Context: Infusion of insulin acutely stimulates leptin production and chronic insulin treatment is associated with elevated serum leptin levels and body weight in subjects with type 2 diabetes.
Objective: The objective of the study was to investigate the relationship between insulin administration, leptin levels, and weight gain in subjects with type 2 diabetes.
Design: This was a post hoc analysis of two randomized, controlled trials.
Setting: The study was conducted at an outpatient clinic.
Subjects: Subjects included 35 (study 1) and 32 (study 2) poorly controlled oral hypoglycemic agent (OHA)-treated type 2 diabetic subjects.
Intervention: Study 1: subjects were investigated during a hyperinsulinemic, euglycemic glucose clamp and 12 months after being randomly allocated to start insulin or continue on OHAs. Study 2: 1 yr treatment with either OHAs and lifestyle intervention or insulin with or without concomitant lifestyle intervention.
Main outcome measure: Changes in serum leptin levels during clamp and during 1 yr of treatment in relationship to changes in body weight.
Results: Study 1: during acute insulin infusion leptin levels increased by 10% (P < 0.001). During 1 yr of insulin therapy, mean body weight increased by 6%, whereas the fasting leptin levels increased by 108% (both P < 0.001). The weight gain observed at 1 yr correlated with the increase in leptin levels observed during the clamp (r = 0.62, P = 0.003). Study 2: mean body weight increased by 4% (P < 0.01), whereas leptin levels increased by 56% (P < 0.001) during 1 yr of insulin treatment and the increase in leptin preceded the increase in body weight.
Conclusions: Significant correlations were observed between insulin’s effect on serum leptin levels and the increase in weight that accompanied insulin therapy.
Long-term insulin treatment in type 2 diabetes is typically accompanied by a mean weight gain of 4–5 kg the first year of treatment (1–3), and the underlying mechanisms for this weight gain are only partly understood.
Leptin is secreted from fat cells in response to nutritional status. Leptin acts on the hypothalamus and induces an anorectic response. In leptin-deficient obese mice and humans, administration of leptin decreases food intake and increases energy expenditure resulting in weight loss (4, 5). However, most obese individuals have elevated circulating levels of leptin as a consequence of their large fat mass, but they seem to have an inadequate response to the increased leptin levels. This underresponsiveness to leptin in common obesity has given rise to the idea that obesity is associated with or even caused by a state of relative leptin resistance (6–8).
In some, but not all, studies, infusion of insulin acutely stimulated leptin production (9, 10), and chronic insulin treatment was associated with elevated leptin serum levels and body weight in subjects with type 2 diabetes (11, 12). We hypothesized that insulin may be instrumental in the increase in leptin-levels observed during insulin treatment and that this hyperleptinemia may induce leptin resistance, which again may contribute to weight gain.
Subjects and Methods
This article is based on post hoc analyses of data from two previously published studies. We used stored sera from these studies to measure serum leptin levels and included all subjects from whom sera for leptin measurements were available.
Study 1
Subjects were recruited to participate in a prospective study of sulfonylurea (SU) vs. insulin treatment (Oslo Comparative Trial of Peroral vs. Insulin Treatment in Type 2 Diabetes study). The present data were obtained during the run-in phase and the first year of the prospective study. Subjects were randomly allocated to start insulin therapy or continue treatment with glibenclamide. Inclusion criteria were age 40–70 yr, body mass index (BMI) less than 35 kg/m2, duration of diabetes more than 2 yr without insulin treatment, glucagon-stimulated C-peptide greater than 0.7 nmol/liter, and glycosylated hemoglobin (HbA1c) greater than 7.5%. After the initial investigations, one group of subjects stopped treatment with oral antihyperglycemic agents (OHAs) and started with 8 IU Nuetral Protamine Hagedorn (NPH) insulin before breakfast and at bedtime. The other group continued on glibenclamide only. Insulin doses were increased subsequently with the aim to reach HbA1c levels less than 7.5%. The baseline findings and early follow-up data were previously published (2, 13).
Of the 53 subjects originally randomized to start insulin therapy or continue treatment with glibenclamide, in the present studies, we included the following three cohorts in whom sera for leptin measurements were available: 1) 35 subjects with repeated measurements during a euglycemic glucose clamp at baseline; 2) 21 of these subjects who were randomized to start insulin treatment and in whom weight changes during 1 yr follow-up were recorded; and 3) 13 of these subjects randomized to start insulin and 13 of the subjects who continued on glibenclamide in whom leptin and weight measurements before and after 12 months on the respective treatments were available.
At baseline, the subjects were treated with diet alone (two subjects) or diet in addition to glibenclamide. Four subjects were also using metformin. The 35 subjects (20 males) included in the baseline clamp analysis had a mean ± sd age of 59 ± 6.2 yr and a mean duration of diabetes of 7 ± 2.9 yr, BMI of 26.5 ± 3.9 kg/m2, and HbA1c 8.7 ± 1.5%.
The baseline characteristics of the subjects included in the analysis of leptin changes during 12 months treatment with insulin or glibenclamide are presented in Table 1. Mean age of these subjects was 60 ± 6.2 yr with a mean duration of diabetes of 8 ± 3.0 yr.
Mean ± sd baseline and 1 yr characteristics of population in study 1
| Insulin treateda n = 13 (5 males, 8 females) | OHA treatedb n = 13 (9 males, 4 females) | |||
|---|---|---|---|---|
| Start | 12 months | Start | 12 months | |
| Fasting plasma glucose (mmol/liter) | 11.7 ± 2.8 | 7.5 ± 3.0c | 11.5 ± 2.2 | 11.3 ± 2.3 |
| HbA1c (%) | 9.1 ± 1.7 | 8.1 ± 1.3c | 8.7 ± 1.5 | 9.3 ± 1.8 |
| Weight (kg) | 74.2 ± 13.2 | 78.8 ± 12.3d | 75.8 ± 12.3 | 75.5 ± 12.7 |
| BMI (kg/mb) | 26.3 ± 2.9 | 28.0 ± 2.8d | 26.2 ± 4.2 | 26.0 ± 4.2 |
| Leptin (pmol/liter)e | 405 (253–650) | 844 (545–1307)d | 301 (189–480) | 296 (183–478) |
| Fasting insulin (pmol/liter) | 100 ± 110 | 132 ± 72 | 101 ± 88 | 101 ± 60 |
| GDR (mg/kg/min) | 4.4 ± 2.1 | 5.0 ± 2.2 | 4.9 ± 2.6 | 4.6 ± 2.1 |
| Insulin treateda n = 13 (5 males, 8 females) | OHA treatedb n = 13 (9 males, 4 females) | |||
|---|---|---|---|---|
| Start | 12 months | Start | 12 months | |
| Fasting plasma glucose (mmol/liter) | 11.7 ± 2.8 | 7.5 ± 3.0c | 11.5 ± 2.2 | 11.3 ± 2.3 |
| HbA1c (%) | 9.1 ± 1.7 | 8.1 ± 1.3c | 8.7 ± 1.5 | 9.3 ± 1.8 |
| Weight (kg) | 74.2 ± 13.2 | 78.8 ± 12.3d | 75.8 ± 12.3 | 75.5 ± 12.7 |
| BMI (kg/mb) | 26.3 ± 2.9 | 28.0 ± 2.8d | 26.2 ± 4.2 | 26.0 ± 4.2 |
| Leptin (pmol/liter)e | 405 (253–650) | 844 (545–1307)d | 301 (189–480) | 296 (183–478) |
| Fasting insulin (pmol/liter) | 100 ± 110 | 132 ± 72 | 101 ± 88 | 101 ± 60 |
| GDR (mg/kg/min) | 4.4 ± 2.1 | 5.0 ± 2.2 | 4.9 ± 2.6 | 4.6 ± 2.1 |
The median dose of insulin at 12 months was 40 U in the insulin-treated group.
The median dose of glibenclamide was 10.5 mg at start and 12 months in the OHA-treated group, and one subject received additional metformin at 12 months.
P < 0.01, paired sample t test within treatment group, changes from baseline to 12 months.
P < 0.001, paired sample t test within treatment group, changes from baseline to 12 months.
Leptin values were ln transformed before calculating the geometric mean and 95% confidence interval.
Mean ± sd baseline and 1 yr characteristics of population in study 1
| Insulin treateda n = 13 (5 males, 8 females) | OHA treatedb n = 13 (9 males, 4 females) | |||
|---|---|---|---|---|
| Start | 12 months | Start | 12 months | |
| Fasting plasma glucose (mmol/liter) | 11.7 ± 2.8 | 7.5 ± 3.0c | 11.5 ± 2.2 | 11.3 ± 2.3 |
| HbA1c (%) | 9.1 ± 1.7 | 8.1 ± 1.3c | 8.7 ± 1.5 | 9.3 ± 1.8 |
| Weight (kg) | 74.2 ± 13.2 | 78.8 ± 12.3d | 75.8 ± 12.3 | 75.5 ± 12.7 |
| BMI (kg/mb) | 26.3 ± 2.9 | 28.0 ± 2.8d | 26.2 ± 4.2 | 26.0 ± 4.2 |
| Leptin (pmol/liter)e | 405 (253–650) | 844 (545–1307)d | 301 (189–480) | 296 (183–478) |
| Fasting insulin (pmol/liter) | 100 ± 110 | 132 ± 72 | 101 ± 88 | 101 ± 60 |
| GDR (mg/kg/min) | 4.4 ± 2.1 | 5.0 ± 2.2 | 4.9 ± 2.6 | 4.6 ± 2.1 |
| Insulin treateda n = 13 (5 males, 8 females) | OHA treatedb n = 13 (9 males, 4 females) | |||
|---|---|---|---|---|
| Start | 12 months | Start | 12 months | |
| Fasting plasma glucose (mmol/liter) | 11.7 ± 2.8 | 7.5 ± 3.0c | 11.5 ± 2.2 | 11.3 ± 2.3 |
| HbA1c (%) | 9.1 ± 1.7 | 8.1 ± 1.3c | 8.7 ± 1.5 | 9.3 ± 1.8 |
| Weight (kg) | 74.2 ± 13.2 | 78.8 ± 12.3d | 75.8 ± 12.3 | 75.5 ± 12.7 |
| BMI (kg/mb) | 26.3 ± 2.9 | 28.0 ± 2.8d | 26.2 ± 4.2 | 26.0 ± 4.2 |
| Leptin (pmol/liter)e | 405 (253–650) | 844 (545–1307)d | 301 (189–480) | 296 (183–478) |
| Fasting insulin (pmol/liter) | 100 ± 110 | 132 ± 72 | 101 ± 88 | 101 ± 60 |
| GDR (mg/kg/min) | 4.4 ± 2.1 | 5.0 ± 2.2 | 4.9 ± 2.6 | 4.6 ± 2.1 |
The median dose of insulin at 12 months was 40 U in the insulin-treated group.
The median dose of glibenclamide was 10.5 mg at start and 12 months in the OHA-treated group, and one subject received additional metformin at 12 months.
P < 0.01, paired sample t test within treatment group, changes from baseline to 12 months.
P < 0.001, paired sample t test within treatment group, changes from baseline to 12 months.
Leptin values were ln transformed before calculating the geometric mean and 95% confidence interval.
There were no significant clinical important differences between baseline variables in the subgroups who were included in the different subanalyses and those who were not.
Study 2
This study included 38 OHA-treated, overweight (BMI > 25 kg/m2) type 2 diabetic subjects with poor glycemic control (HbA1c 8–10.5%). They were originally randomly assigned to one of three treatment modalities for 1 yr: group 1 continued OHAs combined with lifestyle intervention (n = 11). The remaining subjects stopped OHAs and started insulin therapy alone (group 2, n = 12) or in combination with lifestyle intervention (group 3, n = 9). Subjects randomized to insulin treatment started with two doses of 8 IU NPH insulin in the morning and at bedtime. The doses were increased one to two times weekly, and regular insulin was added if necessary. The treatment goal was HbA1c less than 7.5%. Mean dose of insulin at 12 month was 62 (32–292) IU. The lifestyle intervention program focused on diet and exercise changes that were expected to produce weight loss and improve metabolic control. The participants were given dietary advice in 14 group sessions and two individual consultations throughout the year the intervention lasted. The exercise program consisted of group based aerobic exercise of moderate intensity for 1 h twice a week. The subjects received maximal tolerated doses of combination therapy with SU and metformin before the intervention started. The intervention took place in 1998–2000, and the main results from this study are previously published (1, 14).
In the present paper, the subjects were studied according to whether they received insulin treatment (n = 21) or not (n = 11) because previous analyses of the same material showed that both insulin-treated groups had a similar leptin response, regardless of whether the insulin treatment was combined with lifestyle intervention (14). The baseline characteristics of the two groups were similar (Table 2). Measurements of serum leptin (s-leptin) levels and body weight changes were available in 32 of the 38 subjects (18 males). There were no significant differences between baseline variables in the subgroup who were included in the present analysis and those who were not.
Mean (sd) baseline and 1 yr characteristics of population in study 2
| Insulin treateda n = 21 (12 males, 9 females) | Lifestyle and OHA treatedb n = 11 (6 males, 5 females) | |||
|---|---|---|---|---|
| Start | 12 months | Start | 12 months | |
| Fasting plasma glucose (mmol/liter) | 11.8 ± 3.0 | 8.5 ± 2.5c | 11.1 ± 2.0 | 8.6 ± 2.8 |
| HbA1c (%) | 9.2 ± 1.1 | 8.0 ± 1.0c | 8.7 ± 0.7 | 8.0 ± 1.2d |
| Weight (kg) | 88.7 ± 13.9 | 92.3 ± 14.4e | 94.2 ± 7.7 | 91.0 ± 6.9e |
| BMI (kg/mb) | 30.7 ± 4.2 | 31.9 ± 4.2e | 31.9 ± 4.8 | 30.8 ± 4.6e |
| Fat mass (kg) | 28.8 ± 7.6 | 31.1 ± 7.6e | 34.0 ± 11.3 | 32.6 ± 10.5d |
| Leptin (pmol/liter)f | 601 (458–788) | 939 (717–1230)c | 773 (459–1303) | 654 (359–1189) |
| Insulin treateda n = 21 (12 males, 9 females) | Lifestyle and OHA treatedb n = 11 (6 males, 5 females) | |||
|---|---|---|---|---|
| Start | 12 months | Start | 12 months | |
| Fasting plasma glucose (mmol/liter) | 11.8 ± 3.0 | 8.5 ± 2.5c | 11.1 ± 2.0 | 8.6 ± 2.8 |
| HbA1c (%) | 9.2 ± 1.1 | 8.0 ± 1.0c | 8.7 ± 0.7 | 8.0 ± 1.2d |
| Weight (kg) | 88.7 ± 13.9 | 92.3 ± 14.4e | 94.2 ± 7.7 | 91.0 ± 6.9e |
| BMI (kg/mb) | 30.7 ± 4.2 | 31.9 ± 4.2e | 31.9 ± 4.8 | 30.8 ± 4.6e |
| Fat mass (kg) | 28.8 ± 7.6 | 31.1 ± 7.6e | 34.0 ± 11.3 | 32.6 ± 10.5d |
| Leptin (pmol/liter)f | 601 (458–788) | 939 (717–1230)c | 773 (459–1303) | 654 (359–1189) |
The median dose of insulin at 12 months was 62 U in the insulin-treated group.
At start, all subjects in the OHA-treated group except one received a combination of metformin and SU. The median doses and number of subjects who used the respective therapies were: metformin, 1250 mg (n = 10); glibenclamide, 10.5 mg (n = 8); glipizide, 17.5 (n = 2); glimeperide, 6.0 mg (n = 1).
P < 0.001, paired sample t test within treatment group, changes from baseline to 12 months.
P < 0.05, paired sample t test within treatment group, changes from baseline to 12 months.
P < 0.01, paired sample t test within treatment group, changes from baseline to 12 months.
Leptin values were ln transformed before calculating the geometric mean and 95% confidence interval.
Mean (sd) baseline and 1 yr characteristics of population in study 2
| Insulin treateda n = 21 (12 males, 9 females) | Lifestyle and OHA treatedb n = 11 (6 males, 5 females) | |||
|---|---|---|---|---|
| Start | 12 months | Start | 12 months | |
| Fasting plasma glucose (mmol/liter) | 11.8 ± 3.0 | 8.5 ± 2.5c | 11.1 ± 2.0 | 8.6 ± 2.8 |
| HbA1c (%) | 9.2 ± 1.1 | 8.0 ± 1.0c | 8.7 ± 0.7 | 8.0 ± 1.2d |
| Weight (kg) | 88.7 ± 13.9 | 92.3 ± 14.4e | 94.2 ± 7.7 | 91.0 ± 6.9e |
| BMI (kg/mb) | 30.7 ± 4.2 | 31.9 ± 4.2e | 31.9 ± 4.8 | 30.8 ± 4.6e |
| Fat mass (kg) | 28.8 ± 7.6 | 31.1 ± 7.6e | 34.0 ± 11.3 | 32.6 ± 10.5d |
| Leptin (pmol/liter)f | 601 (458–788) | 939 (717–1230)c | 773 (459–1303) | 654 (359–1189) |
| Insulin treateda n = 21 (12 males, 9 females) | Lifestyle and OHA treatedb n = 11 (6 males, 5 females) | |||
|---|---|---|---|---|
| Start | 12 months | Start | 12 months | |
| Fasting plasma glucose (mmol/liter) | 11.8 ± 3.0 | 8.5 ± 2.5c | 11.1 ± 2.0 | 8.6 ± 2.8 |
| HbA1c (%) | 9.2 ± 1.1 | 8.0 ± 1.0c | 8.7 ± 0.7 | 8.0 ± 1.2d |
| Weight (kg) | 88.7 ± 13.9 | 92.3 ± 14.4e | 94.2 ± 7.7 | 91.0 ± 6.9e |
| BMI (kg/mb) | 30.7 ± 4.2 | 31.9 ± 4.2e | 31.9 ± 4.8 | 30.8 ± 4.6e |
| Fat mass (kg) | 28.8 ± 7.6 | 31.1 ± 7.6e | 34.0 ± 11.3 | 32.6 ± 10.5d |
| Leptin (pmol/liter)f | 601 (458–788) | 939 (717–1230)c | 773 (459–1303) | 654 (359–1189) |
The median dose of insulin at 12 months was 62 U in the insulin-treated group.
At start, all subjects in the OHA-treated group except one received a combination of metformin and SU. The median doses and number of subjects who used the respective therapies were: metformin, 1250 mg (n = 10); glibenclamide, 10.5 mg (n = 8); glipizide, 17.5 (n = 2); glimeperide, 6.0 mg (n = 1).
P < 0.001, paired sample t test within treatment group, changes from baseline to 12 months.
P < 0.05, paired sample t test within treatment group, changes from baseline to 12 months.
P < 0.01, paired sample t test within treatment group, changes from baseline to 12 months.
Leptin values were ln transformed before calculating the geometric mean and 95% confidence interval.
The mean age was 57 ± 2.7 yr and the mean duration of diabetes was 8 ± 4.7 yr. All subjects were treated with OHAs when recruited. In the group that continued OHA treatment together with lifestyle intervention, all were using SU and metformin in combination except one subject who did not tolerate metformin and one subject who also used acarbose. The doses of OHAs were kept stable during the intervention year, except in two subjects who reduced their doses of metformin and/or SU due to hypoglycemic episodes.
In both studies eligible subjects completed a consent form approved by the regional medical ethics committee. All subjects went through a clinical examination including measurements of body weight and height in light clothes, and BMI was calculated. In study 2, body composition was also estimated at baseline and at 12 months by dual-energy x-ray absorptiometry scanning (DPX-l; Lunar Radiation Corp., Madison, WI). Subjects were seen every third month in the outpatient clinic, body weight and insulin doses were recorded, and fasting venous blood samples were drawn for the measurements of HbA1c and glucose values. Before start and after 12 months, serum was frozen at −40 C for later analyses.
Euglycemic, hyperinsulinemic glucose clamp
In study 1 a euglycemic, hyperinsulinemic glucose clamp was performed as previously described (15). Briefly, subjects met after an overnight fast, and a Teflon catheter was placed in a hand vein and connected to an artificial pancreas (Betalike; Esacontrol, Genova, Italy), which measured blood glucose concentration and controlled the infusion rates of glucose. Insulin (3.6 U/ml) was infused at a fixed rate of 1 mU/kg · min. When the target glycemic level of 5 mmol/liter was reached, glucose (240 mg/liter) was infused in a rate to keep the euglycemia for 2 h. Euglycemia was reached in all subjects within 40–210 min after the start of the insulin infusion. Blood was sampled before the start and every 5 min for the measurement of glucose levels and for the measurement of serum insulin and leptin levels before the start, 1 h after start of insulin infusion, and thereafter when the glucose level reached 5 mmol/liter and 1 and 2 h later. The glucose disposal rate (GDR) was calculated as the amount of glucose infused in the steady state during the last 20 min of the clamp. Average insulin concentration during the last 20 min was 655 ± 209 pmol/liter.
Laboratory analyses
Glucose was analyzed with the glucose oxidase method using a Glucose Analyzer II (Beckman Instruments, Fullerton, CA), and HbA1c was determined using HPLC in study 1 with Diamat and in study 2 with Variant (both from Bio-Rad, Richmond, CA). Insulin was measured with an in-house RIA method with intraassay coefficient of variation less than 8% as previously described (15). Leptin was measured with a commercially available RIA (Linco Research Inc., St. Charles, MO).
Statistical methods
Values are given as mean ± sd. s-leptin levels were ln transformed to obtain normal distribution, and values are given as geometric mean (95% confidence interval) if not stated otherwise. To test the significance of changes from baseline to 12 months within the groups, the paired-sample t test was used. To test correlations between changes in s-leptin and changes in body weight, we used Spearman’s rho because leptin values were not evenly distributed. The difference between groups concerning continuous variables was analyzed by Student’s t test for unpaired samples. Linear mixed model was used to test whether there was a significant trend in leptin changes between or within the groups during the intervention in study 2. To test whether the changes in body weight confounded the trends observed in leptin, we did additional trend analyses with changes in BMI added as covariates in the model. Finally, additional multiple regression analyses were done to find predictors of change in weight and change in s-leptin.
Statistical analyses were performed with SPSS for Windows 14.0.2 (SPSS Inc., Chicago, IL). Two-sided P <0.05 was considered significant.
Results
Study 1
During 4.2 ± 1.0 h of insulin infusion, the mean leptin levels increased by 10% from 383 (287–510) pmol/liter at start to 422 (315–563) pmol/liter at the end of the insulin infusion (P < 0.001, n = 35). Figure 1 shows the changes in serum levels of insulin and leptin and plasma levels of glucose measured at different time points during the clamp.
Changes in s-insulin (geometric mean, 95% confidence interval), p-glucose (mean, 95% confidence interval), and s-leptin (geometric mean, 95% confidence interval) during baseline hyperinsulinemic, euglycemic glucose clamp (n = 35). p-glucose, Plasma glucose.
During 1 yr of insulin treatment, mean leptin levels increased markedly by 108% from 405 (253–650) pmol/liter to 844 (545–1307) pmol/liter (P < 0.001, n = 13). During the same period, body weight increased by 6% (P < 0.001), and HbA1c was significantly reduced (Table 1). In the group that continued oral treatment for their diabetes, the HbA1c slightly increased (P = 0.15, n = 13), and there were no significant changes in weight or fasting leptin levels (Table 1).
In bivariate analysis, the percent weight gain observed during 1 yr insulin therapy correlated significantly with the insulin stimulated increase in leptin during clamp (Fig. 2). The 21 subjects shown in Fig. 2 are those with a complete set of leptin values during baseline clamp that also had follow-up data after 12 months. In the insulin-treated subjects, the Spearman’s nonparametric correlation coefficient calculated between Δ-leptin (percent) levels during acute insulin infusion and Δ-weight (percent) from start of insulin treatment to 12 months was 0.62 (P = 0.003). The weight gain at 12 months also tended to correlate with the increase in fasting leptin levels during the same period in the insulin-treated subjects (n = 13, Spearman’s rho = 0.53, P = 0.061).
Relationship between the changes observed in s-leptin levels from start to end of baseline clamp and the changes in body weight from start of insulin therapy to 12 months in the 21 subjects with available leptin measurements during clamp and weight recordings after 12 months (study 1). Line, Linear regression line.
There was no correlation between GDR at baseline and leptin change during clamp (r = −0.07, P = 0.78) or during 1 yr of insulin treatment (r = 0.12, P = 0.71). Neither was there any correlation between GDR at baseline and body weight change after 1 yr of insulin treatment (r = 0.28, P = 0.36). However, there was a highly significant inverse correlation between fasting s-leptin levels and GDR at baseline (Spearman’s rho = −0.58, P = 0.002). In a multivariate regression analysis, changes in leptin levels from baseline to 12 months correlated significantly with weight gain (P < 0.001), whereas changes in fasting insulin levels did not (P = 0.71).
Study 2
During 1 yr of insulin treatment, the mean levels of leptin increased by 56% from 601 (458–788) pmol/liter to 939 (717–1230) pmol/liter (P < 0.001, n = 21). During the same period, body weight increased by 4.1% (P = 0.005), and HbA1c was significantly reduced (Table 2). In the group that continued treatment with OHAs and received lifestyle intervention, there was a mean reduction in body weight of 3.4% (P = 0.005, n = 11) and a significant reduction in HbA1c, but there was no significant change in s-leptin levels (Table 2).
The increase in s-leptin preceded the increase in body weight in the insulin-treated subjects (Fig. 3). After 3 months of insulin treatment, there was a 38% increase in s-leptin levels in the insulin-treated group, whereas there was no significant increase in body weight (0.24 kg, P = 0.54). After 6 months of treatment, body weight had increased by 2.4% and the mean s-leptin level had increased by 49% from baseline.
Mean (95% confidence interval) changes in body weight and s-leptin from baseline to 12 months in study 2. White circles and dotted line, Insulin therapy (n = 21), black circles and solid line, no insulin therapy (n = 11).
The changes in leptin levels were significantly different between the groups also by trend analysis (P < 0.0001). There was a significant increase in leptin levels in the insulin-treated group [β = 30 (95% confidence interval 20–39) P < 0.001], whereas the changes observed in the lifestyle+OHA group was not significant [β = −7 (−20 to 6), P = 0.32]. When adjusted for changes in BMI during the intervention, the change in leptin levels within the insulin-treated groups remained significant but was less marked [β = 20 (10–30), P < 0.001].
In the insulin-treated subjects, the percent change in s-leptin from baseline to 3 months tended to correlate with changes observed in body weight (r = 0.42, P = 0.057) and significantly with the percent change in total body fat mass from baseline to 12 months (Spearman’s rho = 0.49, P = 0.025). This correlation was particularly strong in women (r = 0.80, P = 0.010, n = 9).
After 1 yr of insulin treatment, the regression curve between BMI and s-leptin levels was shifted upward in both studies when evaluated separately and combined. At the same BMI level, s-leptin levels were higher than what was observed before insulin treatment (Fig. 4).
The relationship between BMI and s-leptin at baseline and after 1 yr of insulin therapy in all insulin-treated subjects in study 1 (n = 13) and study 2 (n = 21) combined (n = 34) (A) and all OHA-treated subjects in study 1 (n = 12) and study 2 (n = 11) combined (n = 23) (B). Open circles and dotted line, Baseline values with corresponding linear regression line; black circles and solid line, 12-month values and corresponding linear regression line after 1 yr of insulin therapy. Open triangles and dotted line, Baseline values with corresponding linear regression line; black triangles and solid line, 12-month values and corresponding linear regression line after 1 yr of OHA therapy.
In separate multiple regression analyses performed for studies 1 and 2, with change in leptin levels between start and 12 months as the dependent variable, the associated independent variables were change in body weight (study 1: P = 0.069 and study 2: P = 0.010) and treatment group (P = 0.031 and P = 0.093), whereas change in HbA1c was not significantly associated (P = 0.25 and P = 0.43).
Likewise, changes in HbA1c were not significantly correlated to changes in body weight (dependent variable) in a multiple regression analysis with leptin change and treatment group as independent variables in any of the two studies, in neither the total study population (study 1: P = 0.62, study 2: P = 0.34) nor the insulin-treated individuals (study 1: P = 0.14, study 2: P = 0.30).
Discussion
The main finding in the present study was the significant correlation between the increase in leptin levels after acute insulin infusion and the weight gain that accompanied insulin therapy during 12 months of treatment. After 1 yr of insulin treatment, the relative increase in fasting leptin was substantially higher than could be expected from the increase in body weight or fat mass and the increase in s-leptin levels preceded the weight gain. Based on these findings, we speculate that insulin treatment may induce or aggravate resistance to the action of leptin and thus contribute to body weight increase in this population of patients with type 2 diabetes.
Saad et al. (16) reported that the insulin-induced increase in leptin was proportionally lower in obese, insulin-resistant men compared with lean healthy men. However, Malmstrom et al. (9) found no difference in response between type 2 diabetic subjects and nondiabetic controls other than a prolonged leptin response from insulin (8.5 h before increase compared with 6 h). We found that the most insulin-resistant subjects (judged from the glucose disposal rate during baseline clamp) had the highest levels of fasting s-leptin, but there was no correlation between GDR at baseline and changes in leptin during clamp or during 1 yr of insulin treatment. This suggests that even though insulin-resistant subjects started with elevated levels of s-leptin, insulin resistance was not important for changes in leptin levels in response to insulin, acute or long term.
Previous reports found that long-term insulin therapy in type 2 diabetic subjects increases body weight and plasma leptin levels and that the increase in leptin levels were significantly greater than would have been expected from weight change alone (11, 12). Furthermore, both boys and girls with type 1 diabetes had higher leptin levels adjusted for percent body fat than controls (17, 18). This was related to insulin dose and greater increase in fat mass among girls in one study (17), whereas in the other study, a significant positive linear correlation between leptin and daily insulin dosage (units per kilogram) was found in all type 1 diabetic subjects (18). In another study, initiation of insulin treatment in newly diagnosed type 1 diabetic children produced a rapid increase in s-leptin levels independent of body weight change. Serum leptin levels correlated with insulin doses at 3–5 d of insulin treatment (19).
Although we cannot exclude that improvement in glycemic control may have increased leptin levels during our 1-yr studies, we did not observe significant correlations between change in leptin levels and changes in HbA1c levels when tested in multiple regression models. Furthermore, we cannot completely exclude a leptin-suppressing effect of OHAs in our studies; however, most subjects were already treated with OHAs at start and had no or only a modest increase in dose during the study. Data from the United Kingdom Prospective Diabetes Study (12) showed no difference in leptin levels between subjects treated with OHAs or diet alone, whereas insulin-treated individuals had relatively higher leptin levels, and Sivitz et al. (20) found that glibenclamide therapy, but not metformin, increased fasting leptin levels after a period of discontinuation of OHA treatment in type 2 diabetic subjects.
There might be a gender difference in the leptin response to insulin because women showed a considerable higher increase in leptin levels compared with men in one study (21). Utriainen et al. (10) found no significant difference in plasma leptin concentrations during clamp between men and women when the levels were compared using the percentage of fat as the covariant. We did not observe gender differences in leptin response to insulin during clamp, but during 1 yr of insulin therapy women showed a 3-fold higher mean increase in serum leptin compared with men.
To our knowledge, no other study has shown a relationship between leptin increase during clamp and subsequent weight increase during insulin treatment. The observation that the increase in leptin occurs before body weight increase in insulin treatment, indicate that insulin per se, and not increased fat mass, is responsible for the increase in leptin levels. It may also explain why the leptin increase is much greater than what could be expected from increase in fat mass alone. The fact that the regression curve between BMI and s-leptin levels was shifted upward in both studies after 1 yr of insulin treatment further strengthen our hypothesis that insulin may increase leptin levels over and above what could be expected from the body weight increase.
There is evidence in the literature indicating that an overstimulation of the leptin signaling pathway might induce leptin resistance and hence attenuate leptin’s protective effect against weight gain (6, 7, 22). Chronic overexpression of central leptin in rodents induces a leptin resistance that mimics many of the characteristics associated with diet-induced obesity including reduced leptin receptors, diminished signaling, and impaired responsiveness to exogenous leptin (8).
Alternatively, s-leptin may simply be a marker of propensity for weight gain. Subjects who exhibit the highest increase in leptin may be those with a greater number of adipocytes, and as a consequence these subjects may have a greater capacity to gain body weight. If this is the case, the increase in body weight gain is simply a result of the anabolic/energy storing effects of insulin (23).
The increase in body weight could also be due to improvement in glucometabolic control and less loss of energy in glucosuria, but changes in HbA1c were not significantly correlated to changes in body weight in a multiple regression analysis with leptin change and treatment group as independent variables in any of the two studies. Neither did we observe any change in energy intake in the insulin-treated subjects in study 2, based on a 5-d food record before and at the end of the intervention. The combined OHA and lifestyle intervention group reduced their energy intake by 225 kcal/d and their body weight by 3 kg (1).
Understanding the role of leptin in insulin-induced weight gain may be important in unraveling linkages between insulin resistance, hyperinsulinemia, and obesity and its relationship to cardiovascular disease.
A limitation of the current study is that it represents a post hoc analysis of two studies that were not designed to address the question of insulin’s effect on serum leptin and consequences for body weight changes. We set out to investigate the effect of different treatment options on metabolic control and body weight. Leptin was investigated as a possible mediator that could explain weight gain during insulin treatment of type 2 diabetes. The small sample size raises the question of the generalizability of the findings. The population used in this investigation consisted of a random selection of the complete study population in study 1. In study 2 missing subjects were those who did not complete the intervention study. Selection bias could conceivably have influenced our results. However, we have little a priori reason to believe that the association between exogenous insulin and leptin and body weight should be stronger or weaker among those who did not complete the intervention.
In conclusion, acute, continuous insulin infusion increased serum leptin levels, and the increase was positively correlated to subsequent weight increase after chronic insulin therapy in subjects with type 2 diabetes. Long-term insulin treatment with NPH insulin increased serum leptin levels inappropriately according to the increase in body mass, and the increase in leptin levels preceded the increase in body weight. Further studies exploring these mechanisms are recommended.
Acknowledgements
This work was supported by research grants from the Norwegian Foundation for Health and Rehabilitation and the Norwegian Diabetes Association.
Disclosure Summary: K.F.H. and P.M.T. have nothing to declare. A.-M.A., J.P.B., and K.I.B. have received lecture fees from Novo Nordisk Norway AS and Eli Lilly Norway AS.
We acknowledge Dr. Jan Falch (Aker University Hospital) for performing dual-energy x-ray absorptiometry for estimation of body composition in study 2.
Abbreviations
- BMI
Body mass index
- GDR
glucose disposal rate
- HbA1c
glycosylated hemoglobin
- OHA
oral hypoglycemic agent
- s-leptin
serum leptin
- SU
sulfonylurea.
