This study is part of the FluvaBAT trial (clinicaltrials.gov ID NCT03189511) which prospectively investigated the effect of fluvastatin on human BAT (10). The Ethics Committee for Northwestern and Central Switzerland (EKNZ, Number 2017-00455) approved the study and all participants provided written informed consent. Healthy, male volunteers were screened for presence of significant CIT and 17 participants were enrolled in the trial. Sixteen participants completed the trial. One participant was excluded from the analysis after showing no significant BAT activity in the PET. An overview of the trial flow is provided in Figure 1A.

The study was designed as an open-label before-and-after trial assessing the effect of 40 mg of fluvastatin twice daily on glucose tolerance, CIT and BAT activity. Participants attended a screening visit and two study visits, each before and after two weeks of treatment. At the screening visit CIT was assessed and only participants with a CIT > 5% were enrolled into the study. Within two weeks of the screening visit, an oral glucose tolerance test (oGTT) was performed and on the following day BAT activity was assessed by [¹⁸F]-FDG-PET/MRI after cold-stimulation. Thereafter, the participants took fluvastatin for 14 days and identical visits were performed at the end of the treatment period.

Participants arrived fasted (no caloric intake for at least six hours) in the morning of the study day. They were asked to refrain from strenuous exercise for at least 24 hours prior to the study visit. After an initial resting phase of at least 15 minutes an indirect calorimetry was performed with the participant lying on a hospital bed. After a baseline blood sampling oGTT was performed: All participants drank 75 g of glucose dissolved in 250 ml tap water. Blood was sampled at 30, 60, 90 and 120 minutes after ingestion of glucose. Note, that for the first seven patients’ blood was only sampled at 0, 60 and 120 minutes. EE was measured three times during visit 1 and visit 3: just before, as well as 30 to 60 and 90 to 120 minutes after the intake of 75 g glucose, by indirect calorimetry (Figure 1B).

We determined plasma glucose, sodium, potassium, kidney and liver parameters, TSH, fT4 and vitamin D at the screening visit to rule out metabolic disorders and to check the kidney function in order to ensure the safety of the study participants. Total cholesterol, LDL and triglycerides were measured before and after two weeks of fluvastatin in order to monitor compliance with the study drug. Glucose and insulin values were determined during the oGTT at the time points specified above. All routine analysis were carried out at the central laboratory of the University Hospital Basel. Cholesterol, HDL, triglycerides and glucose levels were measured using enzymatic methods on a COBAS 8800 analyzer (Roche Diagnostics, Mannheim, Germany). LDL levels were calculated using the Friedewald formula. Insulin levels were measured using an immuno-assay (ECLIA, Roche Diagnostics).

Fibroblast growth factor 21 (FGF21) levels were measured by enzyme-linked immunosorbent assay with a sensitivity of 8.69 pg/mL and an intra-assay coefficient of variation (CV) of 2.9 to 3.9% and an inter-assay CV of 5.2 to 10.9% (range given in product data sheet, Human FGF-21 Quantikine ELISA Kit, R&D Systems, Minneapolis, MN). In all 16 participants who completed the study, samples were taken at 0 min and 120 min after ingestion of 75 g glucose.

Indirect calorimetry was performed over a period of 30 minutes with the participants resting comfortably in a supine position on a hospital bed using a Quark RMR (Cosmed, Rome, Italy). All measurements took place in an air-conditioned examination room at a constant temperature of 24°C. The subjects were asked to fast for at least 6 h before the visit, and to abstain from exercise 24 h before, as these factors are known to affect the sensitivity of the measurement system. The sensitivity of an indirect calorimeter to detect significant changes in energy expenditure is approximately 1.5% (23). The calorimeter sensors were calibrated prior to each study visit and the system has previously been shown to perform well in comparison to “gold-standard” calorimeters (24).

CIT was calculated as the difference between resting energy expenditure (REE) at warm ambient temperatures and EE after a mild cold stimulus of two hours duration, as described previously (25). To determine CIT at the screening visit, participants were resting comfortably on a hospital bed, wearing t-shirts and shorts and were covered with a fleece blanket. After baseline indirect calorimetry and measurement of temperature, the blanket was removed and a water-circulating cooling system (Hilotherm Clinic, Hilotherm GmbH, Germany) was placed around the waist. Every 2 min the water temperature was reduced by 1°C, starting from 25°C to a minimum of 10°C. Cooling time was 120 minutes and care was taken to ensure that subjects did not shiver. When shivering occurred, the water temperature was raised by 2°C and the participant was covered with a fleece blanket until the shivering stopped. During the last 30 min of cold exposure, a second measurement of EE was performed and temperature was taken again.

As the β3-adrenoreceptor agonist mirabegron had been recently demonstrated to activate human BAT (26), we decided to combine cold exposure and β3 adrenergic stimulation as to assess the maximum metabolic activity of BAT. Subjects arrived in a fasted state in the morning of the study day. After the intake of 4 x 50 mg Mirabegron, the participants were instructed to wait for 1.5 hours, followed by 2 hours of standardized cooling as described above. Before start of the dynamic [¹⁸F]-FDG-PET/MRI Scan, 75MBq [¹⁸F]-FDG was injected through an intravenous line. Total scan time was 40min with a partial body scan from head to the upper abdomen after 30 minutes (Figure 1C).

For the quantification of glucose uptake into BAT, we used the 3D Slicer software, version 4.11.20 (National Institutes of Health, Bethesda, MD). In order to reliably segment supraclavicular adipose tissue (AT) for measurement of fat-fraction (FF) and glucose uptake, we used a step-wise segmentation approach (27). In brief, it consisted of a rough anatomic segmentation of the supraclavicular AT depot, which was further refined by thresholding based on the FF in the MRI volume and the [¹⁸F]-FDG uptake in the PET volume.

The suspected area containing the AT was outlined in the neck and shoulder region with the sphere brush tool of 3D slicer. In a next step, we used the FF volume and included all voxels with a FF of 400-1000‰ as further segmentation. The generated segments were intersected to create the volume of interest, MRI and PET volumes were co-registered (PET/MRI) and therefore the corresponding PET image could be overlaid.

Data analysis was performed using GraphPad Prism Version 9 (GraphPad Inc., La Jolla, CA). Pairwise comparisons were calculated, using paired t tests. Repeated measures were analyzed by two-way ANOVA. Correlations were first analyzed by simple linear regression. Mixed-effects models were built in R Version 4.1.0 (28) using the “nlme” library (29) to investigate co-variables influencing insulin sensitivity. As levels of FGF21 are usually not normally distributed, they were log transformed prior to statistical analysis.

We saw 16 healthy male volunteers both before and after administration of fluvastatin 2x40 mg b.i.d. for 14 days. Baseline characteristics of the participants are summarized in Table 1.

As expected, the concentration of low-density lipoprotein (LDL) and cholesterol dropped significantly after two weeks of statin intervention. Baseline total cholesterol levels fell from 3.96 mmol/l to 3.08 mmol/l (p <0.001) (Figure 2A). The concentration of LDL cholesterol decreased from 2.16 mmol/l to 1.35 mmol/l (p <0.001) (Figure 2B). The triglyceride concentration remained stable (Figure 2C) indicating that the study medication was taken as prescribed. The study medication was well tolerated by all study participants and no adverse events related to the study medication were observed.

In order to assess the effect of fluvastatin on glucose tolerance and insulin sensitivity, we performed an oGTT, both before and after two weeks of fluvastatin intake. The administration of fluvastatin had a significant effect on the blood glucose concentration during oGTT, as determined by area-under-the-curve (AUC). The AUC during oGTT increased from baseline 717 ± 125 to 796 ± 177 mmol/l x min (p=0.02). Mixed-effects analysis of the glucose concentrations showed a significant effect of time after ingestion of glucose (p<0.0001) and for fluvastatin (p=0.045). There was no significant interaction between time and fluvastatin (p=0.23) (Figure 3A). To corroborate our analysis, we also performed a two-way ANOVA, excluding subjects with missing values and obtaining virtually the same result. Effect of time on glucose concentration (p<0.0001) and administration of fluvastatin (p=0.036) were still significant, while the interaction between time and fluvastatin remained insignificant (p=0.13). The AUC for insulin was 4069 ± 1992 mU/l x min at baseline and increased to 4518 ± 2511 mU/l x min after two weeks of treatment, which was not statistically significant (p=0.26). In the mixed-effects model analysis, insulin concentration during oGTT showed a significant interaction between glucose administration and fluvastatin (p=0.04). The time after ingestion of glucose had a significant effect on insulin concentration (p<0.0001), whereas fluvastatin did not (p=0.22) (Figure 3B).

Fasting glucose levels remained unchanged at 4.7 mmol/l before and 4.8 mmol/l (p=0.73) after fluvastatin treatment. Fasting insulin levels also increased only insignificantly from 5.1 mmol/l at baseline to 5.5 mmol/l (p=0.51) after two weeks of fluvastatin intake.

Next, we calculated the Matsuda index, a parameter of insulin sensitivity derived from the oGTT data (30). The Matsuda index was lower after the intake of fluvastatin (p=0.09) (Figure 3C), but this was not statistically significant. HOMA-IR, a surrogate parameter of insulin-resistance and β-cell function, did not change significantly (p=0.65) (Figure 3D). The insulinogenic index (IGI), a frequently used index of β-cell function (31), did not change significantly (p=0.59) (Figure 3E).

We also measured levels of FGF21, which has been shown to be associated with insulin resistance (32) and BAT activity (33). FGF21 increased significantly after administration of 75g of glucose (p=0.01 for time), but fluvastatin intake did not significantly affect FGF21 levels (p=0.2) (Figure 3F).

As DIT contributes to the overall energy balance, we investigated whether fluvastatin affected the increase in EE after intake of glucose. EE increased from 5.55 ± 0.48 kJ/min to 6.09 ± 0.49 kJ/min (one hour) and to 5.68 ± 0.45 kJ/min (two hours post glucose uptake) before exposure to fluvastatin. After two weeks of treatment with fluvastatin EE at baseline was 5.49 ± 0.38 kJ/min and increased to 5.93 ± 0.50 kJ/min (1h) and 5.64 ± 0.47 kJ/min (2h). As expected, the effect of glucose was highly significant (p<0.0001) while fluvastatin had no significant influence on EE (p=0.57); there was no significant interaction between glucose administration and fluvastatin (p=0.21) (Figure 4A). In addition, the respiratory exchange ratio (RER) increased from 0.78 ± 0.06 to 0.84 ± 0.06 (1h) and 0.87 ± 0.07 (2h) before administration of fluvastatin. After treatment, RER was 0.81 ± 0.08 at baseline and increased to 0.86 ± 0.07 (1h) and 0.89 ± 0.08 (2h). As with EE, the effect of glucose administration was significant (p<0.0001) while fluvastatin did not affect RER (p=0.28) and did not interact with glucose administration (p=0.56) (Figure 4B). Multiple comparison revealed that the effect of glucose from baseline to 1h and baseline to 2h was significant (p<0.0001), while the effect of glucose on RER from 1h to 2h was not significant (p=0.18).

REE remained practically unchanged from 334 kJ/h ±29 kJ/h to 329 ± 23 kJ/h (p=0.44) (Figure 4C). Baseline RER was lower before than after fluvastatin, 0.79 and 0.81, respectively, but the difference was not statistically significant (p=0.21) (Figure 4D). DIT decreased from baseline 754 ± 363 kJ to 682 ± 297 kJ after exposure to fluvastatin, but the difference was not significant (p=0.27) (Figure 4E).

Interestingly, at baseline the Matsuda index and RER correlated inversely (R^{2 =} 0.44, p=0.005) (Figure 5A). After intake of fluvastatin the tendency was still present, but the correlation was no longer significant (R^{2 =} 0.14, p=0.16) (Figure 5B). HOMA-IR did not show any relation to RER before (R^{2 =} 0.11, p=0.22) and after (R^{2 =} 0.03, p=0.5) fluvastatin (Figures 5C, D).

As part of the main study protocol and based on the established BARCIST (Brown Adipose Reporting Criteria in Imaging Studies) criteria (34), we evaluated BAT activity after stimulation with cold and the β3-agonist mirabegron. The mean standardized uptake value (SUV_mean), volume of BAT, and total lesion glycolysis were used to evaluate BAT function in FDG-PET. Both SUV_mean of [¹⁸F]-FDG into BAT and volume of BAT did not differ significantly before and after treatment, 2.54 g/ml vs. 2.77 g/ml (p=0.12) (Figure 6A) and 80.9 cm³ vs. 75.8cm³ (p=0.49) (Figure 6B), respectively. Also, BAT total lesion glycolysis was similar 182 vs. 189 g (p=0.74) (Figure 6C). We next investigated whether parameters of insulin sensitivity were associated with BAT activity. There was no correlation between HOMA-IR and SUV_mean (Figures 6D, E). At baseline, BAT activity (SUV_mean) was inversely related to the Matsuda Index (R² = 0.44, p=0.005) (Figure 6F). After two weeks of treatment with fluvastatin however, this inverse correlation (SUV_mean/Matsuda Index) was no longer present (R² = 0.08, p=0.29) (Figure 6G).

As BAT is an important contributor to temperature homeostasis, we also measured the core temperature before and after cooling at each visit. During cooling, temperature dropped from 36.56 ± 0.29°C to 36.34 ± 0.21°C before fluvastatin administration and decreased also from 36.48 ± 0.27°C to 36.28 ± 0.2°C after two weeks of fluvastatin. Cooling had a significant effect on core temperature (p = 0.005), whereas the effect of fluvastatin was not significant (p=0.052). Fluvastatin and cold exposure did not interact (p=0.65).