Papoz and colleagues at the Hotel Dieu, Paris conducted a randomized, doubleblind trial of a sulfonylurea (S) (glibenclamide, 2 mg twice daily), biguanide (B) (dimethylbiguanide, 0.85 g twice daily), alone and in combination (S+B) in a 2 X 2 factorial design with placebo202. Men aged 25-55 years (x = 45 yrs) who had borderline glucose tolerance (most would have IGT by current criteria) were randomized from 1969 to 1971 and tested for glucose and insulin levels every six months for two years.
There were 28% drop-outs, with similar levels across treatment groups. At two years, there were no significant differences in glucose or insulin levels in any group, though B, S+B and placebo groups all lost about 4 kg of weight, more than the 2 kg in the S group. Worsening to diabetes was not reported. The trial could have detected as significant drop of 21 mg/dl (1.2 mmol/L) in 2 h glucose with 95% power, though actual differences were very small (0-2 mg/dl, 0-0.1 mmol/L) between groups. This study was well conducted, even by modern criteria, and indicates no benefit of either drug to lower glucose levels or enhance insulin secretion. In addition, the authors note that OGTT testing was conducted while on medication, except for the last test, which was done 15 days after stopping the drug. Even the interim results showed essentially no differences in glucose and insulin levels while on medication.
Jarrett and co-workers identified a group of Whitehall civil service workers with borderline glucose tolerance from a large (n = 20000) survey204. Men aged 40-64 years were asked to participate in an RCT of dietary carbohydrate restriction and the biguanide phenformin (50 mg/day) for five years. Progression to diabetes was confirmed by OGTT or development of intercurrent symptoms and elevated glucose levels.
Overall, 27 of the 181 men (14.9%) completing the trial developed diabetes, with no significant differences between the treatment groups. Combining the phenformin groups across diet or no diet, the RR was 0.90 (CI 0.45-1.80) for the development of diabetes on drug. In logistic regression analyses adjusting for baseline imbalances at randomization, no effect of either diet or drug were seen. Only the level of fasting glucose predicted deterioration.
In a related report, no benefit (or harm) of phenformin on vascular or total mortality was noted205 using either intention-to-treat or efficacy (per protocol)
analysis. This study, like that of Papoz et al. 202, offers no evidence that phen-formin will prevent diabetes for up to five years.
Phenformin (50 mg) plus diet was compared to diet alone in Polish subjects starting in 1916 and continuing for five years as reported by Kasperska et at.206. Seventy-three people aged 13-14 (x = 54 years) were alternately allocated to interventions. At baseline, approximately 60% had 'borderline' glucose intolerance by older EASD criteria, similar to current IGT levels. Only 61% completed the five year trial, and over all subjects there was no benefit of phenformin in preventing worsening of glucose intolerance (RR = 0.89, CI 0.15-3.14). Among those with 'IGT', small differences in glucose and insulin levels were noted, which were not significant. This study provides no support for the use of phenformin for prevention of diabetes. The high rate of lactic acidosis with phenformin led to its removal from the market.
French investigators conducted two studies using the second-generation biguanide metformin, to determine the impact of this agent on components of the insulin resistance syndrome. In the first study, BIGuanides and Prevention of Risks of Obesity, BIGPRO-1)201,208 they selected 451 people with high waist-hip ratio (men >0.95, women >0.80) aged 35-60 (x = 49 years), randomly allocated them to 850 mg of metformin twice daily or placebo, and followed them every three months for one year. Both groups were given diet and physical activity instruction.
The metformin group showed small (-1 kg) but significant weight loss, less rise in fasting glucose, and marginally greater fall in fasting insulin, without changes in 1 h glucose or insulin. Lower LDL cholesterol, but no change in BP, triglycerides or HDL cholesterol were found. Of a number of hemostatic factors explored, the metformin group showed decreases in tissue plasminogen activator (tPA) antigen, and vonWillebrand factor109, but no change in plasminogen activator inhibitor-1 (PAI-1) activity or antigen not accounted for by weight loss. Five in the placebo group developed diabetes, versus none in the met-formin group (exact p = 0.06, post-hoc analysis). The results are difficult to interpret since there was a 18% and 30% drop-out rate in the two groups at one year. While subjects who dropped out were similar to those remaining in the trial, unexplained bias could have accounted for the results.
The investigators undertook a confirmatory study (BIGPRO-1.1) among 168 men who had slightly higher BP and triglyceride levels110,111. A similar randomization procedure and dose of metformin were used, though men were followed only for three months. Results were quite consistent with those of BIGPRO-1, in that fasting glucose, insulin, LDL cholesterol, ApoB (Apolipoprotein B) and tPA antigen declined, withouts change in BP or triglycerides. No diabetes occurred in either group and weight loss was only -0.5 kg.
In both studies, metformin produced more diarrhea, nausea and vomiting, but no hypoglycemic episodes108. The authors concluded that metformin would be an acceptable intervention in a type 1 diabetes prevention trial, but, with minimal changes in important cardiovascular risk factors, it would be less suitable for a CVD prevention study.
Li and colleagues evaluated the use of low-dose metformin (250 mg three times a day) compared to placebo to prevent the development of diabetes212. Subjects with IGT were identified during screening of a large workforce in Bejing, People's Republic of China, and 90 were randomly assigned to metformin or placebo.
After one year of follow-up, three people on metformin developed diabetes (7.1%), compared to six on placebo (14.0%) (RR = 0.51,CI 0.02-1.91, ARR = 6.9/100 person-years, post-hoc intention-to-treat analysis) There were also fewer people remaining with IGT and more reverting to normal glucose tolerance (p = 0.091). Using an efficacy analysis (including only the 70 persons who were compliant and continued in follow-up), the reduction in risk was naturally greater (RR = 0.19, CI 0.02-1.47), but it was not statistically significant unless the lower frequency of people with IGT were also included (p = 0.011).
This is one of the few pharmacological prevention studies using randomization, a placebo and a double-blind design where the primary end-point was conversion to diabetes using modern criteria. While the results suggest that a 50% or greater reduction in diabetes incidence occurred, the duration of follow-up (one year) and the number of subjects were both limited. Glucose tolerance testing was conducted quarterly and conversion on a single test was considered diabetes, so the relatively high conversion rates may be due in part to frequent testing. Nonetheless, both placebo and intervention groups had an equal testing frequency, and intention-to-treat analyses were presented. The rate of drop out for gastrointestinal side-effects was 4.4% among people randomized to metformin versus 0% on placebo, and the non-compliance rate (which could have included people with mild side-effects) was also higher among metformin subjects (15.6% versus 11.1%).
The Diabetes Prevention Program (DPP) also included metformin (850 twice daily) or placebo with simple annual lifestyle advice, in a double-blind design70. The design and intensive lifestyle results were discussed earlier in this chapter. The primary end-point was clinical or OGTT diabetes, confirmed on at least two tests. Randomization procedures and drug assignments were carefully masked to investigators, and the metformin and placebo groups were extremely similar at baseline on all risk factors measured71.
Metformin significantly reduced the incidence of diabetes by 31% compared to placebo (95% CI = 17-43%) (average annual incidence of 7.8/100 person-years versus 11.0/100 person-years in placebo; ARR = 3.2/100 person-years)72. Seventy-two percent of participants reported >80% adherence to the drug. In contrast to consistent results in subgroups for the DPP lifestyle intervention, metformin was less effective in persons with lower BMI, and in persons aged 60 years or greater. The lifestyle intervention was 39% more effective than metformin (p <0.001). Based on three-year lifetable cumulative estimates of diabetes development (28.9%, 21.7% and 14.4% in placebo, metformin and lifestyle groups respectively), the number needed to treat (NNT) was 6.9 to prevent one case using lifestyle, and was 13.9 for metformin. There were no differences in the rates of serious adverse events (hospitalizations, mortality) between groups, though mild gastrointestinal symptoms were significantly more common in the metformin group72.
The DPP is the first large, carefully randomized study to show that met-formin will delay or prevent diabetes in high-risk IGT subjects. The effect was smaller than that seen in the study in China by Li et al.212 Interestingly, in the DPP, fasting plasma glucose levels were reduced similarly in the metformin and lifestyle groups, though two-hour glucose levels were reduced much more by lifestyle intervention, consistent with metformin's action on hepatic glucose production. It is not known whether the reduction in incidence is due to an acute metabolic effect of treatment of hyperglycemia, or a more fundamental change in glucose homeostasis. A washout study is underway to explore this question. The long-term duration of the metformin effect is also not known.
Other published studies have been excluded since they included only a small number of subjects213 or were not randomized214.
Was this article helpful?