Skip to main content
PMC home page
Diabetes Care logoLink to Diabetes Care
. 2020 Apr 23;43(7):1462–1470. doi: 10.2337/dc19-2391

Hospitalization for Lactic Acidosis Among Patients With Reduced Kidney Function Treated With Metformin or Sulfonylureas

Patricia Y Chu 1,2, Amber J Hackstadt 3, Jonathan Chipman 4, Marie R Griffin 1,2,5, Adriana M Hung 1,2, Robert A Greevy Jr 3, Carlos G Grijalva 1,5, Tom Elasy 1,2, Christianne L Roumie 1,2,✉
PMCID: PMC7305006  PMID: 32327421

Abstract

OBJECTIVE

To compare the risk of lactic acidosis hospitalization between patients treated with metformin versus sulfonylureas following development of reduced kidney function.

RESEARCH DESIGN AND METHODS

This retrospective cohort combined data from the National Veterans Health Administration, Medicare, Medicaid, and the National Death Index. New users of metformin or sulfonylureas were followed from development of reduced kidney function (estimated glomerular filtration rate [eGFR] <60 mL/min/1.73 m2 or serum creatinine ≥1.4 mg/dL [female] or 1.5 mg/dL [male]) through hospitalization for lactic acidosis, death, loss to follow-up, or study end. Lactic acidosis hospitalization was defined as a composite of primary discharge diagnosis or laboratory-confirmed lactic acidosis (lactic acid ≥2.5 mmol/L and either arterial blood pH <7.35 or serum bicarbonate ≤19 mmol/L within 24 h of admission). We report the cause-specific hazard of lactic acidosis hospitalization between metformin and sulfonylureas from a propensity score–matched weighted cohort and conduct an additional competing risks analysis to account for treatment change and death.

RESULTS

The weighted cohort included 24,542 metformin users and 24,662 sulfonylurea users who developed reduced kidney function (median age 70 years, median eGFR 55.8 mL/min/1.73 m2). There were 4.18 (95% CI 3.63, 4.81) vs. 3.69 (3.19, 4.27) lactic acidosis hospitalizations per 1,000 person-years among metformin and sulfonylurea users, respectively (adjusted hazard ratio [aHR] 1.21 [95% CI 0.99, 1.50]). Results were consistent for both primary discharge diagnosis (aHR 1.11 [0.87, 1.44]) and laboratory-confirmed lactic acidosis (1.25 [0.92, 1.70]).

CONCLUSIONS

Among veterans with diabetes who developed reduced kidney function, occurrence of lactic acidosis hospitalization was uncommon and not statistically different between patients who continued metformin and those patients who continued sulfonylureas.

Introduction

Metformin is considered first-line pharmacologic treatment for type 2 diabetes partly on the basis of the publication of the UK Prospective Diabetes Study (UKPDS) in 1998 (1,2). In addition to reducing glycated hemoglobin (HbA1c) and microvascular complications, metformin users experience weight loss, enhanced insulin sensitivity, and reduced incidence of long-term macrovascular complications compared with sulfonylureas or insulin (2–6).

Metformin was approved in 1994 by the U.S. Food and Drug Administration (FDA) with a black box warning about lactic acidosis, and it was considered contraindicated for patients with serum creatinine ≥1.5 mg/dL in males or ≥1.4 mg/dL in females (7). The metformin label also listed heart failure and other hypoxic states under warnings and precautions because of an increased risk of lactic acidosis (7). These concerns surrounding metformin-associated lactic acidosis were based on the clinical experience with phenformin and buformin, other medications in the biguanide class (8,9). By the 1970s, there was evidence that phenformin and buformin use was associated with lactic acidosis, and they were withdrawn from the U.S. market in 1978 (9). On the basis of accumulating observational evidence on metformin safety, the FDA directed the metformin label to be revised in 2016 such that the contraindication was limited to severe kidney impairment defined as an estimated glomerular filtration rate (eGFR) <30 mL/min/1.73 m2. While there have been prospective studies evaluating lactic acidosis in patients with normal kidney function taking metformin, the evidence supporting the safety of metformin use among patients with reduced kidney function is limited to studies with a small number of events or lack of laboratory confirmation of lactic acidosis (2,10). The aim of the current study was to compare the association of continued use of metformin or sulfonylureas with lactic acidosis hospitalization among patients with type 2 diabetes who developed mild to moderate kidney disease.

Research Design and Methods

Study Design and Data Sources

We assembled a retrospective cohort of Veterans Health Administration (VHA) patients. Pharmacy data included dispensed prescriptions, medication name, date filled, days supplied, and dose. Demographic, diagnostic, and procedure information identified inpatient and outpatient encounters. We collected laboratory results and vital signs data from clinical sources. For Medicare or Medicaid enrollees, we obtained enrollment, claims files, and prescription (Part D) data. We obtained dates and causes of death from vital status and National Death Index files. The institutional review board of VHA Tennessee Valley Healthcare System approved this study.

Study Population

The population included veterans age ≥18 years who received regular VHA care, defined as having at least one medical encounter every 365 days for ≥2 years before cohort entry. We identified patients who were new users of metformin, glipizide, glyburide, or glimepiride. New users were patients who filled a first hypoglycemic prescription without any diabetes drug fill in the 180 days before that first fill. Patients were required to persist on this hypoglycemic medication with medication gaps no larger than 180 days until they reached the date of cohort entry. The date of cohort entry was the date of reaching a reduced kidney function threshold (Supplementary Fig. 1), defined as either an eGFR of <60 mL/min/1.73 m2 or serum creatinine of ≥1.5 mg/dL for males or 1.4 mg/dL for females. Cohort entry was restricted to the period between 1 January 2002 and 30 December 2015 to allow sufficient collection of baseline data and follow-up. We excluded patients who added or switched medications, had a single episode of dialysis, had an organ transplant, or enrolled in hospice at or within the 2 years before reaching the reduced kidney function threshold.

Exposure

The study exposures were continued metformin or sulfonylurea use after reaching the reduced kidney function threshold. Sulfonylurea use included use of second-generation sulfonylureas glyburide, glipizide, or glimepiride. Follow-up began on the date when the reduced kidney function (eGFR <60 mL/min/1.73 m2 or serum creatinine of ≥1.5 mg/dL for males or 1.4 mg/dL for females) was fulfilled. The follow-up period ended when the outcome of lactic acidosis hospitalization occurred (defined below). Follow-up also ended if there was a censoring event or a competing risk event. The censoring events, which are not influenced by the treatment, were the 181st day of no contact with any VHA facility (inpatient, outpatient, or pharmacy use) and end of study (31 December 2016). The competing risk events, which are potentially influenced by treatment, were death and therapy nonpersistence, defined as 90 days without the antidiabetic agent or filling an antidiabetic drug other than the current therapy.

The daily dose of metformin or sulfonylureas at the time of reaching kidney function decline was calculated using the World Health Organization criteria for defined daily dose criteria. Daily dose was determined by multiplying the number of pills dispensed by the dose per pill prescribed divided by the recorded days of supply. If drug dose was missing or was a liquid formulation (<2%), the dose was considered missing. If days of supply was missing, it was assumed to be 90 days because 70% of the population received 90-day prescriptions (11). One defined daily dose was 2,000 mg for metformin, 10 mg for glipizide, 10 mg for glyburide, and 2 mg for glimepiride (12).

Outcome: Hospitalization for Lactic Acidosis

The primary outcome was hospitalization for lactic acidosis, which was defined as a composite of two event types. The first event was a hospitalization with a primary discharge diagnosis code for lactic acidosis using ICD-9 or ICD-10 before or after 1 October 2015. Codes included ICD-9 276.2 or ICD-10 E87.2 (13,14). The second event was a hospitalization with laboratory evidence of lactic acidosis defined as a measured lactic acid ≥2.5 mmol/L in combination with either a pH <7.35 on arterial blood gas or a serum bicarbonate ≤19 mmol/L within 24 h before or after the hospital admission. The outcome date was the date of hospital admission. Composite event types were mutually exclusive, and if a patient had both a discharge diagnosis code and a laboratory-confirmed diagnosis of lactic acidosis, the laboratory-confirmed event was taken as the event type.

Covariates

Study covariates were included as the closest to cohort entry and were measured up to 720 days before the date of reaching the reduced kidney function threshold and included age, sex, race, fiscal year, number of months from initial antidiabetic medication start to reaching the reduced kidney function threshold, and Veterans Integrated Service Networks of care. Physiologic variables were evaluated as the most recent measure in the 720 days before reduced kidney function threshold and included BMI, blood pressure, HbA1c, LDL levels, hemoglobin, presence of proteinuria, and creatinine (both historical and level at cohort entry). Creatinine was used to calculate eGFR using the Chronic Kidney Disease Epidemiology Collaboration equation (14,15). Health care utilization was measured in the year before the reduced kidney function threshold (hospitalization, nursing home, number of outpatient visits or medications, and additional insurance use, including Medicare or Medicaid). Selected medications filled within 180 days of cohort entry and identified through pharmacy claims, smoking, and other comorbidities defined in Supplementary Table 1 were covariates.

Statistical Analyses

The analysis used a propensity score–weighted model. The propensity score modeled the probability of metformin or sulfonylurea use at the time of the reduced kidney function threshold and included covariates, Veterans Integrated Service Networks, and an indicator of missing covariates. Missing covariates were handled with multiple imputations using 20 iterations of chained imputations and adjusting for canonical variates. We used matching weights derived from the propensity score to balance both exposure groups on observed covariates (16) (Supplementary Description of Propensity Score Model and Weighting, Supplementary Table 2, and Supplementary Figs. 2–4). Standardized mean differences (SMDs) were calculated as the difference between groups in number of SDs because this is the preferred measure of covariate balance when dealing with large sample sizes. Smaller SMD values indicate less difference between groups, with 0 indicating perfect balance in mean or proportion.

Cox proportional hazards models estimated the cause-specific hazard ratios (HRs) for metformin versus sulfonylureas (referent) in the weighted cohort and adjusted for the above covariates, including the mean metformin dose, to allow for possible dose effects on the hazard of lactic acidosis hospitalization. Statistical significance for the two-sided P value was set at 0.05. Fulfillment of the proportional hazards assumptions was verified through examination of Schoenfeld residuals over time (17). The Cox model estimates the relative increase in hazard of lactic acidosis associated with metformin exposure compared with sulfonylureas. In the presence of competing risks, it should not be used to estimate an absolute effect. Therefore, cumulative incidence plots used the Aalen-Johansen estimator. The outcome (lactic acidosis) and the competing risks (nonpersistence and death) were treated as terminal states to allow estimation of the cumulative incidence, while the underlying population at risk was changing over time because of nonpersistence and death. End of study and no VHA contact remained as censoring events. The cause-specific HR and cumulative incidence plot were compared for consistency regarding the overall clinical interpretation of the associations with metformin versus sulfonylureas.

Sensitivity and Subgroup Analyses

Sensitivity analysis varied the definition of the outcome or the population. First, the laboratory-confirmed lactic acidosis definition was made more specific. We maintained the requirement for arterial blood pH <7.35 or serum bicarbonate <19 mmol/L within 24 h of admission but required a lactate value ≥5 mmol/L (18). Second, we restricted the population to those with an eGFR <60 mL/min/1.73 m2 and required a second eGFR <60 mL/min/1.73 m2 within 30–180 days of the index reduced kidney function event and evaluated outcomes in this restricted cohort with chronic kidney disease. Follow-up started at 180 days after reaching the reduced kidney function threshold. Finally, the population excluded patients who were enrolled in Medicare Advantage during the baseline period and censored patients’ follow-up upon enrollment in Medicare Advantage programs. In this sensitivity analysis, Medicare Advantage (Part C) individuals were excluded because their claims may be missing or incomplete during the time frame of the study. In addition, we conducted subgroup analyses and tested for effect modification by stratifying by the following covariates: age (≥65, <65 years) and race (black, nonblack). We also examined eGFR at the kidney function threshold date. We chose to examine eGFR subgroups stratified at ≥45 and <45 mL/min/1.73 m2 to have sufficient numbers of patients and outcomes in each group. Because of the small number of outcome events in subgroup analyses, weighted unadjusted estimates are reported. Analyses were conducted using R (www.r-project.org).

Data and Resource Availability

The protocol, statistical code, and deidentified and anonymized data set are available from C.L.R. with a written request.

Results

Study Cohort and Patient Characteristics

The study identified 67,381 new users of metformin and 28,801 new users of sulfonylureas who developed reduced kidney function (Fig. 1). This cohort of persistent new users represented 55.6% of 174,882 new users who had a baseline creatinine and reached the reduced kidney function threshold. There were 18,651 who reached the kidney threshold outside the prespecified time range, 49,755 whose regimens changed before or on the day that kidney function threshold was reached, 9,184 who had no active diabetes medications or had not accessed the health care system for at least 6 months before reaching the kidney function threshold, and 117 with data errors. There were an additional 993 who met exclusion criteria: 206 with organ transplant, 219 in hospice, 193 with a dialysis episode within the past 2 years, and 375 with a documented lactic acidosis hospitalization before reaching the reduced kidney function threshold. After propensity score calculation and weighting, the cohort included 24,542 metformin users and 24,662 sulfonylurea users (54% glipizide, 45% glyburide, and 1% glimepiride).

Figure 1.

Figure 1

Open in a new tab

Flow of eligible patients in the VHA diabetes kidney disease cohort. *Weighted number uses matching weights derived from the propensity score to balance both exposure groups on observed covariates.

Cohort veterans were 96.5% male and 82.6% white. Metformin and sulfonylurea patients had similar baseline characteristics. However, metformin users were younger than sulfonylurea users (median age 67 vs. 71 years). After weighting, patient characteristics were similar between metformin and sulfonylureas, including age 70 years (interquartile range [IQR] 63, 78) vs. 70 years (63, 78), HbA1c 6.5% (6.1, 7.1) vs. 6.6% (6.1, 7.2) (HbA1c 48 mmol/mol [43, 54] vs. 49 mmol/mol [43, 55]), and eGFR 55.8 mg/min/1.73 m2 (51.7, 58.2) vs. 55.8 mg/min/1.73 m2 (51.5, 58.2) at the time of reduced kidney function threshold (Table 1). The median observed follow-up in the weighted cohort was 1.1 years (0.4, 2.6) for patients taking metformin and 1.2 years (0.5, 2.7) for sulfonylureas.

Table 1.

Patient characteristics

Full unweighted cohort Weighted cohort
Metformin Sulfonylureas Metformin Sulfonylureas SMD†
Patients, n 67,381 28,801 24,542 24,662
Age (years)* 67 (62, 74) 71 (63, 79) 70 (63, 78) 70 (63, 78) 0.001
Male, n (%) 64,571 (95.8) 28,288 (98.2) 24,056 (98.0) 24,172 (98.0) <0.001
Race, n (%)
 White 56,104 (83.3) 23,394 (81.2) 20,082 (81.8) 20,184.6 (81.9) 0.001
 Black 9,807 (14.6) 4,883 (17.0) 4,004 (16.3) 4,014.1 (16.3)
 Other 1,470 (2.2) 524 (1.8) 456 (1.9) 460 (1.9)
Medication start to kidney threshold (months)* 16.2 (6.5, 35.0) 13.6 (5.9, 29.0) 14 (5.8, 30.2) 14.0 (6.0, 30.3) 0.013
Year of reaching kidney threshold, n (%) 0.027
 2002–2003 3,158 (4.7) 4,880 (16.9) 2,913 (11.8) 2,907 (11.8)
 2004–2005 5,770 (8.6) 5,706 (19.8) 4,463 (18.2) 4,423 (17.9)
 2006–2007 9,041 (13.4) 6,068 (21.1) 5,184 (21.2) 5,413 (22.0)
 2008–2009 9,905 (14.7) 4,026 (14.0) 3,856 (15.7) 3,870 (15.7)
 2010–2011 12,146 (18.0) 3,318 (11.5) 3,340 (13.6) 3,267 (13.2)
 2012–2013 12,773 (19.0) 2,600 (9.0) 2,629 (10.7) 2,580 (10.4)
 2014–2015 14,588 (21.6) 2,203 (7.7) 2,157 (8.8) 2,198 (8.9)
Laboratory variables
 HbA1c (%)* 6.5 (6.1, 7.0) 6.6 (6.1, 7.3) 6.5 (6.1, 7.1) 6.6 (6.1, 7.2) 0.005
 HbA1c (mmol/mol)* 48 (43, 53) 49 (43, 56) 48 (43, 54) 49 (43, 55)
 Missing HbA1c measure 2,758 (4.1) 1,127 (3.9) 1,003.3 (4.1) 987.5 (4.0) 0.004
 eGFR before threshold (mL/min/1.73 m2)* 70.4 (65.1, 78.6) 69.2 (64.5, 76.4) 69.6 (64.7, 77.0) 69.6 (64.7, 77.0) 0.001
 eGFR at kidney threshold (mL/min/1.73 m2)* 55.9 (51.7, 58.2) 55.8 (51.5, 58.2) 55.8 (51.7, 58.2) 55.8 (51.6, 58.2) 0.002
 Hemoglobin (g/dL)* 14.0 (12.9, 15.0) 14.1 (13.0, 15.2) 14.1 (13.0, 15.1) 14.1 (13.0, 15.2) 0.003
 Missing hemoglobin measure, n (%) 3,625 (5.4) 1,709 (5.9) 1,510.8 (6.2) 1,505.3 (6.1) 0.002
 LDL (mg/dL)* 85 (67, 106) 89 (72, 111) 88.0 (70, 110) 88 (71, 110) 0.001
 Missing LDL measure, n (%) 1,312 (1.9) 1,133 (3.9) 790.6 (3.2) 792.4 (3.2) <0.001
 MACR stage, n (%)
  A1 (<30 mg/g, normal to mild increased albuminuria) 29,514 (43.8) 10,577 (36.7) 9,426.4 (38.4) 9,485.9 (38.5) 0.003
  A2 (30–300 mg/g, moderate increased albuminuria) 7,345 (10.9) 3,055 (10.6) 2,659.5 (10.8) 2,659.3 (10.8)
  A3 and positive unable to quantify (>300 mg/g, severely increased albuminuria) 1,801 (2.7) 925 (3.2) 764.0 (3.1) 757.7 (3.1)
 Missing MACR measure, n (%) 28,721 (42.6) 14,244 (49.5) 11,691.8 (47.6) 11,756.2 (47.7)
 Proteinuria by urinalysis, negative, n (%) 32,812 (48.7) 13,441 (46.7) 11,589 (47.2) 11,644 (47.2) 0.002
  Urine protein trace or 1+ 9,971 (14.8) 4,137 (14.4) 3,536 (14.4) 3,565 (14.5)
  Proteinuria present at 2+ 2,150 (3.2) 976 (3.4) 794 (3.2) 800 (3.2)
  Proteinuria present at 3+ or 4+ 622 (0.9) 479 (1.7) 343 (1.4) 344 (1.4)
 Missing urine protein measure, n (%) 21,826 (32.4) 9,768 (33.9) 8,280 (33.7) 8,306 (33.7)
Clinical variables
 Systolic blood pressure (mmHg)* 129 (118, 140) 131 (120, 143) 131 (119, 142) 131 (119, 142) 0.002
 Diastolic blood pressure (mmHg)* 73 (65, 80) 71 (64, 80) 72 (64, 80) 72 (64, 80) <0.001
 BMI (kg/m2)* 31.1 (27.7, 35.2) 30.1 (26.9, 34.1) 30.4 (27.1, 34.4) 30.3 (27.1, 34.4) 0.004
 Missing BMI measure, n (%) 11,471 (17.0) 5,707 (19.8) 4,591 (18.7) 4,617 (18.7) <0.001
Baseline comorbidities, n (%)‡
 Malignancy 7,118 (10.6) 3,486 (12.1) 2,867 (11.7) 2,884 (11.7) <0.001
 Liver disease 1,087 (1.6) 786 (2.7) 571 (2.3) 568 (2.3) 0.001
 HIV 231 (0.3) 112 (0.4) 92 (0.4) 94 (0.4) 0.001
 Congestive heart failure 5,419 (8.0) 4,154 (14.4) 2,939 (12.0) 2,960 (12.0) 0.001
 Cardiovascular disease 17,525 (26.0) 9,726 (33.8) 7,729 (31.5) 7,798 (31.6) 0.003
 Stroke 1,877 (2.8) 1,017 (3.5) 822 (3.4) 819 (3.3) 0.002
 Transient ischemic attack 704 (1.0) 406 (1.4) 318 (1.3) 328 (1.3) 0.003
 Serious mental illness 16,446 (24.4) 5,755 (20.0) 4,988 (20.3) 5,063 (20.5) 0.005
 Smoking 8,654 (12.8) 3,515 (12.2) 3,032 (12.4) 3,054 (12.4) 0.001
 Chronic obstructive pulmonary disease 10,161 (15.1) 5,203 (18.1) 4,146.7 (16.9) 4,181.6 (17.0) 0.002
 History of respiratory failure 1,884 (2.8) 927 (3.2) 763.6 (3.1) 762.0 (3.1) 0.001
 History of sepsis 893 (1.3) 482 (1.7) 370 (1.5) 378 (1.5) 0.002
 History of pneumonia 2,111 (3.1) 1,389 (4.8) 1,027 (4.2) 1,045 (4.2) 0.003
 Arrhythmia 9,383 (13.9) 5,408 (18.8) 4,244 (17.3) 4,272 (17.3) 0.001
 Cardiac valve disease 1,869 (2.8) 1,181 (4.1) 888 (3.6) 895 (3.6) 0.001
 Parkinson disease 495 (0.7) 309 (1.1) 227 (0.9) 230 (0.9) 0.001
 Urinary tract infection 2,197 (3.3) 1,346 (4.7) 1,011 (4.1) 1,021 (4.1) 0.001
 Osteomyelitis 299 (0.4) 192 (0.7) 151 (0.6) 149 (0.6) 0.002
 Osteoporosis 468 (0.7) 238 (0.8) 195 (0.8) 200 (0.8) 0.002
 Falls 143 (0.2) 73 (0.3) 55 (0.2) 56 (0.2) 0.001
 Fractures 1,236 (1.8) 668 (2.3) 538 (2.2) 538 (2.2) 0.001
 Amputation 224 (0.3) 168 (0.6) 115 (0.5) 118 (0.5) 0.002
 Retinopathy 502 (0.7) 394 (1.4) 286 (1.2) 287 (1.2) <0.001
Use of medications, n (%)
 ACE inhibitors 42,993 (63.8) 18,703 (64.9) 15,881 (64.7) 16,002 (64.9) 0.004
 Angiotensin II receptor blockers 8,643 (12.8) 3,095 (10.7) 2,800 (11.4) 2,792 (11.3) 0.003
 β-Blockers 33,084 (49.1) 14,674 (50.9) 12,414 (50.6) 12,485 (50.6) 0.001
 Calcium channel blockers 19,585 (29.1) 8,611 (29.9) 7,333 (29.9) 7,367 (29.9) <0.001
 Thiazide and potassium-sparing diuretics 29,836 (44.3) 11,508 (40.0) 10,051 (41.0) 10,143 (41.1) 0.004
 Loop diuretics 10,190 (15.1) 6,544 (22.7) 4,900 (20.0) 4,925 (20.0) <0.001
 Other antihypertensive medications 18,349 (27.2) 7,783 (27.0) 6,676 (27.2) 6,685 (27.1) 0.002
 Statin lipid-lowering drugs 49,632 (73.7) 18,566 (64.5) 16,456 (67.1) 16,605 (67.3) 0.006
 Nonstatin lipid-lowering agents 13,087 (19.4) 4,639 (16.1) 4,222 (17.2) 4,250 (17.2) 0.001
 Antiarrhythmics digoxin and inotropes 4,346 (6.4) 3,114 (10.8) 2,239 (9.1) 2,249 (9.1) <0.001
 Anticoagulants, platelet inhibitors 5,961 (8.8) 3,068 (10.7) 2,465 (10.0) 2,471 (10.0) 0.001
 Nitrates 7,752 (11.5) 4,671 (16.2) 3,595 (14.6) 3,631 (14.7) 0.002
 Aspirin 14,232 (21.1) 6,476 (22.5) 5,305 (21.6) 5,351 (21.7) 0.002
 Platelet inhibitors not aspirin 6,201 (9.2) 3,067 (10.6) 2,551 (10.4) 2,569 (10.4) 0.001
 Antipsychotics 5,344 (7.9) 1,960 (6.8) 1,714 (7.0) 1,735 (7.0) 0.002
 Oral glucocorticoids 4,988 (7.4) 2,106 (7.3) 1,771 (7.2) 1,787 (7.2) 0.001
Indicators of health care utilization, n (%)
 Hospitalized within year (VHA) 8,809 (13.1) 4,394 (15.3) 3,475 (14.2) 3,528 (14.3) 0.004
 Hospitalized within year (Medicare/Medicaid) 5,563 (8.3) 3,560 (12.4) 2,745 (11.2) 2,762 (11.2) <0.001
 Hospitalized in 30 days (VHA) 2,385 (3.5) 1,144 (4.0) 900 (3.7) 919 (3.7) 0.003
 Hospitalized in 30 days (Medicare/Medicaid) 965 (1.4) 569 (2.0) 433 (1.8) 443 (1.8) 0.003
 Nursing home encounter in last year 187 (0.3) 131 (0.5) 89 (0.4) 95 (0.4) 0.003
 Number medications* 7 (5, 11) 7 (4, 10) 7 (4, 10) 7 (4, 10) 0.003
 Outpatient visits in past year* 6 (3, 11) 7 (4, 12) 6 (4, 11) 6 (4, 11) 0.001
 Medicare insurance use in last year 21,311 (31.6) 10,486 (36.4) 8,768 (35.7) 8,772 (35.6) 0.003
 Medicaid insurance use in last year 659 (1.0) 428 (1.5) 319 (1.3) 327 (1.3) 0.002
 Medicare Advantage use 10,208 (15.1) 4,319 (15.0) 3,755 (15.3) 3,768 (15.3) 0.001
Open in a new tab

MACR, microalbumin-to-creatinine ratio.

*

Median (IQR).

†

SMDs are the absolute difference in means or percent divided by an evenly weighted pooled SD or the difference between groups in number of SDs. In the weighted cohort, all standardized differences were not statistically significant (see Supplementary Fig. 3 for the plot of the mean standardized differences of the prematched and matched cohorts).

‡

Definitions of comorbidities in Supplementary Table 1.

Primary Outcome: Hospitalization for Lactic Acidosis

In the weighted cohort, there were 193 composite lactic acidosis events in the metformin group and 180 in the sulfonylurea group, yielding 4.18 (95% CI 3.63, 4.81) vs. 3.69 (3.19, 4.27) composite lactic acidosis events per 1,000 person-years, respectively (Table 2). The cause-specific adjusted HR (aHR) was 1.21 (95% CI 0.99, 1.50) among patients using metformin compared with those using sulfonylureas. HRs were consistent for each component of the composite primary outcome, including lactic acidosis defined by primary discharge diagnosis code (aHR 1.11 [0.87, 1.44]) and laboratory-confirmed lactic acidosis hospitalizations (1.25 [0.92, 1.70]) (Table 2). Figure 2 depicts that the cumulative probability of lactic acidosis for patients persisting on metformin versus sulfonylureas at 5 years was 0.74% (95% CI 0.66, 0.83) vs. 0.71% (0.61, 0.82) and, at 10 years, was 0.85% (0.76, 0.95) vs. 0.79% (0.68, 0.90). These estimates accounted for the competing risks of nonpersistence (84.3% metformin vs. 82.7% sulfonylureas) and death (4.6% vs. 6.7%).

Table 2.

Rates and aHRs (95% CIs) for lactic acidosis hospitalizations among patients with reduced GFR who used metformin vs. sulfonylureas in weighted cohort

Metformin Sulfonylureas
Number at risk matched weighted 24,542 24,662
 Primary outcome: lactic acidosis hospitalization 193 180
  Person-years 46,197 48,748
  Unadjusted rate/1,000 person-years (95% CI) 4.18 (3.63, 4.81) 3.69 (3.19, 4.27)
  aHRb (95% CI) 1.21 (0.99, 1.48) Reference
 Laboratory-confirmed lactic acid hospitalization 84 75
  Person-years 46,283 48,860
  Unadjusted rate/1,000 person-years (95% CI) 1.81 (1.46, 2.24) 1.54 (1.23, 1.93)
  aHRb (95% CI) 1.25 (0.92, 1.70) Reference
 Primary discharge diagnosis of lactic acidosis hospitalization 122 121
  Person-years 46,250 48,785
  Unadjusted rate/1,000 person-years (95% CI) 2.63 (2.20, 3.14) 2.49 (2.08, 2.97)
  aHRb (95% CI) 1.11 (0.87, 1.44) Reference
Sensitivity analysis: requiring lactate ≥5 mmol/L
 Number at risk matched weighted 24,542 24,662
  Composite lactic acidosis hospitalizations 146 140
  Person-years 46,238 48,769
  Unadjusted rate/1,000 person-years (95% CI) 3.16 (2.69, 3.72) 2.89 (2.45, 3.41)
  aHRb (95% CI) 1.15 (0.91, 1.46) Reference
Sensitivity analysis: population with second reduced eGFR
 Number at risk matched weighted 3,851 3,872
  Composite lactic acidosis hospitalizations 22 27
  Person-years 7,160 8,487
  Unadjusted rate/1,000 person-years (95% CI) 3.07 (2.02, 4.64) 3.17 (2.18, 4.60)
  aHRb (95% CI) 1.09 (0.64, 1.84) Reference
Sensitivity analysis: excluding Medicare Advantage
 Number at risk matched weighteda 20,787 20,893
  Composite lactic acidosis hospitalizations 163 154
  Person-years 37,216 39,507
  Unadjusted rate/1,000 person-years (95% CI) 3.89 (3.33, 4.56) 4.37 (3.75, 5.10)
  aHR (95% CI)b 1.18 (0.94, 1.47) Reference
Open in a new tab
a

Primary analysis considers patients persistent on regimen until they do not have oral antidiabetic medications for 90 days.

b

Cox proportional hazards model for time to event. Adjusted for mean-centered metformin dose, demographics, clinical information derived from the electronic health record, comorbidities, use of medications, and health care utilization (see Supplementary Table 1). All continuous variables were modeled as restricted cubic splines.

Figure 2.

Figure 2

Open in a new tab

Aalen-Johansen cumulative (Cum) incidence demonstrating lactic acidosis events with the competing risks of nonpersistence and death in the weighted cohort. Non-Persist Mono, nonpersistent monotherapy.

We conducted a post hoc evaluation of the primary diagnosis among patients who met the laboratory-confirmed lactic acidosis event criteria (84 metformin events and 75 sulfonylurea events). The sepsis code (ICD-9 038.x or ICD-10 A41) was the most common primary discharge diagnosis associated with laboratory-confirmed lactic acidosis (13.1% of metformin events vs. 19.6% of sulfonylurea events). The second most frequent diagnosis associated with laboratory-confirmed lactic acidosis was acute kidney injury (7.1% of metformin events vs. 2.7% of sulfonylurea events).

Impact of Dose

Patients were more likely to stop both metformin and sulfonylureas at reaching reduced kidney function than to have a dose change. The median (IQR) defined daily dose at the time of kidney function decline was 0.5 (0.5, 0.9) for metformin (1,000 mg) and 0.5 (0.3, 1.0) for sulfonylureas (5 mg for both glyburide and glipizide, which represented 99% of sulfonylureas). At 6 months after reaching the reduced kidney threshold, only 55% and 61% of metformin and sulfonylurea users, respectively, persisted on their regimens, and the median defined daily dose remained 0.5 (0.5, 0.9) for metformin and 0.5 (0.3, 1.0) for sulfonylureas. At 12 months after reaching the reduced kidney threshold, 43% vs. 48% persisted on therapy, and the median doses remained the same. We further evaluated metformin dose among patients who had an event stratified by eGFR. At the time of the lactic acidosis hospitalization, there were 46 patients with an eGFR <45 mL/min/1.73 m2, and the median metformin dose at the event was 1,000 mg (1,000, 2,000). There were 146 patients with an eGFR ≥45 mL/min/1.73 m2, and the median metformin dose at the event was 1,000 mg (1,000, 2,000). After accounting for the main effect of metformin use, the independent effect of metformin dose associated with lactic acidosis hospitalization was negligible.

Sensitivity and Subgroup Analysis

Results were similar in all sensitivity analyses. Sensitivity analyses that required a lactate value ≥5 mmol/L resulted in 146 events among metformin users (116 ICD and 30 laboratory-confirmed lactic acidosis diagnoses) and 140 events among sulfonylurea users (111 ICD and 29 laboratory-confirmed lactic acidosis diagnoses). For the analysis that required a second reduced eGFR within 30–180 days of the index date results, the median number of days to the second confirmatory eGFR was 112 (IQR 72, 147), and the median eGFR at that time was 54 mL/min/1.73 m2 (48,57). Results were consistent with the main results (Table 2). Excluding patients with Medicare Advantage and censoring upon enrollment in Advantage plans were also similar. Subgroups stratified by age, race, and eGFR at time of kidney function decline were consistent with the main analysis, with no evidence of effect modification (Supplementary Table 3). The 95% CIs were wide for some groups, including for those with eGFR in the lowest categorization. Among patients with eGFR <45 mL/min/1.73 m2, point estimates of lactic acidosis rates were numerically greater among metformin users versus sulfonylurea users, with lactic acidosis rates of 12.89 (95% CI 9.14, 18.17) vs. 8.62 (6.04, 12.29) per 1,000 person-years (weighted HR 1.31 [0.85, 2.01]).

Conclusions

In this national evaluation of patients with diabetes, there was no statistically significant association between hospitalizations for lactic acidosis among patients who continued to use metformin versus sulfonylureas after they reached a reduced kidney threshold. Lactic acidosis hospitalization was uncommon and comparable between those using metformin and sulfonylureas, and the cumulative probability of lactic acidosis for patients persisting on metformin and sulfonylureas at 5 years was 0.74% vs. 0.71%. On the basis of our estimates, this risk difference of 0.03% over 5 years translates into one additional lactic acidosis hospitalization per 3,300 patients treated with metformin for 5 years. Our prior work estimated that a single year of metformin versus sulfonylurea treatment in this same cohort of patients with diabetes and reduced kidney function was associated with prevention of 167 major adverse cardiovascular events (19). Taken together, in patients with type 2 diabetes and kidney disease, the association of lactic acidosis hospitalization with metformin is small and unequivocally outweighed by the prevention of major adverse cardiovascular events.

Some studies estimated that as many as 1 million U.S. patients with diabetes and eGFR between 31 and 89 mL/min/1.73 m2 who could take metformin are not taking this medication (20). A major reason metformin is discontinued in patients is the development of kidney disease and the concern for increased risk of lactic acidosis, despite guidance recommending the continuation of metformin at reduced eGFR (21,22). While there is laboratory evidence that metformin does increase lactic acid levels, the clinical significance of these laboratory abnormalities remains unclear (23,24). Multiple studies have demonstrated that the association of metformin with lactic acidosis is low and comparable to that associated with sulfonylurea use. Often, instances of lactic acidosis occurred in the context of another critical illness that was the more likely cause of lactic acidosis (24,25). Furthermore, a recent publication noted no statistically significant increase in the FDA’s Adverse Event Reporting System reports for metformin and lactic acidosis compared with any other diabetes medications after the 2016 change in FDA prescribing guidance (26).

This study adds to the growing body of evidence of a low risk for lactic acidosis hospitalizations in patients with type 2 diabetes and mild to moderate reduction in kidney function. While metformin may increase lactic acid levels, it was not associated with a significantly increased risk of lactic acidosis hospitalizations compared with sulfonylureas in our cohort. Inzucchi et al. (25) conducted a systematic review of published clinical trials and observational studies, with most appearing to confirm the drug’s overall safety profile, finding lactic acidosis rates not significantly different in patients with type 2 diabetes and reduced kidney function compared with the general population on metformin. The range of incident metformin-associated lactic acidosis was estimated at ∼3–10 per 100,000 person-years and not significantly different from those in the general population of patients treated with alternate agents. Lazarus et al. (27) performed a retrospective cohort study of 75,413 patients that used billing and discharge diagnosis data from the Geisinger Health System. They found no significant difference in rates of incident lactic acidosis between metformin users and sulfonylurea users who had an eGFR between 30 and 60 mL/min/1.73 m2. The association of lactic acidosis diagnosis codes in metformin users ranged between 4 and 10 per 1,000 person-years for patients with eGFR between 30 and 89 mL/min/1.73 m2, which is similar to the rate of 4.18 (95% CI 3.63, 4.81) events per 1,000 person-years reported in this study. Notably, Lazarus et al. did find an increased hazard of lactic acidosis associated with metformin compared with sulfonylurea users in the subgroup of patients with eGFR <30 mL/min/1.73 m2 (20 per 1,000 person-years, aHR 2.07 [95% CI 1.33, 3.22]). The current study adds to the work of Lazarus et al. by incorporating laboratory data and reporting both highly sensitive and specific definitions of laboratory-confirmed lactic acidosis hospitalizations.

A major strength of our study was the large sample size. With 180,000 person-years of follow-up and an average event rate of 4 per 1,000 person-years, if the true risk of metformin was double that of sulfonylureas, a study of this size would have detected it with near certainty. Our study yielded precise estimates of the HR for lactic acidosis with 95% CI widths of ∼0.4. Nevertheless, although our study cannot completely rule out smaller increases in risk associated with metformin use, the study’s results can be used to rule out HRs >1.5 when comparing metformin with sulfonylureas in this cohort. Therefore, our study findings support the recent change in the FDA label for metformin to include patients with mild to moderate kidney dysfunction. This is especially important in light of recent studies demonstrating that metformin is associated with a reduced risk of major cardiovascular events in patients with reduced kidney function (19).

This study has several limitations. First, persistence on incident therapy through the time of reaching the kidney function threshold was required, excluding many patients who discontinued, added, or switched initial medications at or before reaching the kidney threshold. While reducing sample size, this design choice allowed evaluation of patients with deteriorating kidney function. Furthermore, a competing risk model was used to address concerns that glucose-lowering medication nonpersistence or death would bias the assessment of lactic acidosis events. Second, veterans may not receive all their care at VHA facilities, and some lactic acidosis hospitalizations may have been missed despite the linkage to Medicare and Medicaid data. Furthermore, events captured in Medicare or Medicaid were not eligible for laboratory confirmation since laboratory data from these hospitalizations are not systematically available. Still, we have no reason to believe that this potential underreporting would be differentially distributed among our exposures. Third, cohort entry and the start of follow-up was either an elevated serum creatinine or a reduced eGFR <60 mL/min/1.73 m2. It is possible that for some patients, this kidney threshold may represent an acute kidney injury event rather than progression of chronic kidney disease. We believe, however, that providers made clinical decisions on the basis of the patient reaching this reduced kidney function threshold. At the time of the reduced kidney threshold, the cohort patients had continued their antidiabetic medication for a median of 14 months, and within 6 months of reaching the threshold, ∼50% of this cohort switched their antidiabetic medication or became nonpersistent (>90 days without a medication fill). This pattern of drug nonpersistence may be more consistent with a trajectory of chronic reduction in kidney function rather than a transient acute kidney injury. Fourth, although propensity score weighting and direct covariate adjustment were used to reduce concerns about confounding, residual confounding may exist. Fifth, we do not have serum metformin levels at the time of lactic acidosis hospitalizations, thereby complicating the assessment of a causal relationship. Furthermore, the hospitalization for lactic acidosis may be related to another underlying condition. Given the concern about metformin causing lactic acidosis, there may also be surveillance bias, such that lactic acid levels may be obtained more often in metformin patients. This type of bias would, however, tend to exaggerate risk. Finally, the study population was mostly elderly white men and may not be representative of the larger population of patients with diabetes and reduced kidney function. This should be considered when generalizing the study results to other populations.

The current study, like previous smaller cohort studies, finds that in patients with type 2 diabetes and reduced kidney function metformin is not associated with a statistically significant increased rate of lactic acidosis compared with sulfonylureas. Our findings support the 2016 FDA metformin label change, which indicates that metformin can be used in patients with mild to moderate reduced kidney function with appropriate dosage reductions and continued monitoring and assessments of kidney function.

Article Information

Funding. This project was funded by Veterans Affairs Clinical Science Office of Research and Development investigator-initiated grant CX000570-07 (C.L.R.). P.Y.C., A.J.H., T.E., and C.L.R. were supported in part by National Institute of Diabetes and Digestive and Kidney Diseases Center for Diabetes Translation Research grant P30-DK-092986. Support for Veterans Affairs/Centers for Medicare & Medicaid Services data was provided by the Department of Veterans Affairs, Veterans Affairs Health Services Research and Development Service, and Veterans Affairs Information Resource Center (project numbers SDR 02-237 and 98-004).

The Veterans Affairs Clinical Science Office of 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 of this article do not represent the views of the U.S. Department of Veterans Affairs or the U.S. government.

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

Author Contributions. P.Y.C. drafted the manuscript. P.Y.C., A.J.H., J.C., M.R.G., A.M.H., R.A.G., C.G.G., T.E., and C.L.R. contributed to the critical revision of the manuscript. M.R.G., A.M.H., R.A.G., T.E., and C.L.R. contributed to the study design. A.J.H., J.C., M.R.G., R.A.G., and C.L.R. contributed to the study conduct and data collection. A.J.H., J.C., and R.A.G. contributed to the data analysis. A.J.H. and C.L.R. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Footnotes

This article contains supplementary material online at https://doi.org/10.2337/figshare.12042783.

References

  • 1.American Diabetes Association Standards of Medical Care in Diabetes—2017: summary of revisions. Diabetes Care 2017;40(Suppl. 1):S4–S5 [DOI] [PubMed] [Google Scholar]
  • 2.UK Prospective Diabetes Study (UKPDS) Group Effect of intensive blood-glucose control with metformin on complications in overweight patients with type 2 diabetes (UKPDS 34). Lancet 1998;352:854–865 [PubMed] [Google Scholar]
  • 3.Rojas LBA, Gomes MB. Metformin: an old but still the best treatment for type 2 diabetes. Diabetol Metab Syndr 2013;5:6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Holman RR, Paul SK, Bethel MA, Matthews DR, Neil HAW. 10-year follow-up of intensive glucose control in type 2 diabetes. N Engl J Med 2008;359:1577–1589 [DOI] [PubMed] [Google Scholar]
  • 5.Roumie CL, Hung AM, Greevy RA, et al. . Comparative effectiveness of sulfonylurea and metformin monotherapy on cardiovascular events in type 2 diabetes mellitus: a cohort study. Ann Intern Med 2012;157:601–610 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Diabetes Prevention Program Research Group Long-term safety, tolerability, and weight loss associated with metformin in the Diabetes Prevention Program Outcomes Study. Diabetes Care 2012;35:731–737 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Glucophage final printed labeling, Food Drug Administration [Internet] , 2001. Available from https://www.accessdata.fda.gov/drugsatfda_docs/nda/2000/20357S019_Glucophage_prntlbl.pdf. Accessed 5 January 2015
  • 8.Assan R, Heuclin C, Girard JR, LeMaire F, Attali JR. Phenformin-induced lactic acidosis in diabetic patients. Diabetes 1975;24:791–800 [DOI] [PubMed] [Google Scholar]
  • 9.Luft D, Schmülling RM, Eggstein M. Lactic acidosis in biguanide-treated diabetics: a review of 330 cases. Diabetologia 1978;14:75–87 [DOI] [PubMed] [Google Scholar]
  • 10.Cryer DR, Nicholas SP, Henry DH, Mills DJ, Stadel BV. Comparative outcomes study of metformin intervention versus conventional approach: the COSMIC Approach Study. Diabetes Care 2005;28:539–543 [DOI] [PubMed] [Google Scholar]
  • 11.Greevy RA Jr, Huizinga MM, Roumie CL, et al. . Comparisons of persistence and durability among three oral antidiabetic therapies using electronic prescription-fill data: the impact of adherence requirements and stockpiling. Clin Pharmacol Ther 2011;90:813–819 [DOI] [PubMed] [Google Scholar]
  • 12.World Health Organization. ATC/DDD index [Internet] , 2018. Available from https://www.whocc.no/atc_ddd_index/. Accessed 10 June 2018
  • 13.U.S. Health Care Financing Administration International Classification of Diseases, Ninth Revision, Clinical Modification. Washington, DC, U.S. Department of Health and Human Services, Public Health Service, Health Care Financing Administration, 1988 [Google Scholar]
  • 14.Levey AS, Stevens LA, Schmid CH, et al.; CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration) . A new equation to estimate glomerular filtration rate. Ann Intern Med 2009;150:604–612 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Skali H, Uno H, Levey AS, Inker LA, Pfeffer MA, Solomon SD. Prognostic assessment of estimated glomerular filtration rate by the new Chronic Kidney Disease Epidemiology Collaboration equation in comparison with the Modification of Diet in Renal Disease Study equation. Am Heart J 2011;162:548–554 [DOI] [PubMed] [Google Scholar]
  • 16.Li L, Greene T. A weighting analogue to pair matching in propensity score analysis. Int J Biostat 2013;9:215–234 [DOI] [PubMed] [Google Scholar]
  • 17.Grambsch PM, Therneau TM. Proportional hazards tests and diagnostics based on weighted residuals. Biometrika 1994;81:515–526 [Google Scholar]
  • 18.Luft D, Deichsel G, Schmülling RM, Stein W, Eggstein M. Definition of clinically relevant lactic acidosis in patients with internal diseases. Am J Clin Pathol 1983;80:484–489 [DOI] [PubMed] [Google Scholar]
  • 19.Roumie CL, Chipman J, Min JY, et al. . Association of treatment with metformin vs sulfonylurea with major adverse cardiovascular events among patients with diabetes and reduced kidney function. JAMA 2019;322:1–11 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Flory JH, Hennessy S. Metformin use reduction in mild to moderate renal impairment: possible inappropriate curbing of use based on food and drug administration contraindications. JAMA Intern Med 2015;175:458–459 [DOI] [PubMed] [Google Scholar]
  • 21.Lalau JD, Kajbaf F, Bennis Y, Hurtel-Lemaire AS, Belpaire F, De Broe ME. Metformin treatment in patients with type 2 diabetes and chronic kidney disease stages 3A, 3B, or 4. Diabetes Care 2018;41:547–553 [DOI] [PubMed] [Google Scholar]
  • 22.Executive Committee KDIGO clinical practice guideline on diabetes management in chronic kidney disease [Internet]. Available from https://kdigo.org/wp-content/uploads/2018/03/KDIGO-Diabetes-Management-in-CKD_Public-Review.pdf. Accessed 12 February 2020
  • 23.DeFronzo R, Fleming GA, Chen K, Bicsak TA. Metformin-associated lactic acidosis: current perspectives on causes and risk. Metabolism 2016;65:20–29 [DOI] [PubMed] [Google Scholar]
  • 24.Lalau JD, Race J-M. Lactic acidosis in metformin-treated patients. Prognostic value of arterial lactate levels and plasma metformin concentrations. Drug Saf 1999;20:377–384 [DOI] [PubMed] [Google Scholar]
  • 25.Inzucchi SE, Lipska KJ, Mayo H, Bailey CJ, McGuire DK. Metformin in patients with type 2 diabetes and kidney disease: a systematic review. JAMA 2014;312:2668–2675 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Flory JH, Hennessy S, Bailey CJ, Inzucchi SE. Reports of lactic acidosis attributed to metformin, 2015–2018. Diabetes Care 2020;43:244–246 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Lazarus B, Wu A, Shin J-I, et al. . Association of metformin use with risk of lactic acidosis across the range of kidney function: a community-based cohort study. JAMA Intern Med 2018;178:903–910 [DOI] [PMC free article] [PubMed] [Google Scholar]

Articles from Diabetes Care are provided here courtesy of American Diabetes Association

RESOURCES