The Long-term Effects of Metformin on Patients With Type 2 Diabetic Kidney Disease

Diabetes is the leading cause of chronic kidney disease (CKD) (1,2). According to the American Diabetes Association care guidelines, metformin is considered a first-line treatment for type 2 diabetes because of its efficacy, low cost, weight neutrality, and benefits regarding cardiovascular outcomes (3–5). In patients with CKD, however, the use of metformin is not recommended due to the risk of lactic acidosis (6,7).

The risk of lactic acidosis and its fatal consequences have resulted in the withdrawal of biguanide, phenformin and buformin, from the market (8,9). However, decades of clinical experience have provided clinicians with insights into the low incidence of lactic acidosis resulting from the use of metformin (10,11). A recent Cochrane review reported a lack of evidence that metformin treatment increases the incidence of lactic acidosis compared with other antidiabetic drugs (6).

Kidney Disease: Improving Global Outcomes (KDIGO) recommended the continuation of the use of metformin in people with an estimated glomerular filtration rate (eGFR) ≥45 mL/min/1.73 m² (eGFR categories G1–G3a), a review of its use in patients with an eGFR between 30 and 44 mL/min/1.73 m² (eGFR category G3b), and its discontinuation in people with an eGFR <30 mL/min/1.73 m² (eGFR categories G4–G5) (12). Moreover, the U.S. Food and Drug Administration allows the use of metformin in individuals with an eGFR ≥45 mL/min/1.73 m² but still restricts its use in patients with an eGFR <30 mL/min/1.73 m² (13). Metformin use in patients with an eGFR between 30 and 45 mL/min/1.73 m² is controversial (12–14).

Several recent studies revealed no difference in the number of lactic acidosis events between patients with CKD who were using metformin and those using other antidiabetic drugs (15–19). However, few studies have estimated the long-term advantages of metformin use, and the results are controversial (16,19–21). We performed a retrospective study with the hypothesis that metformin administration to advanced CKD patients can be beneficial in terms of all-cause mortality and incident end-stage renal disease (ESRD) and cannot increase the incidence of lactic acidosis.

We performed a retrospective observational cohort study of patients with type 2 diabetes who were followed at the nephrology clinics of two tertiary hospitals in South Korea (Seoul National University Hospital and Seoul National University Boramae Medical Center). Deidentified patient data retrieved from electronic medical records were used, including the date of birth, sex, BMI, diagnostic codes according to the ICD-10-Clinical Modification, drug prescriptions, and laboratory results. The follow-up period for each patient was defined as the interval between the first and last dates of creatinine measurements. From 1 January 2001 to 31 December 2016, 11,677 patients were investigated (ICD-E11, -E13, and -E14). We excluded patients with missing serum creatinine levels (n = 10), patients with short follow-up periods (fewer than 90 days of follow-up, n = 306) and patients who received renal replacement (including hemodialysis, peritoneal dialysis, and kidney transplantation) before or within 30 days of the first visit (n = 499). Finally, 10,862 patients were included (Supplementary Fig. 1).

A metformin user was defined as a patient who was prescribed metformin for longer than 90 days during the follow-up period. The start day was defined as the first prescription date, and the stop date was defined as the last prescription date plus the last prescribed period. The definition criteria were also applied to sulfonylurea and insulin users. The eGFR was calculated from the serum creatinine level using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation (22), and patients were divided into three groups based on the cutoff eGFR values of 30 and 45 mL/min/1.73 m². HbA_1c levels were obtained to assess glycemic control.

This study was approved by the institutional review board of Seoul National University Hospital (no. 20180105/10–2018–4/021), and the requirement for informed consent was waived due to the study’s retrospective design. All clinical investigations were conducted in accordance with the guidelines of the 2013 Declaration of Helsinki.

The primary outcomes were all-cause mortality and progression to ESRD. The information on death before ESRD was obtained from the National Statistical Office of Korea. ESRD was defined as a requirement of dialysis longer than 3 months (either hemodialysis or peritoneal dialysis) or preemptive renal transplantation. The 3-month limit was designed to rule out acute renal failure that temporarily requires dialysis. The incident ESRD information was ascertained through the Korean Society of Nephrology Database and electronic medical records of enrolled hospitals. Both primary outcomes were censored by the date of the last serum creatinine measurement.

The secondary outcome was drug-induced acidosis events. We defined a lactic acidosis event when a serum lactate concentration >5.0 mmol/L and serum pH <7.35 were simultaneously recorded. If more than 1 month had elapsed between two lactic acidosis events, the second event was considered a separate event. After reviewing patient charts for lactic acidosis events, we excluded patients who had other simultaneous causes of lactic acidosis (e.g., sepsis, cardiogenic shock, and hepatic failure) for drug-induced lactic acidosis events.

We used the χ² test for categorical variables and unpaired Student t tests for continuous variables to compare the baseline characteristics. We report the categorical variables as percentages of all patients and continuous variables as the means ± SDs. A negative binomial regression analysis was conducted to compare adverse events due to overdispersion. Multivariate Cox proportional hazards models were used to calculate the hazard ratios (HRs) and 95% CIs for all-cause mortality and renal outcomes. Additionally, we compared the primary outcomes among the three groups divided by eGFR using Kaplan-Meier curves and multivariate Cox proportional hazards models. A penalized spline curve was used to summarize the effect of the metformin administration duration on metformin users compared with nonmetformin users. We used a multivariate Cox regression to analyze the penalized spline curve with full adjustment (i.e., age, sex, BMI, hypertension, liver disease, initial eGFR, initial HbA_1c level, presence of proteinuria, and medication usage).

The propensity scores were estimated using logistic regression analyses and were adjusted for the patient’s age, sex, BMI, hypertension, liver disease, initial eGFR, initial HbA_1c level, presence of proteinuria, and medication usage (sulfonylurea, insulin, angiotensin II receptor blocker, and ACE inhibitor). We performed propensity score matching (PSM) using the MatchIt package by applying the nearest-neighbor algorithm with 1:1 matching and discarding in both groups. Statistically significant differences in each variable after matching were tested using the χ² test and unpaired Student t tests. We also assessed the standardized difference of each covariate, which defined the difference of the means or proportions.

All statistical analyses were performed using R, version 3.5.0 (Comprehensive R Archive Network: https://cran.r-project.org) and SPSS, version 22.0 (Armonk, NY). In all analyses, P < 0.05 was considered statistically significant.

According to the basic criterion of metformin usage for 90 days, 4,597 metformin users and 6,265 nonmetformin users were included in the current study. The median follow-up period was 7.3 ± 4.8 years in the whole population: 8.7 ± 4.5 years for metformin users and 6.3 ± 4.7 years for nonmetformin users (Table 1).

Before matching, the two groups had substantially different characteristics. Compared with nonmetformin users, metformin users had a higher mean BMI, higher initial and mean HbA_1c levels, and better initial mean eGFR. Metformin users took more sulfonylurea than nonmetformin users, but no difference in insulin and renin-angiotensin system (RAS) blockade usage was observed between the two groups. After PSM for age, sex, BMI, hypertension, liver disease, initial eGFR, initial HbA_1c level, the presence of proteinuria according to the dipstick tests, and other medication usage, the differences in baseline characteristics disappeared (Table 1).

In total, 634 patients (13.8%) in the metformin group and 1,678 patients in the nonmetformin group (26.8%) died during the follow-up period. All-cause mortality was significantly lower in the metformin group than in the nonmetformin group according to the multivariate Cox analysis (adjusted hazard ratio [aHR] 0.58; 95% CI 0.52–0.64; P < 0.001) (Table 2). Also, 522 patients (11.4%) in the metformin group and 1,533 patients (24.5%) in the nonmetformin group progressed to ESRD. Compared with the nonmetformin group, the metformin group was less likely to develop ESRD (aHR 0.66; 95% CI 0.59–0.76; P < 0.001) (Table 2).

We performed a subgroup analysis stratified according to renal function. We divided the patients into three groups based on eGFR cutoff values of 30 and 45 mL/min/1.73 m², which represented CKD stages better than 3A and 3B and worse than 4. The fully adjusted multivariate Cox regression analysis was conducted for all eGFR groups. Compared with the nonmetformin users, metformin users had lower all-cause mortality rates: eGFR ≥45 mL/min/1.73 m² (aHR 0.61; 95% CI 0.54–0.70; P < 0.001); 30 mL/min/1.73 m² ≤ eGFR < 45 mL/min/1.73 m² (aHR 0.58; 95% CI 0.45–0.75; P < 0.001); and eGFR ≤30 mL/min/1.73 m² (aHR 0.49; 95% CI 0.35–0.69; P < 0.001) (Table 2) after full adjustment. Additionally, metformin users had a significantly lower incidence of ESRD in the CKD stage than in the 3A stage: eGFR ≥45 mL/min/1.73 m² (aHR 0.62; 95% CI 0.53–0.73; P < 0.001); 30 mL/min/1.73 m² ≤ eGFR < 45 mL/min/1.73 m² (aHR 0.77; 95% CI 0.58–1.00; P = 0.052); and eGFR ≤30 mL/min/1.73 m² (aHR 0.88; 95% CI 0.71–1.10; P = 0.257) (Table 2).

In the Kaplan-Meier analysis, the metformin group had a lower incidence of all-cause mortality and ESRD progression than the nonmetformin group (Fig. 1 and Supplementary Figs. 2 and 3).

To balance the differences in baseline characteristics, we performed PSM. Metformin users still had lower all-cause mortality (aHR 0.65; 95% CI 0.57–0.73; P < 0.001) and progression to ESRD (aHR 0.67; 95% CI 0.58–0.77; P < 0.001) (Table 2). Metformin usage was associated with low all-cause mortality in all three subgroups even after a fully adjusted multivariate Cox regression model: eGFR ≥45 mL/min/1.73 m² (aHR 0.70; 95% CI 0.60–0.82; P < 0.001); 30 mL/min/1.73 m² ≤ eGFR < 45 mL/min/1.73 m² (aHR 0.64; 95% CI 0.47–0.86; P = 0.003); and eGFR ≤30 mL/min/1.73 m² (aHR 0.55; 95% CI 0.37–0.81; P = 0.003) (Table 2). Metformin users with an eGFR >30 mL/min/1.73 m² had a significantly lower incidence of ESRD progression; eGFR ≥45 mL/min/1.73 m² (aHR 0.62; 95% CI 0.51–0.76; P < 0.001); 30 mL/min/1.73 m² ≤ eGFR < 45 mL/min/1.73 m² (aHR 0.73; 95% CI 0.54–0.99; P = 0.049); and eGFR ≤30 mL/min/1.73 m² (aHR 0.87; 95% CI 0.67–1.12; P = 0.278) (Table 2).

When performing a subgroup analysis of the PSM cohort, the metformin group still had a lower risk of all-cause mortality and ESRD progression (Supplementary Figs. 4 and 5). Metformin usage showed more risk reduction in obese patients in all-cause mortality and ESRD progression. Patients who used metformin and sulfonylurea at the same time benefited more than those who used metformin alone. In contrast, compared with metformin therapy alone, the simultaneous use of metformin with insulin or RAS blockers had less benefit.

We additionally constructed a penalized spline to investigate the trend between the predicted risks of each outcome according to the duration of metformin administration. In the whole population without classification according to eGFR, the overall mortality rate was decreased in patients treated with metformin for more than 2.7 years, and the ESRD progression rate decreased in patients treated with metformin for more than 2.5 years (Fig. 2A and E). When a subgroup analysis stratified by an eGFR of 30 mL/min/1.73 m² and 45 mL/min/1.73 m² was performed, longer metformin usage was associated with a lower risk of all-cause mortality and ESRD progression, with the exception of the risk of ESRD progression in the group with an eGFR <30 mL/min/1.73 m². Additionally, patients with more advanced CKD tended to experience more benefit from a shorter metformin treatment duration in both primary outcomes. The correlations between the daily mean metformin dose and the predicted risks of all-cause mortality and ESRD progression are presented in Supplementary Fig. 6.

Overall, 249 lactic acidosis events were recorded for 228 patients, regardless of the cause of lactic acidosis. When we reviewed all 249 events, only one patient experienced a metformin-induced lactic acidosis event. The patient was 80 years old, and she had CKD stage 4 (eGFR of ∼25 mL/min/1.73 m²) at the time of drug-induced lactic acidosis. She had started metformin 20 months before the event, when her eGFR was 40 mL/min/1.73 m².

Because only one drug-induced lactic acidosis event was recorded, we additionally analyzed the risk of lactic acidosis events from all causes for patients receiving metformin, sulfonylurea, and insulin treatment. In all patients without PSM, metformin and sulfonylurea did not increase the risk of lactic acidosis events from all causes: metformin (aHR 0.92; 95% CI 0.668–1.276; P = 0.629) and sulfonylurea (aHR 1.25; 95% CI 0.924–1.697; P = 0.147) (Supplementary Table 1). In contrast, insulin usage was associated with a greater risk (aHR 5.13; 95% CI 3.543–7.421; P < 0.001). Even when analyzed in a PSM cohort, metformin did not increase the risk of lactic acidosis from all causes: metformin (aHR 0.80; 95% CI 0.546–1.162; P = 0.238), sulfonylurea (aHR 1.06; 95% CI 0.710–1.580; P = 0.777), and insulin (aHR 5.37; 95% CI 3.222–8.954; P < 0.001) (Supplementary Table 1). Additionally, there was no difference in the maximal lactate level between metformin users and nonmetformin users (P = 0.440).

Because glycemic control can affect patient outcomes, we conducted two sensitivity analyses of glycemic control. First, we adjusted HbA_1c levels as a time-varying covariate based on a fully adjusted multivariate Cox regression model. Metformin users with an eGFR >30 mL/min/1.73 m² still had low all-cause mortality and low ESRD incidence (Supplementary Tables 2 and 3).

Second, we investigated severe hypoglycemic events that caused an emergency department visit. A total of 535 events were recorded in the whole cohort, and 294 events were recorded in the PSM cohort. Metformin did not increase the number of severe hypoglycemic events for either the whole population (HR 0.83; 95% CI 0.653–1.051; P = 0.121) or the PSM cohort (HR 0.83; 95% CI 0.627–1.109; P = 0.211) (Supplementary Table 4).

In the current study, metformin usage was associated with a lower risk of all-cause mortality and ESRD progression in CKD patients, particularly those with an eGFR >30 mL/min/1.73 m². Because baseline characteristics were significantly different between metformin users and nonmetformin users, we performed subgroup analyses and PSM to close the gap. In patients whose eGFR was >30 mL/min/1.73 m², metformin users still had lower risks at both primary outcomes. To the best of our knowledge, this study is the first to show the beneficial effects of metformin on a large Asian population with a long-term follow-up period.

An observational study of patients with type 2 diabetes and stage 5 CKD by Hung et al. (20) showed that metformin usage did not increase the risk of metabolic acidosis. Moreover, metformin users were less likely to develop ESRD even after accounting for competing risks of mortality. However, because the authors showed a dose-dependent increase in all-cause mortality, they did not recommend a metformin prescription for patients whose serum creatinine concentration is >530 μmol/L (6.00 mg/dL). Although this well-organized study presented meaningful results, a limitation in its clinical application exists, as the authors only focused on patients with stage 5 CKD. Because we analyzed patients with various stages of CKD, the current study is likely to be more clinically informative. Although this Taiwanese study and many other studies (15–19) have shown a low incidence of metformin-associated lactic acidosis, a debate about metformin usage in patients with advanced CKD persists. Our study showed a low incidence of metformin-associated acidosis events. Moreover, compared with other antidiabetic drugs, metformin usage did not increase the incidence of all-cause lactic acidosis events. These results imply that compared with other antidiabetic agents, metformin may not increase the risk of lactic acidosis in CKD patients.

Our results support the findings from a previous nationwide observational study in Sweden suggesting protective effects of metformin on cardiovascular disease (CVD) and all-cause mortality in patients with an eGFR of 45–60 mL/min/1.73 m², without increases in acidosis or serious infection (16). Moreover, there were no increased risks of all-cause mortality, CVD, acidosis, or serious infection in patients with an eGFR of 30–45 mL/min/1.73 m². As a nationwide cohort study, the authors presented subanalysis results about drug usage (metformin, other oral hypoglycemic drugs, insulin, and combinations of drugs) and outcomes (CVD, acidosis, infection, and all-cause mortality), although the median follow-up period was short (3.9 years). In the subgroup analysis presented in the Swedish study, patients who used metformin alone had lower risks of all-cause mortality and CVD than patients treated with monotherapy of other drugs (other oral hypoglycemic agents and insulin), and metformin monotherapy had a better outcome than the combination therapy with other drugs.

However, as an observational study, the baseline characteristics of each treatment group were unbalanced. The metformin monotherapy group presented a lower initial HbA_1c concentration, higher eGFR, and lower microalbuminuria than the other groups. For a balanced comparison, a subgroup analysis was performed in the PSM group in the current study. Compared with the patients who received combination treatment with insulin or RAS blockers, those who received metformin-only treatment had a reduced risk of all-cause mortality and ESRD progression. In contrast, combination therapy with metformin and sulfonylurea reduced the risks further than the metformin-only treatment. This different result of combination therapy with metformin might come from remaining biases even after PSM, and further investigations are needed.

Recently, the Trial to Reduce Cardiovascular Events with Aranesp (darbepoetin α) Therapy (TREAT) study group reported that metformin usage independently reduced the risks of all-cause mortality, cardiovascular death, cardiovascular composite, and kidney disease composite (ESRD or death) in patients with an eGFR of 20–60 mL/min/1.73 m² (23). In the TREAT study, the combined end point was lower in metformin users than in nonusers. However, metformin use did not significantly reduce the risk of the ESRD-only outcome (aHR 1.01; 95% CI 0.65–1.55; P = 0.98), which is contrary to the current study. Considering the results of our sensitivity analysis of metformin prescription duration, we thought that the duration could explain this opposite result about ESRD progression in the TREAT study and the current study. At an eGFR of 30–45 mL/min/m², treatment with metformin for more than 2.6 years (when considering the 95% CI, it was 4.5 years) was needed to reduce the risk. The mean follow-up period was 2.5 years, and the maximal follow-up period was 4.5 years in the TREAT study, which might be too short to benefit from metformin usage based on the current study. Moreover, our results from patients with an eGFR <30 mL/min/1.73 m² (Fig. 2D and H) suggest a possible effect of metformin usage on reducing the risks of all-cause mortality and ESRD progression, although the significance was very low due to the small number of patients (n = 208).

Because a prospective randomized cohort study is difficult to perform, researchers have used many different approaches to clarify the benefits of metformin use. In addition, many researchers are wary of accepting the benefits of metformin because the exact mechanism of action of metformin has not been clearly elucidated. However, many in vitro and in vivo models have recently revealed the pleiotropic beneficial action of metformin. Currently, CKD is understood to be a process of glomerulosclerosis and tubulointerstitial fibrosis involving the epithelial-to-mesenchymal transition, regardless of cause (24,25). Metformin has the potential to attenuate tubulointerstitial fibrosis and the epithelial-to-mesenchymal transition by activating AMPK and downregulating transforming growth factor-β1 (26–30). Additionally, Neven et al. (31) reported that metformin not only prevented the development of severe CKD but also preserved calcium phosphorus homeostasis, which is related to CKD mineral and bond disorder. When they treated adenine-induced and warfarin-induced CKD rats with 200 mg/kg/day metformin compared with vehicle treatment as a control, the metformin-treated rats showed lower serum creatinine, phosphorus, and parathyroid hormone concentrations. Additionally, the metformin group showed less renal cellular infiltration, renal fibrosis, vascular calcification, and progression to high bone turnover status at pathology.

After a novel assay was introduced to calculate the metformin concentration, a few studies regarding metformin concentration in CKD patients were conducted (31–35). Recently, Lalau et al. (34) reported an interesting study about metformin dose and safety validation in advanced CKD patients. In the dose-finding study, there was a significant inverse correlation between eGFR and metformin levels. On the basis of this result, they selected a daily metformin dose of 1,500 mg in CKD stage 3A, 1,000 mg in CKD stage 3B, and 500 mg in CKD stage 4 for a 4-month validation. In the validation period, only one case of lactic acidosis was defined, but the patient had myocardial infarction. Also, the peak metformin concentration in all patients was quite below the U.S. Food and Drug Administration’s maximum plasma safety concentration of 5 mg/L. Additionally, there was no correlation between metformin and lactate concentration. Even though more validations are needed, this study presents the possibility of metformin usage in advanced CKD patients with a reduced dose.

Although our results are informative, our study has several limitations. First, as a retrospective cohort study, the collection and summarization of baseline characteristics and prescriptions during the follow-up period was difficult. Additionally, the prescription and actual drug history could be different. Second, even though we tried to balance the baseline characteristics by PSM, we could not balance unmeasured characteristics and confounders, which is the most important limitation of PSM. Additionally, this study cannot escape the possibility of initial selection bias induced by PSM. Third, although we tried to present duration- and dose-dependent effects of metformin on reducing risks, the cumulative metformin dose in each person was difficult to calculate. To overcome these limitations and to reflect this result in real-world practice, further well-organized randomized controlled trials are necessary. However, the current study is meaningful in that it conveyed a need for further randomized controlled trials about metformin usage in advanced CKD patients, which has been a controversial topic among clinicians for a long time.

In this retrospective study of Korean DKD patients, we support a recent trend that metformin can be considered in CKD 3B patients because of its association with decreasing all-cause mortality and delaying ESRD progression and because of its association with a low incidence of lactic acidosis. However, further randomized controlled trials are needed to change real-world practice.

Funding. This research project was supported by a grant from the Korea Health Technology R&D Project through the Korea Health Industry Development Institute funded by the Ministry of Health and Welfare, Republic of Korea (grant number HC15C1129).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. All authors contributed to this study and made substantial contributions to the conception and design of the study, acquisition of data, or analysis and interpretation of data. All authors participated in drafting the article and revising it critically for important intellectual content. All authors provided their final approval of this version of the manuscript for publication. J.P.L. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented as an oral presentation at the American Society of Nephrology's Kidney Week, San Diego, CA, 23–28 October 2018.