SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Impaired glucose tolerant (IGT) adults are at elevated risk for cardiovascular disease (CVD). Exercise or metformin reduce CVD risk, but the efficacy of combining treatments is unclear.

Objective:

To determine the effects of exercise training plus metformin (EM), compared with each treatment alone, on CVD risk factors in IGT adults.

Design and Methods:

Subjects were assigned to placebo (P), metformin (M), exercise training plus placebo (EP), or EM (8/group). In a double-blind design, P or 2,000 mg/d of M were administered for 12 weeks and half performed aerobic and resistance training 3 days/week for ∼60 min/day at 70% pretraining heart rate peak. Outcomes included adiposity, blood pressure (BP), lipids, and high sensitivity C-reactive protein (hs-CRP). Z-scores were calculated to determine metabolic syndrome severity.

Results:

M and EM, but not EP, decreased body weight compared with P (P < 0.05). M and EP lowered systolic blood pressure by 6% (P < 0.05), diastolic blood pressure by 6% (P < 0.05), and hs-CRP by 20% (M: trend P = 0.06; EP: P < 0.05) compared with P. Treatments raised high-density lipoprotein cholesterol (P < 0.05; EM: trend P = 0.06) compared with P and lowered triacyglycerol (P < 0.05) and metabolic syndrome Z-score compared with baseline (EP; trend P = 0.07 and EM or M; P < 0.05).

Conclusions:

Although exercise and/or metformin improve some CVD risk factors, only training or metformin alone lowered hs-CRP and BP. Thus, metformin may attenuate the effects of training on some CVD risk factors and metabolic syndrome severity in IGT adults.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Individuals with impaired glucose tolerance (IGT) are at elevated risk for cardiovascular disease (CVD ref. 1,2) and approximately half of these individuals have metabolic syndrome (i.e., hypertensive, hyperglycemic, and dyslipidemic; ref. 3). CVD risk is largely explained by insulin resistance and excess body weight (4,5). Treatments that raise insulin sensitivity or lower body weight may lower CVD risk in individuals with IGT.

The mechanisms by which exercise lowers CVD risk likely involve increasing insulin sensitivity (6-,7,8) and lowering circulating lipids (e.g., triacylglycerol (TAG), low-density lipoproteins, etc.), blood pressure (BP), and C-reactive protein (9-,10,11). Metformin treatment can also increase insulin sensitivity (12,13) and in addition, reduces fasting hyperglycemia and body weight (14,15). Because the actions of exercise and metformin are potentially additive, it has been suggested that individuals with IGT and at least 1 CVD risk factor (e.g., hypertension, elevated TAG, low high-density lipoproteins (HDL), fasting hyperglycemia, etc.) be considered for metformin treatment while participating in a regular exercise program (16). However, the interactions between exercise and metformin on CVD risk factors have not been systematically evaluated (17,18). Adding to the uncertainty, we recently showed that metformin blunted the effects of training on insulin sensitivity in men and women with IGT (19). Since insulin resistance is believed to be a key underlying factor causing metabolic syndrome (formerly syndrome X), it is important to understand the effects of metformin on CVD risk after exercise training (20). Therefore, the purpose of this study was to determine the effects of combining metformin with exercise training, compared with either treatment alone, on reducing CVD risk factors in men and women with IGT. Given the opposing effects on insulin sensitivity we observed in this same study group (19), we hypothesized that the combined treatment would oppose the reduction in CVD risk compared with either treatment alone.

Methods and Procedures

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Overview

The effects of exercise training, metformin, or the combined treatment on CVD risk factors were compared in 32 otherwise healthy men and women with IGT. Using a double-blind, placebo-controlled design, individuals were tested before and after 12 weeks of placebo (P), metformin (M), exercise training plus placebo (EP), or exercise training plus metformin (EM). Outcome measures included body weight, waist circumference (WC), BP, and concentrations of blood lipids, and high sensitivity C-reactive protein (hs-CRP) as well as metabolic syndrome severity.

Subjects

These subjects were the same individuals who served as the study population in our prior study on whole-body and hepatic insulin sensitivity (19). Subjects were recruited via newspaper advertisements and flyers posted in the local community. Individuals (Table 1) were nonsmoking, weight stable (<5% weight change over last 3 months) and free of CVD or type 2 diabetes. All subjects had IGT. Individuals who were using medications to manage BP (M = 1, EP = 1, and EM = 3) or cholesterol (M = 1, EP = 1, and EM = 1) were enrolled and continued treatment throughout the study. Subjects were excluded from the study if they had any contraindications to metformin use (e.g., respiratory disease, heart failure, renal/hepatic disease). All subjects were verbally briefed about the study and signed informed consent documents approved by the Institutional Review Board at the University of Massachusetts Amherst.

Table 1. Subject characteristics
inline image

Screening

A 75-g oral glucose tolerance test was used to screen individuals for IGT after a minimum 5-h fast. Individuals with glucose concentrations between 7.8 and 11.1mmol/l (140-199 mg/dl) 120 min after consuming glucose met the criterion for IGT. Approximately half of the subjects with IGT also had impaired fasting glucose, i.e., fasting glucose concentrations between 5.5 and 6.9 mmol/l (100-125 mg/dl). Most of the individuals in each treatment group (M = 6; EP = 7, and EM = 6) also met ATP III criteria for metabolic syndrome (21). Only 2 participants in the P group had metabolic syndrome. To minimize the potential for type 1 and type 2 errors, we restricted the presentation of results to the three treatment groups.

Exercise testing

VO2 peak was used to characterize cardiorespiratory fitness, and 1-repetition max tests were performed to determine strength as previously described (19).

Metformin or placebo protocol

A member of the research group, blinded to the protocol, administered pills to the subjects. All subjects were instructed to take their pills with food to minimize potential side effects. Subjects started treatment with 500 mg/d of metformin. The dose was increased 500 mg/d each week until a standard clinical dose of 2,000 mg/d was reached by week 4. This dose was maintained for the ensuing 8 weeks of the protocol.

Exercise training

Subjects underwent supervised exercise sessions 3 days/week for 60-75 min per session (∼190 min/week total). Subjects cycled for 45 min at 70% of their pretraining heart rate peak 3 days/week, and performed whole-body resistance exercise at 70% of 1-repetition max of the subject 2 days/week. Subjects performed two sets of 12 repetitions for: chest press, latissimus pull down, leg press, bicep curl, triceps pushdown, and upright rows. Pedometers (Omron HJ112, Lake Forest, IL) were provided to all subjects and worn around the waistband for 7 consecutive days at week 0, 6, and 12. These data were averaged to characterize habitual ambulation. Habitual ambulation did not change in any group after the intervention, nor were there differences across groups (data not shown).

Body weight, composition, and WC

Body weight was recorded without shoes, to the nearest 0.1 kg on a calibrated scale, every 2 weeks at consistent times of day. Subjects were weighed in the fasted state for the first and last measurements. As previously described (19) body fat and central fat (defined as fat mass in the region bounded by the last floating rib to the top of the iliac crest/total fat mass) were assessed by dual-energy X-ray absorptiometry (Lunar Prodigy, Madison, WI). WC was measured to the nearest 0.25 cm with a plastic tape measure in the standing position ∼2 cm above the umbilicus.

Food intake

Food intake was assessed using 3-day food records at week 0, 6, and 12. Subjects selected 3 days of the week at the start of the study and the same 3 days of the week were used each time food intake was assessed. Results were averaged across the 3 days to generate a single value for energy intake at each time point. Participants were given verbal and written instructions to accurately record the types and quantities of all food and beverages consumed, including brand names, and methods of food preparation. Food records were analyzed (by the same investigator S.C.) using a commercial software program (Fitday, El Segundo, CA).

BP, lipids, and high sensitivity CRP

Subjects were provided isocaloric meals (∼55% carbohydrate, 30% fat, and 15% protein multiplied by 1.3 activity factor) 24 h before baseline and week 12 measurements. Energy intake was based on resting metabolic rate measurements for 30 min in the supine position by indirect calorimetry (SensorMedics 800, Yorba Linda, CA). On the day of testing, subjects reported to the laboratory after an overnight fast. Systolic blood pressure (SBP) and diastolic blood pressure (DBP) was determined once in the left arm using an automated system (Mark of Fitness, Shrewsbury, NJ) after at least 5-10 min of quiet sitting. Next, indwelling catheters were placed in a superficial vein of the forearm and fasting blood samples were collected in 3 ml syringes. All vacutainers were kept on ice before to blood collection. Blood samples for the analysis of glucose were transferred to vacutainers containing sodium fluoride to inhibit glycolysis. Blood samples for the analysis of total cholesterol (TC), HDL, and TAG were collected in vacutainers containing the anticoagulant EDTA. Serum samples for the analysis of hs-CRP were collected in vacutainers containing a serum separator (SST) and allowed to clot for 15 min. Blood samples were centrifuged at room temperature for 10 min at 1550g. Samples were immediately aliquoted and transferred to a −80 °C freezer until analysis.

Biochemical analyses

Fasting plasma glucose (FPG), TC, and HDL concentrations were determined enzymatically (GL5 Analyzer; Analox Instruments, Lunenberg, MA). TAG concentrations were measured by colorimetric assay (Sigma Aldrich, St. Louis, MO). Serum hs-CRP concentrations were measured using high-sensitivity ELISA kit (Diagnostic Systems Laboratory, Webster, TX). Calculations: mean arterial pressure (MAP) was calculated as: MAP = (2/3 DBP) + (1/3 SBP). The cardiac risk ratio was calculated as TC divided by HDL (19). Low-density lipoprotein was calculated using the Friedwald equation (Low-density lipoprotein = TC−HDL−(TAG × 0.2; ref. 20). Metabolic syndrome Z-scores and ATP III criteria were calculated to describe the efficacy of each treatment on metabolic syndrome severity and total number of risk factors (18). Sex specific Z-scores were calculated as: male Z-score = ((40−HDL)/10.7) + ((TAG− 150/88.5)) + ((FPG−100)/11.9) + ((WC−102)/14.3) + ((MAP− 100)/8.8), and female Z-score = ((50−HDL)/11.0) + ((TAG−150/88.5)) + ((FPG−100)/11.9) + ((WC−88)/13.6) + ((MAP−100)/8.8).

Statistical analysis

Data were analyzed using the R statistical software package (version 2.4.0; The R foundation, Vienna, Austria, 2006). Baseline subject characteristics were measured across conditions by one-way analysis of variance (ANOVA). There was no statistical difference in any baseline value except DBP. Using baseline DBP as a covariate did not affect the response to the treatments. All other condition means were analyzed before and after the intervention using 2-way (condition by test) repeated measures ANOVA. Because baseline hs-CRP concentrations were not normally distributed, the data were log transformed for statistical analysis. When there was a significant condition x test interaction, Tukey post hoc analysis was used to detect differences across conditions. Paired t tests were used to assess pre- to post differences within a condition, including caloric intake, metabolic syndrome Z-score, and ATP III score. McNemar test was used to assess metabolic syndrome prevalence (i.e., yes/no) after each treatment (22). Pearson's correlation was used to examine relationships between weight loss, cardiorespiratory fitness, CVD risk factors, and insulin sensitivity (19). Significant differences were accepted as α ≤ 0.05 and trends were reported as 0.05 < α < 0.10.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The raw data for fitness, body composition, fasting glucose and insulin have been published previously (19). However, given the obvious relevance of those data to risk for CVD they are summarized here for clarity.

Cardiorespiratory fitness and strength

EM and EP increased VO2 peak by 10-20% compared with baseline and strength by ∼15% compared with baseline (17).

Weight change, body composition, and food intake

Metformin (M) and EM, but not EP, decreased body weight more than P (P < 0.05; Figure 1a). EM and EP also lowered percent body fat (P = −0.8 ± 0.6%; M = 0.1 ± 0.5%; EP = −2.2 ± 0.8%; and EM = −1.5 ± 0.7%) and central body fat (P = −0.6 ± 0.1%; M = −0.1 ± 0.6%; EP = −1.4 ± 0.5%; EM = −1.9 ± 0.6%) by similar amounts. WC was 2 to 3% lower after all three treatments compared with baseline (time effect: P < 0.01; Figure 1b), but was not different across groups. Caloric intake was decreased by ∼15% from baseline for M (2,107.3 ± 103.2 vs. 1,819.8 kcal; P < 0.05) and EM (2,251.0 ± 135.7 vs. 1,777.3 ± 130.2 kcal; P < 0.05). Both EP (2,292.0 ± 177.8 vs. 2,149.2 ± 138.1 kcal; P = 0.57) and P (1,923.22 ± 135.6 vs. 1,861.3 ±144.5; P = 0.69) had no effect on intake.

thumbnail image

Figure 1. Effect of treatments on body mass. (a) Change in body weight over the 12-week protocol. ˆSignificant effect of M and EM compared with P (P < 0.05). Significant effect of M(*) and EM(‡) compared with baseline (P < 0.05). Data are mean ± s.d. (b) Change in waist circumference over the 12-week intervention. *Significant effect of time (P < 0.05). Data are mean ± s.d. EM, exercise training plus metformin; EP, exercise training plus placebo; M, metformin; P, placebo.

Download figure to PowerPoint

Fasting plasma glucose and insulin

Exercise training and/or metformin had no effect on fasting glucose concentrations (P = −0.1 ± 0.2 mmol/l; M = −0.2 ± 0.3 mmol/l; EP = 0.0 ± 0.3 mmol/l; EM = −0.4 ± 0.2 mmol/l). All 3 treatments lowered fasting insulin concentrations (P = 1.5 ± 2.3 uU/ml; M = −5.2 ± 1.9 uU/ml; EP = −1.7 ± 0.7 uU/ml; EM = −2.9 ± 1.5 uU/ml), but there were no group differences.

Blood pressure

Compared with baseline, SBP, and DBP increased with placebo (P; P < 0.05). M and EP lowered SBP (P < 0.01), but all three treatments lowered DBP compared with P (P < 0.05; Table 2). Both M and EP decreased MAP compared with P (P < 0.05; Table 2), but there was no change after EM.

Table 2. Effect of exercise and/or metformin on blood pressure
inline image

High sensitivity C-reactive protein

EP and M, but not EM, lowered CRP by ∼20% compared with P (EP: P < 0.05 and M: trend P = 0.06; Table 3).

Table 3. Effect of exercise and/or metformin on blood lipids and inflammation
inline image

TAG and cholesterol

Compared with baseline, exercise and/or metformin lowered TAG by ∼13% (time effect: P < 0.05; Table 3), but there were no group differences. All three treatments raised HDL cholesterol concentrations by 8-13% compared with P (M or EP; P < 0.05 and EM; trend P = 0.06; Table 3). Each treatment also decreased the cardiac risk ratio (P < 0.05; Table 3).

Metabolic syndrome severity and prevalence

Exercise training and/or metformin reduced the metabolic syndrome Z-score (i.e., severity) compared with baseline (EP; trend P = 0.07 and EM or M; P < 0.05; Figure 2a). Although each treatment reversed metabolic syndrome prevalence (M = 6 to 1, EP = 7 to 4, and EM 6 to 3; P < 0.05; Figure 2b), only EP and M reduced ATP III criteria (P < 0.05) compared with baseline (Figure 2c).

thumbnail image

Figure 2. Effect of treatments on metabolic syndrome criteria. (a) Metabolic syndrome severity before and after the 12-week intervention. *Pre to post was statistically different by paired t test; P < 0.05. $P = 0.07. Data are mean ± s.e.m. (b) Metabolic syndrome prevalence before and after the 12-week intervention. Prevalence refers to the fraction of individuals with metabolic syndrome pre and post. *Pre to post was statistically different by McNemar; P < 0.05. Data are frequency. (c) ATP III score before and after the 12-week intervention. *Pre to post was statistically different by paired t test; P < 0.05. Data are mean ± s.e.m. EM, exercise training plus metformin; EP, exercise training plus placebo; M, metformin.

Download figure to PowerPoint

Correlation analysis

Increased cardiorespiratory fitness was significantly correlated with lower fasting glucose concentrations (r = −0.38; P < 0.05), but not with any other CVD risk factor. Weight loss was correlated with lower fasting insulin concentrations (r = 0.40; P < 0.05), caloric intake (r = 0.46; P < 0.05), and WC (r = 0.60; P < 0.05). Reductions in TAG concentrations were correlated with increased insulin sensitivity (r = −0.60; P < 0.05). Lower hs-CRP concentrations were correlated with decreased SBP (r = 0.46; P < 0.05).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Metformin is not only used to treat type 2 diabetes, but it is also suggested to reduce CVD risk factors in adults with IGT (23). Previous work suggested that exercise with metformin reduced CVD risk factors more than the respective treatments alone (17,18), however, we observed no additive effect of combining exercise with metformin to reduce CVD risk factors IGT men and women. There are several possible differences between studies that may explain this. First, previous work did not systematically compare the combined treatment to either exercise (17) or metformin alone (18), thereby limiting the ability to know the effectiveness of the combined treatment on CVD risk factors. Second, recommendations to increased physical activity were not structured or supervised (18). Third, differences between HIV infected patients and IGT pathophysiology (17) complicate comparing previous work to ours. Last, dietary manipulations (i.e., low-fat diet) confound the ability to identify the independent effect of exercise with metformin on CVD risk factors (18). Consistent with our previous work (19), the current findings suggest that metformin may blunt some of the beneficial effects of training to lower SBP, hs-CRP, and metabolic syndrome criteria.

Weight loss (i.e., fat loss) is associated with lower CVD risk. Although exercise training with ad libitum food intake has mixed effects on weight loss (24,25), combining it with metformin has been shown to cause greater weight loss (26-,27,28). We found that combining metformin with exercise reduced body weight by ∼4 kg more than exercise alone, however, this was not different than metformin alone. Neither metformin, alone or with exercise, affected CVD risk factors (i.e., blood lipids, BP, and hs-CRP) more than exercise alone. Comparable reductions in central body fat between the exercise groups may explain the similar effects of exercise at reducing blood lipids (e.g., TAG and TC (19)). However, metformin decreased blood lipids equivalently to either exercise group without a change in central fat (measured by dual-energy x-ray absorptiometry and WC). Although we report reductions in WC with metformin in this study, more direct measures of central fat (e.g., magnetic resonance imaging or computed tomography scan) would be needed to confirm a true change in visceral fat. Despite no strong correlation between weight loss or WC and CVD risk factors, we cannot rule out the possibility that fat loss explains at least some of the metformin group outcomes. In either case, these results suggest that each treatment has distinctive mechanisms for opposing CVD development.

BP is often lowered after exercise training (11), but the effects of metformin are less consistent (13,29-,30,31). We found that metformin yielded little added benefit to the effects of exercise training on lowering BP and this outcome is consistent with other work (18). Our data suggest that the interaction of metformin with exercise training blunts reductions in SBP, and these results may be related to changes in cardiorespiratory fitness. Although there were no statistical differences in VO2 peak between the exercise groups, the attenuated rise in VO2 peak (EM = 10% vs. EP = 20%) is worth considering since it is consistent with our previous work in healthy adults (32). The smaller rise in VO2 peak is likely a result of either alterations in mitochondrial adaptations or vascular blood flow. Although we did not directly measure endothelial function, we found that exercise training or metformin alone, but not the combined treatment, reduced the vascular inflammatory marker hs-CRP by ∼20% compared with placebo. We observed a significant correlation between the change in hs-CRP and SBP, suggesting that reduced SBP may be related to decreased vascular inflammation. Interestingly, metformin alone reduced both SBP and hs-CRP, but had no effect on VO2 peak compared with placebo. This result strengthens the hypothesis that there is a unique interaction between metformin and exercise on the aforementioned outcomes that could have opposing effects on cardiovascular health. From this study, we cannot determine the mechanism by which the combined treatment blunted reductions in SBP or serum hs-CRP concentrations, but it may be related to the interaction between nitric oxide synthesis and adenosine monophosphate-activated protein kinase activation (33,34). Exercise combined with metformin was previously shown to blunt adenosine monophosphate-activated protein kinase activation in skeletal muscle (35). Since adenosine monophosphate-activated protein kinase activation is important for nitric oxide production, it is possible that the combined treatment blunted nitric oxide production and affected SBP and hs-CRP concentrations.

Although combining metformin with lifestyle modification lowered TAG concentrations more than lifestyle modification alone in overweight insulin resistant adolescents (26,28), there was no additive effect in IGT adults (18). Consistent with Andreadis et al.(18), we found that exercise training and/or metformin had similar effects at lowering TAG concentrations, despite exercise increasing VO2 peak and reducing central body fat. The decline in TAG concentrations across treatments was associated with increased insulin sensitivity (r = −0.60; P < 0.05). Although we cannot prove causality or directionality from this correlation, a possibility is that reductions in circulating TAG lead to improved peripheral glucose uptake (36). Fasting nonesterified fatty acids concentrations were elevated after exercise training with metformin, which could potentially explain the blunted rise in insulin sensitivity (19). The incongruent finding between lower TAG and higher nonesterified fatty acid concentrations with exercise plus metformin may relate to decreased nonesterified fatty acids uptake in skeletal muscle and liver (37,38) or increased lipolytic activity (39). We are not able to discern which of the mechanisms account for elevated nonesterified fatty acid concentrations, but our results support the beneficial effects of exercise to lower TAG and increase insulin sensitivity (40). Whether exercise and/or metformin treatments have a direct impact on reducing CVD incidence in individuals with IGT remains to be seen.

Approximately 50% of individuals with IGT have metabolic syndrome (3). Lifestyle modification or metformin decrease metabolic syndrome prevalence (3), and the combined treatment may have additive effects (18). In this study, we showed that structured exercise training and/or metformin had similar effects on reducing metabolic syndrome severity (i.e., Z-score; Figure 2a). The initial ATP III score was lower in the exercise with metformin group, compared with either treatment alone. The lower baseline may have made it more difficult to detect a significant improvement from baseline, and, only exercise training or metformin alone reduced the number of subjects meeting ATP III criteria (Figure 2c). In addition, although not statistically different, metformin decreased metabolic syndrome prevalence more than the combined treatment (Figure 2b). Since the combined treatment did not lower SBP, an ATP III criteria factor, our findings suggest that the interaction of metformin and exercise may oppose the reversal of metabolic syndrome in IGT adults compared with metformin alone. We recognize that further work is needed given the modest sample size in this study and our inability to compare treatment groups to placebo because only two subjects in that group had metabolic syndrome. However, the novel observation from this study is that combining exercise with metformin may have opposing effects on cardiovascular health, which could be clinically significant (16).

In conclusion, we observed that combining metformin with structured exercise training did not have additive effects on any CVD risk factor measured in this group of IGT adults. Metformin may blunt some of the effects of exercise training on lowering SBP and hs-CRP and consequently contribute to preventing a decrease in the number of metabolic syndrome risk factors. If our results are reproduced and generalizable to men and women with IGT then the clinical utility of recommending the combination of exercise and metformin for people with IGT may need to be re-evaluated. Despite metformin alone reducing CVD risk factors comparably to exercise alone, the improvements in cardiorespiratory fitness after exercise is one of the best predictors of CVD and early mortality. As a result, our findings highlight the overall positive effects of exercise on lowering CVD risk in adults with IGT.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

S.K.M., S.R.C., and B.B. contributed to the study design and data collection. S.K.M. was primarily responsible for data analysis and statistical integrity. J.N. contributed to data entry and data analysis. S.C. was solely responsible for food data analysis. S.K.M., S.R.C., and B.B. wrote the manuscript and J.N. and S.C. reviewed/edited the manuscript. The authors would like to thank Kirsten Granados, Richard Viskochil, and Robert Gerber for technical assistance and helpful discussion. B.B. receives consulting fees from Pfizer Global. We thank all the dedicated participants and research assistants for assistance on this project. This research was supported by NIH 5 R56 DK081038 grant.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References