The Effect of a Novel Low-Volume Aerobic Exercise Intervention on Liver Fat in Type 2 Diabetes: A Randomized Controlled Trial

Type 2 diabetes is a polygenic disorder involving interactions between genetic and environmental risk factors, resulting in hepatic and muscle insulin resistance and pancreatic β-cell dysfunction (1,2). Ectopic fat is a term used to describe the excess accumulation of fat in nonadipose tissue (2,3), with ectopic fat in the liver being closely linked to the development, progression, and severity of diabetes (4). Individuals with type 2 diabetes have increased amounts of ectopic fat that cannot be simply explained by excess body weight (5). Furthermore, insulin resistance plays a critical role in the pathogenesis and progression of nonalcoholic fatty liver disease, which is highly prevalent in individuals with type 2 diabetes (6,7). Consequently, preventing and reducing liver fat accumulation has emerged as an important therapeutic target for the management of cardiometabolic health in type 2 diabetes.

Systematic reviews and meta-analyses have shown that regular aerobic exercise significantly reduces liver fat in nondiabetic cohorts (8,9), and this may be similar in type 2 diabetes, although there is a relative lack of data (10). Importantly, while exercise studies often vary in prescribed volume, intensity, and training frequency, the majority have used exercise interventions resembling those described in current physical activity guidelines (11). Although efficacious, the current physical activity guidelines may be too burdensome for time-poor individuals (12), and as a result, low-volume high-intensity interval training (HIIT) has become increasingly investigated.

Prior studies have suggested that HIIT may be comparable to traditional higher-volume moderate-intensity exercise for reducing visceral fat and improving glucose control (13,14). However, a study comparing the effects of low-volume HIIT versus high-volume HIIT reported similar improvements in cardiorespiratory fitness (CRF) and glucose control in both intervention groups, but only high-volume HIIT improved body composition (15).

Currently, there are limited studies that have investigated the efficacy of low-volume HIIT for improving liver fat in adults with type 2 diabetes. Therefore, the aim of this study was to examine the effect of regular aerobic exercise involving either moderate-intensity continuous training (MICT) or a novel low-volume HIIT approach versus a sham placebo (PLA) control on liver fat in inactive adults with obesity and type 2 diabetes. The effect of these interventions on other cardiometabolic risk factors, such as glycemia and CRF, were also assessed. It was hypothesized that only MICT would lead to improvements in liver fat and that both exercise interventions would improve CRF and glycemia.

All assessments were undertaken at the Charles Perkins Centre at The University of Sydney except for proton MRS (¹H-MRS), which was undertaken at a separate imaging facility. Liver fat percentages, CRF, anthropometric measures, fasting blood biochemistry, glycosylated hemoglobin (HbA_1c), and fasting blood lipids were assessed before and after 12 weeks of exercise or placebo interventions. Participants were asked to abstain from caffeine, alcohol, tobacco, and strenuous physical activity for 48 h before baseline and postintervention testing, and postintervention blood analysis was performed ≥48 h after the last scheduled exercise session. Participants were also asked to report medications being taken at the time of commencement and whether any change in medication type, dosage, or frequency occurred throughout the trial. After baseline measures were completed, participants were randomized into one of the three groups (MICT, HIIT, or PLA) by equally distributed, pregenerated lists of permuted blocks (http://www.randomization.com). Participants were given a sealed opaque envelope containing group allocation. The study protocol conformed to the ethical standards of the 1975 Declaration of Helsinki and was approved by the human research ethics committee of The University of Sydney and the Royal Prince Alfred Research Ethics and Governance Office (Camperdown, New South Wales, Australia) and was registered prospectively with the Australian New Zealand Clinical Trials Registry.

Inactive (exercising <3 days/week) adult men and women with obesity (BMI >30 kg/m²) and type 2 diabetes were recruited between July 2015 and February 2019. Participants provided written informed consent before being medically cleared to participate in the study. Participants were excluded if they had uncontrolled hypertension and/or hyperglycemia; were pregnant; reported claustrophobia; had a rapidly progressive disease; reported a high alcohol intake (>30 and >20 g/day for men and women; respectively); had a secondary cause of steatohepatitis, including alcoholic liver disease; had viral hepatitis; or could not attend three exercise sessions per week for 12 weeks at the Charles Perkins Centre at The University of Sydney.

Liver fat percentages were quantified by ¹H-MRS using a Philips Intera Achieva 1.5T MRI system (Philips Medical Systems, Best, the Netherlands). Image-guided localized ¹H-MRS spectra were acquired using single-voxel (3.0 × 2.0 × 2.0 cm) point-resolved spectroscopy through the anterior and posterior arrays and Q-body transmit coil (repetition time 5,000 ms, echo time 45 ms, 1,024 sample points), with a total of 32 measurements acquired. Participants laid supine, and spectra were acquired in a respiratory-gated manner. Fully automated high-order shimming was performed to ensure maximum field homogeneity. Spectral data were postprocessed through magnetic resonance user interface software (jMRUI version 5.2; www.jmrui.eu [16,17]) by an assessor (N.A.J.) who was blinded to participant details, including group allocation and time point. Liver water signal amplitudes were quantified using Hankel Lanczos squares singular values decomposition, and a five-resonance model was used to fit the lipid peaks (18).

CRF was determined through a graded maximal exercise test on an electronically braked cycle ergometer (Corival; Lode, Groningen, the Netherlands) under the supervision of an accredited exercise physiologist. VO_2peak was measured through breath-by-breath analysis using a metabolic cart (Medical Graphics Corporation, St. Paul, MN). All tests incorporated a 3-min warm-up at 35 W for women and 65 W for men. The workloads were incrementally adjusted by 25 W every 150 s until volitional fatigue. The test was terminated when the participant could not sustain the pedaling rate of 60–70 revolutions/min for >10 s or if the participant ceased exercise. VO_2peak was determined by averaging the last 30 s of oxygen consumption before termination of the test.

Height was measured with a stadiometer (seca Model 220; seca, Hamburg, Germany) to the nearest 0.5 cm. Waist circumference (WC) was measured to the nearest 0.5 cm in triplicate at the horizontal plane, midway between the inferior margin of the ribs and the superior border of the iliac crest after expiration and before inspiration. Body weight was measured with an electronic digital platform scale (Tanita BC-418 Body Composition Analyzer; Tanita Corporation, Tokyo, Japan).

After an overnight fast (>10 h), venous blood was collected from the antecubital vein and processed as previously described (18). Blood analysis for HbA_1c, fasting blood glucose (FBG), fasting insulin, triglycerides (TG), total cholesterol (TC), HDL cholesterol (HDL-C), LDL cholesterol (LDL-C), ALT, AST, and hs-CRP was performed in a commercial laboratory on the same day as collection. Concentration of plasma free fatty acids (FFAs) was measured using stored plasma. To undertake HOMA of steady-state stable pancreatic β-cell function (HOMA2-%β), insulin sensitivity (HOMA2-%S), and insulin resistance (HOMA2-IR) on the basis of fasting blood sampling, measures were calculated using the HOMA2 calculator released by the Diabetes Trial Unit, University of Oxford, and as described elsewhere (19).

Participants were informed that the purpose of the study was to determine the effect of exercise on health in individuals with type 2 diabetes. All exercise sessions were supervised by an accredited exercise physiologist, and heart rate, rating of perceived exertion, and blood pressure were measured at regular intervals throughout the exercise sessions. The exercise sessions took place between July 2015 and May 2019. Objectively quantified physical activity was assessed (as detailed in Supplementary Material S.1.1), and participants were asked at commencement and reminded every 2 weeks to not alter their dietary or physical activity behaviors throughout the trial.

Participants in the HIIT group performed three exercise sessions for 12 weeks consisting of a minimum of 19 min of cycling. Each exercise session included 4 min of cycling at a work rate equivalent to 90% VO_2peak and a 10-min warm-up and 5-min cool-down at a work rate equivalent to 50% VO_2peak. Participants were required to progressively increase the duration of the HIIT bout from 1 to 4 min per session by week 4.

Participants in the MICT group performed three exercise sessions for 12 weeks with each consisting of 40–55 min of continuous cycling. The exercise duration was progressed from 30 min in the 1st week to 45 min by the 4th week and involved continuous cycling at a work rate equivalent to 60% VO_2peak. All exercise sessions included a 5-min warm-up and cool-down at 50% VO_2peak.

Participants in the PLA group were prescribed a fitball exercise and upper and lower body stretches. Participants performed supervised sessions every 2 weeks involving a 5-min cycle at low intensity (25 W) to maintain familiarity with the cycle ergometer, and total session times were <25 min. The PLA intervention was designed to not elicit cardiometabolic improvements but to control for potential confounders, such as attention and participation in a lifestyle intervention.

An a priori, two-tailed power calculation at an α of 0.05 and β of 0.8 gave an actual power of 0.813 for a sample size of 11 in each group (G-Power software; University of Trier, Trier, Germany). This calculation was determined using the effect size of 1.28 of a similar exercise intervention from a previous study, which detected significant improvements in liver fat within groups (18).

Exercise session compliance was calculated as the total number of sessions attended/total number of available sessions × 100. An intention-to-treat analysis was used, with group mean change scores imputed for dropouts and where data were missing (20) (Table 2). The primary analyses explored the effects between the PLA and exercise groups for change in liver fat, which was assessed by ANCOVA using the baseline value as the covariate. Least significant difference post hoc comparisons were used to locate significant treatment differences. Baseline differences in primary, secondary, and physical activity measures were compared using one-way ANOVA between the PLA and exercise groups. Effect size values were calculated as standardized differences in the means and expressed as Hedges g. Relationships among changes in liver fat, HbA_1c, and other outcome measures were determined through Pearson correlation coefficient. Separate hierarchical linear regression analyses were performed for change in liver fat. Baseline values of the variables significantly correlated with change in liver fat were entered into the regression at block 1 of the model. Change scores of variables significantly correlated with change in liver fat were entered into block 2 of the model. The change in VO_2peak (mL/kg/min) was entered into block 3 of the model. Data were analyzed using SPSS version 24.0 software (IBM Corporation, Armonk, NY). Statistical significance was determined as P < 0.05. Data are reported as mean (SE) from the mean.

Thirty-five eligible participants (19 men and 16 women) underwent initial assessment and subsequent randomization (Fig. 1). Baseline participant characteristics are detailed in Table 1. The study participants had a mean BMI of 35.9 ± 0.9 kg/m² and a mean age of 54.6 ± 1.4 years at baseline. Thirty-one participants reported taking oral glycemia-lowering medications, 20 reported taking lipid-lowering medications, and 17 reported taking antihypertensive medications (Supplementary Table 2.2). There were no significant between-group differences in primary or secondary outcomes at baseline. All participants in HIIT completed the training intervention, whereas two participants in MICT and one participant in PLA did not complete the intervention (Fig. 1). Postintervention values were determined using an intention-to-treat analysis for noncompleters; however, per-protocol analyses are included in Supplementary Table 2.3. Exercise intervention compliance was 93%, 98%, and 63% for participants who completed postintervention assessments in the MICT, HIIT, and PLA groups, respectively. One participant in HIIT suffered from an acute illness in the last 2 weeks of the training intervention and gained >5% body weight. Consequently, postintervention outcome scores for this participant were imputed using an intention-to-treat analysis. Intention-to-treat analysis was also used for one participant in PLA who began lipid-lowering medication during the course of the intervention (prescribed by a nonstudy physician). In total, intention-to-treat analysis was used for two participants in MICT, one in HIIT, and two in PLA.

There was a significant between-group difference for change in liver fat, which reduced in MICT (−0.9 ± 0.7%) and HIIT (−1.7 ± 1.1%) but increased in PLA (1.2 ± 0.5%) (P = 0.046) (Fig. 2). There were no significant differences between exercise interventions for change in liver fat (P > 0.05).

Baseline and postintervention data for CRF are summarized in Table 2. There was a significant between-group difference for change in CRF, which increased in MICT (2.3 ± 1.2 mL/kg/min) and HIIT (1.1 ± 0.5 mL/kg/min) but decreased in PLA (−1.5 ± 0.9 mL/kg/min) (P = 0.006) (Table 2). There were no significant differences between exercise interventions for change in CRF (P > 0.05).

Baseline and postintervention data for anthropometric measurements are summarized in Table 2. There was a significant between-group difference for change in waist circumference (P = 0.029). There were no significant differences between the exercise interventions for reducing waist circumference (P > 0.05). There were no significant between-group differences for change in BMI or body weight (P > 0.05).

Baseline and postintervention data for blood lipids and biochemistry are summarized in Table 2. There was a significant between-group difference for change in HbA_1c, which decreased in MICT (−0.3 ± 0.3%) and HIIT (−0.3 ± 0.3%) but increased in PLA (0.4 ± 0.2%) (P = 0.014). There were no significant differences between exercise interventions for change in HbA_1c (P > 0.05). The exercise interventions led to a significant between-group reduction in FBG and an increase in fasting insulin relative to PLA (P = 0.035 and P = 0.042, respectively). Derived indices using HOMA to explore changes in pancreatic β-cell function and insulin sensitivity showed no significant between-group differences in HOMA2-%β, HOMA2-%S, or HOMA2-IR (P = 0.056, P = 0.143, and P = 0.126, respectively). There were also no significant between-group differences for change in fasting AST, ALT, hs-CRP, TC, HDL-C, LDL-C, TG, or FFAs (P > 0.05 for all).

The change in liver fat percentage was positively associated with baseline AST (r = 0.390, P = 0.021) and ALT (r = 0.426, P = 0.011) and changes in body weight (r = 0.552, P = 0.001), AST (r = 0.445, P = 0.006), ALT (r = 0.560, P < 0.001), LDL-C (r = 0.571, P < 0.001), and TG (r = 0.588, P < 0.001). The change in liver fat percentage was inversely associated with change in VO_2peak (mL/kg/min) (r = −0.389, P = 0.021).

The change in HbA_1c was positively associated with changes in WC (r = 0.346, P = 0.042), FBG (r = 0.819, P < 0.001), and CRP (r = 0.356, P = 0.036). The change in HbA_1c was inversely associated with baseline HbA_1c (r = −0.420, P = 0.012), hs-CRP (r = −0.380, P = 0.025), and change in HOMA2-%β (r = −0.454, P = 0.013).

Baseline levels of ALT accounted for 27% of the variance in change in liver fat percentage (P = 0.002). The inclusion of changes in body weight, LDL-C, TG, and ALT in block 2 increased the prediction of change in liver fat percentage to 61% (P = 0.001). The inclusion of change in VO_2peak in block 3 further increased the prediction of change in liver fat percentage to 70% (P = 0.009) (Table 3).

This study compared the effect of a novel low-volume HIIT or MICT exercise intervention versus PLA on ¹H-MRS–quantified liver fat and glycemia in inactive adults with obesity and type 2 diabetes. The results show that when compared with PLA, a MICT or a low-volume HIIT approach involving 4 min of vigorous-intensity exercise can reduce liver fat and improve glycemia. The benefits of exercise on liver fat and glycemia were independently associated with improvements in CRF and were evident in the absence of a significant change in body weight.

Ectopic fat contributes to inflammation, mitochondrial dysfunction, β-cell dysfunction, and insulin resistance, thereby increasing the risk for cardiovascular disease (4,21). Similarly, excess liver fat directly contributes to the development and progression of type 2 diabetes (1). Weight loss of ∼5–15% produced by lifestyle intervention (22) or bariatric surgery (23) has been shown to reduce liver fat in individuals with type 2 diabetes and has, in some instances, led to the remission of type 2 diabetes. However, these improvements depend on sustained weight loss, which is often an unlikely outcome of lifestyle interventions (24). Exercise training has previously been shown to modestly reduce liver fat (8,10,25) independent of weight loss in nondiabetic populations. The results of this study extend these findings by showing that, when compared with a PLA intervention, aerobic exercise improves liver fat without significant weight loss. Although significant weight loss was not achieved, the change in liver fat was positively associated with body weight change, suggesting that even small amounts of exercise-induced weight loss may lead to beneficial reductions in liver fat. Furthermore, there were no significant differences between the exercise training groups for the reduction of liver fat, which is remarkable given the low volume/time commitment in the HIIT intervention.

Technological advancements in noninvasive techniques for quantifying liver fat, including ¹H-MRS, have aided research into ectopic fat. ¹H-MRS is currently considered the gold standard noninvasive approach for quantifying liver fat (26,27). While ¹H-MRS can accurately quantify liver fat percentage, it does not assess severity or stage of liver disease, which can only be assessed accurately using liver biopsy (28). Consequently, this study is unable to draw definite conclusions regarding exercise-induced changes in liver fat on hepatic inflammation or steatohepatitis. While the observed relative reductions in FBG in this study were expected by the exercise interventions compared with PLA control, the relative increases in fasting insulin were not. Fasting glucose commonly reduces and fasting insulin appears less susceptible to change in exercise studies of type 2 diabetes or metabolic syndrome (29). In the current study, we could not readily explain the relative increase in fasting insulin by the diabetes glycemic drugs used during the study, even though the use of sulfonylurea (insulin secretagogue–specific) treatment was overall more common in the exercise groups than PLA (Supplementary Table 2.2) and because this regimen did not appreciably change after baseline. We cannot exclude the possibility that some exercise events that occurred acutely (within 48 h) of blood sampling led to the observed increase in fasting blood insulin compared with the nonexercising PLA control group. Furthermore, while both exercise training interventions appear to be effective for improving liver fat when compared with PLA, mean postintervention liver fat levels were still considered abnormal (>5.5% [30]); consequently, it is unclear whether these types of changes are clinically meaningful (31). Again, notwithstanding the absence of cause-effect data, the observed glycemic improvements associated with these changes may provide some evidence of their clinical relevance. Future exercise studies incorporating more direct measures and sensitive indices of pancreatic β-cell function, including those undertaken nonfasting, are required to determine whether exercise may change the rate of attrition in pancreatic β-cell function in the short or long term.

Exercise has long been considered a cornerstone therapy for the management of type 2 diabetes. Previous studies have shown that structured aerobic and/or resistance exercise leads to significant improvements in glucose control and insulin resistance (32,33). The results of this study demonstrated a significant between-group difference between the exercise groups and PLA control group for HbA_1c, thus expanding the current understanding of the relationship between exercise dose and glucose control. A joint position statement by the American College of Sports Medicine and American Diabetes Association recommended that individuals with type 2 diabetes accumulate ≥150 min of moderate- to vigorous-intensity exercise per week to improve glucose control (11). While both exercise interventions led to improvements in HbA_1c by similar magnitudes, the HIIT intervention did so with only 12 min of vigorous aerobic exercise per week. Because exercise adherence is a major challenge for individuals with type 2 diabetes (12), it is important to determine minimum thresholds for exercise-related benefits. The findings from this study demonstrate that even low volumes of aerobic exercise training may lead to significant improvements in glucose control in type 2 diabetes. Although there are no physical activity guidelines related to ectopic fat reduction, the results of this study also show that low-volume HIIT and MICT at volumes below current exercise recommendations led to reductions in liver fat. This suggests that future physical activity guidelines should also consider the metabolic benefits of low-volume approaches to exercise in adults with type 2 diabetes. Importantly, these findings were, in part, explained by improvements in CRF, which is increasingly recognized as a therapeutic target for individuals with or at risk for obesity-related disease (34).

This study has limitations that should be considered before interpreting the results. First, while participants were instructed to not alter their diet, change in habitual energy intake was not monitored, which could have affected the results. However, the absence of a between-group difference for change in body weight suggests that any confounding effect should be negligible and affect all groups equally. Similarly, while participants were instructed to not change their physical activity behaviors for the duration of the intervention and attempts were made to monitor change in habitual energy expenditure, there were insufficient postintervention data to be able to confirm that significant changes in habitual energy expenditure did not occur (Supplementary Table 2.1). Second, despite including adjusted R² values, which facilitate a more comprehensive evaluation of the regression analysis, the relatively small sample size should be considered when interpreting these results. Third, this project was not designed to determine whether exercise-induced reductions in liver fat mediate insulin sensitivity and HbA_1c; however, the main novel findings in this work establish the rationale for undertaking such detailed mechanistic studies in the future, including clamp studies and glucose tolerance tests with derived indices. Finally, while group mean scores for baseline and postintervention liver fat were reported, individuals varied in their response to the interventions, with some individuals in the exercise groups experiencing an increase in liver fat. Therefore, the generalizability of these results may be limited, and further research is warranted.

In conclusion, the novel data produced in this study suggest that as little as 12 min of HIIT per week may be an effective strategy for reducing liver fat in inactive adults with type 2 diabetes, including in the absence of significant weight loss. These changes were accompanied by improvements in CRF and HbA_1c. However, further studies are required to elucidate the relationship between exercise-induced reductions in liver fat and improvements in glycemia and to establish valid noninvasive techniques for quantification of ectopic fats in lifestyle intervention studies.

Funding. This research was supported by funding from the Collaborative Research Network for Advancing Exercise & Sports Science (CRN-AESS)/Research Capacity Building Seeding Grant Scheme.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. K.L.W., S.E.K., M.K.B., and N.A.J. contributed to the conception and design of the study. A.S., K.L.W., R.N.S., J.A.G., and N.A.J. contributed to data acquisition and/or analysis. A.S., M.K.B., S.M.T., and N.A.J. contributed to the interpretation of data. A.S. drafted the manuscript. All authors critically revised the manuscript and gave final approval. A.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.