Participant characteristics
A flowchart of participant enrolment is depicted in ESM Fig. 2. Baseline participant characteristics are presented in Table 1. The median Morningness-Eveningness Questionnaire Self-Assessment (MEQ-SA) score amounted to 59.5 (range 41–72). Only one volunteer was identified as an extreme morning type but was included in the study as the intervention did not interfere with his habitual day–night rhythm.
Table 1 Baseline characteristics of participants
Adherence
Volunteers did not indicate any changes in diabetes medication throughout the study. Volunteers recorded their daily food intake and sleep habits for, on average, 17 days during TRE and 18 days during CON. Based on these data, the eating window averaged 9.1±0.2 h in TRE vs 13.4±0.1 h in CON (p<0.01). Sleep–wake patterns were similar in both interventions, with a mean sleep duration of 8.1±0.2 h during TRE and 8.0±0.2 h during CON (p=0.17). Body weight at the start of each intervention was comparable between TRE and CON (89.1±3.7 vs 89.2±3.8 kg, respectively, p=0.62). Although volunteers were instructed to remain weight stable, a small but significant weight loss occurred in response to TRE (−1.0±0.3 kg, p<0.01) but not CON (−0.3±0.3 kg, p=0.22). The weight loss with TRE was significantly greater than the weight change observed with CON (p=0.02). Body composition determined on day 19 was comparable between TRE and CON (TRE vs CON: fat mass 37.4±2.7 vs 37.9±2.9 kg, p=0.58; and fat-free mass 50.7±2.6 vs 51.0±2.6 kg, p=0.60).
Hepatic glycogen and lipid content
Approximately half-way through each intervention period, hepatic glycogen levels were assessed in the morning following a 14 h (TRE) and 10 h (CON) night-time fast. Hepatic glycogen did not differ significantly between TRE vs CON (0.16±0.03 vs 0.17±0.02 arbitrary units [AU], respectively, p=0.43). At the end of each intervention, hepatic glycogen levels were also assessed after a standardised overnight fast of 11 h for both TRE and CON but did not reveal an altered hepatic glycogen content with TRE compared with CON (0.15±0.01 vs 0.15±0.01 AU, respectively, p=0.88). We also assessed hepatic lipid content; neither the amount of lipids nor the composition of the hepatic lipid pool was altered with TRE vs CON (respectively: total lipid content 9.0±2.0 vs 8.6±1.6%, p=0.47; polyunsaturated fatty acids 17.0±1.3 vs 16.2±1.2%, p=0.41; mono-unsaturated fatty acids 40.6±0.9 vs 42.9±1.4%, p=0.19; and saturated fatty acids 42.4±1.2 vs 40.9±1.5%, p=0.41).
Insulin sensitivity and glucose homeostasis
A hyperinsulinaemic–euglycaemic two-step clamp with a glucose tracer and indirect calorimetry was performed to assess insulin sensitivity. No differences in M value were found when comparing TRE and CON (19.6±1.8 vs 17.7±1.8 μmol kg−1 min−1, respectively, p=0.1). Hepatic insulin sensitivity was not affected by TRE, as exemplified by a similar endogenous glucose production (EGP) with TRE and CON in the fasted state and in the low- and high-insulin-stimulated states (p=0.83, p=0.38 and p=0.30, respectively; Fig. 1a). Suppression of EGP was also similar when comparing TRE with CON upon low- and high-insulin infusion (p=0.67 and p=0.47; Fig. 1a). NEFA suppression upon low insulin exposure was not different between TRE and CON (−365.2±41.6 vs −359.1±43.2 mmol/l, p=0.8). However, absolute levels of NEFAs were lower with TRE during the low- and high-insulin phase (p=0.02 and p=0.04; Fig. 1b), which may hint at an improved adipose tissue insulin sensitivity.
Peripheral insulin-stimulated glucose disposal, reflected by the change in rate of disappearance (Rd) from basal to high insulin, remained unchanged with TRE (p=0.25; Fig. 1c). However, we observed a larger insulin-stimulated non-oxidative glucose disposal (NOGD, difference from baseline to high insulin) with TRE than with CON (4.3±1.1 vs 1.5±1.7 μmol kg−1 min−1, respectively, p=0.04; Fig. 1d) reflecting an increased ability to form glycogen. Conversely, insulin-stimulated carbohydrate oxidation from basal to high insulin appeared to be lower with TRE than with CON (4.7±0.9 vs 6.2±0.9 μmol kg−1 min−1, respectively) but this difference was not statistically significant (p=0.07). Consistently, insulin-induced suppression of fat oxidation from basal to high-insulin was lower with TRE than with CON (−1.3±0.3 vs −1.8±0.2 μmol kg−1 min−1, p=0.04; Fig. 1e). Energy expenditure did not differ between TRE and CON during the basal, low-insulin and high-insulin phase of the clamp. These results indicate that while peripheral insulin sensitivity is unchanged with TRE, glucose uptake is more directed towards storage compared with oxidation. Both hepatic and peripheral insulin sensitivity, as well as levels of hepatic glycogen, were additionally analysed in volunteers who only used metformin as diabetes treatment (n=7) and this did not alter the outcomes.
To examine the effect of TRE on glucose homeostasis, CGM data from the last 4 days in the free-living situation (days 15–18) were analysed for both interventions. Four volunteers presented incomplete CGM data due to technical issues, hence statistics were performed on CGM data from ten volunteers. Mean 24 h glucose levels were lower in TRE compared with CON (6.8±0.2 vs 7.6±0.3 mmol/l, p<0.01; Fig. 2a–f). Nocturnal glucose levels were consistently lower in TRE vs CON (Fig. 2a–d). Furthermore, volunteers spent more time in the normal glucose range upon TRE compared with CON (15.1±0.8 vs 12.2±1.1 h per day, p=0.01; Fig. 2f). Concomitantly, time spent in the high glucose range was less in TRE compared with CON (5.5±0.5 vs 7.5 0.7 h per day, p=0.02) whereas no differences between eating regimens were found for time spent in hyperglycaemia (2.3±0.4 vs 3.7±0.8 h per day, p=0.24), time spent in the low glucose range (0.5±0.1 vs 0.4±0.1 h per day, p=1.00) or time spent in hypoglycaemia (0.7±0.3 vs 0.1±0.0 h per day, p=0.48).
Additionally, fasting plasma metabolites were assessed on day 20 and day 21 of each intervention. On day 20, blood samples were taken after a 10 h (CON) or 14 h (TRE) overnight fast. Plasma glucose on day 20 was lower after TRE (7.6±0.4 vs 8.6±0.4 mmol/l, respectively, p=0.03) whereas plasma insulin, triglycerides (TG), and NEFA levels were comparable between conditions (Table 2). On day 21, when overnight fasting time was similar for both interventions (11 h), plasma glucose levels remained lower in TRE than in CON (8.0±0.3 vs 8.9±0.5 mmol/l, respectively, p=0.04), whereas no differences were detected in plasma insulin, TG and NEFA levels (Table 2).
Table 2 Blood plasma biochemistry
Twenty-four-hour energy and substrate metabolism
On day 19, volunteers resided in a respiration chamber for 36 h for measurement of energy expenditure and substrate oxidation. Twenty-four-hour energy expenditure was similar for TRE and CON (9.57±0.22 vs 9.68±0.29 MJ/day, respectively, p=0.22; Fig. 3a), as was the 24 h respiratory exchange ratio (RER) (0.86±0.01 vs 0.86±0.01, respectively, p=0.13). Nonetheless, 24 h carbohydrate oxidation was lower in TRE vs CON (260.2±7.6 vs 277.8±10.7 g/day, respectively, p=0.04; Fig. 3b), whereas 24 h fat oxidation (91.9±6.6 vs 93.5±5.5 g/day, respectively, p=0.72; Fig. 3c) was unaffected. Twenty-four-hour protein oxidation seemed higher upon TRE but the difference did not reach statistical significance (72.8±7.2 vs 58.5±5.4 g/day, respectively, p=0.18; Fig. 3d). Sleeping metabolic rate appeared to be lower with TRE compared with CON (4.66±0.14 vs 4.77±0.18 kJ/min, respectively), although this decrease was not statistically significant (p=0.05; Fig. 3e). There was no change in carbohydrate or fat oxidation during sleep in response to TRE vs CON (RER 0.84±0.01 vs 0.84±0.01, p=0.50; Fig. 3f).
On day 21, muscle biopsies were obtained to assess ex vivo mitochondrial oxidative capacity by means of high-resolution respirometry. In total, paired biopsies from 13 out of 14 volunteers were analysed. Mitochondrial respiration did not differ between TRE and CON (Table 3).
Table 3 Mitochondrial oxidative capacity