications at or before reaching the kidney threshold and limited follow-up for patients who changed their medications or died so that our outcomes could be attributed to the drug exposure. Furthermore, many factors influenced the choice of diabetes medication (sulfonylurea v. metformin monotherapy) at onset of disease during the study period, which could potentially be confounding. The study years were before 2016 when guidance suggested stopping metformin if patients reached a creatinine level of 1.4–1.5 mg/dL.34 During the same time period, guidance suggested stopping glyburide at a serum creatinine level over 2.0 mg/dL. These time trends were accounted for, but we noted that medication nonpersistence and early changes to medications were common and limited the sample size available for analysis. Veterans may not receive all their care at VHA facilities, and some events were likely missed. The kidney event relies solely on VHA-collected laboratory data. It is possible those patients older than 65 years were less likely to receive their care or bloodwork within VHA as they are eligible for care through Medicare coverage across many health care systems, leading to a systematic exclusion of some events. The kidney threshold may represent an acute kidney injury event rather than progression of chronic kidney disease. Although we used propensity score weighting to reduce concerns about confounding, this was an observational study and residual confounding may exist. Finally, the study population was mostly older white men, and may not be representative of the larger population of patients with diabetes and reduced kidney function. This should be considered when extrapolating the study results to other populations including women.

### Conclusion

Treatment with metformin in the first 360 days of reduced kidney function was associated with a lower incidence of kidney event or death in patients with diabetes, compared with sulfonylureas. Persistent treatment with metformin beyond 361 days was associated with fewer kidney events (including eGFR decline and end-stage renal disease) or deaths, compared with sulfonylureas. Furthermore, our study provides reassurance that continued use of metformin in patients with reduced kidney function supports the use of metformin as the first-line therapy for patients with mild-to-moderate kidney disease.

## Footnotes

*   **Competing interests:** Outside the submitted work, Adriana Hung reports funding from Veterans Affairs and the Vertex trial; consulting fees from the National Heart, Lung, and Blood Institute; and speaking fees from Stanford Medical Grand Rounds. Amber Hackstadt reports participation with the data safety and monitoring board for the PILOT trial, and with the independent monitoring committee for a study on telehealth and pain. Carlos Grijalva reports funding from Campbell Alliance, Syneos Health, Sanofi, the United States Centers for Disease Control and Prevention, the Agency for Healthcare Research and Quality, the National Institutes of Health and the US Food and Drug Administration. He also reports consulting fees from Merck, Pfizer and Sanofi-Pasteur. Christianne Roumie reports funding from the Center for Diabetes Translation Research. No other competing interests were declared.

*   This article has been peer reviewed.

*   **Contributors:** Adriana Hung, Amber Hackstadt, Carlos Grijalva, Robert Greevy Jr. and Christianne Roumie contributed to study conception and design, as well as data collection. Adriana Hackstadt, Robert Greevy Jr. contributed to data analysis. Adriana Hung and Christianne Roumie drafted the manuscript. All of the authors revised it critically for important intellectual content, gave final approval of the version to be published and agreed to be accountable for all aspects of the work.

*   **Funding:** This project was funded by a Veterans Affairs Clinical Science Research and Development investigator-initiated grant CX000570.

*   **Data sharing:** The protocol, statistical code, and deidentified and anonymized data set are available from the corresponding author with a written request.

*   **Disclaimer:** The Veterans Affairs Clinical Science Research and Development had no role in the design and conduct of the study; collection, management, analysis and interpretation of the data; preparation, review or approval of the manuscript; and decision to submit the manuscript for publication. The contents do not represent the views of the United States Department of Veterans Affairs or the United States Government.

*   **Supplemental information:** For reviewer comments and the original submission of this manuscript, please see [www.cmajopen.ca/content/11/1/E77/suppl/DC1](http://www.cmajopen.ca/content/11/1/E77/suppl/DC1).

This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY-NC-ND 4.0) licence, which permits use, distribution and reproduction in any medium, provided that the original publication is properly cited, the use is noncommercial (i.e., research or educational use), and no modifications or adaptations are made. See: [https://creativecommons.org/licenses/by-nc-nd/4.0/](https://creativecommons.org/licenses/by-nc-nd/4.0/)

## References

1.  United States Renal Data System (USRDS) (2018) (2018) USRDS Annual Data Report: epidemiology of kidney disease in the United States (National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases, Bethesda (MD)) Available[https://www.usrds.org/annual-data-report/](https://www.usrds.org/annual-data-report/). accessed 2021 Oct. 12.
    
    

2.  UK Prospective Diabetes Study (UKPDS) Group (1998) Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33). Lancet 352:837–53.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1016/S0140-6736(98)07019-6&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=9742976&link_type=MED&atom=%2Fcmajo%2F11%2F1%2FE77.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000075857300006&link_type=ISI) 

3.  (2017) Standards of Medical Care in Diabetes-2017: summary of revisions. Diabetes Care 40(Suppl 1):S4–5.
    
    [FREE Full Text](http://www.cmajopen.ca/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiRlVMTCI7czoxMToiam91cm5hbENvZGUiO3M6NzoiZGlhY2FyZSI7czo1OiJyZXNpZCI7czoxODoiNDAvU3VwcGxlbWVudF8xL1M0IjtzOjQ6ImF0b20iO3M6MjA6Ii9jbWFqby8xMS8xL0U3Ny5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 

4.  Holman RR, Paul SK, Bethel MA, et al. (2008) 10-year follow-up of intensive glucose control in type 2 diabetes. N Engl J Med 359:1577–89.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1056/NEJMoa0806470&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=18784090&link_type=MED&atom=%2Fcmajo%2F11%2F1%2FE77.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000259903100006&link_type=ISI) 

5.  Huizinga MM, Roumie CL, Greevy RA, et al. (2010) Glycemic and weight changes after persistent use of incident oral diabetes therapy: a Veterans Administration retrospective cohort study. Pharmacoepidemiol Drug Saf 19:1108–12.
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=20878643&link_type=MED&atom=%2Fcmajo%2F11%2F1%2FE77.atom) 

6.  Hung AM, Roumie CL, Greevy RA, et al. (2013) Kidney function decline in metformin versus sulfonylurea initiators: assessment of time-dependent contribution of weight, blood pressure, and glycemic control. Pharmacoepidemiol Drug Saf 22:623–31.
    
    

7.  Fihn SD, Francis J, Clancy C, et al. (2014) Insights from advanced analytics at the Veterans Health Administration. Health Aff (Millwood) 33:1203–11.
    
    [Abstract/FREE Full Text](http://www.cmajopen.ca/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6OToiaGVhbHRoYWZmIjtzOjU6InJlc2lkIjtzOjk6IjMzLzcvMTIwMyI7czo0OiJhdG9tIjtzOjIwOiIvY21ham8vMTEvMS9FNzcuYXRvbSI7fXM6ODoiZnJhZ21lbnQiO3M6MDoiIjt9) 

8.  Levey AS, Stevens LA, Schmid CH, et al., CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration) (2009) A new equation to estimate glomerular filtration rate. Ann Intern Med 150:604–12.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.7326/0003-4819-150-9-200905050-00006&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=19414839&link_type=MED&atom=%2Fcmajo%2F11%2F1%2FE77.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000265903800004&link_type=ISI) 

9.  van de Velden M, Takane Y (2012) Generalized canonical correlation analysis with missing values. Comput Stat 27:551–71.
    
    

10. Inker LA, Lambers Heerspink HJ, Mondal H, et al. (2014) GFR decline as an alternative end point to kidney failure in clinical trials: a meta-analysis of treatment effects from 37 randomized trials. Am J Kidney Dis 64:848–59.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1053/j.ajkd.2014.08.017&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=25441438&link_type=MED&atom=%2Fcmajo%2F11%2F1%2FE77.atom) 

11. KDOQI (2007) KDOQI clinical practice guidelines and clinical practice recommendations for diabetes and chronic kidney disease. Am J Kidney Dis 49(Suppl 2):S12–154.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1053/j.ajkd.2006.10.014&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=17276798&link_type=MED&atom=%2Fcmajo%2F11%2F1%2FE77.atom) 

12. Levey AS, Stevens LA, Coresh J (2009) Conceptual model of CKD: applications and implications. Am J Kidney Dis 53(Suppl 3):S4–16.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1053/j.ajkd.2008.07.048&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=19231760&link_type=MED&atom=%2Fcmajo%2F11%2F1%2FE77.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000263844200002&link_type=ISI) 

13. White IR, Royston P, Wood AM (2011) Multiple imputation using chained equations: issues and guidance for practice. Stat Med 30:377–99.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1002/sim.4067&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=21225900&link_type=MED&atom=%2Fcmajo%2F11%2F1%2FE77.atom) 

14. D’Agostino R Jr., Rubin D (2000) Estimating and using propensity scores with partially missing data. J Am Stat Assoc 95:749–59.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.2307/2669455&link_type=DOI) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000165591200011&link_type=ISI) 

15. Franklin JM, Eddings W, Austin PC, et al. (2017) Comparing the performance of propensity score methods in healthcare database studies with rare outcomes. Stat Med 36:1946–63.
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=28208229&link_type=MED&atom=%2Fcmajo%2F11%2F1%2FE77.atom) 

16. Grambsch P, Therneau T (1994) Proportional hazards tests and diagnostics based on weighted residuals. Biometrika 81:515–26.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1093/biomet/81.3.515&link_type=DOI) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=A1994PP36700006&link_type=ISI) 

17. Fine JP, Gray RJ (1999) A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc 94:496–509.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.2307/2670170&link_type=DOI) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000081058500019&link_type=ISI) 

18. Aalen OO, Johansen S (1978) An empirical transition matrix for non-homogeneous Markov chains based on censored observations. Scand J Stat 5:141–50.
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=A1978GW62000002&link_type=ISI) 

19. Putter H, Fiocco M, Geskus RB (2007) Tutorial in biostatistics: competing risks and multi-state models. Stat Med 26:2389–430.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1002/sim.2712&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=17031868&link_type=MED&atom=%2Fcmajo%2F11%2F1%2FE77.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000246161400007&link_type=ISI) 

20. Andersen PK, Geskus RB, de Witte T, et al. (2012) Competing risks in epidemiology: possibilities and pitfalls. Int J Epidemiol 41:861–70.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1093/ije/dyr213&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=22253319&link_type=MED&atom=%2Fcmajo%2F11%2F1%2FE77.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000306417300032&link_type=ISI) 

21. Azoulay L, Suissa S (2017) Sulfonylureas and the risks of cardiovascular events and death: a methodological meta-regression analysis of the observational studies. Diabetes Care 40:706–14.
    
    [Abstract/FREE Full Text](http://www.cmajopen.ca/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6NzoiZGlhY2FyZSI7czo1OiJyZXNpZCI7czo4OiI0MC81LzcwNiI7czo0OiJhdG9tIjtzOjIwOiIvY21ham8vMTEvMS9FNzcuYXRvbSI7fXM6ODoiZnJhZ21lbnQiO3M6MDoiIjt9) 

22. Hung AM, Roumie CL, Greevy RA, et al. (2012) Comparative effectiveness of incident oral antidiabetic drugs on kidney function. Kidney Int 81:698–706.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1038/ki.2011.444&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=22258320&link_type=MED&atom=%2Fcmajo%2F11%2F1%2FE77.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000301847900012&link_type=ISI) 

23. Morales AI, Detaille D, Prieto M, et al. (2010) Metformin prevents experimental gentamicin-induced nephropathy by a mitochondria-dependent pathway. Kidney Int 77:861–9.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1038/ki.2010.11&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=20164825&link_type=MED&atom=%2Fcmajo%2F11%2F1%2FE77.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000277169800006&link_type=ISI) 

24. Ye J, Zhu N, Sun R, et al. (2018) Metformin inhibits chemokine expression through the AMPK/NF-κB signaling pathway. J Interferon Cytokine Res 38:363–9.
    
    

25. Lin C-X, Li Y, Liang S, et al. (2019) Metformin attenuates cyclosporine A-induced renal fibrosis in rats. Transplantation 103:e285–96.
    
    

26. Satriano J, Sharma K, Blantz RC, et al. (2013) Induction of AMPK activity corrects early pathophysiological alterations in the subtotal nephrectomy model of chronic kidney disease. Am J Physiol Renal Physiol 305:F727–33.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1152/ajprenal.00293.2013&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=23825068&link_type=MED&atom=%2Fcmajo%2F11%2F1%2FE77.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000324101000014&link_type=ISI) 

27. Ramos E, Sahinoz M, Pena CO, et al. (2020) Metformin improves vascular function in CKD patients with metabolic syndrome [abstract]. J Am Soc Nephrol 31:21.
    
    

28. Kidney Disease: Improving Global Outcomes (KDIGO) Diabetes Work Group (2020) KDIGO 2020 clinical practice guideline for diabetes management in chronic kidney disease. Kidney Int 98:S1–115.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1016/j.kint.2020.06.019&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fcmajo%2F11%2F1%2FE77.atom) 

29. Tonelli M, Muntner P, Lloyd A, et al., Alberta Kidney Disease Network (2012) Risk of coronary events in people with chronic kidney disease compared with those with diabetes: a population-level cohort study. Lancet 380:807–14.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1016/S0140-6736(12)60572-8&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=22717317&link_type=MED&atom=%2Fcmajo%2F11%2F1%2FE77.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000308396300029&link_type=ISI) 

30. Hung AM, Roumie CL, Greevy RA, et al. (2016) Comparative effectiveness of second-line agents for the treatment of diabetes type 2 in preventing kidney function decline. Clin J Am Soc Nephrol 11:2177–85.
    
    [Abstract/FREE Full Text](http://www.cmajopen.ca/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6ODoiY2xpbmphc24iO3M6NToicmVzaWQiO3M6MTA6IjExLzEyLzIxNzciO3M6NDoiYXRvbSI7czoyMDoiL2NtYWpvLzExLzEvRTc3LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 

31. Roumie CL, Chipman J, Min JY, et al. (2019) Association of treatment with metformin vs sulfonylurea with major adverse cardiovascular events among patients with diabetes and reduced kidney function. JAMA 322:1167–77.
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fcmajo%2F11%2F1%2FE77.atom) 

32. Roumie CL, Hung AM, Greevy RA, et al. (2012) Comparative effectiveness of sulfonylurea and metformin monotherapy on cardiovascular events in type 2 diabetes mellitus: a cohort study. Ann Intern Med 157:601–10.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.7326/0003-4819-157-9-201211060-00003&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=23128859&link_type=MED&atom=%2Fcmajo%2F11%2F1%2FE77.atom) 

33. Charytan DM, Solomon SD, Ivanovich P, et al. (2019) Metformin use and cardiovascular events in patients with type 2 diabetes and chronic kidney disease. Diabetes Obes Metab 21:1199–208.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1111/dom.13642&link_type=DOI) 

34. FDA Drug Safety Communication (Aug 4, 2016) FDA revises warnings regarding use of the diabetes medicine metformin in certain patients with reduced kidney function [news] (US Food and Drug Administration, Silver Spring (MD)) updated April 2017.
    
    

*   © 2023 CMA Impact Inc. or its licensors