Provider Decisions and Patient Outcomes After Premature Metformin Discontinuation

A retrospective chart review of electronic medical records was performed, with a date range of 1 January 1999 through 30 September 2013, as approved by the VANTHCS institutional review board. This study was designed as a single-subject study, with each patient serving as his or her own control. Patients with type 2 diabetes were included if they were on metformin therapy for at least 3 months before its discontinuation for increased SCr (>1.5 mg/dL in males and >1.4 mg/dL in females). Patients were excluded if they were <18 years of age, if their metformin was discontinued for a reason other than declining renal function in CKD, if they were not followed for at least 6 months after metformin discontinuation, or if their chart had incomplete or undocumented data, including patients followed primarily by a non–Veterans Affairs physician. Patient outcomes were compared before and after discontinuation of metformin, with patient records followed for at least 6 months and up to 17 months after metformin discontinuation.

The primary endpoint was change in peak glycemic control at 1 year defined by A1C and was analyzed by a two-tailed, paired t test. Secondary endpoints were weight, monthly antihyperglycemic medication cost, hypoglycemic events, and alternative antihyperglycemic medication–related adverse drug events (ADEs). The secondary outcomes of weight and cost were analyzed using a two-tailed, paired t test. Hypoglycemic events were analyzed using a two-tailed McNemar’s test. ADEs were collected as descriptive data only. Subgroup analyses of insulin-containing therapy versus alternative oral antihyperglycemic regimens were also performed.

The initial patient cohort identified 584 patients with a discontinued metformin prescription followed by a new prescription for an alternative antihyperglycemic agent. After application of inclusion and exclusion criteria, 179 patients were found to meet study criteria. The top reasons for exclusion included patient primarily followed by non-VA physician (n = 83), missing or lack of required data (n = 75), patient not followed for an appropriate time period surrounding metformin discontinuation (n = 46), metformin discontinuation due to suboptimal glycemic control (n = 34), and metformin discontinuation due to congestive heart failure (CHF) (n = 30). Of note, three patients were discontinued from metformin in the acute setting of lactic acidosis; in all cases, lactic acidosis was determined to have a cause other than metformin therapy, and SCr was within normal limits in all three cases.

Baseline characteristics of the study cohort were as follows: 98% male, average age 67.6 years (range 36–85 years), mean duration of metformin therapy before discontinuation 37 months (range 3–114 months), mean metformin daily dose 1,712 mg/day (range 500–3,000 mg/day), mean baseline SCr 1.73 mg/dL, and mean baseline A1C of 7.7% at the time of metformin discontinuation. All but 16 patients would have been candidates for continuation of metformin using the eGFR calculation. It was not clear why one patient received a metformin prescription that exceeded the maximum approved dosage.

The primary outcome revealed that patients discontinued from metformin reach a significantly higher peak A1C within the follow-up period (P <0.001) compared to their final A1C before discontinuation of metformin (Table 1). By the end of the 1-year follow-up period (range 6–17 months) after dose increases or the addition of new antihyperglycemic medication, glycemic control was restored, and there was no longer a significant difference between A1C values. There was also no difference in mean A1C at metformin discontinuation compared to mean A1C at the end for the subgroup of patients who were maintained on oral antihyperglycemic agents. Significant weight gain on an alternative regimen of an average 3.81 kg (P <0.001) was also noted. This remained true even for the subgroup of patients on alternative oral agents only at 1 year, who had an average weight gain of 3.7 kg (P <0.001). Cost of therapy was also significantly greater—nearly double—on alternative regimens (P <0.0001). The risk of hypoglycemic events was significantly higher on an alternative regimen than on a metformin-containing regimen (P <0.001). ADEs from alternative regimens are described in Table 2, with the most common being hypoglycemia and edema, which were experienced by 24 and 6.1% of patients, respectively.

Baseline antihyperglycemic regimens before metformin discontinuation and initial antihyperglycemic regimens chosen by providers after metformin discontinuation are detailed in Table 3. Of note, 84.4% of patients had dose increase(s) or additional antihyperglycemic agent(s) added at some point during the follow-up period. The majority of patients (73.2%) were on insulin therapy by the end of the follow-up period: 14 were on insulin therapy before metformin discontinuation, 54 were prescribed insulin as initial alternative therapy (± oral antihyperglycemics), and 63 were prescribed insulin later in the follow-up period. A minority of 48 patients (26.8%) remained on an oral antihyperglycemic regimen, with a thiazolidinedione (TZD) being the most common oral agent added (45.8%).

Based on the results of this retrospective study, discontinuing metformin results in weight gain, increase in medication costs, increase in risk for hypoglycemia, reversible worsening of glycemic control, and potential ADEs from alternative therapy. It is noted that, although discontinuing metformin initially resulted in deterioration of glycemic control, providers were able to use additional antihyperglycemic medications or increase the doses of current medications, and, at 1 year, glycemic control, was restored to the level seen before metformin discontinuation. However, this caused a transient increase in A1C, which can exacerbate the complications of diabetes and lead to patient and provider frustrations such as more frequent visits needed to titrate new medications.

In this study, 73.2% of patients were on insulin therapy 1 year after metformin discontinuation. The use of insulin may be one explanation for the significant amount of weight these patients gained after discontinuing metformin. Interestingly, patients who discontinued metformin and remained on oral antihyperglycemic agents only also gained a significant amount of weight, possibly as a result of metformin’s association with weight neutrality or weight loss. Weight gain in this population increases the risk for further metabolic complications, including the promotion of greater insulin resistance and the potential for reduced compliance, especially with insulin-based therapies (25).

The increase in hypoglycemic events after metformin discontinuation also places patients at an increased risk for further complications, including mortality (26). Both insulin and some alternative oral agents carry this risk.

The increased monthly medication cost found after metformin discontinuation has the potential to affect both patients and the health system. Within the VA health care system, most veterans are charged a copayment for each medication, and at least 15% of the patients in this study required more than one medication to attain similar glycemic control after metformin discontinuation. An additional medication often would require a veteran to pay an additional copayment of up to $8 for a 30-day supply.

Finally, the patients’ quality of life is likely to be affected from glycemic control deterioration requiring dose adjustments or additional medications, weight gain, ADEs such as hypoglycemia and edema from alternative therapy, and increased cost.

The risk of lactic acidosis in patients on metformin with mild to moderate renal impairment has not been demonstrated in studies (22,27). Available evidence (16,18–19) suggests that metformin can be used with benefit because of its low risk of adverse events in this population, as long as sensible dosing and monitoring are implemented. The eGFR offers a more reliable estimate of renal function than SCr, and recent clinical guideline updates support the use of metformin in patients with mild to moderate stable renal dysfunction with eGFR-based dosing.

In a study by Rachmani et al. (28), patients with a relative contraindication to the use of metformin were randomized to continue metformin or have it discontinued. These patients had either chronic obstructive pulmonary disease, CHF, abnormal liver function, acute coronary syndrome, or renal dysfunction. Patients whose metformin was discontinued experienced an average weight gain of 1.8 kg during the first year and additional weight gain of 1.8 kg during the next 3 years (P <0.001). Patients who remained on metformin experienced only a minor weight gain of 0.9 kg over the 4-year period. There were no cases of lactic acidosis in either group. Patients’ glycemic control significantly deteriorated over time if they discontinued metformin.

There are many similarities between the study by Rachmani et al. and ours, including the weight gain that patients experienced when taken off metformin and placed on other hypoglycemic medications. Additionally, our study demonstrated a loss of glycemic control, although this loss was not as durable as in the Rachmani study, perhaps because of a strong emphasis on controlling diabetes within the VA health care system. Although these two studies were similar, there were important differences, including additional evaluation in our study of the adverse events that may occur in patients who are required to start other hypoglycemic agents in lieu of metformin. Additionally, the cost of alternative medication to attain glycemic control was not evaluated by Rachmani et al. Finally, the focus on patients with renal dysfunction in the current study emphasizes one of the most common reasons for metformin discontinuation.

At the VANTHCS, the Veterans Integrated Service Network (VISN) 17 Pharmacy and Therapeutics Committee approved the eGFR-based dosing recommendations detailed in Table 4. These recommendations were adapted from the American Diabetes Association/European Association for the Study of Diabetes position statement on the topic (8). VISN 17 recommends that providers document acknowledgment of diminished but stable renal function and specific plans for long-term patient monitoring. Of note, the VISN 17 approval of eGFR-based dosing does not override the FDA-approved labeling, and providers may continue to adhere to the metformin dosing restrictions based on SCr, as described in the package insert. We expect this recommendation to increase the number of patients who are candidates for metformin.

This study was a retrospective design; thus, there is potential for missing data such as providers not documenting ADEs or hypoglycemic events. Also, retrospective studies carry the potential for confounding variables; patients may have had documented or undocumented factors such as lifestyle changes that affected one or more of the outcomes assessed.

Additionally, although electronic records in the VA health care system provide an abundance of data, there is not a suitable method for assessing patient compliance to medication therapy in a retrospective study. The scope of this study also did not include an assessment of individual patients’ A1C goals or whether those goals were met on metformin-containing therapy versus alternative therapy.

Because this study was conducted at a VA facility, its results may not be applicable to a private institution. VA providers are restricted to formulary agents, and the majority of included patients were older males. Additionally, the cost data for the agents added to replace metformin were based on costs at the time of switching, and many of these medications are now available in generic form.

Our study did not collect data on patients’ race. We calculated the eGFR assuming all patients were non-black, which likely underestimated subjects’ renal function. Finally, although 16 patients would not have been able to continue metformin based on eGFR calculations at the time of their discontinuation, it is possible that adjustment of medications and increasing fluid intake could have resulted in an eGFR >30 mL/min, which may have made them candidates to continue metformin at a subsequent visit.

The results of this study promote the use of metformin in patients with type 2 diabetes in the setting of stable CKD to minimize worsening of glycemic control and risk of hypoglycemic events and avoid increased medication cost and potential ADEs from other antihyperglycemic therapies. To our knowledge, this is the first study that has evaluated the consequences of discontinuation of metformin in patients who could have continued if eGFR-based dosing were used. Providers should be educated about and implement current clinical practice guidelines and consider dosing metformin based on eGFR, rather than following the recommendations of the package insert, which state that providers should discontinue metformin when a patient’s SCr reaches the defined threshold in the setting of chronic but stable renal dysfunction.

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