Correspondence
Metformin-Associated Mortality in U.S. Studies
N Engl J Med 1996; 334:1612-1613June 13, 1996
- Article
To the Editor:
The data submitted for the approval of metformin for use in the United States were from two 29-week clinical trials involving patients with non-insulin-dependent diabetes mellitus (NIDDM), reported by DeFronzo et al. (Aug. 31 issue),1 and one 2-year, unreported, open-enrollment study.2 One of the clinical trials was placebo-controlled, with 143 patients assigned to receive metformin and 146 to receive placebo. In the second trial, 210 patients were assigned to monotherapy with metformin, 213 to metformin plus glyburide, and 209 to glyburide alone.
Six hundred two of these patients chose to enroll in the open study to receive metformin with or without a sulfonylurea drug: 75 patients from the placebo group, 142 from the glyburide group, 217 from the metformin groups, and 168 from the metformin-plus-glyburide group. With the open study, the total duration of metformin treatment during all U.S. open and controlled studies was increased to 1136 patient-years.
There was one death in the controlled trials, and there were six additional deaths in the open study. All seven deaths occurred among the patients who had initially been assigned to metformin therapy; no deaths occurred among those initially assigned to glyburide or placebo. A comparison of the survival distributions in the treatment groups by the log-rank test revealed a significant difference from the expected distribution of 4.4 deaths in the metformin group and 2.6 deaths in the control group (P = 0.04). Moreover, all seven deaths occurred among 423 patients (1.7 percent) randomly assigned to therapy with metformin in the second study. An analysis of survival in this subgroup revealed no statistically significant difference from an expected distribution of 4.9 deaths in the metformin group and 2.1 in the control group (P = 0.09). The shortest duration of treatment resulting in a death was 97 days, and the longest was 825 days (mean [±SD], 463±242 days).
Five of the seven deaths were from cardiovascular causes (including one case of lactic acidosis and another of sudden death after newly developed glomerular dysfunction). Another death was ascribed to suicide. A 1994 review of 255 cases of metformin-associated lactic acidosis2 revealed a significant proportion of cases due to suicidal overdose (4.7 percent; 95 percent confidence interval, 2.5 to 8.1 percent; P<0.01). The seventh death was ascribed to pulmonary fibrosis in a patient with a small-cell carcinoma of the lung.
Morbidity and mortality in patients with NIDDM are primarily from cardiovascular complications, which have not been found to be correlated with glucose control in any well-designed studies.3 Because of the substantial morbidity and mortality from cardiovascular causes in patients with this disorder, it is difficult to discern drug-induced cardiovascular toxicity. Wishing to monitor potential metformin-associated mortality, and without the benefit of the data on excess metformin-associated mortality noted above, the Endocrinologic and Metabolic Drugs Advisory Committee of the Food and Drug Administration (FDA) recommended on March 18, 1994, that a registry be established for all patients given metformin in the United States, should the drug be approved.4 So far, the FDA has not acted on this recommendation. A one-year study of 8000 patients receiving metformin alone or combined with a sulfonylurea drug and 2000 control patients has been requested by the FDA. This marketing study should allow the detection of differences in the rates of clinically important adverse events but will not detect small differences in overall mortality or mortality from cardiovascular causes.
4 ReferencesRonald Jay Innerfield, M.D.
Dorothy Bullock Memorial National Diabetes Center, Olney, MD 20832-10021
DeFronzo RA, Goodman AM, Multicenter Metformin Study Group
Full Text | Web of Science | Medline2
Innerfield RJ. Medical officer safety review of an NDA submission. Metformin. (NDA 20-357.) (On file at FDA, Bethesda, Md., May 1994 [FOI].)
3
Lebovitz HL . The DCCT and its implications for NIDDM. Clin Diabetes 1994;12:3-44
FDC Reports. Glucophage Rx registry to track adverse events recommended by FDA cmte. The Pink Sheet. March 21, 1994.
Author/Editor Response
The authors and the FDA reply:
To the Editor: Dr. Innerfield is concerned about a possible increase in mortality, particularly from cardiovascular causes, in patients with NIDDM treated with metformin on the basis of his analysis of data from our two double-blind, placebo-controlled, 29-week trials and an open-label, 2-year follow-up study involving many of the same patients. The patients who completed all the scheduled visits in the two 29-week trials were eligible for the open-label study, in which all patients initially received metformin as monotherapy. Dr. Innerfield failed to mention that glyburide was added if hyperglycemia was not satisfactorily controlled after one month of maximal treatment with metformin. Of the 602 patients enrolled in the open study, 87 percent concomitantly received glyburide for all but one month of the two-year period.
Of the seven deaths, five were related to cardiovascular causes, one was due to non-small-cell lung cancer, and one was due to suicide by ingestion of chemicals (not ingestion of metformin). Dr. Innerfield incorrectly states that all five patients who died from cardiovascular causes had “initially been assigned to metformin therapy” during the 29-week, double-blind trials. Only one of these patients was assigned to the metformin-alone group; the other four were assigned to the metformin-plus-glyburide group. He also fails to note that all five of these patients had previously received treatment with a sulfonylurea drug for an average of 10 years. The average duration of diabetes in these five patients was 14 years, all five were obese, two had dyslipidemia, two were smokers, and three had a history of hypertension. On the basis of these factors, we think Dr. Innerfield's comparison of the survival distributions has no scientific validity. His arbitrary, post hoc assignment of patients to metformin and control groups is erroneous and unjustified. The use of the log-rank test (or other, similar statistical tests) is based on the assumption that at the end of the study, patients remain in the group to which they were originally assigned. Dr. Innerfield's analysis violates this basic principle of statistics.
Dr. Innerfield also does not appear to be familiar with data on mortality in patients with NIDDM. In three large prospective studies of middle-aged white patients with NIDDM, the average annual death rates were 4.6, 5.2, and 5.7 percent.1 These rates are 10 times higher than the average death rate among the patients with NIDDM cited by Dr. Innerfield (7 of 602 patients died in a 2.6-year period, for a rate of 0.45 percent per year). On the basis of these numbers, it is difficult to believe that treatment with metformin or a sulfonylurea drug was associated with any increase in mortality.
Further support for the safety of metformin comes from the United Kingdom Prospective Diabetes Study.2 Since 1977, 753 patients with NIDDM have been randomly assigned to therapy with metformin alone or diet alone. A major safety end point of this study is cardiovascular events. To date, the data-monitoring committee has recommended no change in the protocol after approximately 7000 patient-years of experience (Turner RC: personal communication). The study is scheduled to be completed in 1998 and should provide useful data on mortality and morbidity from cardiovascular causes in patients with NIDDM treated with metformin.
2 ReferencesRalph A. DeFronzo, M.D.
University of Texas Health Science Center, San Antonio, TX 78284-7886Anita M. Goodman, M.D.
Lipha Pharmaceuticals, Inc., New York, NY 100191
National Diabetes Data Group. Diabetes in America. 2nd ed. Bethesda, Md.: National Institutes of Health, 1995:233-52. (NIH publication no. 95-1468.)
2
United Kingdom Prospective Diabetes Study Group
CrossRef | Web of Science | Medline
Author/Editor Response
The two 29-week controlled clinical trials of metformin were followed by a 2-year uncontrolled extension study in which all patients who chose to participate were treated with metformin. No deaths were reported among the 289 patients in the first trial, during either the trial or the extension study. One death was reported during the second trial, and six during the extension study. However, there was full reporting of deaths only through the end of the 29-week clinical trials. Of the 523 patients who completed the second trial, only 442 enrolled in the extension study, and 434 of these patients dropped out during the course of the two years. Vital status throughout the extension study was not reported for the patients who did not enroll or who dropped out.
We believe the lack of information about vital status throughout the extension study in most of the patients who had been enrolled in the clinical trials greatly limits any conclusions that can be drawn from the extension study about mortality rates, both for the patients who were treated with metformin in the clinical trials and for those who were not. Nevertheless, it is clear that more information is needed about the safety of metformin, as compared with the safety of other currently approved treatments for NIDDM. To meet this need, a large study is now under way in the United States, which is designed to compare total mortality, mortality from cardiovascular causes, and other outcomes for patients with NIDDM receiving metformin or usual care. The protocol for this study was developed in consultation with members of the FDA's Endocrinologic and Metabolic Drugs Advisory Committee who are experts in diabetes-related medicine and biostatistics. This study replaces the initial idea of a registry for patients receiving metformin, because a registry alone would not provide comparative data on outcomes with metformin and with other currently approved treatments for NIDDM.
Bruce V. Stadel, M.D., M.P.H.
John Gueriguian, M.D.
G. Alexander Fleming
Food and Drug Administration, Rockville, MD 20857-1706- Citing Articles (2)
Citing Articles
1
William Guy Herrington, Jeremy B. Levy. (2008) Metformin: effective and safe in renal disease?. International Urology and Nephrology 40:2, 411-417
CrossRef2
A. Holstein, D. Nahrwold, S. Hinze, E. -H. Egberts. (1999) Contra-indications to metformin therapy are largely disregarded. Diabetic Medicine 16:8, 692-696
CrossRef
