Potential Impact of Prescribing Metformin According to eGFR Rather Than Serum Creatinine

Delphine S. Tuot; Feng Lin; Michael G. Shlipak; Vanessa Grubbs; Chi-yuan Hsu; Jerry Yee; Vahakn Shahinian; Rajiv Saran; Sharon Saydah; Desmond E. Williams; Neil R. Powe


Diabetes Care. 2015;38(11):2059-2067. 

In This Article


This study has three key findings. First, although metformin use has increased in the past decade for treatment of the general population with type 2 diabetes, it remains underused among individuals with diabetes and mild kidney disease. Second, implementing eGFR or CrCl rather than sCr thresholds to determine individual eligibility for metformin could considerably expand the population eligible for its use, particularly among non-Hispanic blacks and men. Third, while various GFR- and CrCl-estimating equations identify different populations of individuals eligible/ineligible for metformin and all expand the population of people with diabetes for whom metformin is likely safe, they also identify a large population of individuals for whom metformin safety remains unclear based on current U.S. recommendations.

The trend toward greater metformin use over the past decade is positive, as metformin remains the antidiabetes medication associated with the highest efficacy, the best cardiovascular profile, and the fewest unwanted side effects.[2,4,18] Nationally representative data have recently demonstrated a decrease in diabetes complications over the same period.[19] These improvements reflect advances in acute clinical care as well as chronic disease care and risk factor control. While speculative, it is possible that greater use of diabetes medications, and metformin in particular, for tighter glycemic control in the late 1990s and early 2000s may have contributed at least in part to these important public health gains.

Despite potential benefits, metformin remains underused among individuals with diabetes and mild kidney disease, who are at even greater risk of cardiovascular morbidity and mortality compared with the general population with diabetes.[20] Creatinine thresholds are problematic for defining CKD. Creatinine production correlates with muscle mass and can underestimate or overestimate kidney function among individuals with muscle mass that differs from the population average. Estimates of GFR based on sCr, race, age, and sex are more clinically useful measures of kidney function, though they, too, must be cautiously interpreted among patients at anthropomorphic extremes. Nevertheless, these equations are recommended for medication dosing by several national and international nephrology societies. The Kidney Disease Outcomes Quality Initiative and Kidney Disease Improving Global Outcomes, for example, both recommend using metformin as a first-line agent among individuals with an eGFR of ≥45 mL/min/1.73 m2and to discontinue metformin definitively among individuals with an eGFR <30 mL/min/1.73 m2. [8,21] While our data do not allow us to ascertain clinician behavior, the sharp drop in metformin use among individuals with an eGFR <45 mL/min/1.73 m2 after 2007 compared with those with an eGFR ≥45 mL/min/1.73 m2 allows us to speculate that the recommended eGFR thresholds are gaining importance in determining metformin eligibility.

Replacing sCr thresholds with eGFR thresholds could expand the pool of patients for whom metformin is likely safe without creating substantial safety concerns. Notably, 18% of individuals newly eligible for metformin in our study had an A1C >9% (75 mmol/mol). While uncontrolled diabetes is associated with more rapid renal function decline, studies have quoted rates of renal function decline ranging from −1.26 mL/min/1.73 m2 per year to −3.24 mL/min/1.73 m2 per year.[22] Assuming at least yearly or more frequent monitoring of renal function among these patients, these rates of decline do not likely pose safety concerns and should not be impediments to metformin prescription. Thus, our study suggests that the number of individuals eligible for metformin in the U.S. can be expanded by at least 104,000, if using MDRD eGFR to calculate kidney function. This is a conservative estimate, as it does not take into account individuals who might be eligible for metformin with an MDRD eGFR 30–44 mL/min/1.73 m2.

Approximately 50% of individuals with an sCr above the conventional threshold of metformin eligibility and 1.7% of individuals with an sCr below the conventional threshold had an MDRD eGFR between 30–44 mL/min/1.73 m2. Individuals with diabetes and this level of kidney dysfunction are at higher risk of hypoglycemia, CKD progression, and mortality compared with individuals with less severe CKD, and may particularly benefit from metformin rather than a sulfonylurea or thiazolidinedione.[23,24] Given the lack of robust data, current guidelines do not provide much guidance about metformin use in this population, though a few studies suggest its safety among individuals with a stable eGFR >30 mL/min/1.73 m2. [25] A randomized controlled trial is needed to clarify whether use of metformin in this subgroup would be safe and efficacious.

Importantly, the expanded pool of individuals for whom metformin is likely safe was predominantly male and non-Hispanic black. Prior European studies have documented that replacing creatinine thresholds with eGFR thresholds can minimize the number of males denied treatment with metformin.[26,27] Our work builds upon these studies and identifies the potential impact of eGFR on metformin eligibility by race/ethnicity in addition to sex. Racial/ethnic disparities with respect to diabetes health outcomes are well recognized. Non-Hispanic black Americans with diabetes have worse glycemic control than their non-Hispanic white counterparts and have been demonstrated to shoulder a greater burden of diabetes complications, such as end-stage renal disease, retinopathy, neuropathy, and nontraumatic lower-extremity amputations.[28–31] Paradoxically, recent studies have not shown differences in receipt of routine A1C testing, nephropathy screening, or monofilament foot examination between non-Hispanic blacks and non-Hispanic whites when accounting for individual patient and facility variables.[32] Non-Hispanic blacks generally have higher sCr than individuals of other race/ethnicities.[33] Relying on eGFR rather than creatinine thresholds to determine metformin eligibility and safety may thus help bridge the gap between the aforementioned process and outcome measures.[34]

The ideal method for estimating kidney function is an area of active research, as all kidney function–estimating formulas have inherent shortcomings compared with the gold standard of measured GFR using urinary or plasma clearance of exogenous filtration markers.[35] CG estimates of CrCl are frequently used by pharmacists to determine medication dosing.[36] However, CrCl is not readily available to clinicians who prescribe metformin. The National Kidney Disease Education Program reports that most laboratories use the four-variable MDRD and recommends it to determine medication safety.[37] The newer CKD-EPIcr[12] generally has less bias than the four-variable MDRD and is slowly gaining traction among U.S. nephrologists and clinical laboratories;[38] however, some studies suggest that it may perform less well than the four-variable MDRD equation when estimating GFR among individuals with type 2 diabetes.[39] Recent data suggest that cystatin C–based equations may reclassify individuals into less severe stages of CKD and are more highly correlated with health outcomes than creatinine-based eGFR among patients with CKD.[40] CKD-EPIcys has thus been recommended for confirmation of CKD status for elderly individuals in whom creatinine-based equations may not be accurate.[41] Discrepancies in medication dosing using different kidney function–estimating equations have been well documented, particularly for elderly patients.[42–44] However, to our knowledge, only one study has demonstrated the potential impact of these discrepancies on health outcomes.[45] In our study, while the CG equation expanded the number of individuals eligible for metformin the most, it also appeared to be the most conservative equation, reclassifying even more individuals to subpopulations for whom metformin is not safe or indeterminate. This is consistent with data demonstrating that CG underestimates GFR among patients with type 2 diabetes and overt diabetic nephropathy.[46,47] Without hard outcomes, it is difficult to identify which kidney function–estimating equation is optimal to use to guide clinical decision making. Prospective studies should clarify the role of each equation for evaluation of safety and efficacy of medication dosing, including metformin, among CKD patients.

There are several limitations to this study, notably that NHANES is not a clinical database and includes community-dwelling individuals who do not seek medical care. However, we restricted our study population to participants who self-reported a routine site for health care and found similar results when restricting the study population to individuals who were aware of their diabetes. We could not ascertain the reasoning behind low levels of metformin use. Specifically, we could not determine whether this was due to patient nonadherence or lack of provider prescription, perhaps owing to nonrenal clinical conditions that contraindicate the use of metformin, such as liver disease. Additionally, NHANES relies on single measurements of eGFR and urinary albumin, leading to possible misclassification.

In summary, we demonstrate that metformin use may be expanded among adults with diabetes and mild CKD by focusing on eGFR rather than sCr thresholds for prescribing purposes, per recent national and international recommendations. In so doing, we may help mitigate racial/ethnic disparities in diabetes management and outcomes for non-Hispanic blacks. Additional research is needed to identify the best kidney function–estimating equation for optimal use and dosing of metformin at point of care. Lastly, it is important to identify the safety and efficacy of metformin among individuals with eGFR 30–44 mL/min/1.73 m2, as this represents another potential avenue to further enhance diabetes care for adults at high risk of cardiovascular complications.