Objective: Lifestyle intervention with diet modification and increase in physical activity is effective for reducing hepatic steatosis in patients with non-alcoholic fatty liver disease (NAFLD). However, for a similar weight loss, there is a large variability in the change in liver fat. We hypothesised that cardiorespiratory fitness may predict the response to the intervention.
Design: Longitudinal study with increase in physical activity and diet modification.
Setting: University teaching hospital.
Patients: 50 adults with NAFLD and 120 controls at risk for metabolic diseases.
Main outcome measures: Total-, subcutaneous abdominal- and visceral adipose tissue by magnetic resonance tomography, liver fat by 1HMR spectroscopy and cardiorespiratory fitness (VO2,max) by a maximal cycle exercise test at baseline and after 9 months of follow-up.
Results: In all subjects total-, subcutaneous abdominal- and visceral adipose tissue decreased and fitness increased (all p<0.0001) during the intervention. The most pronounced changes were found for liver fat (−31%, p<0.0001). Among the parameters predicting the change in liver fat, fitness at baseline emerged as the strongest factor, independently of total- and visceral adipose tissue as well as exercise intensity (p = 0.005). In the group of subjects with NAFLD at baseline, a resolution of NAFLD was found in 20 individuals. For 1 standard deviation increase in VO2,max at baseline the odds ratio for resolution of NAFLD was 2.79 (95% confidence interval, 1.43–6.33).
Conclusions: Cardiorespiratory fitness, independently of total adiposity, body fat distribution and exercise intensity, determines liver fat content in humans, suggesting that fitness and liver fat are causally related to each other. Moreover, measurement of fitness at baseline predicts the effectiveness of a lifestyle intervention in reducing hepatic steatosis in patients with NAFLD.
Statistics from Altmetric.com
Non-alcoholic fatty liver disease (NAFLD) is the most common chronic liver disease in westernised countries, with its prevalence continuing to increase worldwide in parallel with the epidemic of obesity.1 2 3 4 NAFLD is thought to be the predominant present-day cause of cryptogenic cirrhosis.5 Furthermore, it represents a risk factor for the development of cardiovascular events, insulin resistance and type 2 diabetes.1 2 6 7 8 9 10 11 12 13 14 Consequently, strategies to prevent and treat hepatic steatosis are the focus of intense research efforts. Various pharmacological agents have been tested. Hitherto only thiazolidinediones15 16 17 have been found to be relatively effective in reducing liver fat, although contraindications and side effects call for use of these drugs with caution.
Strategies aimed at weight reduction have been tested repeatedly and are now considered to constitute the most appropriate initial treatment for NAFLD. The beneficial effect of weight loss on NAFLD, in terms of a decrease in steatosis and/or serum aminotransferases, particularly in obese subjects, is undoubted.18 19 20 21 However, there is still a large variability in change in liver fat under weight loss and the precise mechanisms by which weight loss affects change in liver fat are not fully understood. Perhaps the most plausible mechanism includes a decrease in visceral fat leading to a reduced portal free fatty acid supply and favourable changes in secretion of adipocytokines, particularly adiponectin.22 However, changes in visceral or total fat did not correlate with changes in liver fat in all studies,23 24 suggesting that factors beyond weight loss may be operative. Furthermore, it is still a matter of discussion whether there is a positive effect of exercise, when included in an intervention, on liver fat.24 25 26 27 In contrast, for cardiorespiratory fitness, data are more consistent. The association of maximal aerobic capacity (VO2,max), an estimate of cardiorespiratory fitness, with liver fat was investigated in three cross-sectional studies. While in a relatively small study no significant difference in VO2,max between subjects with high versus low liver fat was found,28 two larger studies showed a close relationship of fitness, both with liver fat29 and the prevalence of NAFLD.30 These relationships were not independent of visceral adiposity, suggesting that fitness and hepatic steatosis are not causally related with each other. However, in a recent study involving a cohort of 2603 adults who were followed up for a mean of 12 years, fitness and body mass index predicted mortality, independently of several established risk factors. Most notably, the effect of fitness was independent of total and abdominal adiposity.31 As discussed, the effect of low fitness on mortality may have been mediated by high liver fat.32
To our knowledge, no study has determined the effect of fitness on liver fat in a longitudinal setting. In the present study, we determined the impact of cardiorespiratory fitness as well as total adiposity and body fat distribution, precisely quantified by magnetic resonance (MR) imaging, on the reduction in liver fat, measured by 1HMR spectroscopy, during an intervention with diet modification and increase in physical activity.
Materials and methods
Caucasians from the southern part of Germany participated in the ongoing TUebingen Lifestyle Intervention Program (TULIP). This study was designed to find parameters that predict the effect of a lifestyle intervention with diet and moderate increase in aerobic physical activity to improve prediabetes phenotypes and the cardiovascular risk profile. Individuals were included in the study when they fulfilled at least one of the following criteria: a family history of type 2 diabetes, a BMI>27 kg/m2, a previous diagnosis of impaired glucose tolerance and/or of gestational diabetes. They were considered healthy according to a physical examination and routine laboratory tests. The participants had no history of liver disease and did not consume more than two alcoholic drinks per day. Serum aminotransferase levels were lower than two times the upper limit of normal. A total of 170 subjects met the aforementioned requirements, had measurements of body fat distribution and liver fat content using magnetic resonance techniques at baseline and at follow-up and were included in the present study. Informed written consent was obtained from all participants.
After the baseline measurements, individuals underwent dietary counselling and had up to 10 sessions with a dietician. Counselling was aimed to reduce body weight, intake of calories and particularly intake of calories from fat and to increase intake of fibre. Diet composition was estimated with a validated computer program using two representative days of a 3 day diary (DGE-PC 3.0; Deutsche Gesellschaft für Ernährung, Bonn, Germany). Furthermore, all subjects completed a standardised self-administered and validated questionnaire to measure physical activity and a habitual physical activity score was calculated.33 Individuals were asked to perform at least 3 h of moderate sports per week. Aerobic endurance exercise (eg, walking, swimming) with an only moderate increase in the heart rate was encouraged. Participants were seen by the staff on a regular basis to ensure that these recommendations were accomplished.
Total body fat and body fat distribution
Measurements of total body adipose tissue, subcutaneous abdominal adipose tissue and visceral adipose tissue were performed by an axial T1-weighted fast spin echo technique with a 1.5 T whole-body imager (Magnetom Sonata; Siemens Medical Solutions, Erlangen, Germany) as previously described.34
Liver fat was measured by localised proton magnetic resonance (1HMR) spectroscopy as previously described.35 NAFLD was defined as liver fat content >5.56%.36 Liver fat measured by this method correlated well with histomorphometric findings.37 38 39 Also, in our hands liver fat content estimated by 1HMR spectroscopy correlated well with the histological features in a separate group of 15 subjects who underwent abdominal surgery (see supplementary figure).
Maximal aerobic capacity (VO2,max)
The individuals underwent a continuous, incremental exercise test to volitional exhaustion using a cycle ergometer. The cycle ergometer test was performed on an electromagnetically braked cycle ergometer (Ergometrics 800 S; Ergoline, Bitz, Germany). Oxygen consumption was measured using a spiroergometer (MedGraphics System Breese Ex 3.02 A; MedGraphics, St Paul, Minnesota, USA). VO2,max is expressed as VO2 (ml/min) per kg lean body mass (ml/min/kg).
Oral glucose tolerance test
All individuals underwent a 75 g oral glucose tolerance test and venous blood samples were obtained at 0, 30, 60, 90 and 120 min for determination of plasma glucose and insulin. Blood glucose was determined using a bedside glucose analyser (glucose oxidase method; YSI, Yellow Springs Instruments, Yellow Springs, Colorado, USA). For measurements of insulin, blood was placed on ice after drawing, immediately transferred to the lab and subsequently analysed. Plasma insulin was determined by microparticle enzyme immunoassay (Abbott Laboratories, Tokyo, Japan).
Unless otherwise stated, data are given as mean with the standard error (SE). Data that were not normally distributed (Shapiro–Wilk W test) were logarithmically transformed and a normal distribution of these parameters was achieved. Differences between baseline and follow-up were tested using the matched pairs t test. Univariate associations between parameters were tested using Pearson’s correlation analyses. To adjust the effects of relevant covariates (gender, age, total and visceral adipose tissue) multivariate linear regression analyses were used. Logistic regression was applied to determine the odds ratio for achieving a resolution of NAFLD. A p value ⩽0.05 was considered statistically significant. The statistical software package JMP (version 4.0; SAS Institute) was used.
Demographics, anthropometric and metabolic characteristics of the subjects
The characteristics of the 170 subjects (70 men and 100 women) at baseline are shown in table 1. All parameters covered a wide range, eg, age 19–68 years; body mass index (BMI) 19.36–41.47 kg/m2; waist circumference 63–123 cm; total adipose tissue 3.96–58.78 kg; subcutaneous abdominal adipose tissue 1.95–24.79 kg; visceral adipose tissue 0.38–7.94 kg and liver fat 0.16–28.22%. A total of 50 subjects were diagnosed with NAFLD. Also shown in table 1 are the characteristics of the subjects with and without NAFLD at baseline.
Univariate correlations between liver fat, total body fat, body fat distribution and VO2,max in the 170 subjects are presented in table 2. Of note there was a large variability in the strength of the correlations, possibly due to the variability in the parameters age and body composition of the subjects. Nevertheless, the correlations were all statistically significant. Liver fat was positively correlated with total-, subcutaneous abdominal- and visceral adipose tissue and negatively with VO2,max.
In multivariate linear regression analyses, liver fat, adjusted for gender, age and total adipose tissue, was significantly associated with VO2,max (r = −0.20, p = 0.011), and the model explained 27% of the variability in liver fat. Similar results were obtained when total adipose tissue was replaced by subcutaneous abdominal adipose tissue in the model; there was a significant association of liver fat with VO2,max (r = −0.17, p = 0.025), and the model explained 29% of the variability in liver fat. When visceral adipose tissue was included into the model instead of total- or subcutaneous abdominal adipose tissue, 38% of the variability in liver fat was explained, and the association of liver fat with VO2,max was rendered non significant (p = 0.17). Similarly, in the subjects with NAFLD, VO2,max did not correlate significantly with liver fat independently of age, gender and visceral fat (p = 0.35).
In the whole group, the mean duration of follow-up was 8.7 (SD 1.8) months. Data on energy intake and diet composition in subjects in whom complete information from questionnaires were available, revealed that energy intake and intake of saturated fat decreased and intake of fibre increased. Habitual physical activity also increased, both in subjects without and with NAFLD at baseline (table 1), indicating that subjects were compliant with the recommendations regarding exercise intensity. In the whole group VO2,max increased and alanine aminotransferase (ALT), aspartate transaminase (AST), fasting and 2 h glycsemia as well as insulinsemia decreased (table 1). Among fat compartments total- (−8.7%), subcutaneous abdominal- (−5.7%) and visceral adipose tissue (−14.4%) decreased. The largest decrease was seen in liver fat (−31%). Similar decreases in fat compartments were observed in the group of subjects having NAFLD at baseline (total adipose tissue −8.5%; subcutaneous abdominal adipose tissue −7.4%; visceral adipose tissue −12.6%; liver fat −35.3%). Table 3 presents the univariate correlations of change in liver fat with change in fat compartments and VO2,max during the intervention.
A resolution of NAFLD was found in 20 individuals. The characteristics of subjects who responded to the intervention with resolution of NAFLD (converters) and subjects who still had NAFLD at follow-up (non-converters) are shown in table 4. At baseline converters had less visceral- and liver fat and a trend for less total body fat and higher VO2,max compared to non-converters (table 4). The habitual physical activity (HPA) score was not different between the groups.
During the intervention the HPA score increased by 11% in the converters and by 9% in the non-converters. These changes were not significantly different from each other (p = 0.86). Regarding diet, changes in intake of calories, carbohydrates, fat, protein and alcohol were also not different among these two groups (all p>0.12).
Out of 120 subjects without NAFLD at baseline five developed NAFLD during follow-up. In these subjects the HPA score did not increase and body weight did not decrease, suggesting that they were not compliant with the recommendations of the study.
To study the predictive effect of parameters at baseline on change in liver fat during the intervention, multivariate linear regression models were performed with change in liver fat as the dependent variable and liver fat at baseline, gender, age and the parameter of interest as independent variables. As shown in table 5, total adipose tissue (model 1), as well as visceral adipose tissue (model 2) at baseline were associated with change in liver fat. VO2,max at baseline turned out to be an even stronger determinant of this change (model 3). When all three aforementioned parameters were included in the model, VO2,max at baseline remained a predictor of change in liver fat, independently of total- and visceral adipose tissue (model 4). The relationship of VO2,max with change in liver fat (p = 0.005) was not affected by additional inclusion of the HPA score at baseline into the model. Power analyses revealed that we had a power of 81% to detect the effect of VO2,max in this model and the least significant number of subjects to detect the effect at an α level of 0.05 was 82.
Furthermore, additional inclusion of total- and visceral adipose tissue at follow-up in model 4 had only a moderate effect on the relationship of VO2,max with change in liver fat (p = 0.01).
To better depict the impact of VO2,max at baseline on change in liver fat, quartiles of VO2,max at baseline were plotted against changes in total- and visceral adipose tissue and liver fat. VO2,max at baseline was not associated with changes in total- (fig 1A) or visceral adipose tissue (fig 1B). In contrast, the decrease in liver fat was larger with increasing VO2,max quartile (fig 1C). Of note, habitual physical activity at baseline (p = 0.85) and increase in habitual physical activity (p = 0.80) were not different among the quartiles of VO2,max, suggesting that subjects in all four groups were equally active.
We then tested the predictive effects of these parameters on change in liver fat in the group of 50 subjects with NAFLD. In univariate analyses VO2,max at baseline (p = 0.02), but not age, gender, or the baseline measurements of liver fat, total adipose tissue or visceral adipose tissue (all p>0.11), were associated with change in liver fat. In the forward stepwise regression analysis including the aforementioned parameters, VO2,max at baseline (estimate, −0.81; p = 0.008) emerged as the strongest predictor of change in liver fat, followed by liver fat at baseline (p = 0.052). Power analyses revealed that we had a large power of 78% to detect the effect of VO2,max on change in liver fat (least significant number of subjects to detect the effect at an α level of 0.05: n = 27), in this smaller group. A similar forward stepwise regression analysis in subjects without NAFLD (n = 120) showed that higher VO2,max at baseline predicted larger decrease in liver fat also in this group (estimate −0.61, p = 0.026).
Furthermore, the odds ratios for resolution of NAFLD were calculated. For 1 standard deviation (6.26 ml/min/kg) increase in VO2,max at baseline the odds ratio for resolution of NAFLD was 2.79 (95% confidence interval, 1.43 to 6.33). Subjects in the highest compared to the lowest quartile of VO2,max at baseline had an odds ratio of 8.0 (95% confidence interval, 1.5 to 54.4) to achieve this goal.
Lifestyle intervention programmes including diet and exercise proved to be effective in reducing liver fat. The beneficial outcome is probably largely mediated by weight loss and particularly, by decrease in total- and visceral fat mass. However, a reduction of total- or visceral adiposity is not always associated with a reduction of liver fat,23 24 suggesting that the effect of a lifestyle intervention on liver fat is not exclusively mediated by changes in total- and visceral adiposity.
Physical exercise and cardiorespiratory fitness are factors that may directly determine change in liver fat, beyond their effect on weight loss. In particular, cardiorespiratory fitness is closely associated with mitochondrial function, an important determinant of lipid oxidation.40 Therefore, in the present study, we investigated the effect of cardiorespiratory fitness on liver fat in the setting of a lifestyle intervention. Furthermore, we determined the dependency of the effect on total- and visceral fat. At baseline, VO2,max was negatively associated with liver fat, both in univariate and multivariate analyses. This effect was independent of total fat, but inclusion of visceral fat into the model rendered the association non-significant. These results are in line with previous studies reporting negative cross-sectional relationships of liver fat with cardiorespiratory fitness28 30 or habitual physical activity,41 42 43 which were independent of BMI, but mostly not independent of visceral obesity.29 30 41 However, we found a strong predictive effect of high fitness at baseline on reduction in liver fat during the intervention. This effect was not only greater than the impact of total- and visceral fat, but it was also independent of these parameters. This is the largest study applying the precise techniques of measurements of total adiposity as well as body fat distribution by MRT and liver fat by 1HMR spectroscopy under a longitudinal setting and an intervention. The analyses were first performed in all 170 subjects who underwent these measurements, allowing for a large variability in the parameters of interest which is necessary to investigate independent relationships under an appropriate statistical power. When we separated the groups we observed a predictive effect of fitness at baseline on change in liver fat in subjects without NAFLD and also in the smaller group of subjects with NAFLD, who are at high risk for fatty liver-induced complications. More importantly, high fitness at baseline predicted a resolution of NAFLD in these patients. This finding is important in two aspects; first, the observed large variability in the outcome of weight loss studies in patients with NAFLD may have been brought about by the existence of large differences in fitness, thus explaining why some individuals did not experience benefits from the intervention.18 19 20 21 Second, future lifestyle interventions in patients with NAFLD may be effective particularly when a relatively high level of fitness can be documented prior to the intervention. Otherwise, pharmacological support may become necessary. Mechanisms explaining the finding of an independent relationship between fitness and hepatic steatosis possibly include factors regulating hepatic lipid oxidation.44 45 46 47 48 49 50 The primary sites for the oxidation of fatty acids are the mitochondria, which occupy about 18% of the liver cell volume.51 Thus, mitochondrial function, a strong determinant of fitness,40 may also directly affect lipid oxidation in the liver. Indeed, we found that genetic variability in peroxisome proliferator-activated receptor (PPAR)γ coactivator 1α (PGC1α) and PPARδ genes regulate mitochondrial function, the response of fitness to physical activity46 and liver fat content.52 Thus, the aforementioned findings, together with our present data suggest that mitochondrial function may directly regulate liver fat.
Cardiorespiratory fitness is an objective reproducible measure that reflects the functional consequences of genetics45 and recent physical activity habits.49 Furthermore, there is an interaction between the intensity of physical activity with genetics on cardiorespiratory fitness which is supported by findings on marked inter-individual differences in the response of fitness phenotypes to regular exercise. These differences are not randomly distributed but aggregate in families.45 Because in the present study physical exercise per se could influence liver fat, we further investigated whether exercise intensity was a major factor underlying the strong relationship between fitness and change in liver fat. The increase in habitual physical activity correlated with the reduction in liver fat; however, fitness was associated with change in liver fat independently of habitual physical activity. Furthermore, our data are in line with previous findings showing an effect of fitness but not exercise intensity on mortality31 or only a moderate effect of exercise intensity on various metabolic parameters.53
Diet composition represents another factor that was found to affect liver fat content.54 To investigate whether diet may have influenced the results we determined the relationships between macronutrient and alcohol intake with fitness in individuals in whom data from dietary protocols were available. Total energy intake, carbohydrate, fat and protein intake as well as alcohol intake were not different among the quartiles of fitness (data not shown). These data further support our main finding, that fitness has direct effects on hepatic steatosis.
A limitation of the study is that we have no information about effects of the lifestyle intervention and the impact of fitness on liver histology such as steatohepatitis and liver fibrosis. As shown in children a lifestyle intervention with diet and increased physical activity is associated with a significant improvement in liver fat and liver histology55 suggesting that in our population reduction in liver fat may also have been accompanied by histological improvement.
In conclusion, with the present study we provide novel evidence that cardiorespiratory fitness at baseline is a strong and independent predictor of the decrease in liver fat during a lifestyle intervention. These findings support the hypothesis that there are specific effects of cardiorespiratory fitness on hepatic lipid metabolism. Moreover, our data suggest that measurement of fitness could be helpful for identifying patients with NAFLD who are more or less likely to respond to a lifestyle intervention and, thus, may help to tailor the intervention towards an individualised prevention and treatment.
Funding The study was supported by grants from the Deutsche Forschungsgemeinschaft (KFO 114 and a Heisenberg-Grant to NS, STE 1096/1-1) and the European Community’s FP6 EUGENE2 (LSHM-CT-2004-512013). The supporters had no influence on the study design and on the collection, analysis, and interpretation of data.
Competing interests None.
Ethics approval The local medical ethics committee approved the protocol on 20 December 2002.
▸ A supplementary figure is published online only at http://gut.bmj.com/content/vol58/issue9
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.