Metformin as an adjunctive treatment to control body weight and metabolic dysfunction during olanzapine administration: A multicentric, double-blind, placebo-controlled trial

Introduction

There is a growing concern about the excessive body weight gain (BWG) and metabolic dysfunction associated with the use of some second generation antipsychotic drugs (SGAD) ( Newcomer and Haupt, 2006). In particular, olanzapine, a highly effective agent in schizophrenia and other psychotic disorders ( Lieberman et al., 2005), is linked to clinically significant BWG ranging from 0.9 kg/month up to 6 to 10 kg or more after 1 year of treatment ( ^{Nemeroff, 1997 Lieberman et al., 2005} ). The CATIE study, a major prospective project funded by the US National Institute of Health reported that 30% of olanzapine-treated schizophrenia patients gained 7% or more of their baseline body weight (BW) ( Lieberman et al., 2005). Hence, effective pharmacological and nonpharmacological strategies are urgently needed to assist an optimal BW control during olanzapine treatment (Faulkner et al., 2007). Olanzapine and clozapine administration is also linked to an elevated frequency of carbohydrate and lipid metabolic dysfunction and perhaps to a procoagulant state ( Hagg and Spigset, 2002).

Limited evidence exists that nizatidine (an H ₂ histamine receptor blocker), amantadine (a NMDA receptor antagonist), reboxetine (a noradrenergic antidepressant) sibutramine (a noradrenergic and serotonergic reuptake blocker) and topiramate (an anticonvulsant) induce a moderate degree of weight loss in the short-term in olanzapine-treated patients. However, the efficacy of these pharmacological treatments is inconclusive due to the limited number of randomized clinical trials (RCT) and small sample size in most studies ( Faulkner et al., 2007). The antidiabetic agent metformin is a promising agent since it improves glycemic control and promotes a moderate weight loss in diabetic and non-diabetic populations ( ^{Seufert et al., 2004 Wiernsperger and Bailey, 1999} ).

Four published studies have evaluated the effects of metformin in psychiatric populations. The first one was a placebo-controlled, crossed-over pilot study in which metformin was ineffective to decrease BW in obese women undergoing prolonged treatment with conventional antipsychotics ( Baptista et al., 2001). The second was an open study conducted in a child-adolescent population receiving diverse psychotropic medication; a significant BW loss was observed after metformin administration ( Morrison et al., 2002). The third was a RCT conducted in 40 schizophrenia inpatients starting olanzapine administration. The weight increment was similar in the metformin (5.5 kg) and the placebo group (6.3 kg) ( p= 0.4) after 14 weeks of treatment ( Baptista et al., 2006). The most recent study was a 16-week double-blind, placebo-controlled conducted in 39 subjects, ages 10–17, whose BW had increased by more than 10% during less than 1 year of olanzapine, risperidone, or quetiapine administration. Weight was stabilized in subjects receiving metformin, while those receiving placebo continued to gain weight at a rate of 0.31 kg/week ( Klein et al., 2006).

We aimed here to extend previous studies by assessing the effects of metformin on BW in adult patients under prolonged olanzapine administration. Besides evaluating the effects of the antidiabetic on serum glucose, glycated hemoglobin (HbA1c) insulin and lipids, we measured the levels of leptin, fibrinogen, c-reactive protein (CRP) cortisol and growth hormone (GH) as additional markers of metabolic dysfunction ( ^{Baptista and Beaulieu, 2002 Festa et al., 1999 Festa et al., 2000} ).

Methods

Assessments

Biochemical variables

Discussion

Prevention and treatment of the SGAD-induced excessive weight gain and metabolic dysfunction are a priority in current clinical practice. Nutritional advice, programmed physical activity ( ^{Centorrino et al., 2006Menza et al., 2004} ) and personalized SGAD selection are key factors in this endeavor. Additional drug assistance will often be necessary in non-compliant patients and/or in subjects with high risk of metabolic dysfunction. Among the potential pharmacological agents, metformin displays a favorable profile given its safety and effectiveness for metabolic control in diabetic and non-diabetic populations ( Glueck et al., 2006a). In addition, metformin has a neutral effect on BW or induces a moderate weight loss in non-psychiatric patients (Fontbonne et al., 1996).

In this study, we recruited a heterogeneous set of patients highly motivated to prevent or lose excessive BWG during olanzapine administration and avert carbohydrate and lipid dysregulation. In fact, most subjects were very compliant with the provided healthy life style recommendations. In consequence, the placebo group maintained a stable BW (− 0.18 ± 2.8 kg), leptin and lipid levels along the study. However, the HOMA-IR significantly increased at the consequence of insulin level elevation. Metformin administration was associated with a modest but significant BW decrease of − 1.40 ± 3.2 kg and a concomitant, marginally significant decrement in leptin levels. Importantly, glucose and insulin levels, and hence, the HOMA-IR remained stable under metformin. Unexpectedly, the HDL cholesterol levels significantly increased whereas the other lipid fractions remained even. The BW change in both groups was not significantly affected by gender, age and basal BMI. Unfortunately, few subjects had a BMI above 30 kg/m ²: 4 and 9 patients in the metformin and placebo groups respectively.

Previous studies of metformin administration in psychiatric patients do not allow a proper comparison with our results regarding the magnitude of BW change, since they were conducted in subjects receiving conventional antipsychotics ( Baptista et al., 2001), child-adolescent populations ( ^{Klein et al., 2006 Morrison et al., 2002} ) or adult schizophrenia patients just starting olanzapine administration (Baptista et al., 2006). Comparisons are not possible either with studies that used nizatidine and reported similar BWG as placebo after 12 weeks (0.7 vs. 1.1 kg) ( Assuncao et al., 2006) or favoring nizatidine after 16 weeks (placebo: 4.18 kg; nizatidine 150 mg BID: 3.56 kg; nizatidine 300 mg BID: 3.29 kg) ( Cavazzoni et al., 2003). Another study reported a significantly higher BW loss after sibutramine up to 15 mg/day (3.7 kg) than after placebo (0.8 kg) in a 12-week trial (Henderson et al., 2005). The difference in BWG between metformin and placebo in this study is consistent with a recent analysis of effectiveness of pharmacological management in schizophrenia reporting a modest prevention of weight gain ( n= 274, 6 Randomized Clinical Trials, Weight Mean Decrease:− 1.16 kg CI − 1.9 to −0.4) (Faulkner et al., 2007). No study has been published so far considering BW control with other antidiabetic agents during SGAD. Among the agents most frequently used in current clinical practice, the insulin secretagogues (sulfonylureas and non-sulfonylureas) may rather increase BW and additionally induce hypoglycemia, which is troublesome in psychotic patients. The thiazolidinediones can promote additional BWG but may provide supplementary benefits for endothelial protection ( Baptista et al., 2004).

Out of the psychiatric field, metformin has been used in BW control at least since 1965 ( ^{Pedersen, 1965 Pedersen and Olesen, 1968} ). Most studies have been conducted in people with type 2 diabetes, glucose intolerance, hyperinsulinemia or polycystic ovary syndrome (PCOS) and a minority in subjects with primary obesity. Importantly, most trials lasted more than 12 weeks and some of them extended up to 4 years. The magnitude of BW loss in PCOS ranged from 1.5–3.9 kg after 8 months of treatment ( Harborne et al., 2005); 2.84 kg after 6 months ( Tang et al., 2006); 7.7% and 8% of baseline BW after 14 months and 4 years respectively ( ^{Glueck et al., 2006a Glueck et al., 2006b} ); and 3.6 kg in adolescents after 1 year ( Glueck et al., 2006c). In a study in women with midlife weight gain and hyperinsulinemia, BW loss was 3.4 and 8.06 kg after 3 and 12 months respectively of metformin (1500 mg/day + diet) and 4.4 and 15.1 kg after 3 and 12 months respectively of metformin (2000 mg/day + diet) ( Mogul et al., 2001). In morbidly obese adolescents, metformin and a low-calorie diet induced a BW loss of 6.5% of basal BW after 8 weeks ( Kay et al., 2001). In general, it appears that the impact of metformin on BW is modest if it is not accompanied of diet. In addition, its lack of effect in reducing abdominal obesity has led some experts to affirm that definitively it is not a BW loss drug ( Lord et al., 2006).

The BW decrease after metformin (− 1.4 kg), while modest, may be relevant given the remarkably stable BW in the placebo group (− 0.18 kg). Specifically, it appears that these patients had reached a plateau in BWG with the assistance of healthy life style recommendations. Hence, metformin-induced BW loss occurred in subjects with an otherwise stable BW and might be even more robust in other clinical populations, for instance in subjects with elevated BMI, as it has been shown in subjects with PCOS ( Glueck et al., 2006a). Unfortunately our study comprised few subjects in this category. Age was another potential variable influencing the effects of metformin, since this agent displays less positive effects on glucose metabolism in older people ( Crandall et al., 2006). However, no significant correlation was observed between BW change and age after metformin ( p= 0.6), but a significant negative correlation was observed after placebo ( p= 0.02). As previously discussed, treatment duration was prolonged in those studies where metformin displayed a more robust effect on BW ( ^{Harborne et al., 2005 Glueck et al., 2006a Glueck et al., 2006b Mogul et al., 2001 Tang et al., 2006} ). Hence, it must be considered that a 12-week trial might be insufficient to achieve a clinically significant BW loss.

In order to identify biochemical variables associated with a positive effect of metformin, we correlated BW with those variables at baseline and their change after treatment. Only change in leptin levels and BW correlated positively in both treatment groups. However, this is expected, given the well demonstrated stimulating effect of BW gain on blood and tissue leptin levels ( Baptista and Beaulieu, 2002).

Interestingly enough, a marginally significant, negative correlation was observed between the change in BW and in BPRS scores in the metformin group ( p= 0.061). A positive correlation between BWG and clinical improvement under antipsychotic drug-treatment has been suggested ( Czobor et al., 2002), but it has not been definitively confirmed. In fact, we did not observe it in the placebo group. Hence, the clinical significance of that finding is unclear.

The HOMA-IR index remained stable during metformin administration and significantly increased after placebo. Hence, similarly to psychotropic drug-free subjects, metformin decreased insulin resistance ( Glueck et al., 2006a). Glycated hemoglobin (Hb1c) increased in both groups, but only reached statistical significance after metformin. The Hb1c increase occurred in 72% and 61% of patients in the metformin and placebo group respectively. This contradictory finding agrees with the notion that the HOMA-IR index and the Hb1c are assessing different aspects of carbohydrate metabolism. In addition, it suggests that in spite of metformin, olanzapine continued to impact metabolism. To further explore this issue, the subjects in each treatment group were divided according to the change in Hb1c (increase vs. decrease or no change). Then, we compared the change in the anthropometric and biochemical variables between these subgroups in each treatment group. Only BW loss was numerically higher in subjects in the metformin group whose Hb1c increased, but the difference was marginally significant: − 1.9 ± 3.5 kg vs.− 0.11 ± 1.9 kg; t(33) = 1.4, p= 0.1.

A secondary objective in our study was to assess cortisol and growth hormone (GH) levels, which are factors involved in insulin sensitivity regulation and, in the case of cortisol in the stress response. In addition, a significant decrease in cortisol has been observed after olanzapine and other SGAD administration ( ^{Baptista et al., 2006 Baptista et al., 2007} ). Cortisol increased in both groups but only reached marginal significance, with no between group differences. Limitations in blood sample size prevented us to quantify GH levels at baseline, and post-treatment values did not differ between the groups. Hence, the insulin sensitivity deterioration observed in the placebo group and its stability after metformin assessed through the HOMA-IR index is not directly explained by differences in cortisol and GH blood levels.

Another secondary objective was to assess the impact of metformin on markers of endothelial dysfunction (CRP) and procoagulant state (fibrinogen). This latter variable increased in both treatment groups, which suggests that olanzapine may promote a procoagulant state. Fibrinogen levels were not decreased by metformin in any sex (gender analysis not shown). The same lack of effect was recently observed in subjects with impaired glucose tolerance ( Haffner et al., 2005). CRP were not affected by metformin and rather tended to decrease after placebo, with no between group differences. Since some studies (Haffner et al., 2005) but not others ( Caballero et al., 2004) in subjects with impaired glucose tolerance showed that metformin decreases CRP mainly in women, we conducted an additional analysis discriminating by sex. In fact, metformin did not affect CRP (mg/dL) in men ( p= 0.44) but significantly decreased it in women (before treatment: 5.4 ± 3.1: after treatment: 3.7 ± 2.5; t(8) = 3.1, p= 0.016). However, the same trend was observed in the placebo group, even though the effect in women was not significant (men: p= 0.46; women:p= 0.068).

In summary, in a heterogeneous psychiatric population regarding basal BMI, olanzapine treatment duration and BWG before the study, we found that metformin induced a small but significant BW loss, leptin level decrease and a stable HOMA-R after 12 weeks. The Hb1c levels tended to deteriorate in both treatment groups, but unexpectedly only reached significance in the metformin group. The LDL and HDL cholesterol profile also deteriorated after metformin, whereas fibrinogen levels significantly increased in both groups. CRP levels tended to decrease in both groups but reached statistical significance in metformin-treated women.

Inconsistencies in the changes observed in the biochemical variables are probably related to a very important limitation of our study, which is the heterogeneity of basal BMI and previous BWG in our patients.

There is no any comparable study published so far, but our findings appears to agree with the conclusion that metformin is not a potent BW loss drug ( Crandall et al., 2006). However, given the very stable BW in the comparison group suggests that metformin-induced BW loss is clinically relevant. In addition, metformin significantly prevented an additional deterioration in the insulin resistance index.

Since the excessive BWG observed during olanzapine administration is probably related to increased appetite ( Baptista et al., 2004) and metformin appears to be devoid of direct anorectic effects, it is possible that a more prolonged administration is required before a robust impact on BW and positive effects on the lipid profile effect can be observed. This has been reported in some pivotal trials in non-psychiatric patients ( Glueck et al., 2006a).