Obese Older Type 2 Diabetes Mellitus Patients with Muscle Insulin Resistance Benefit from an Enriched Protein Drink during Combined Lifestyle Intervention: The PROBE Study

(1) Background: Recent research showed that subtypes of patients with type 2 diabetes may differ in response to lifestyle interventions based on their organ-specific insulin resistance (IR). (2) Methods: 123 Subjects with type 2 diabetes were randomized into 13-week lifestyle intervention, receiving either an enriched protein drink (protein+) or an isocaloric control drink (control). Before and after the intervention, anthropometrical and physiological data was collected. An oral glucose tolerance test was used to calculate indices representing organ insulin resistance (muscle, liver, and adipose tissue) and β-cell functioning. In 82 study-compliant subjects (per-protocol), we retrospectively examined the intervention effect in patients with muscle IR (MIR, n = 42) and without MIR (no-MIR, n = 40). (3) Results: Only in patients from the MIR subgroup that received protein+ drink, fasting plasma glucose and insulin, whole body, liver and adipose IR, and appendicular skeletal muscle mass improved versus control. Lifestyle intervention improved body weight and fat mass in both subgroups. Furthermore, for the MIR subgroup decreased systolic blood pressure and increased VO2peak and for the no-MIR subgroup, a decreased 2-h glucose concentration was found. (4) Conclusions: Enriched protein drink during combined lifestyle intervention seems to be especially effective on increasing muscle mass and improving insulin resistance in obese older, type 2 diabetes patients with muscle IR.


Introduction
A key therapeutic goal in the treatment of obese type 2 diabetes (mellitus) patients is weight loss. Several studies using energy restriction and weight loss interventions showed beneficial effects on the metabolic state, including reduced HbA1c and lipoprotein levels, reduced hepatic and pancreatic

Study Design
The study was a randomized, controlled, double blind 13-wk trial [14]. Subjects were randomized into either the test group or control group and followed a hypocaloric diet of 600 kcal below estimated energy needs (based on measured resting energy expenditure and estimated physical activity level using an accelerometer). Subjects received six individual dietary counseling sessions with a dietitian and six group sessions on alternating weeks together under supervision of a dietitian to facilitate a 600 kcal energy deficit with their food intake. The exercise program, consisted of three 1-h group training sessions per week together at a fitness center, and included resistance exercise and high intensity interval training (HIIT). The training sessions were supervised by a sports instructor. During the combined lifestyle intervention, the test group received a study drink 10 times a week: one serving daily with breakfast and one extra after each training session. Data of subjects was obtained at baseline and after 13 weeks of intervention.

Study Products
In the study, drinks were provided by Danone Nutricia Research, Utrecht, The Netherlands. The test drink was a whey protein drink enriched with leucine and vitamin D, containing 21 g leucine-enriched whey protein (3 g total leucine), 9 g carbohydrates, 3 g fat, 800 IU cholecalciferol (vitamin D3), and a mixture of vitamins, minerals, and fibers, referred to as protein + (FortiFit ® ). The control product contained 25 g carbohydrates and 6 g fat. Both products were similar in taste and appearance, isocaloric (150 kcal per serving), and were dissolved in 150 mL water just before consumption. The supplement was consumed daily and after each training session, so a total of ten consumptions per week. On average the protein intake for the protein + users therefore increased with 28 g a day, or 112 kCal, representing about 6% of the daily energy intake. A detailed description of the study products is provided in Appendix A.

Clinical Chemistry and Glycemic Control
To standardize the OGTTs subjects were instructed before the OGTTs: not to participate in any vigorous exercise 48 h; to refrain from alcohol for 24 h; have a standardized dinner the evening; not to eat, drink and smoke for 12 h; drinking some water was allowed till two hours before the OGTT. Upon overnight fasting an OGTT was performed at baseline and after 13 weeks of intervention. Venous blood samples were taken prior to intake (t = 0), and at 30, 60, 90 and 120 min after intake of a 300 mL Nutrients 2020, 12, 2979 4 of 16 75 g glucose solution (Added Pharma). Plasma heparin samples were used for measurements of glucose (mmol/L) and insulin (pmol/L). Plasma samples taken prior to OGTT were also assessed for HbAc1 (mmol/mol) and plasma lipids (FFA, triglycerides, High-Density Lipoprotein, Low-Density Lipoprotein, and total cholesterol) and serum calcidiol (nmol/L) was assessed according to standard procedures in the laboratory of the Vrije Universiteit Medical Centre, Amsterdam, The Netherlands. Leucine analysis was based on the AccQTag method of Waters measured with HP-LC from a fasting dry blood spot sample at TNO, Zeist, The Netherlands.

Subtyping of Type 2 Diabetes
To define if a subject had muscle IR as the diabetic subtype, the muscle insulin sensitivity index (MISI) [18] was calculated from the OGTT. Additionally, the Matsuda index [19], hepatic insulin resistance index (HIRI) [18], disposition index (DI) [20], and the adipose tissue insulin resistance index (Adipo-IR) [21] were also calculated as measures for respectively systemic insulin sensitivity, hepatic insulin resistance, β-cell function (BCF), and adipose tissue insulin resistance (Table 1). Units used for calculations of the indices were in mg/dL for glucose and U/L for insulin. Using the data of DiOGenes [22], CORDIOPREV [23], and PhenFlex [24,25] studies, cut-off values for MISI were determined. These values were calculated and validated by using data of more than 1000 subjects with different health status: healthy subjects, subjects with prediabetes, and patients with undiagnosed and clinically diagnosed type 2 diabetes. The cut-off used for MISI was −1; subjects with values ≤ −1 had no MIR, subjects with values > −1 had MIR. In the present analysis, we calculated the MISI of all subjects at baseline and categorized them based on their MISI value in the MIR or no-MIR group, respectively. Abbreviations used: fG = fasting plasma glucose; fI = fasting plasma insulin; mG = mean plasma glucose; mI = mean plasma insulin; AUC = area under the curve; ∆G = delta glucose; ∆t = delta time; FFA = free fatty acids. Units used: for glucose mg/dL; for insulin U/L; for FFA mmol/L.

Physical Measurements
Body weight was measured using a weighing scale (Life Measurement, Inc., Concord, CA, USA), body height using a wall mounted stadiometer (De Grood DGI 250D; De Grood metaaltechniek, Nijmegen, The Netherlands), and waist circumference using measuring tape. Body weight and height were used to calculate the body mass index (BMI). Appendicular skeletal muscle mass (ASMM) (kg), fat mass (kg), and visceral adipose tissue (VAT; cm 2 ) were measured by dual-energy X-ray absorptiometry (DXA; Discovery A, Hologic), using a whole body scan.
Three exercise tests were conducted to measure physical performance. The 10-RM leg press strength (kg) was measured using a leg press machine (Technogym Selection; Technogym, The Netherlands) Knee extension power (Watt) was measured using a leg extension machine (Technogym Selection; Technogym (Humac 360; CSMi) Technogym, The Netherlands) [26]. VO 2 peak (L/min) was measured using a steep ramp test performed on a cycle ergometer (Quark RMR/CPET; Cosmed, Rome, Italy) [27].

Statistics
The number of subjects in the randomized clinical trial (RCT) was based on the study of Verreijen et al. [13] evaluating the effect of the same test drink during intentional weight loss in older adults with obesity. A sample size of 44 per arm provided 80% power to detect an absolute difference of 0.92 kg leg muscle mass with a standard deviations (SD) of 1.51 kg and p < 0.05 (2-sided). Assuming a Nutrients 2020, 12, 2979 5 of 16 dropout rate of 25%, 118 subjects were needed for the study. Since subjects were enrolled in 5 different clusters, we aimed to enroll approximately 24 subjects per cluster. Finally, 123 subjects were enrolled in the study [14].
The data from the PROBE study were analyzed retrospectively, comparing the effect of test versus control in the MIR subgroup and no-MIR subgroup separately. To ensure high quality data, of the 123 subjects only those subjects were selected who consumed at least 7 out of 10 weekly study drinks (test or control), attended at least 2 out of 3 weekly training sessions (94 subjects), and completed the OGTT at start and end of the study (82 subjects) (per protocol analysis).
Observations having a residual >3 times the root mean square error of the model were removed as statistical outlier. Before statistical analysis, variables visually showing increased residual variation with higher fitted values were transformed by taking their natural logarithm, while the MISI was logarithmically transformed after substituting the '0' values by −0.1 and multiplying all values by −1 for proper statistical evaluation (due to non-linear distribution of the data log-transformation was needed).
Mixed linear models were used to analyze each dependent outcome variable with 'treatment' (test vs. control) and 'time' (0 vs. 13 weeks) and their interaction as fixed independent effects, and 'subject' as random independent factor. This was performed for the MIR and the no-MIR subgroup separately, since baseline characteristics of the two subgroups were different for their diabetic status ( Table 2) not allowing for a direct comparison between subgroups. No other subgroups were examined because the numbers of patients was unbalanced and too low. Interaction effects of time x treatment with p < 0.05 were further evaluated in a post hoc analysis to investigate whether a 'time' effect occurred in the test, the control group, or both. A post-hoc (two-sided student t-test after Bonferroni multiple comparison correction) effect with p < 0.05 was considered significant. Mean and SD were calculated for all outcomes. Statistical analyses were performed using R version 3.6.0 (www.r-project.org).

Data and Resource Availability
The data sets generated during and/or analysed during the current study are available from the corresponding author upon reasonable request.

Study Population
Eighty-two out of 123 subjects, who were compliant with respect to the PROBE study protocol and had complete OGTT data at baseline (week 0) and at the end of the intervention (week 13), were used for the retrospective type 2 diabetes subgroup analysis. The descriptive characteristics as presented in Table 2 were similar for the 82 subjects included as for the 41 excluded subjects of this retrospective analysis (data not shown). The subgroups MIR and no-MIR had similar baseline characteristics (see Table 2).
Based on the MISI, 42 subjects were found to have muscle IR (MIR) against 40 subjects that did not (no-MIR). Table 2 shows some descriptive characteristics of the MIR and no-MIR subgroups, specified by protein + and control group. The protein + and control groups were comparable in terms of baseline characteristics within MIR as well as in no-MIR (NS). Although equal numbers of participants were found for MIR (n = 42) and no-MIR (n = 40), some main characteristics were different between the two subgroups. The MIR group has more males (75 vs. 55%), more prediabetics (26 vs. 5%) and related to that a lower diabetes duration (average median of 45 months vs. average median of 100 months) as compared to the no-MIR subgroup. Table 3 presents the intervention effect of the MIR subgroup. A significantly greater change over time was seen for the protein + drink than for the control drink on Calcidiol (p < 0.001), fasting plasma glucose (FPG) (p = 0.020), fasting plasma insulin (FPI) (p = 0.029), Matsuda index (p = 0.002), HIRI (p = 0.019), Adipo-IR (p = 0.043), and ASMM (p = 0.043). Post-hoc analysis revealed that these effects were all explained by a significant improvement in the protein + group alone: Calcidiol (+17.5 nmol/L,   Table 4 presents the intervention effect of the no-MIR subgroup. A significantly greater change over time was seen for the protein + drink on calcidiol (p = 0.001), HbA1c (p = 0.024), DI (p = 0.045), and 10-RM leg press (p = 0.026). Post-hoc evaluation showed increased calcidiol levels in the protein + group (+19.1 nmol/L, p = 0.002) and improvements in both intervention groups, but stronger in the control group: HbA1c (control: −7.69 mmol/mol, p < 0.001; test: −2.92 mmol/mol, p = 0.040), was reduced after 13 weeks of intervention, while DI (control: +0.12, p = 0.024; test: −0.01, NS) Nutrients 2020, 12, 2979 7 of 16 and 10-RM leg press (control: +63.1, p < 0.001; test: +36.9, p < 0.001) were increased after 13 weeks of intervention.  In Figure 1 the changes of the glycemic and physiological parameters are presented for the test and the control group of the MIR and the no-MIR group. In Figure 2 the changes in the IR indices and 10-RM are shown for the test and control group for the MIR and the no-MIR group.
Nutrients 2020, 12, x FOR PEER REVIEW 9 of 18 In Figure 1 the changes of the glycemic and physiological parameters are presented for the test and the control group of the MIR and the no-MIR group. In Figure 2 the changes in the IR indices and 10-RM are shown for the test and control group for the MIR and the no-MIR group.  So besides similar effects of lifestyle on both subgroups, the systolic Blood Pressure (BP), HbA1c, 10-RM leg press and VO 2 peak were specific lifestyle effects found for the MIR subgroup. In the no-MIR group, the HbA1c and 10-RM leg press lifestyle effectiveness could not be determined due to the interaction effect with the protein + drink. A specific lifestyle effect only seen in the no-MIR subgroup was the 2hr plasma glucose level, which was reduced. The lifestyle specific effects for the MIR and no-MIR subgroups on 2-h glucose, systolic blood pressure and VO 2 peak are shown in Figure 3. We were also interested to evaluate the effect of the lifestyle intervention itself within the two subgroups. The effect of diet and exercise for both subgroups is presented in Tables 3 and 4 as 'Time effect'. In the MIR subgroup (Table 3), a significant time effect was found for body weight (−2.7 kg, p < 0.001), waist circumference (−3.0 cm, p < 0.001), BMI (−0.9 kg/m 2 , p < 0.001), systolic blood pressure (−8 mmHg, p = 0.05), HbA1c (−3.0 mmol/mol, p = 0.04), fat mass (−2.8 kg, p < 0.001), VAT (−22 cm 2 , p < 0.01), 10-RM leg press (+50 kg, p < 0.001), knee extension power (+20 Watt, p = 0.01), and VO2peak (+0.19 L/min, p < 0.001). All these parameters improved after 13 weeks of combined lifestyle intervention within the MIR subgroup, independent of the study drink (Table 3). Plasma cholesterol and triglyceride concentrations, Disposition Index (DI) and MISI did not change over time.
In the no-MIR subgroup, presented in Table 4, a significant time effect was found for body weight (−2.7 kg, p < 0.001), waist circumference (−3.6 cm, p < 0.001), BMI (−0.9 kg/m 2 , p < 0.001), 10-RM leg press and VO2peak were specific lifestyle effects found for the MIR subgroup. In the no-MIR group, the HbA1c and 10-RM leg press lifestyle effectiveness could not be determined due to the interaction effect with the protein + drink. A specific lifestyle effect only seen in the no-MIR subgroup was the 2hr plasma glucose level, which was reduced. The lifestyle specific effects for the MIR and no-MIR subgroups on 2-h glucose, systolic blood pressure and VO2peak are shown in Figure  3.

Discussion
The present study indicates that obese, older type 2 diabetes patients with muscle IR who use the protein drink enriched with leucine and vitamin D during a combined lifestyle intervention, show benefits on appendicular skeletal muscle mass, fasting plasma glucose, fasting plasma insulin, systemic insulin sensitivity and adipose tissue insulin resistance.
Earlier publications showed specific effects of interventions on type 2 diabetes subgroups [17,28]. We hypothesized that especially subjects with muscle insulin resistance may benefit from additional whey protein intake as provided by the protein drink, since exercise specifically improves insulin sensitivity of the muscle [29,30] and the protein + drink may increase the muscle glucose uptake capacity via muscle mass increase. Also of vitamin D beneficial effects for muscle growth and function have been reported [31]. The enriched protein supplement clearly resulted in increased Calcidiol levels, as a marker of vitamin D, after 13 weeks of intake in both subgroups. The role of vitamin D on insulin resistance has been reported, but recent reviews stressed that the relation is still not clear [32] and that there is no evidence that vitamin D supplementation has a beneficial effect on peripheral insulin sensitivity in people with or at risk of insulin resistance [33], as is stressed by our results as well. Mechanisms to elucidate these test product-driven beneficial effects specifically in the MIR subgroup, may therefore be related to the protein + intake. The increased appendicular muscle mass of the MIR subgroup that consumed the protein + drink suggests that the anabolic process is stimulated in this subgroup. Insulin not only stimulates glucose uptake in muscle, it is also an anabolic hormone, stimulating uptake of amino acids for muscle growth [34][35][36][37]. The MIR subgroup supplemented with the protein + drink (whey protein) is becoming more insulin sensitive (tendency p = 0.08, see Table 3). The availability of the proper amino acids, as well as the increased training stimuli and the tendency for improved insulin sensitivity, are all factors favoring muscle growth [35]. The muscle IR of the no-MIR group did not improve at all and was already in the healthy range what may explain that no significant muscle growth was seen in the no-MIR group using protein + drink.
Hepatic insulin sensitivity also improved significantly upon consumption of the protein + drink in the MIR subgroup, suggesting that the relative increase in dietary protein and maybe the relative decrease in carbohydrate intake, also plays a role in hepatic insulin sensitivity. The lower relative carbohydrate content of subjects' diets in the protein + group as compared to control (daily about 20 g of carbohydrates less in the protein + group) may have contributed to the reduction in fasting plasma glucose and insulin, adipose tissue insulin resistance, and the increase in systemic insulin sensitivity [38,39]. While we observed these beneficial effects of the protein + drink compared to the control drink in the MIR subgroup, we did not observe them in the no-MIR subgroup, suggesting that these protein + drink effects may be MIR subgroup-specific. Whether the results found are a consequence of increased protein or decreased carbohydrate intake needs further investigation. In all groups the HbA1c level decreased during the intervention and a clear decline in the HbA1C level in the no-MIR subgroup on the protein + and the control product was observed, whereas the decline in the no-MIR subgroup on control product was more pronounced and significantly different from the protein + group. Also, the Disposition Index was significantly different between protein + and control of the no-MIR group, where Disposition Index changed in the beneficial direction for control. Taken together, the clear beneficial effects of the protein + drink in the MIR group were not observed in the no-MIR group. We cannot explain these findings and, therefore, this subgroup analysis needs to be confirmed and conducted in future research focusing on efficacy of protein supplementation in obese older type 2 diabetes subjects.
In the present study, beneficial effects of the combined lifestyle intervention itself were seen in both the MIR and the no-MIR subgroup, with clinically relevant improvements in body weight, waist circumference, BMI, fat mass, visceral fat, hyperglycemia and knee extension power. Such effects are commonly reported for combined lifestyle interventions [40,41], with the exercise component being important to sustain health improvements obtained [42]. Remarkably, systolic blood pressure (SBP) improved (from 144 to 137 mm Hg, p = 0.02) due to the combined lifestyle intervention in the MIR subgroup only, independent of treatment. It is known that exercise training reduces SBP [43], probably due to improved endothelial function [44]. Our observation that this was only present in the MIR and not in the no-MIR subgroup suggests that MIR subgroup-specific effects of the lifestyle intervention occurred, as was also observed for the effects on VO 2 peak. A possible underlying explanation could be improved vasodilation and improved muscle perfusion in MIR subgroup only [45,46]. Due to the combined lifestyle intervention muscle blood flow may increase and result in the increased VO 2 peak. An increased blood flow by improved vasodilation could also explain the health benefits observed in the MIR subgroup when supplemented with a protein + drink, like the increased appendicular skeletal muscle mass.
The muscle insulin resistance we focused on in the present study is suggested to be related to mitochondrial dysfunction [47,48]. It is known that exercise can improve mitochondrial oxidative capacity and has the potential to improve insulin sensitivity of the muscle [49]. In the MIR subgroup the combined lifestyle intervention had the tendency to improve muscle insulin sensitivity independent of the study drink (p = 0.08). It is suggested that disturbed muscle fat oxidative capacity is important for the development of muscle insulin resistance. This impaired oxidative capacity of the mitochondria in the muscle results in accumulation of intermediates of lipolysis, which could, in turn, interfere with insulin signaling and lead to inhibition of insulin-mediated glucose uptake [28,50]. Multiple physiological aspects of fat metabolism in muscle like ectopic fat storage, skeletal muscle fatty acid handling, and lipotoxicity may play a role in muscle insulin resistance. Scientific evidence is available, showing that disturbed fat metabolism in muscle is related to insulin resistance [50][51][52][53][54][55]. Whether this is the cause or consequence of diabetes development is still under debate.
In the present retrospective subgroup analysis we had a balanced distribution of MIR and no-MIR subjects, with protein + and control (20-22 subjects), enabling us to perform the statistical subgroup analyses. The subgroups appeared to be less balanced when examining possible confounding variables like type 2 diabetes status (pre-diabetic or type 2 diabetes) and duration of type 2 diabetes (see Table 2). It was hypothesized that the MIR subgroup would be representative of type 2 diabetes patients with the most severe diabetes status. However, more pre-diabetics were present in the MIR (11 subjects) versus the no-MIR group (2 subjects). This could also partially explain the shorter duration of diabetes in the MIR subgroup (around 45 months for the MIR group and around 100 months in no-MIR group). This may have confounded the MIR versus no-MIR subtype intervention effects, which may relate to the progression of the disease rather than to the subtype itself. Patients with a shorter duration of type 2 diabetes may be more responsive to the combined lifestyle intervention and protein + drink, which has been observed for type 2 diabetes remission with lifestyle [56]. In our study, the more severe type 2 diabetes patients were more often characterized as no-MIR. At this stage, it is unclear how type 2 diabetes disease progression is related to muscle insulin resistance and this requires further research.
The diabetes subgroups shown in the present study, were based upon the calculation of the muscle insulin sensitivity index. The use of organ-specific IR indices for subgroup analysis has been done by others as well [52][53][54]57]. They stress, like we do, that this in-depth phenotyping of subjects provides more insight into disturbed underlying disease mechanisms. It is furthermore suggested that the use of organ-specific IR indices may be interesting and useful in personalization of type 2 diabetes treatment, and treatment effect evaluation [52][53][54].
In our study, the following strengths and limitations are present. A strength of the present study is the use of data of a well-performed randomized-controlled, double-blind trial, the PROBE study [14]. The number of subjects present in our subgroups was nicely balanced and enabled the within-group comparison. Limitations of the study can also be identified. The small group sizes and relatively large standard deviations for some parameters may have limited the ability to obtain statistical significance. Multiple testing correction was not applied because of the exploratory nature of this work. The subgroup analysis performed was done retrospectively, thus could not be sufficiently powered and balanced by rules of controlled trials and a risk of type II errors with the statistical analysis is present. Follow-up studies are needed to confirm the promising effect of the enriched protein supplement in combination with lifestyle intervention for subgroups of type 2 diabetic patients.

Conclusions
The results of our retrospective study indicate that especially obese, older type 2 diabetes patients with muscle IR may benefit from the use of a vitamin D and leucine enriched whey protein drink during combined lifestyle intervention, by showing improved systemic insulin sensitivity, adipose tissue insulin resistance, and appendicular muscle mass. The combined lifestyle intervention itself was also more effective in the MIR subgroup. Awaiting confirmation by future controlled trials, this study indicates subgroup-specific lifestyle intervention responses and reiterates the potential for a personalized lifestyle and diet approach to support patients with type 2 diabetes.
Author Contributions: W.J.P., R.G.M., J.d.V.-V.d.B., P.J.M.W. and S.W. contributed to the conception and design of the study. R.G.M. was responsible for data collection. W.J.v.d.B. was responsible for data analysis. W.J.P., R.G.M. and S.W. were responsible for interpretation of the data. W.J.P. 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. M.P.V.B. and W.J.P. drafted the manuscript. W.J.P., R.G.M., J.d.V.-V.d.B., P.J.M.W. and S.W. contributed to reviewing and editing the manuscript. W.J.P., R.G.M. and S.W. gave approval of the final version of the manuscript and take responsibility for the integrity of the data and the accuracy of the data analysis. All authors have read and agreed to the published version of the manuscript.
Funding: This study was supported by a grant from Topsector Agri & Food, The Netherlands. RGM received a PhD grant from the Dutch Research Council (NWO). Danone Nutricia Research provided study products and financial support for conducting the study.

Acknowledgments:
We would like to thank the interns Mike Vosskühler and Maarten Costerus for their analysis of the OGTT response data.