Accepted Manuscript Title: Volume-dependent effect of supervised exercise training on fatty liver and visceral adiposity index in subjects with type 2 diabetes The Italian Diabetes Exercise Study (IDES) Author: Stefano Balducci Patrizia Cardelli Luca Pugliese Valeria D’Errico Jonida Haxhi Elena Alessi Carla Iacobini Stefano Menini Lucilla Bollanti Francesco G. Conti Antonio Nicolucci Giuseppe Pugliese PII: DOI: Reference:
S0168-8227(15)00263-6 http://dx.doi.org/doi:10.1016/j.diabres.2015.05.033 DIAB 6409
To appear in:
Diabetes Research and Clinical Practice
Received date: Revised date: Accepted date:
3-12-2014 26-2-2015 2-5-2015
Please cite this article as: S. Balducci, P. Cardelli, L. Pugliese, V. D’Errico, J. Haxhi, E. Alessi, C. Iacobini, S. Menini, L. Bollanti, F.G. Conti, A. Nicolucci, G. Pugliese, for the Italian Diabetes Exercise Study (IDES) Investigators, ¨ 1 Volume-dependent effect of supervised exercise training on fatty liver and visceral adiposity index in subjects with type 2 diabetes The Italian Diabetes Exercise Study (IDES), Diabetes Research and Clinical Practice (2015), http://dx.doi.org/10.1016/j.diabres.2015.05.033 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Highlights
In subjects with type 2 diabetes, supervised exercise is effective in reducing fatty liver index (FLI) and visceral adiposity index (VAI), two validated markers of the presence and severity of NAFLD,
The effect of exercise appears to be mediated by improvements in central obesity and atherogenic
cr
dyslipidemia, but also by exercise-specific mechanisms at the muscular level.
The extent of improvements in FLI and VAI is dependent on volume of physical activity exercise and changes in fitness parameters.
te
d
M
an
The impact of intensity on NAFLD markers is elusive.
Ac ce p
us
ip t
respectively, whereas it has no significant effect on liver enzymes.
1 Page 1 of 26
Volume-dependent effect of supervised exercise training on fatty liver and visceral adiposity index in subjects with type 2 diabetes The Italian Diabetes Exercise Study (IDES)
ip t
Running head: Supervised exercise training and surrogate NAFLD markers.
cr
Stefano Balducci, MD 1,2,3, Patrizia Cardelli, PhD 1,4, Luca Pugliese, MD 1,5, Valeria D’Errico, MD 1,2,3, Jonida
us
Haxhi, MD 3,6, Elena Alessi, MD 1,3, Carla Iacobini, PhD 1, Stefano Menini, PhD 1, Lucilla Bollanti, MD 1,2, Francesco G. Conti, MD, PhD 1,2, Antonio Nicolucci, MD, PhD 7, and Giuseppe Pugliese, MD, PhD 1,2; for the
an
Italian Diabetes Exercise Study (IDES) Investigators †
Department of Clinical and Molecular Medicine, “La Sapienza” University, Rome, Italy; 2 Diabetes Unit and
4
Laboratory of Clinical Chemistry, Sant’Andrea Hospital, Rome Italy; 3 Metabolic Fitness Association,
M
1
d
Monterotondo, Rome, Italy; 5 Radiology Unit, Sant’Andrea Hospital, Rome Italy; 6 Department of Human
te
Movement and Sport Sciences, ‘‘Foro Italico’’ University, Rome, Italy; and 7 Department of Clinical Pharmacology and Epidemiology, Consorzio Mario Negri Sud, S. Maria Imbaro, Chieti, Italy. A complete list of the IDES Investigators can be found as on-line Supplemental Material.
Ac ce p
†
Corresponding Author:
Giuseppe PUGLIESE, M.D., Ph.D.
Department of Clinical and Molecular Medicine, “La Sapienza” University of Rome Via di Grottarossa, 1035-1039 - 00189 Rome, Italy Phone: +39-0633775440; Fax: +39-0633776327; E-mail: [email protected]
Word count: abstract: 195; text (including tables and figure legends): 3,921. Number of tables and figures: Tables: 2 (+ 2 Supplemental Table); Figures: 3 (+ 2 Supplemental Figure). Trial registration number: ISRCTN-04252749, www.ISRCTN.org. 2 Page 2 of 26
Abstract Aims: This study evaluated the effect of supervised exercise training on liver enzymes and two surrogate measures of non-alcoholic fatty liver disease (NAFLD) in subjects with type 2 diabetes.
ip t
Methods: Sedentary patients from 22 outpatient diabetes clinics were randomized by center, age and
treatment to twice-a-week supervised aerobic and resistance training plus structured exercise counseling
cr
(exercise group, EXE; n=303) versus counseling alone (control group, CON; n=303) for 12 months. EXE
us
participants were further randomized to low-to-moderate (n=142) or moderate-to-high (n=161) intensity training of equal energy cost. Baseline and end-of-study levels of liver enzymes, fatty liver index (FLI) and
an
visceral adiposity index (VAI) were obtained.
Results: Enzyme levels did not change, whereas FLI and VAI decreased significantly in EXE, but not CON
M
participants. Physical activity (PA) volume was an independent predictor of both FLI and VAI reductions, the extent of which increased from the 1st to the 4th quintile of PA volume and baseline to end-of-study changes
d
in fitness parameters. Differences in the effect of LI versus HI training were negligible.
te
Conclusions: Data from this large cohort of subjects with type 2 diabetes indicate that FLI and VAI decrease
Ac ce p
with supervised training in a volume-dependent manner. Key words: supervised exercise; exercise volume; FLI; VAI; liver enzymes.
3 Page 3 of 26
1. Introduction Nonalcoholic fatty liver disease (NAFLD) is the most common chronic liver disease worldwide [1]. Its prevalence is rapidly increasing along with the epidemics of obesity and type 2 diabetes, which are risk
ip t
factors for NAFLD [2,3]. It encompasses various disease conditions, from simple steatosis to nonalcoholic steatohepatitis (NASH), cirrhosis, and possibly hepatocellular carcinoma [4]. NAFLD is recognized as an
cr
independent predictor of insulin resistance [5], the metabolic syndrome [6], and cardiovascular disease [3]. Diagnosis of NAFLD requires noninvasive imaging, which however do not reliably distinguish nonalcoholic
us
steatohepatitis (NASH) from simple steatosis, and liver biopsy is therefore needed [7]. However, since imaging techniques are expensive and biopsy is an invasive procedure with significant sampling error [8],
an
several biochemical markers have been proposed for identifying the presence and severity of NAFLD. Among these, fatty liver index (FLI) has been shown to predict NAFLD [9,10], whereas visceral adiposity
with NAFLD [11] and chronic hepatitis C [12].
M
index (VAI), a marker of visceral fat distribution and function, has been correlated with histology in subjects
d
Available data indicate that body weight reduction with a combination of diet and exercise is the most
te
effective and safe intervention, though more than 50% of patients fail to achieve target weight loss [13].
Ac ce p
However, it is difficult to discriminate between the effects of diet and exercise, despite the finding of an inverse relation between NAFLD and physical activity (PA) [14] or fitness levels [15]. A systematic review and meta-analysis of randomized controlled trials has shown a benefit of exercise on liver fat, independent of weight loss, but not on serum enzyme levels. Unfortunately, the small size and short duration of the studies do not allow drawing conclusions regarding optimal exercise type and dose [16]. Another issue was exercise intensity, which varied among the trials considered in this meta-analysis. A retrospective study reported an association of PA intensity, but not volume and duration, with NAFLD histology, though PA was assessed using a self-reported questionnaire [17]. These data indicate the need for RCTs of adequate size and duration to confirm the benefit from exercise independent of diet and to identify the most effective exercise modality in patients with NAFLD.
4 Page 4 of 26
This study aimed at verifying, in the large cohort of the Italian Diabetes and Exercise Study (IDES), whether supervised aerobic and resistance training on-top of exercise counseling is effective in reducing liver enzymes and surrogate measures of NAFLD in subjects with type 2 diabetes, as compared with counseling alone. In addition, this pre-specified analysis aimed at assessing the role of PA/exercise volume
Ac ce p
te
d
M
an
us
cr
ip t
and intensity in changes in liver outcomes.
5 Page 5 of 26
2. Materials and Methods The study protocol conforms to the ethical guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval by the institutions’ human research committees and participants gave written informed
ip t
consent. The research protocol and CONSORT checklist are available as on-line Supplemental Material. 2.1. Setting and participants
cr
The IDES involved 22 outpatient diabetes clinics throughout Italy between October 1st, 2005 and March
us
31st, 2006. Each diabetes clinic was connected with a Metabolic Fitness Center, a dedicated facility where patients trained under the supervision of an exercise professional. [18,19]. Sedentary Caucasian patients
an
with type 2 diabetes and the metabolic syndrome were eligible for this study, whereas subjects having any condition limiting or contraindicating PA were excluded. Exclusion criteria included also history of hepatitis
M
B or C virus infection and self-reported alcohol consumption above 30 and 20 g/day in males and females, respectively [20]. Of the 691 eligible patients, 85 were excluded for various reasons [19] and 606 were
d
recruited from October 1st, 2005 to March 31st, 2006 and randomized to supervised training plus structured
te
exercise counseling (exercise, EXE, group; n=303) versus counseling alone as part of standard care (control, CON, group; n=303) for 12 months. Randomization was stratified by center and, within each center, by age
Ac ce p
and type of diabetes treatment, using a permuted-block randomization software. Participants in the EXE group were further allocated to either low-to-moderate intensity (LI; n=142) versus moderate-to-high intensity (HI; n=161) training of equal energy cost (i.e. exercise volume), within the intensity range indicated by the guidelines of the time [21]. Allocation was made according to the center, which was preliminary assigned to train patients at either LI or HI [22]. The study flow-chart is reported in Figure 1. Physicians and patients were not blinded to randomization whereas sample blinding at central laboratory was achieved using bar codes. 2.2. Interventions The supervised training program for the exercise group consisted of twice a week supervised sessions of mixed (aerobic and resistance) exercise, based on recent evidence [23] and guidelines [21] recommending both types of exercise. 6 Page 6 of 26
Aerobic training was performed using treadmill, step, elliptical, arm or cycle-ergometer. Resistance training consisted of 4 resistance exercises, i.e. thrust movement on the transverse plane (chest press or equivalent), traction movement on the frontal plane (lateral pull down or equivalent), squat movement (leg press or equivalent), and trunk flexion for the abdominals, plus three stretching positions [18,19].
ip t
Individuals in the LI subgroup performed aerobic training at 55% of predicted maximal oxygen
consumption (VO2max) and resistance training at 60% of predicted 1-Repetition Maximum (RM), i.e. the
cr
maximum amount of weight one can lift in a single repetition for a given exercise. Subjects in the HI
us
subgroup performed aerobic training at 70% of predicted VO2max and resistance training at 60% of predicted 1-RM (see below for methods of estimation of VO2max and 1-RM). Duration of aerobic training and number
an
of series of resistance training were varied to obtain the same caloric expenditure per kg body weight in the LI and HI subgroups, independent of intensity [22]. Intensity was adjusted according to improvements in
M
predicted VO2max and 1-RM, as assessed monthly throughout the study. In addition, caloric expenditure was increased progressively by 0.1 kcal/kg body weight/session every month from 3.0 to 4.1 kcal/kg body
d
weight/session [18,19].
te
Subjects from both groups received a structured individualized counseling, aimed at achieving the currently recommended amount of PA by encouraging any type of commuting, occupational, home and
Ac ce p
leisure-time PA. Both EXE and CON participants received also standard medical care consisting of a treatment regimen aimed at achieving optimal glycemic, lipid, blood pressure (BP) and body weight targets and including glucose-, lipid- and BP-lowering agents as needed [18,19]. 2.3. Outcomes
Here, we present the results of a pre-specified analysis of (a) the effect of intervention on liver enzymes and two surrogate markers of NAFLD, FLI and VAI; and (b) the relationship of changes in these measures with PA/exercise volume and intensity, physical fitness and modifiable cardiovascular risk factors. The main analyses on the primary and secondary endpoints, i.e. changes in hemoglobin (Hb) A1c and other modifiable cardiovascular risk factors, physical fitness, and quality of life, have been reported elsewhere [19,22,24,25]. 2.4. Measurements
7 Page 7 of 26
2.4.1. Cardiovascular risk factors. The following parameters were evaluated at baseline and end-of-study: HbA1c, fasting blood glucose and serum insulin, Homeostasis Model Assessment-Insulin Resistance (HOMAIR) index, body mass index (BMI), waist circumference, BP, triglycerides, total, HDL and LDL cholesterol, and high sensitivity-C-reactive protein (hs-CRP). Biochemical tests were performed at central laboratory by the
ip t
use of standard methods [18,19].
2.4.2. Liver enzymes and indexes. Baseline and end-of-study levels of aspartate aminotransferase (AST),
cr
alanine aminotransferase (ALT), and -glutamyl-transpeptidase (-GT) were measured at the central
us
laboratory by the use of standard methods (VITROS 5,1 FS Chemistry System, Ortho-Clinical Diagnostics Inc, Raritan, NJ). At end-of-study, samples were collected at least three days after completing the training
an
program, in order to exclude an acute effect of exercise on liver enzymes [26]. Values of FLI was calculated from -GT, BMI, waist circumference and triglycerides, whereas VAI was computed from BMI, waist
M
circumference, triglycerides and HDL cholesterol, as previously reported (Supplemental Table 1) [9,27]. The following cut-off levels for liver enzymes were used (corresponding to the upper limit of laboratory range
d
and normal range, respectively): 40 and 30 IU/L for AST, 45 and 40 IU/L for ALT, and 60 and 45 IU/L for -
te
GT. FLI and VAI values were considered abnormal if >60 and 1.9, respectively [9,27].
Ac ce p
2.4.3. Volume of PA/exercise. The amount of non-supervised PA was prospectively evaluated throughout the study by asking patients to fill in a preliminary validated daily diary, which considered the list of PAs coded in the Minnesota leisure-time PA questionnaire [28]. Volume was calculated by multiplying the metabolic equivalent (MET) scores of each Minnesota code by time spent in each activity and expressed as METs-hr·wk-1. Energy expenditure from aerobic exercise during supervised sessions was calculated automatically by the machines from workload (i.e. the combination of speed and slope for treadmill, steps per minute for step and power for ergometer), whereas, for resistance exercise, an estimate of 3 METs-hr was established [18,19]. 2.4.4. Physical fitness. Parameters of physical fitness were evaluated at baseline, end-of-study, and, in the EXE group, also during the study period, in order to adjust training loads [18,19,24]. Assessment of cardiorespiratory fitness consisted of a sub-maximal VO2max test, i.e. at 80% of the predicted maximal heart rate,
8 Page 8 of 26
performed at the treadmill with direct measurement of oxygen consumption using a gas exchange analyzer (FitMate, Cosmed, Rome, Italy). For strength assessment, a maximal repetition (or 5-8 RM) test was used, then predicted 1-RM was calculated using the Brzycki formula. For hip and trunk flexibility assessment, a standard bending test was performed.
ip t
2.5. Statistical analysis
cr
As previously reported, sample size was calculated a priori for comparing EXE and CON participants
[18,19] and post-hoc for assessing differences between HI and LI subjects [22]. Calculation considered a
us
reduction of at least 0.5%in HbA1c levels (primary endpoint), with a standard deviation for baseline values of ~1.5%, and a statistical power of ~90% (=0.05).
an
Patients completing the follow-up (CON=275; EXE=288 [LI=136; HI=152]) (Figure 1) were analyzed. Data were expressed as mean±SD or number and percentage of subjects with values above the normal range.
M
Baseline values of EXE and CON subjects were compared using the unpaired t-test or the Mann Whitney Utest. Within-group end-of-study versus baseline differences were analyzed using the Wilcoxon signed ranks
d
test. Mean differences (95% confidence intervals) between baseline to end-of-study changes in the two
te
groups or in the two EXE subgroups were also reported. The efficacy of EXE versus CON and HI versus LI
Ac ce p
programs was assessed using the unpaired t-test or the Mann Whitney U test, by comparing betweengroups changes from baseline to end-of-study. To identify independent predictors of changes from baseline in FLI and VAI, multiple regression analyses with stepwise variable selection were applied, with baseline FLI or VAI values, study arm, age, gender, statin treatment, and changes in HbA1c, HOMA-IR, HDL (for FLI only) and LDL cholesterol, hs-CRP, VO2max, upper and lower body strength, and volume of PA/exercise as covariates forced in the model. Results were expressed as beta unit coefficients. Values of liver enzymes and markers of NAFLD were also stratified according to quintiles of PA/exercise volume and differences between quintiles were assessed by the use of Kruskal-Wallis test. All analyses were performed on individuals completing the follow-up. Statistical analyses were performed using SPSS version 13.0 (SPSS Inc., Chicago, Illinois, USA).
9 Page 9 of 26
3. Results As previously reported [19], attendance to aerobic and/or resistance sessions was high (median 80.3%, interquartile range 75% to 99%) and drop-out was low (28 [9.24%] CON and 15 [4.95%} EXE participants).
ip t
3.1. Baseline to end-of-study changes in liver enzymes and indexes The percentage of subjects with elevated levels of liver enzymes and indexes decreased throughout the
cr
study for FLI and to a lesser extent VAI and -GT, but not AST and ALT (Supplemental Figure 1). Levels of AST, ALT, -GT, FLI and VAI decreased in 226 (40.1%), 174 (30.9), 263 (46.7%), 314 (55.8%), and 289 (51.3%),
us
respectively, and remained unchanged or increased in the remaining individuals.
Baseline values were similar in EXE and CON subjects except for AST and FLI levels, which were
an
significantly higher (P=0.009 and P=0.02) in EXE and CON groups, respectively. In both groups, no improvement in enzyme levels was observed throughout the study, since AST and ALT increased slightly but
M
significantly (except AST in CON participants), whereas -GT values decreased non-significantly in both groups. As a result, mean differences in baseline to end-of-study changes between the two groups were
d
negligible. Conversely, both FLI and VAI decreased significantly in the EXE but not CON group, with highly
te
significant mean differences between the two groups in favor of EXE participants (Table 1). As previously
Ac ce p
reported [19], of the parameters used for calculation of these surrogate measures of NAFLD, decreases in BMI (mean difference -0.78 kg/m2 [95% confidence interval, -1.07 to -0.49]) and waist circumference (-3.6 cm [-4.4 to -2.9]) and increases in HDL cholesterol (3.7 mg/dl [2.2 to 5.3]) were significantly more marked (P
S0168-8227(15)00263-6 http://dx.doi.org/doi:10.1016/j.diabres.2015.05.033 DIAB 6409
To appear in:
Diabetes Research and Clinical Practice
Received date: Revised date: Accepted date:
3-12-2014 26-2-2015 2-5-2015
Please cite this article as: S. Balducci, P. Cardelli, L. Pugliese, V. D’Errico, J. Haxhi, E. Alessi, C. Iacobini, S. Menini, L. Bollanti, F.G. Conti, A. Nicolucci, G. Pugliese, for the Italian Diabetes Exercise Study (IDES) Investigators, ¨ 1 Volume-dependent effect of supervised exercise training on fatty liver and visceral adiposity index in subjects with type 2 diabetes The Italian Diabetes Exercise Study (IDES), Diabetes Research and Clinical Practice (2015), http://dx.doi.org/10.1016/j.diabres.2015.05.033 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Highlights
In subjects with type 2 diabetes, supervised exercise is effective in reducing fatty liver index (FLI) and visceral adiposity index (VAI), two validated markers of the presence and severity of NAFLD,
The effect of exercise appears to be mediated by improvements in central obesity and atherogenic
cr
dyslipidemia, but also by exercise-specific mechanisms at the muscular level.
The extent of improvements in FLI and VAI is dependent on volume of physical activity exercise and changes in fitness parameters.
te
d
M
an
The impact of intensity on NAFLD markers is elusive.
Ac ce p
us
ip t
respectively, whereas it has no significant effect on liver enzymes.
1 Page 1 of 26
Volume-dependent effect of supervised exercise training on fatty liver and visceral adiposity index in subjects with type 2 diabetes The Italian Diabetes Exercise Study (IDES)
ip t
Running head: Supervised exercise training and surrogate NAFLD markers.
cr
Stefano Balducci, MD 1,2,3, Patrizia Cardelli, PhD 1,4, Luca Pugliese, MD 1,5, Valeria D’Errico, MD 1,2,3, Jonida
us
Haxhi, MD 3,6, Elena Alessi, MD 1,3, Carla Iacobini, PhD 1, Stefano Menini, PhD 1, Lucilla Bollanti, MD 1,2, Francesco G. Conti, MD, PhD 1,2, Antonio Nicolucci, MD, PhD 7, and Giuseppe Pugliese, MD, PhD 1,2; for the
an
Italian Diabetes Exercise Study (IDES) Investigators †
Department of Clinical and Molecular Medicine, “La Sapienza” University, Rome, Italy; 2 Diabetes Unit and
4
Laboratory of Clinical Chemistry, Sant’Andrea Hospital, Rome Italy; 3 Metabolic Fitness Association,
M
1
d
Monterotondo, Rome, Italy; 5 Radiology Unit, Sant’Andrea Hospital, Rome Italy; 6 Department of Human
te
Movement and Sport Sciences, ‘‘Foro Italico’’ University, Rome, Italy; and 7 Department of Clinical Pharmacology and Epidemiology, Consorzio Mario Negri Sud, S. Maria Imbaro, Chieti, Italy. A complete list of the IDES Investigators can be found as on-line Supplemental Material.
Ac ce p
†
Corresponding Author:
Giuseppe PUGLIESE, M.D., Ph.D.
Department of Clinical and Molecular Medicine, “La Sapienza” University of Rome Via di Grottarossa, 1035-1039 - 00189 Rome, Italy Phone: +39-0633775440; Fax: +39-0633776327; E-mail: [email protected]
Word count: abstract: 195; text (including tables and figure legends): 3,921. Number of tables and figures: Tables: 2 (+ 2 Supplemental Table); Figures: 3 (+ 2 Supplemental Figure). Trial registration number: ISRCTN-04252749, www.ISRCTN.org. 2 Page 2 of 26
Abstract Aims: This study evaluated the effect of supervised exercise training on liver enzymes and two surrogate measures of non-alcoholic fatty liver disease (NAFLD) in subjects with type 2 diabetes.
ip t
Methods: Sedentary patients from 22 outpatient diabetes clinics were randomized by center, age and
treatment to twice-a-week supervised aerobic and resistance training plus structured exercise counseling
cr
(exercise group, EXE; n=303) versus counseling alone (control group, CON; n=303) for 12 months. EXE
us
participants were further randomized to low-to-moderate (n=142) or moderate-to-high (n=161) intensity training of equal energy cost. Baseline and end-of-study levels of liver enzymes, fatty liver index (FLI) and
an
visceral adiposity index (VAI) were obtained.
Results: Enzyme levels did not change, whereas FLI and VAI decreased significantly in EXE, but not CON
M
participants. Physical activity (PA) volume was an independent predictor of both FLI and VAI reductions, the extent of which increased from the 1st to the 4th quintile of PA volume and baseline to end-of-study changes
d
in fitness parameters. Differences in the effect of LI versus HI training were negligible.
te
Conclusions: Data from this large cohort of subjects with type 2 diabetes indicate that FLI and VAI decrease
Ac ce p
with supervised training in a volume-dependent manner. Key words: supervised exercise; exercise volume; FLI; VAI; liver enzymes.
3 Page 3 of 26
1. Introduction Nonalcoholic fatty liver disease (NAFLD) is the most common chronic liver disease worldwide [1]. Its prevalence is rapidly increasing along with the epidemics of obesity and type 2 diabetes, which are risk
ip t
factors for NAFLD [2,3]. It encompasses various disease conditions, from simple steatosis to nonalcoholic steatohepatitis (NASH), cirrhosis, and possibly hepatocellular carcinoma [4]. NAFLD is recognized as an
cr
independent predictor of insulin resistance [5], the metabolic syndrome [6], and cardiovascular disease [3]. Diagnosis of NAFLD requires noninvasive imaging, which however do not reliably distinguish nonalcoholic
us
steatohepatitis (NASH) from simple steatosis, and liver biopsy is therefore needed [7]. However, since imaging techniques are expensive and biopsy is an invasive procedure with significant sampling error [8],
an
several biochemical markers have been proposed for identifying the presence and severity of NAFLD. Among these, fatty liver index (FLI) has been shown to predict NAFLD [9,10], whereas visceral adiposity
with NAFLD [11] and chronic hepatitis C [12].
M
index (VAI), a marker of visceral fat distribution and function, has been correlated with histology in subjects
d
Available data indicate that body weight reduction with a combination of diet and exercise is the most
te
effective and safe intervention, though more than 50% of patients fail to achieve target weight loss [13].
Ac ce p
However, it is difficult to discriminate between the effects of diet and exercise, despite the finding of an inverse relation between NAFLD and physical activity (PA) [14] or fitness levels [15]. A systematic review and meta-analysis of randomized controlled trials has shown a benefit of exercise on liver fat, independent of weight loss, but not on serum enzyme levels. Unfortunately, the small size and short duration of the studies do not allow drawing conclusions regarding optimal exercise type and dose [16]. Another issue was exercise intensity, which varied among the trials considered in this meta-analysis. A retrospective study reported an association of PA intensity, but not volume and duration, with NAFLD histology, though PA was assessed using a self-reported questionnaire [17]. These data indicate the need for RCTs of adequate size and duration to confirm the benefit from exercise independent of diet and to identify the most effective exercise modality in patients with NAFLD.
4 Page 4 of 26
This study aimed at verifying, in the large cohort of the Italian Diabetes and Exercise Study (IDES), whether supervised aerobic and resistance training on-top of exercise counseling is effective in reducing liver enzymes and surrogate measures of NAFLD in subjects with type 2 diabetes, as compared with counseling alone. In addition, this pre-specified analysis aimed at assessing the role of PA/exercise volume
Ac ce p
te
d
M
an
us
cr
ip t
and intensity in changes in liver outcomes.
5 Page 5 of 26
2. Materials and Methods The study protocol conforms to the ethical guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval by the institutions’ human research committees and participants gave written informed
ip t
consent. The research protocol and CONSORT checklist are available as on-line Supplemental Material. 2.1. Setting and participants
cr
The IDES involved 22 outpatient diabetes clinics throughout Italy between October 1st, 2005 and March
us
31st, 2006. Each diabetes clinic was connected with a Metabolic Fitness Center, a dedicated facility where patients trained under the supervision of an exercise professional. [18,19]. Sedentary Caucasian patients
an
with type 2 diabetes and the metabolic syndrome were eligible for this study, whereas subjects having any condition limiting or contraindicating PA were excluded. Exclusion criteria included also history of hepatitis
M
B or C virus infection and self-reported alcohol consumption above 30 and 20 g/day in males and females, respectively [20]. Of the 691 eligible patients, 85 were excluded for various reasons [19] and 606 were
d
recruited from October 1st, 2005 to March 31st, 2006 and randomized to supervised training plus structured
te
exercise counseling (exercise, EXE, group; n=303) versus counseling alone as part of standard care (control, CON, group; n=303) for 12 months. Randomization was stratified by center and, within each center, by age
Ac ce p
and type of diabetes treatment, using a permuted-block randomization software. Participants in the EXE group were further allocated to either low-to-moderate intensity (LI; n=142) versus moderate-to-high intensity (HI; n=161) training of equal energy cost (i.e. exercise volume), within the intensity range indicated by the guidelines of the time [21]. Allocation was made according to the center, which was preliminary assigned to train patients at either LI or HI [22]. The study flow-chart is reported in Figure 1. Physicians and patients were not blinded to randomization whereas sample blinding at central laboratory was achieved using bar codes. 2.2. Interventions The supervised training program for the exercise group consisted of twice a week supervised sessions of mixed (aerobic and resistance) exercise, based on recent evidence [23] and guidelines [21] recommending both types of exercise. 6 Page 6 of 26
Aerobic training was performed using treadmill, step, elliptical, arm or cycle-ergometer. Resistance training consisted of 4 resistance exercises, i.e. thrust movement on the transverse plane (chest press or equivalent), traction movement on the frontal plane (lateral pull down or equivalent), squat movement (leg press or equivalent), and trunk flexion for the abdominals, plus three stretching positions [18,19].
ip t
Individuals in the LI subgroup performed aerobic training at 55% of predicted maximal oxygen
consumption (VO2max) and resistance training at 60% of predicted 1-Repetition Maximum (RM), i.e. the
cr
maximum amount of weight one can lift in a single repetition for a given exercise. Subjects in the HI
us
subgroup performed aerobic training at 70% of predicted VO2max and resistance training at 60% of predicted 1-RM (see below for methods of estimation of VO2max and 1-RM). Duration of aerobic training and number
an
of series of resistance training were varied to obtain the same caloric expenditure per kg body weight in the LI and HI subgroups, independent of intensity [22]. Intensity was adjusted according to improvements in
M
predicted VO2max and 1-RM, as assessed monthly throughout the study. In addition, caloric expenditure was increased progressively by 0.1 kcal/kg body weight/session every month from 3.0 to 4.1 kcal/kg body
d
weight/session [18,19].
te
Subjects from both groups received a structured individualized counseling, aimed at achieving the currently recommended amount of PA by encouraging any type of commuting, occupational, home and
Ac ce p
leisure-time PA. Both EXE and CON participants received also standard medical care consisting of a treatment regimen aimed at achieving optimal glycemic, lipid, blood pressure (BP) and body weight targets and including glucose-, lipid- and BP-lowering agents as needed [18,19]. 2.3. Outcomes
Here, we present the results of a pre-specified analysis of (a) the effect of intervention on liver enzymes and two surrogate markers of NAFLD, FLI and VAI; and (b) the relationship of changes in these measures with PA/exercise volume and intensity, physical fitness and modifiable cardiovascular risk factors. The main analyses on the primary and secondary endpoints, i.e. changes in hemoglobin (Hb) A1c and other modifiable cardiovascular risk factors, physical fitness, and quality of life, have been reported elsewhere [19,22,24,25]. 2.4. Measurements
7 Page 7 of 26
2.4.1. Cardiovascular risk factors. The following parameters were evaluated at baseline and end-of-study: HbA1c, fasting blood glucose and serum insulin, Homeostasis Model Assessment-Insulin Resistance (HOMAIR) index, body mass index (BMI), waist circumference, BP, triglycerides, total, HDL and LDL cholesterol, and high sensitivity-C-reactive protein (hs-CRP). Biochemical tests were performed at central laboratory by the
ip t
use of standard methods [18,19].
2.4.2. Liver enzymes and indexes. Baseline and end-of-study levels of aspartate aminotransferase (AST),
cr
alanine aminotransferase (ALT), and -glutamyl-transpeptidase (-GT) were measured at the central
us
laboratory by the use of standard methods (VITROS 5,1 FS Chemistry System, Ortho-Clinical Diagnostics Inc, Raritan, NJ). At end-of-study, samples were collected at least three days after completing the training
an
program, in order to exclude an acute effect of exercise on liver enzymes [26]. Values of FLI was calculated from -GT, BMI, waist circumference and triglycerides, whereas VAI was computed from BMI, waist
M
circumference, triglycerides and HDL cholesterol, as previously reported (Supplemental Table 1) [9,27]. The following cut-off levels for liver enzymes were used (corresponding to the upper limit of laboratory range
d
and normal range, respectively): 40 and 30 IU/L for AST, 45 and 40 IU/L for ALT, and 60 and 45 IU/L for -
te
GT. FLI and VAI values were considered abnormal if >60 and 1.9, respectively [9,27].
Ac ce p
2.4.3. Volume of PA/exercise. The amount of non-supervised PA was prospectively evaluated throughout the study by asking patients to fill in a preliminary validated daily diary, which considered the list of PAs coded in the Minnesota leisure-time PA questionnaire [28]. Volume was calculated by multiplying the metabolic equivalent (MET) scores of each Minnesota code by time spent in each activity and expressed as METs-hr·wk-1. Energy expenditure from aerobic exercise during supervised sessions was calculated automatically by the machines from workload (i.e. the combination of speed and slope for treadmill, steps per minute for step and power for ergometer), whereas, for resistance exercise, an estimate of 3 METs-hr was established [18,19]. 2.4.4. Physical fitness. Parameters of physical fitness were evaluated at baseline, end-of-study, and, in the EXE group, also during the study period, in order to adjust training loads [18,19,24]. Assessment of cardiorespiratory fitness consisted of a sub-maximal VO2max test, i.e. at 80% of the predicted maximal heart rate,
8 Page 8 of 26
performed at the treadmill with direct measurement of oxygen consumption using a gas exchange analyzer (FitMate, Cosmed, Rome, Italy). For strength assessment, a maximal repetition (or 5-8 RM) test was used, then predicted 1-RM was calculated using the Brzycki formula. For hip and trunk flexibility assessment, a standard bending test was performed.
ip t
2.5. Statistical analysis
cr
As previously reported, sample size was calculated a priori for comparing EXE and CON participants
[18,19] and post-hoc for assessing differences between HI and LI subjects [22]. Calculation considered a
us
reduction of at least 0.5%in HbA1c levels (primary endpoint), with a standard deviation for baseline values of ~1.5%, and a statistical power of ~90% (=0.05).
an
Patients completing the follow-up (CON=275; EXE=288 [LI=136; HI=152]) (Figure 1) were analyzed. Data were expressed as mean±SD or number and percentage of subjects with values above the normal range.
M
Baseline values of EXE and CON subjects were compared using the unpaired t-test or the Mann Whitney Utest. Within-group end-of-study versus baseline differences were analyzed using the Wilcoxon signed ranks
d
test. Mean differences (95% confidence intervals) between baseline to end-of-study changes in the two
te
groups or in the two EXE subgroups were also reported. The efficacy of EXE versus CON and HI versus LI
Ac ce p
programs was assessed using the unpaired t-test or the Mann Whitney U test, by comparing betweengroups changes from baseline to end-of-study. To identify independent predictors of changes from baseline in FLI and VAI, multiple regression analyses with stepwise variable selection were applied, with baseline FLI or VAI values, study arm, age, gender, statin treatment, and changes in HbA1c, HOMA-IR, HDL (for FLI only) and LDL cholesterol, hs-CRP, VO2max, upper and lower body strength, and volume of PA/exercise as covariates forced in the model. Results were expressed as beta unit coefficients. Values of liver enzymes and markers of NAFLD were also stratified according to quintiles of PA/exercise volume and differences between quintiles were assessed by the use of Kruskal-Wallis test. All analyses were performed on individuals completing the follow-up. Statistical analyses were performed using SPSS version 13.0 (SPSS Inc., Chicago, Illinois, USA).
9 Page 9 of 26
3. Results As previously reported [19], attendance to aerobic and/or resistance sessions was high (median 80.3%, interquartile range 75% to 99%) and drop-out was low (28 [9.24%] CON and 15 [4.95%} EXE participants).
ip t
3.1. Baseline to end-of-study changes in liver enzymes and indexes The percentage of subjects with elevated levels of liver enzymes and indexes decreased throughout the
cr
study for FLI and to a lesser extent VAI and -GT, but not AST and ALT (Supplemental Figure 1). Levels of AST, ALT, -GT, FLI and VAI decreased in 226 (40.1%), 174 (30.9), 263 (46.7%), 314 (55.8%), and 289 (51.3%),
us
respectively, and remained unchanged or increased in the remaining individuals.
Baseline values were similar in EXE and CON subjects except for AST and FLI levels, which were
an
significantly higher (P=0.009 and P=0.02) in EXE and CON groups, respectively. In both groups, no improvement in enzyme levels was observed throughout the study, since AST and ALT increased slightly but
M
significantly (except AST in CON participants), whereas -GT values decreased non-significantly in both groups. As a result, mean differences in baseline to end-of-study changes between the two groups were
d
negligible. Conversely, both FLI and VAI decreased significantly in the EXE but not CON group, with highly
te
significant mean differences between the two groups in favor of EXE participants (Table 1). As previously
Ac ce p
reported [19], of the parameters used for calculation of these surrogate measures of NAFLD, decreases in BMI (mean difference -0.78 kg/m2 [95% confidence interval, -1.07 to -0.49]) and waist circumference (-3.6 cm [-4.4 to -2.9]) and increases in HDL cholesterol (3.7 mg/dl [2.2 to 5.3]) were significantly more marked (P
