Oral glucose tolerance test.

All participants will undergo frequent sample OGTT according to the technique developed by Matsuda & DeFronzo [32] and Abdul-Ghani et al. (33). This state-of-the-art OGTT was developed to quantify "tissue-specific" insulin resistance that is comparable, and less burdensome, to the hyperinsulinemic euglycemic clamp technique. The procedure entails collection of samples for fasting glucose and insulin at -30, -15, and 0 minutes followed by administration of 75 grams of glucose and collection of samples at 30, 60, 90, and 120 minutes. Plasma glucose will be quantified using a Beckman Glucose Analyzer and serum insulin will be measured using high-sensitivity sandwich enzyme-linked immunosorbent assay (ELISA).

Hepatic and muscle insulin resistance.

The glucose load administered in the OGTT is followed by a rise in plasma glucose concentrations, which, in turn, stimulates secretion of insulin by the pancreatic beta cells. In individuals with normal insulin sensitivity, insulin secretion suppresses hepatic glucose output in the first 30 minutes of the OGTT. This increase in insulin triggers glucose disposal into peripheral tissues, the predominant tissue being the skeletal muscle. Thus, the decline in glucose levels from peak value primarily reflects glucose uptake by the skeletal muscle.

In individuals with hepatic insulin resistance, insulin is unable to suppress hepatic glucose output efficiently. In such subjects, a higher peak of glucose is observed in the first 30 minutes of the OGTT (i.e., a sum of endogenous and exogenous glucose), and the magnitude of the rise in plasma glucose and insulin concentrations during this period is proportional to the degree of the hepatic insulin resistance. If skeletal muscle insulin sensitivity in these individuals is normal, glucose is normally disposed into peripheral tissues, and the glucose peak is followed by a steady decline. In contrast, in individuals with predominantly skeletal muscle insulin resistance, insulin secretion efficiently suppresses hepatic glucose output in the first 30 minutes of the OGTT, but peripheral disposal of glucose is impaired. Thus, these subjects display a slower decline in glucose levels from peak value (Fig 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Anticipated patterns of changes in plasma glucose concentrations according to the primary site of insulin resistance.

* *Based on information from Reference #33.

https://doi.org/10.1371/journal.pone.0281508.g003

In the present study, we will quantify hepatic insulin resistance for each subject using the product of the area-under-the-curve (AUC) for insulin and glucose concentrations during the first 30 minutes (0–30) of the OGTT, according to the equation proposed by Abdul-Ghani et al. [33]:

Skeletal muscle insulin resistance will be assessed based on the rate of decline of plasma glucose with time, calculated as the slope of the least square fit to the decline in plasma glucose concentration from peak to nadir (dG/dt) divided by the mean plasma insulin concentration (I) [33]:

For each of these calculations, the change from baseline to 12 and 24 weeks will be computed at the subject level, and analyses will be performed on the resulting change scores.

Magnetic resonance spectroscopy.

As a noninvasive imaging method for measuring concentrations of intracellular metabolites, MRS offers insight into the fat content in various tissues [34]. In the ADT & Metabolism Study, enrolled participants will undergo MRS for quantification of hepatic and skeletal muscle fat. The MRS will be performed using a 3-Tesla magnetic resonance imaging scanner (Skyra VE11C; Siemens AG, Erlangen, Germany). Hepatic and skeletal muscle MRI and MRS will be acquired with the following steps:

1. Hepatic MRS
2. Subjects will be imaged head first using a phased array torso coil centered at the liver. Plane breath held localizers will be followed by coronal and transverse T2 HAlf fourier Single-shot Turbo spin-Echo (HASTE) for localization of the MRS voxels. Two voxels will be acquired in the posterior and anterior portions of the liver avoiding vasculature and non-hepatic structures. A screenshot will be taken at the baseline scan, which will then be referenced in the follow-up scans to ensure reproducibility of the voxel location between the baseline and follow-up scan. MRS will be acquired using breath-held stimulated echo acquisition mode (STEAM) with an echo time (TE) of 20 ms, repetition time (TR) of 2 seconds, mixing time of 10 ms, 8 averages (16 seconds) with a voxel size of 2x2x2 cm³. Automated shimming will be used but if linewidths are greater than 30 hz, manual shimming will be conducted to minimize linewidths. Spectroscopy data will be exported offline in raw data mode (twix) for post-processing. Two areas of interest will be selected to quantitate hepatic fat. Intensities of the peaks resonating from the protons of hepatic water (4.7 ppm) and protons of methylene groups (-CH2-) in the fatty acid chains (1.3 ppm) will be determined (corrected for T1 and T2 relaxation), and liver fat content will be expressed as a ratio of the signal of the methylene group to the total signal of methylene plus water using LC model. Hepatic lipid content will be determined from the relative percentage of the spectral peak at 1.3 ppm (representing the sum of triglyceride acyl chains) over the spectral peak of protons from hepatic water at 4.7 ppm. To convert the relative quantity of hepatic fat and water into absolute concentrations, the method described by Szczepaniak et al. will be used, which has been widely accepted as the most accurate assessment of liver fat content [35].
3. Skeletal Muscle MRS
4. Muscle fat spectroscopy will be acquired in the same scanner following the hepatic MRS. The patient will lie supine, feet first, and the lower limb will be placed in a 15-channel transmit-receive knee coil. Special care will be taken to ensure that the subject’s leg is parallel with the bore for optimal separation of the intramyocellular lipid content (IMCL) and extramyocellular lipid content (EMCL) resonances. Three plane localizers will be acquired, followed by axial T1-weighted MRI with and without fat suppression. Proton MRS (1H-MRS) single voxel spectra using Point-RESolved Spectroscopy (PRESS) with a TR/TE = 2000/30 ms, 128 averages (4 min, 16 sec) will be acquired from the gastrocnemius and tibialis anterior muscles (1.5 x 1.5 x 2 cm³) for IMCL/EMCL analysis. Likewise, a screenshot will be taken for reference and to reproduce the same location. Scans will be processed using an automated time-domain-based spectral fitting software (LCModel v6.3, Provencher) which will output concentrations of IMCL, EMCL, and creatine.

Measurement of sex steroids, inflammatory cytokines, and adipokines.

Inflammatory cytokines and adipokines will be measured in serum using high-sensitivity sandwich ELISA. Total testosterone will be measured using liquid chromatography-tandem mass spectrometry (sensitivity of 2 ng/dL [0.0693 nmol/L]), and free testosterone will be measured with equilibrium dialysis. Sex hormone binding globulin will be measured by solid phase immunofluorometric assay. Except for sex steroids (which will be measured in real time), inflammatory cytokines and adipokines will be measured on frozen plasma and sera will be banked in a -80°C freezer.

Dual-energy X-ray absorptiometry.

Participants will undergo assessment of body composition using DEXA scans (Hologic QDR 4500A; Hologic, Inc., Bedford, Massachusetts, USA). Both lean body mass and fat mass (subcutaneous and visceral adipose tissue) will be measured. The scanner will be regularly calibrated using a soft-tissue phantom.

Physical activity and nutrition assessment.

The participants’ physical activity will be assessed using the validated Rapid Assessment of Physical Activity (RAPA) questionnaire, a one-page tool that has been used in several clinical trials [36, 37]. Nutrition intake will be assessed using the Harvard Willett Food Frequency Questionnaire, which has been extensively validated in many populations [38–40]. These assessments will be performed in both cohorts at baseline, 12, and 24 weeks.

Muscle performance and physical function.

Maximal voluntary leg extensor muscle strength will be assessed using the one-repetition maximum test during a seated leg-press exercise (Keiser Corporation, Fresno, CA) [41]. Leg extensor muscle power will be assessed using the same equipment and positioning used for the 1-RM assessment, as previously described [42]. Physical function will be assessed by (1) measurement of distance walked during the 6-minute walk test [43], (2) loaded stair climbing power test [44], and (3) self-reported physical function assessed by the physical functioning domain of the 36-item Short-Form Survey (PF-10).

Energy and sexual function.

The participant’s energy (vitality) and sexual desire will be assessed using the validated Hypogonadism Energy Diary (HED) questionnaire and the Sexual Arousal, Interest, and Drive (SAID) scale [45], respectively.

Quantification of primary outcome (Specific Aim 1).

Differentiation between insulin resistance occurring at the liver or skeletal muscle between baseline and 12- and 24 weeks will be obtained from the two-hour OGTT.

Among subjects exhibiting insulin resistance at the liver or skeletal muscle, circulating glucose (G) concentrations are expected to follow roughly the pattern displayed in Fig 3. For participants manifesting resistance predominantly at the liver, glucose concentrations are expected to peak at roughly 30 minutes and then decline steadily. Meanwhile, for participants expressing resistance predominantly at the muscle, concentrations are expected to decline less rapidly. Hepatic insulin resistance is quantified according to lack of suppression of hepatic glucose output in the initial period following start of OGTT. Therefore, following Abdul-Ghani et al. [33], we will quantify hepatic resistance using the product of area-under-the-curve (AUC) for insulin and glucose concentrations during the first 30 minutes of OGTT (glucose_0–30[AUC] X insulin_0–30[AUC]) for each subject. Level of skeletal muscle insulin resistance is in rough (negative) proportion to the rate of decline of plasma glucose as time (t) progresses, dG/dt, during the period between glucose peak until reaching its nadir. We will, therefore, quantify skeletal muscle insulin resistance using estimated dG/dt divided by the mean plasma insulin concentration (I). For each of these, change from baseline to 12 and 24 weeks will be computed at the subject level, and analyses will be performed on the resulting change scores.

Quantification of secondary outcomes (Specific Aim 2).

Skeletal muscle and hepatic fat concentrations will be assessed from MRS (baseline and 24 weeks). Concentrations of inflammatory cytokines and adipokines will be determined using the methodology described above. Body composition will be assessed using DEXA scan (baseline, 12, and 24 weeks).

Overall analytic approach.

Prior to formal analysis, variable distributions will be summarized by cohort (ADT and non-ADT groups). Means, standard deviations, ranges, medians, quartiles, and skewness will be determined. Graphical methods will be utilized to assess overall distributional properties. Scatterplot smoothing and exploratory assessments via generalized additive models will be employed to assess the functional form of associations between covariates and outcomes. All point estimates will be accompanied by 95% confidence intervals. Hypothesis tests will be evaluated at the 0.05 level. As hypotheses are prespecified, no adjustment will be made for multiple comparisons.

Aim-specific analytic plan.

Specific Aim 1. Specific Aim 1 is concerned with determining whether the predominant site of insulin resistance is at the level of the liver and at the skeletal muscle using, respectively, glucose_0–30[AUC] X insulin_0–30[AUC]) (denoted AUC hereafter) and dG/dt / I, as described above. These will be modeled separately. Each subject’s change in AUC, dG/dt and I will be computed, and change at 12 and 24 weeks for each obtained. Primary analyses will estimate change at 12 and 24 weeks simultaneously using a mixed-effects regression analysis with control for cohort membership [46, 47]. The differential increase in AUC with time accompanying ADT as compared to controls will be estimated using a treatment contrast and robust (employing sandwich variance estimators) 95% confidence interval, and quantification of statistical significance obtained using a likelihood ratio test. Analysis of dG/dt / I will proceed in like fashion.

Assuming stable cross-sectional standard deviation (SD) with time, the SD of change in outcomes is given by √2(1 − ρρ), where ρρ is the Pearson correlation between repeated outcome measures, which we anticipate will be high. Under the relatively conservative assumption that ρρ = 0.5, an evaluable sample size of 40 participants in the ADT group and 20 in the non-ADT group will provide 90% power to detect a mean between-group difference in AUC change and dG/dt / I change of 9.03 units and 1.35 units, respectively, and 80% power to detect differences of 7.8 units and 1.17 units, respectively, at either follow-up time. Under these assumptions, a total of 40 evaluable subjects would in turn provide adequate precision to estimate mean change in AUC and dG/dt / I to within 3.2 units and 0.47 units, respectively, using 95% confidence intervals. Based on an assumption of cumulative attrition and missingness of 20%, the trial will therefore enroll 50 participants in the ADT group and 25 in the non-ADT group, or N = 75 participants overall. The proposed sample size is intended to facilitate comparisons of the ADT to non-ADT groups in Specific Aims 1 and 2, as well as providing the necessary precision to estimate change in outcomes in analyses of the ADT group alone in Specific Aim 2.

Specific Aim 2. Specific Aim 2 is concerned with assessment of fat deposition and inflammatory markers.

Analyses of Hepatic and Intramyocellular Fat. These outcome measures will be obtained by MRS at baseline and 24 weeks. Change scores on each will be computed for each participant. In primary analyses, the mean within-subject change in fat measurements will be assessed, and between-group differences in change assessed using a point estimate and associated 95% confidence interval. In secondary analyses, the association between change and subject-level covariates including age, BMI, sex steroid levels, and lipids will be generated using a generalized additive model, which estimates trends and associated confidence regions using penalized likelihood. Where associations are sufficiently linear, multiple linear regression may be employed to generate parameters that are easily interpreted.

Analyses of Inflammatory Cytokines. As noted above, the proposed sample size of 50 participants in the ADT group (twice the size of the non-ADT group) is intended to facilitate evaluation of within-subject changes among participants exposed to ADT. Given projected attrition and missingness of 20%, we anticipate at least 40 evaluable change scores for any outcome and time point. In primary analyses, unadjusted changes in inflammatory cytokines will be estimated using a mixed-effects model as described above, with between-group differences estimated using a treatment contrast and robust confidence interval. Secondary analyses will estimate within-subject changes in each cohort and as a function of baseline age, BMI, and other covariates using generalized additive mixed models.

Additional analyses.

Analyses paralleling those described above will be employed to assess differences in change in measures of physical function and muscle performance. Analyses of 6-minute walk distance and average speed will assess between-group differences using mixed-effects regression models and propensity scoring as described above. We will perform similar analyses to assess between group differences in leg press strength and power. This study will also assess loaded stair climbing power.

Sensitivity analysis and missing data.

We anticipate that outcomes will exhibit reasonable conformity with assumptions, but will assess distributional properties, outlying observations, etc. as described above. In the case of outlying values, models will be re-estimated excluding these observations to ensure that qualitative conclusions are stable and robust. Where conformity with other modeling assumptions is threatened, robust alternatives such as analyses of rank-transformed change in outcome measures will be considered. We will model the missing data process as a repeated-measures binary outcome using the modified Poisson regression model [48].

Statistical computing.

Analyses will be performed using SAS version 9.4 (SAS Institute, Cary, NC) or R version 3.1 or later (R Foundation for Statistical Computing, Vienna). Simultaneous modeling of missing data and repeated-measures outcomes will be performed using MPlus version 7 or later (Muthén and Muthén, Los Angeles, CA).

Ethical considerations and data sharing.

Ethical approval has been obtained from the Institutional Review Board of Dana Farber Cancer Institute (Protocol # 18–442). Patients meeting the research criteria will be enrolled in the study after signing an informed consent. The study will be conducted in accordance with the principles of the Declaration of Helsinki and Good Clinical Practice. Following completion of the project, we will generate a data set suitable for data sharing purposes, which will be deidentified and stripped of protected health information according to the Health Insurance Portability and Accountability Act of 1996 (HIPAA) and Safe Harbor principles. The findings of the study will be published in peer-reviewed academic journals.

Data management plan.

The study will utilize a secure electronic data capture (EDC) system. The database will be hosted on secure, password-protected servers.

The database will be fully HIPAA and 21 CFR part 11 compliant, providing full security and audit capabilities consistent with federal requirements and guidelines.

Status and timeline of the study or current status.

Recruitment of participants commenced in February 2019, and data collection started in March 2019. The enrollment is now complete, and data collection is expected to be completed by April 2023. Data analysis will commence soon thereafter, and manuscripts will be published in 2023 and 2024.