Three weeks of time-restricted eating improves glucose homeostasis in adults with type 2 diabetes but does not improve insulin sensitivity: a randomised crossover trial

Aims/hypothesis Time-restricted eating (TRE) is suggested to improve metabolic health by limiting food intake to a defined time window, thereby prolonging the overnight fast. This prolonged fast is expected to lead to a more pronounced depletion of hepatic glycogen stores overnight and might improve insulin sensitivity due to an increased need to replenish nutrient storage. Previous studies showed beneficial metabolic effects of 6–8 h TRE regimens in healthy, overweight adults under controlled conditions. However, the effects of TRE on glucose homeostasis in individuals with type 2 diabetes are unclear. Here, we extensively investigated the effects of TRE on hepatic glycogen levels and insulin sensitivity in individuals with type 2 diabetes. Methods Fourteen adults with type 2 diabetes (BMI 30.5±4.2 kg/m2, HbA1c 46.1±7.2 mmol/mol [6.4±0.7%]) participated in a 3 week TRE (daily food intake within 10 h) vs control (spreading food intake over ≥14 h) regimen in a randomised, crossover trial design. The study was performed at Maastricht University, the Netherlands. Eligibility criteria included diagnosis of type 2 diabetes, intermediate chronotype and absence of medical conditions that could interfere with the study execution and/or outcome. Randomisation was performed by a study-independent investigator, ensuring that an equal amount of participants started with TRE and CON. Due to the nature of the study, neither volunteers nor investigators were blinded to the study interventions. The quality of the data was checked without knowledge on intervention allocation. Hepatic glycogen levels were assessed with 13C-MRS and insulin sensitivity was assessed using a hyperinsulinaemic–euglycaemic two-step clamp. Furthermore, glucose homeostasis was assessed with 24 h continuous glucose monitoring devices. Secondary outcomes included 24 h energy expenditure and substrate oxidation, hepatic lipid content and skeletal muscle mitochondrial capacity. Results Results are depicted as mean ± SEM. Hepatic glycogen content was similar between TRE and control condition (0.15±0.01 vs 0.15±0.01 AU, p=0.88). M value was not significantly affected by TRE (19.6±1.8 vs 17.7±1.8 μmol kg−1 min−1 in TRE vs control, respectively, p=0.10). Hepatic and peripheral insulin sensitivity also remained unaffected by TRE (p=0.67 and p=0.25, respectively). Yet, insulin-induced non-oxidative glucose disposal was increased with TRE (non-oxidative glucose disposal 4.3±1.1 vs 1.5±1.7 μmol kg−1 min−1, p=0.04). TRE increased the time spent in the normoglycaemic range (15.1±0.8 vs 12.2±1.1 h per day, p=0.01), and decreased fasting glucose (7.6±0.4 vs 8.6±0.4 mmol/l, p=0.03) and 24 h glucose levels (6.8±0.2 vs 7.6±0.3 mmol/l, p<0.01). Energy expenditure over 24 h was unaffected; nevertheless, TRE decreased 24 h glucose oxidation (260.2±7.6 vs 277.8±10.7 g/day, p=0.04). No adverse events were reported that were related to the interventions. Conclusions/interpretation We show that a 10 h TRE regimen is a feasible, safe and effective means to improve 24 h glucose homeostasis in free-living adults with type 2 diabetes. However, these changes were not accompanied by changes in insulin sensitivity or hepatic glycogen. Trial registration ClinicalTrials.gov NCT03992248 Funding ZonMW, 459001013 Graphical abstract Supplementary Information The online version of this article (10.1007/s00125-022-05752-z) contains peer-reviewed but unedited supplementary material.


Procedures
A 3.0 T clinical MRI scanner (Achieva Philips Healthcare, Best, the Netherlands) was used to perform the measurements of hepatic glycogen and hepatic lipid content. Glycogen was measured using 13 C-MRS with a dedicated 13 C / 1 H coil and the volunteer lying in prone position. Power settings were calibrated to achieve a 90-degree pulse in the liver (at 8 cm from the coil) and spectra were acquired without 1H-decoupling. The area under the curve of the glycogen doublet at 100.5 ppm was determined using MATLAB and phantom-based sensitivity maps and MRI image segmentation were used to correct for coil sensitivity in the liver area. For hepatic lipid content, volunteers were positioned in the supine position and a volume of interest was selected within the right lobe of the liver to acquire 1 H-MRS spectra using STEAM (TE: 20 ms, TR: 4500ms, number of averages: 128) [1]. Volunteers were asked to breathe in a rhythm to prevent motion artefacts. The water signal was suppressed by VAPOR water suppression. A spectrum without water suppression was also acquired to quantify the water signal and the ratio of lipid over the sum of lipid and water was determined. From this, absolute values of fat percentage were deduced [2] and hepatic lipid content is given as weight/weight percentage. From the lipid spectrum, the relative contribution of saturated, monounsaturated and polyunsaturated fatty acids were determined according to Roumans et al [1].
To measure mitochondrial oxidative capacity, permeabilized muscle fibres were prepared freshly directly after the muscle biopsy as described previously [4]. Subsequently, the permeabilized muscle fibres (∼2.5 mg wet weight) were analysed for mitochondrial function using an oxygraph (OROBOROS Instruments, Innsbruck, Austria). To prevent oxygen limitation, the respiration chambers were hyper-oxygenated up to ∼400 μmol L -1 O2. Subsequently, two different multi-substrate/inhibition protocols were used in which substrates (malate, octanoyl-carnitine, glutamate, succinate) and other compounds (ADP, oligomycin, FCCP) were added consecutively at saturating concentrations to characterize mitochondrial capacity, as described previously [5]. Measurements were performed in quadruplicate and cytochrome c was added upon maximal coupled respiration (state 3) to assess mitochondrial membrane integrity. If oxygen consumption increased >15% after cytochrome c addition, that particular measurement was excluded from analysis.
The hyperinsulinemic-euglycemic two-step clamp was performed to measure insulin sensitivity, as described previously [3]. Briefly, the clamp started with 120 minutes of primedcontinuous infusion of D-[6,6-2 H2] glucose to determine baseline endogenous glucose production (EGP), glucose appearance (Ra) and glucose disposal (Rd). Afterwards, insulin was infused at 10 mU · m 2 · min -1 to assess hepatic insulin sensitivity reflected by suppression of EGP. After 3 h, insulin infusion was increased to 40 mU · m 2 · min -1 to measure muscle insulin sensitivity. Arterialized blood was drawn every 5-10 minutes to assess glucose levels and glucose (20%) was co-infused to maintain glucose levels at ~5 mmol/l. Energy expenditure and substrate utilization was measured during the last 30 min of every steadystate period (basal, low insulin and high insulin) using indirect calorimetry (Omnical; Maastricht Instruments, Maastricht, the Netherlands).

Biochemical analyses
The ABX Pentra C400 (Horiba, Montpellier, France) was used to enzymatically quantify triglycerides (Sigma, St Louis, USA), free fatty acids (FFAs) (Wako, Neuss, Germany) and glucose concentrations (Horiba, Montpellier, France) in EDTA plasma. Insulin levels were determined using enzyme-linked immunoassay in EDTA plasma (Crystal Chem Inc, Illinois, USA). Nitrogen was assessed in 24-hour urine samples using the Vario Max (Elementar Analysensysteme GmbH, Langenselbold, Germany). Samples from volunteers were analysed in the same run for both interventions.

Data analysis
For determination of glucose homeostasis, continuous glucose monitor data was obtained in the free-living situation of both TRE and CON. Data from the last 4 days (day 15 -18) was combined to account for day-to-day specific effects on glucose excursions. This continuous data was divided into categories defined by the American Diabetes Association [6,7]: hypoglycaemia < 4.0 mmol/l, low blood glucose 4.0 -4.3 mmol/l, normal range 4.4 -7.2 mmol/l, high blood glucose 7.3 -9.9 mmol/l, hyperglycaemia > 10.0 mmol/l. Results were reported as percentage of time spent in the respective categories and differences between TRE and CON were tested using multiple Wilcoxon signed rank tests with a Bonferroni correction.
Sleeping metabolic rate was defined as the lowest 3 h of nocturnal energy expenditure during the first night in the respiration chamber and calculated with the Weir equation [8]. Twenty-four-hour energy expenditure and -substrate utilization were calculated using the equation from Brouwer et al. [9] with data obtained from the last 24 h of the respiration chamber measurement, including 24-hour urine collection to determine protein oxidation. From the indirect calorimetry data collected during the clamp, carbohydrate-and fat oxidation were calculated using the Brouwer equation [9] with protein oxidation being estimated as 12.4% of energy expenditure. In addition, Steele's singe pool non-steady state equations were used to calculate Ra and Rd [10].
Sample size was determined based on the variability in glycogen content measured in the fasted state with 13 C-MRS, which amounted 7 -11% in previous studies [11,12]. For a more conservative estimation of the standard deviation, 11% was used in our calculation. The following equation was used for sample size calculation: N = σ 2 /∆μ 2 * (Z0.8 + Z0.975) 2 with Z0.8 = 0.842, Z0.975 = 1.960, σ 2 = 11%, and ∆μ 2 = 10%. Filling out this equation indicated that we needed to included 10 volunteers in our study. However, since the variation in hepatic glycogen in response to a time restricted eating regime was not investigated at the time of the calculation and might be greater, we decided to include 14 volunteers in our final data analyses. 16 Participation in another biomedical study within 1 month before the first study visit, which would possibly hamper our study results

17
Another medical condition that will preclude the safe performance of the measurements as judged by the medical doctor