Participants were randomly allocated to four study groups: (1) high-intensity interval training (HIIT) and habitual diet, (2) very low-carbohydrate, high-fat diet (VLCHF) and habitual physical activity (no regular exercise training), (3) VLCHF diet and HIIT, and (4) control (habitual diet and physical activity, no regular exercise training) (Table 1). The inclusion criteria were age 20–59 years, non-smokers, overweight/obesity (BMI 25.00–40.00 kg/m²), no specific sports training or regular exercise (low physical activity), no excessive alcohol intake, willingness to accept random assignment, self-reported no evidence of liver, renal, metabolic, and cardiopulmonary disease, and diseases contraindicating physical activity, no cancer, no psychiatric illness, no pregnancy or breast-feeding, no specific diet, PAR-Q pass, body mass stable for the last 2 months (<5% of total body mass), and not on a weight-loss plan, no hypoglycemic, lipid-lowering, antihypertensive, psychiatric medications, or medications known to affect body weight or energy expenditure. The participants had no previous experience with the VLCHF diet or HIIT. The recruitment details and dropouts during the study are shown in Figure 1. Written informed consent was obtained from all study participants. The study design was approved by the local University ethics committee.

It was a randomized, controlled, four-arm, parallel exercise and/or dietary intervention study (ClinicalTrials.gov: NCT03934476). The randomization process was stratified according to age (20–29, 30–39, 40–49, 50–59 years) and sex (male, female). The condition assignments were placed by a principal researcher in envelopes that were drawn by the participants. The study group assignments were randomly permuted within blocks of 8 participants (2 participants randomly allocated to each study arm). There were 91 participants (27 males, 64 females) allocated to the four study groups and these completed a 12 week experimental period (Figure 1). Body composition and fitness level were analyzed before the experimental period (T₀) and after 4, 8, and 12 weeks (T₁, T₂, and T₃).

A control group was included to obtain measures for no intervention. Participants in the control group were advised not to change their habitual diet and physical activity regime. No diet advice was given. Participants were tested at T₀-T₃.

Before starting the intervention, researchers provided detailed instructions on the HIIT program to the participants in both the HIIT and VLCHF+HIIT groups in person verbally and in written form. The participants were told to complete self-performed three sessions per week. Two HIIT sessions were completed during weeks 4, 8, and 12 when the participants visited the laboratory. Each HIIT session was started and finished with 5 min of slow walking. HIIT consisted of a 3 min interval of high-intensity walking (Borg's scale RPE 18–19) followed by a 3 min interval of low-intensity walking (RPE 9–11). There were 4, 6, and 8 high-intensity intervals for the first, second, and third 4 week period. Therefore, the total session time increased from 31 to 43 min and 55 min during each 4 weeks (Figure 2). Training intensity was measured with a heart rate monitor (Polar M430; Polar Electro, Oy, Finland), and training data were uploaded to Polar Flow and analyzed regularly and to track compliance. Participants were obliged to record all additional training sessions of any type in addition to the study protocol.

Participants in both the HIIT and control groups were asked to maintain their habitual dietary intake without restriction. The VLCHF diet was defined as allowing no more than 50 g/day of CHO (33). Neither diet included a specific calorie or energy goal. However, participants in the VLCHF group were advised to compensate for the total energy decrease caused by CHO intake restriction by increasing their natural non-trans fat intake (e.g., cream, butter, olive, and coconut oil). A target protein intake of 1.5 g/kg lean body mass was recommended, and unlike the strict CHO restriction, participants were asked to keep to targets. The use of all sweetened and grain-based products was to be minimized. The recommended food included whole food sources, such as meats, vegetables, non-sweetened products, full-fat dairy items, nuts, and seeds. A dietitian provided detailed dietary advice before and during the study (on request or at least once a month). A handbook was given to participants containing food lists, guidelines for estimating macronutrient amounts, and sample recipes. All foods and quantities consumed were recorded daily in all study groups beginning 7 days before the intervention period (www.kaloricketabulky.cz). Alcoholic beverages were restricted during the intervention period, and dietary supplements were not permitted 1 month before and during the intervention period, while caffeinated beverages were restricted only before the laboratory sessions.

Body mass and composition were determined using dual-energy X-ray absorptiometry (DXA; Hologic Discovery A, Waltham, MA, USA). The DXA measurement shows an excellent reliability for VAT assessment (ICC > 0.98) however it might underestimate longitudinal changes when compared to magnetic resonance imaging (34). Bioelectrical impedance (BI) analyzer (InBody770, Seoul, Korea) was used only for the waist and hip circumference measurements. Whole-body DXA scan and BI were performed on all participants at baseline and every 4 weeks (four times in total; T₀-T₃). Participants were asked to empty their bladder and bowel if possible. They wore only underwear during the evaluation and removed any metal and jewelry before assessment. Participants were fasted and were asked not to drink an excessive amount of fluids 2 h before the measurement.

Before the intervention and at the start of each 4 weeks (four times in total), the participants underwent a laboratory-graded exercise test (GXT) on a motorized treadmill to determine the total time to exhaustion (TTE), peak aerobic power (V∙O_2peak), second ventilatory threshold (VT₂), and respiratory exchange ratio (RER). The Balke-Ware treadmill protocol (35) was used: the inclination increased from 2.0% every minute by 1.0%, while the speed remained at 5.3 km/h until volitional exhaustion. Expired air was continuously monitored to analyze O₂ and CO₂ concentrations during the GXT with a breath-by-breath system (Blue Cherry, Geratherm Medical AG, Germany). The highest average O₂ consumption measured over a 30 s period was used to determine V∙O_2peak. Gas exchange measurements were also used to quantify VT₂. A second sharp increase in ventilation (VE) accompanied by an increase in VE/VO₂ and VE/VCO₂ was used to define VT₂ (36). V∙O_2peak can be determined with the within-subject coefficient of variation (CV) between 4 and 9% (37). Heart rate (HR) was measured using a chest belt monitor (Polar Electro H9, Kempele, Finland). Each participant performed the laboratory sessions at a similar time of the day (±60 min). All sessions were conducted at least 3 h after the participants' last meal and in a thermally controlled laboratory (21°C, 40% relative humidity). Each participant was advised not to participate in any vigorous activity 24 h before the test.

A capillary blood sample was drawn from a finger to measure β-hydroxybutyrate (βHB) (FreeStyle Optium Neo, Oxon, United Kingdom). Participants self-analyzed βHB twice a week (every Monday and Thursday) in a fasting state in the morning.

The categorical variable (sex) was described by frequency ratio, and numerical variables were described by median and interquartile range at each time point. Subsequently, the absolute changes of the monitored variables at time T₁, T₂, and T₃ with respect to the baseline (T₀) were analyzed. The absolute changes were tested for normal distribution using the Shapiro-Wilk test. In some cases, significant deviations from normality were detected such that non-parametric methods of data description (median and interquartile range) and statistical inference were used. Significance of change was tested by 95% confidence interval (CI) of median and two-tailed Wilcoxon signed-rank test for each variable, each group, and each time. The effect size (ES) of the observed changes was specified by the Wilcoxon effect size (r), including its 95% confidence interval. Threshold values for ES were 0.10 to < 0.30 (small), 0.30 to < 0.50 (medium), ≥ 0.50 (large).

Finally, the absolute changes in the given variables for the HIIT, VLCHF, HIIT+VLCHF, and Control groups were compared using the Kruskal-Wallis test at each time point. Dunn's test was used to analyze specific sample pairs for stochastic dominance. The effect size of the observed differences was assessed using the eta squared based on the H-statistic, including its 95% confidence interval. Threshold values for ES eta squared were 0.01 to <0.08 (small), 0.08 to <0.26 (medium), ≥0.26 (large) (38).

An a priori power analysis using GPOWER (39) with power set at 0.80 and significance level set at 0.05 was calculated retrospectively. The power analysis indicated that a total sample of 76 people would be needed to detect large effects (f = 0.40) for this study with 4 groups. A total sample of 180 people would be needed to detect medium effects (f = 0.25) (40). Thus, the sample size was sufficient to reveal that a large effect could not be interpreted as non-significant.

In all cases, statistical significance was set at p < 0.05. Statistical analyses were performed using R Core Team (41).

There were substantial between-group differences in the training characteristics. Total training time in the HIIT and VLCHF+HIIT groups (median 1,424 and 1,452 min, respectively) was substantially higher than in the groups without the HIIT intervention (VLCHF-−124 min, Control 105 min). All training sessions recorded by the HR monitors are presented in Table 2. These results show all the monitored training sessions, including HIIT sessions in the HIIT and VLCHF+HIIT groups.

Total energy intake decreased (p < 0.05) in the HIIT (median [95% CI]: 6.1 [0.2; 13.4] %), VLCHF (19.7 [12.5; 25.2] %), and VLCHF+HIIT (25.8 [20.5; 28.0] %) groups. Carbohydrate intake decreased (p < 0.05) by 81.8 [79.1; 82.9] % and 82.8 [80.4; 85.7] % in the VLCHF and VLCHF+HIIT groups, respectively. Fat intake increased by 44.6 [36.1; 61.7] % and 34.8 [24.6; 47.3] % in the VLCHF and VLCHF+HIIT groups, respectively. Protein intake did not significantly change in any of the study groups. Total energy, protein, and carbohydrate intake did not significantly change in the control group, whereas fat intake decreased (p = 0.023, 6.0 [1.0; 17.5] %) (Figure 3, Supplementary Tables 1, 2).

The β-hydroxybutyrate concentration (βHB) increased substantially in the VLCHF and VLCHF+HIIT groups. The highest βHB concentrations were achieved after 2 weeks of VLCHF diet intervention (Figure 4). βHB concentrations in the HIIT and control groups remained within the range between 0.0 to 0.3 mmol/l for the whole 12 week intervention (data not shown).

There were large between-group differences in visceral adipose tissue (VAT) mass, area and volume changes after 12 weeks (p < 0.001 for all; ES [95% CI]: 0.42 [0.27; 0.58], 0.41 [0.27; 0.58], and 0.42 [0.27; 0.59], respectively). Post-hoc analysis revealed that 12 week VLCHF diet intervention, regardless of combination with HIIT, was effective in decreasing VAT when compared to the HIIT and Control groups (Table 3, Figure 5). The absolute VAT mass (g) reduced by 23.2 % (median; 95% CI: [25.9; 17.0] %) in the VLCHF group and 17.6 % [23.9; 12.0] in the VLCHF+HIIT group while these changes were not significant after 12 weeks in both the HIIT and Control groups. The complete dataset is reported in Supplementary Tables 3, 6. The significant VAT decrease in the VLCHF and VLCHF+HIIT groups was also revealed after 4 and 8 weeks (Supplementary Tables 4, 5, 7, 8, Figure 5).

Total lean mass significantly decreased after 4 weeks by 4.7 [3.9; 5.7] % and 3.9 [3.1; 5.1] % in the VLCHF and VLCHF+HIIT groups, respectively. These changes remained stable over the 8 and 12 week measurements (4.2 [3.1; 5.2] % and 4.9 [3.8; 6.5] %, respectively) (Figure 5). Significant between-group differences in all other monitored body composition variables, with the most pronounced changes in the VLCHF and VLCHF+HIIT groups, were shown after 12 weeks (Table 3, Figure 5).

There were small to large between-group differences in TTE and relative V∙O_2peak changes after 12 weeks (p < 0.05 for both; ES [95% CI]: 0.15 [0.06; 0.33] and 0.13 [0.04; 0.31], respectively). Post-hoc analysis revealed TTE differences between HIIT and VLCHF+HIIT vs. VLCHF and Control groups. TTE increased after 12 weeks significantly by 12.0% (median; 95% CI [5.8; 17.4] %), 16.9 [10.1; 24.7] %, and 5.9 [1.0; 13.3] % in the HIIT, VLCHF+HIIT, and VLCHF diet groups, respectively. The relative V∙O_2peak was significantly different in all intervention groups compared to the Control group after 12 weeks. The relative V∙O_2peak increased after 12 weeks by 4.9 [−1.6; 12.3] %, 11.9 [3.9; 15.9] %, and 3.7 [0.6; 9.6] % in the HIIT, VLCHF+HIIT, and VLCHF groups, respectively. Small to large between group differences after 12 weeks were found also in the RER_peak (p = 0.002; 0.17 [0.07; 0.35]). Post-hoc analysis showed significantly lower RER_peak in the VLCHF and VLCHF+HIIT groups when compared to the HIIT and Control groups (Table 3, Figure 6). The complete dataset, including the outcomes after 4 and 8 weeks, is reported in Supplementary Tables 3–6. VT₂ was not significantly different between all study groups after 4, 8, and 12 weeks (Table 3, Supplementary Tables 7, 8).

A reduced CHO diet is characterized by a CHO intake below 45% of the total energy intake. The VLCHF diet intervention in this study required a CHO intake of <50 g/day (14). Since the purpose of the VLCHF diet intervention in this study was not to reduce the total energy intake, the participants in the VLCHF and VLCHF+HIIT groups were encouraged to compensate for the CHO intake restriction by increasing fat intake while maintaining protein consumption. Nevertheless, fat intake was not sufficient to keep the total energy intake unchanged (Figure 3), which is a common situation in real-life conditions. Notably, total energy intake significantly decreased in the HIIT group despite there being no diet modification. We attribute this effect to an increased interest in a healthy lifestyle when participating in such a research study.

As expected, the βHB concentration increased significantly with CHO restriction, although nutritional ketosis was not primarily intended (Figure 4). These βHB changes in both diet groups are in agreement with our previous results (42, 43) and have also been shown previously (13, 44, 45). However, the pattern of βHB changes was different between the VLCHF and VLCHF+HIIT groups. Higher and more variable βHB concentrations were found in the VLCHF+HIIT group. This was likely caused by glycogen depletion due to regular high-intensity exercise sessions, which increased nutritional ketosis. We assume that the higher βHB fluctuation corresponds to temporary muscle glycogen store depletion in response to HIIT and reduced endogenous glucose synthesis during the early phase of VLCHF diet intervention. Indeed, muscle glycogen stores are not diminished in VLCHF diet-adapted individuals (46). Our study results are unique as they show that nutritional ketosis (if intended) can be supported by high-intensity exercise and not only provoked by strict CHO intake manipulation.

Interestingly, βHB tended to decrease toward the end of the 12 week intervention period, which is in agreement with our previous results (42). This may reflect enhanced cellular uptake after adaptation as a result of enhanced mitochondrial function and capacity for ketone body utilization (47). Another possible explanation for this trend is an increase in gluconeogenetic glucose production (48).

The VLCHF diet, both with and without HIIT, caused significant changes in body composition of these overfat individuals. We found large (ES) between-group differences in the VAT mass changes, which significantly decreased in the VLCHF (by ~23.2) and VLCHF+HIIT (by ~17.6 %) groups (Figure 5). A more substantial reduction in central adiposity was also observed in obese adults when a low-CHO diet was compared to a low-fat diet (9, 49, 50). A low-CHO diet is considered a legitimate and potentially effective form of treatment for patients with obesity (51). A possible explanation for the carbohydrate-restricted diet superiority over a low-fat diet on body composition might be the increasing long-term effect on total energy expenditure (11). However, we cannot relate the presented body composition changes unequivocally to the VLCHF diet intervention because the total energy intake in the VLCHF and VLCHF+HIIT groups was also significantly decreased. On the other hand, the total energy intake significantly decreased in the HIIT group, but without a pronounced effect on body mass and composition.

The total lean mass significantly decreased abruptly in the VLCHF and VLCHF+HIIT groups after 4 weeks, but then remained relatively stable. This can be explained by both the sharp reduction in total energy intake or the macronutrient shift at the beginning of the VLCHF diet intervention. Interestingly, exercise in the VLCHF+HIIT group did not prevent this lean body mass change. Such a decrease in lean body mass, induced by a CHO restricted diet, was previously observed in obese (10) and athletic subjects (52). Although a muscle/liver glycogen reduction and following higher rate of gluconeogenesis from muscle amino acids during exercise may be suggested as a cause of the lean body mass decrease, our data do not support such an assumption since the lean body mass decrease was found similarly in the both VLCHF and VLCHF+HIIT groups. However, we have to consider that lean body mass results gained by DXA can be influenced by hydration status, i.e., DXA interprets body water loss as total lean mass loss (53). Very low CHO diets can be accompanied by body water loss due to accelerated sodium and water excretion and glycogen depletion which may cause also an additional body water loss (10). Although lean body mass decrease ceased after 4 weeks, VAT mass/area/volume continued to decrease progressively throughout the 12 week VLCHF diet intervention in the present study.

The most pronounced TTE increase was revealed in the VLCHF+HIIT group after 12 weeks. Despite the fact that TTE increased also in the HIIT and VLCHF groups, the absolute V∙O_2peak values did not change in any study group. Therefore, we attribute the significant increase of relative V∙O_2peak (ml/kg/min) values in the VLCHF and VLCHF+HIIT groups rather to the rapid body mass reduction than any enhancement of aerobic capacity per se. Although carbohydrate restricted diets combined with HIIT were shown to enhance aerobic capacity in various populations (54) as well as in the obese (55), this rapid body mass reduction effect on relative V∙O_2peak values is not usually considered (56). The walking HIIT protocol and/or 12 week intervention period were probably not sufficient enough to increase aerobic capacity in the present study, although the volume of exercise at high intensity was significantly increased in both HIIT and VLCHF+HIIT groups. However, we showed that regular HIIT sessions over 12 weeks had an obvious beneficial effect on exercise capacity enhancement (i.e., TTE). The increases in physical activity and/or cardiorespiratory fitness can be associated with greater reductions in mortality risk than is intentional body mass loss (20). Therefore, we can still consider HIIT an effective and time saving (30) tool in the prevention and treatment of chronic diseases in overfat individuals, even if the additional effect of the VLCHF diet and HIIT combination on cardiorespiratory fitness was not revealed.

RER_peak significantly decreased in the VLCHF and VLCHF+HIIT groups, as was shown in our previous studies (42, 43). This indicates an increased rate of fat oxidation, which has previously been shown to be potentially associated with impaired high-intensity performance because of a reduction in maximal CHO utilization (57). However, no such exercise capacity reduction was demonstrated in this study. Regarding submaximal exercise, the TTE may vary between individuals despite the lower RER and no significant changes in V∙O_2max following a 4 week eucaloric ketogenic diet (58). Shaw et al. (59) showed that keto-adapted participants with RER <1.0 at V∙O_2max, reduced the TTE in submaximal exercise, while those with RER_max >1.0 preserved the TTE. We cannot confirm these results because the TTE significantly increased in the VLCHF and VLCHF+HIIT groups after 12 weeks (Figure 6). These differences might be explained by the fact that the duration of the VLCHF diet interventions, which is only 3–4 weeks in the aforementioned studies (57, 59), does not allow adequate adaptation to the VLCHF diet by increasing endogenous glucose production (60, 61).

Our study has some limitations. First, this real-life study did not allow us to directly collect some of the data (e.g., daily records of nutrition and exercise sessions). Therefore, the participant's adherence to all study requirements cannot be fully controlled. Second, the study groups were not balanced for sex and, despite the randomization, median age of the HIIT group was slightly higher than in the other study groups. Third, the menstrual cycle and menopause status were not considered within the data collection and analysis.