This is a parallel group, randomized, double-blind, placebo controlled study conducted at University of Mary Hardin-Baylor, TX, USA. Subjects included 54 Caucasians (34 female/20 male), 13 Hispanics (8 female/5 male), 8 African American (4 female/4 male) and 2 Asians (1 female/1 male). Subjects were evaluated with respect to percent fat mass and body fat at baseline and immediately following a 12-week treatment period. Subjects were randomly assigned to receive either placebo (Corn starch,), 2 mg CAPs dosing [100 mg Capsimax providing 2 mg capsaicinoids] or 4 mg CAPs dosing (100 mg × 2 Capsimax providing 4 mg capsaicinoids). Thus, this study employed a pretest – posttest design with three between subject conditions created by random assignment. Baseline measurements were prior to dosing and therefore qualify as covariates free of conflation with treatment. All experimental protocols were approved by the University of Mary Hardin-Baylor Institutional Review Board prior to initiation of research activities (ISRCTN registry #10458693).
The subjects were recruited based on paper advertisements, via flyers, telephone (verbal guide), email, social media, and internet targeting research participants to determine their eligibility and interest. Subjects were recruited who exhibited the following study inclusion characteristics: 1) male or female volunteers ranging between 18 and 56 years of age; 2) healthy; 3) no ergogenic supplement ingestion in the last 6 months; 4) able to comply with required study activities; 5) expressing agreement to avoid strenuous activity 24–48 h prior to study visits; 6) expressing agreement to avoid smoking, caffeine use and tobacco use for 12-h prior to study visits; 6) exhibiting a BMI between of 24.5–29.5 kg/m2; and 7) able to provide a written and dated informed consent for study participation.
Subjects were excluded from the study on the basis of the following characteristics: 1) consumption of ergogenic levels of nutritional supplements that may affect muscle mass or aerobic capacity (e.g., creatine, HMB, etc) or anabolic/catabolic hormones (e.g., androstenedione, DHEA, etc.) within 6 months of study start; 2) presence of any absolute or relative contraindication regarding exercise testing or study prescription as outlined by the ACSM; 3) reporting of any unusual adverse events associated with the study that in consultation with the supervising physician would results in recommended study removal; 4) presence of strong history of food or drug allergy of any kind; 5) ingestion of any dietary supplement (excluding multivitamins) within 1 month of study start; 6) existence of any chronic disease and or condition(s) that the principal investigator believes may jeopardize the study; or 7) existing pregnancy prior to or during the study.
There were 28 placebo subjects, 27 subjects in CAPs 2 mg (low dose) treatment and 22 subjects in CAPs 4 mg (high dose) treatment group completed the study. Seventy five subjects completed the treatments. Inclusion and exclusion criteria were used to screen patients for study entrance but were not used for treatment assignment which was random.
Multiple study parameters were collected over various time points during the study. These variables included a diet log, laboratory values, cardio-metabolic parameters, body composition and anthropometric measurements, adverse events, and QoL (quality of life) indices.
In this study, dietary supplementation of Capsicum for 12 weeks has shown to promote appetite suppression, which translated to reduced self-reported caloric intake after 12 weeks of supplementation. While Capsicum administration resulted in improved body circumferences in a main effects analysis, it did not apparently affect DEXA fat mass or fat-free mass in a statistically significant way . The current study evaluates two measures of fat loss while considering six baseline variables related to fat loss. Baseline measurements of importance in this paper are those used to evaluate body habitus, diet and metabolic propensity. The following parameters at baseline were respectively used for these variable types: lean mass and fat mass (body habitus); protein intake, fat intake and carbohydrate intake (diet); and total serum cholesterol level (metabolic propensity).
It was determined that in the statistical modeling for fat loss in this study, body habitus would be captured as baseline lean mass and baseline fat mass; diet would be captured in three variables, namely, baseline protein intake, fat intake, and carbohydrate intake assessed through food frequency questionnaires; and metabolic tendency would be capture as baseline total serum cholesterol level. The author reasoned that if fat loss could be modeled using these variables along with treatment assignment, the model predicted outcome of weight loss as a function of treatment, while controlling for these background factors, would afford an assessment of the impact of CAPs on fat loss. Without taking background factors into account, evaluation of CAPs would be carried out under suboptimal conditions relative to the available sample size. This paper has both an empirical and a methodologic intent. Studies of economically feasible sizes with the goal of screening a panel of outcome variables for treatment signals are important venues of discovery. These formative studies narrow future investigative windows and generate data based hypotheses. As such, they are an important contribution to the scientific literature. When such studies address complex outcome variables (which often are the variables of greatest interest) such as fat loss, too often they are analyzed with statistical models best suited for large summative trials (e.g., Phase III trials). This paper illustrates the utility of a model based approach to discovery in studies of a moderate size. The value of statistical modeling that accounts for important concomitant factors in fat loss is illustrated in this paper by evaluating fat loss using a simple analysis of variance model without covariates or interactions and a generalized linear model that includes both of these features.
Capsicum extract is a Capsaicinoids enriched standardized product obtained from dried red fruits of Capsicum annuum L. The Capsicum extract is standardized into bead lets form (Capsimax) with food grade carbohydrates that is useful for food applications. Capsimax is a faint pinkish white colored free flowing uniform spheroidal bead lets with spicy odor, characteristics of dried ripe fruits of Capsicum. The product contains a minimum of 2% Capsaicinoids. The product is standardized to 2% Capsaicinoids, of which 1.2–1.35% is Capsaicin, 0.6–0.8% is dihydrocapsaicin, and 0.1–0.2% is nor-dihydrocapsaicin. The final product contains 15–25% extract from capsicum, 45–55% sucrose and 30–35% cellulose gum coatings.
Body composition and blood chemistries
Participants received a whole-body dual x-ray absorptiometry (DEXA) scan for body composition assessment at baseline and 12 weeks (Hologic Wi; Hologic Inc., Bedford, MA). Prior to their study evaluations, subjects fasted overnight. Participants had a venous blood drawn from their arm via standard phlebotomy techniques at baseline visit and 12 weeks. A panel of blood health markers (lipid profile, metabolic health markers and complete blood counts) was assessed by sending samples to a commercial laboratory (Quest Diagnostics, Irving, TX).
A power analysis was done on 25 subjects and 21 subjects per group yield a power of 0.85 and 0.81 in terms of body composition changes.
Prior to implementation of a statistical model to evaluate fat loss it was conceptualized that subject variation on three background dimension should be addressed during the evaluation process. These were baseline body habitus, baseline diet and baseline metabolic propensity. The objective was to parsimoniously capture these dimensions in as few variables as possible. Baseline lean mass and fat mass were used to capture body habitus; baseline carbohydrate intake, fat intake and protein intake were used to capture baseline diet; and baseline total cholesterol level was used to capture baseline metabolic propensity. Once selected, these baseline variables were used as independent variables in the statistical model used to evaluate fat loss. Percent body fat and fat mass were respectively re-expressed as percent change in body fat and change in fat mass by application of formula outcome = (12-week value – baseline value) / baseline value) × 100. Thus, percent change in body fat and fat mass served as dependent variables in the evaluation of CAPs.
After considering potential interactions among the baseline covariates and treatment, a comprehensive evaluation model was defined that expressed percent change in body fat (or fat mass) as a function of the baseline covariates noted previously, the treatment main effect and interactions between treatment and covariates. Interactions were a key component of the model as these were thought to capture the complex interplay between background factors and weight loss. We determined that the baseline variables and interactions could reasonably be expected to impact fat loss. Taking them into account during the evaluation of a nutraceutical designed to facilitate fat loss therefore seemed sensible.
In summary, two identical statistical models were used to respectively assess percent change in body fat and fat mass. Each model contained 1) the treatment effect (placebo, low dose and high dose); 2) six baseline covariates (carbohydrate intake, fat intake, protein intake, fat mass, lean mass and total cholesterol value); and 3) the interaction of each of the six covariates with treatment.
The above generalized linear model was used to estimate mean values and standard errors for percent change in body fat and fat mass. Inferential statistical tests were derived from the model to compare low dose CAPs to placebo and high dose CAPs to placebo. Covariate adjustment was Type III (each effect adjusted for all others); the model was obtained using Restricted Maximum Likelihood Estimation; and the model solution was accomplished using Newton-Raphson iterations [21,22,23]. Individual models were fit to percent change in percent body fat and to percent change in fat mass. Normal model convergence was observed. Missing values were accommodated in a manner that is typical for the generalized linear model. If a covariate value was missing at baseline, the subject was not evaluated in the model. No attempt was made to estimate the missing baseline value. As noted above, there were 77 total subjects in the study. Missing baseline values were concentrated in two subjects. Therefore, the net effect of missing baseline values was that 75 of the 77 total patients were available to each of the two models used to evaluate fat loss. Twenty-six rather than 27 patients were available from the Capsimax 2 mg treatment; all 22 patients were available from the Capsimax 4 mg treatment; and 27 of the 28 patients were available from the placebo group. Otherwise, as is an advantage of the generalized linear model, any missing values in the dependent variable were accommodated by the variance-covariance matrix to obtain model predicted means (i.e., Least Squares (LS) Means) and associated standard errors.
Tables present basic comparisons between treatment groups for the baseline variables serving as covariates. Tables also show the predicted means and standard errors as well as the significance levels for each treatment vs. control contrast obtained from the generalized linear model used to adjust for background factors. Finally, a one-way Analysis of Variance model was used to evaluate three treatments “without” covariate adjustment. These results were used to provide a base of comparison that modeled the data without accounting for the background factors of known importance to fat loss that were included as covariates in the generalized linear model described above.
This paper provides a comparison between the results obtained when important background factors are and are not included in the modeling of outcome data that concern complex physiological response variables such as fat loss. To do this, adjustments for multiplicity of comparisons must be avoided. Then, the unaltered p-values from the model with and the model without covariate adjustment will be available for “direct” comparison. If the more comprehensive model has increased statistical power, the contrast p-value for the comprehensive model should be lower than for the model that eliminates covariates and their interactions with treatment.
Age, height, weight, BMI, waist circumference, hip circumference and waist to hip ratio were compared to show that the treatment groups were similar with respect to variables of obvious relevance to fat loss; systolic blood pressure, diastolic blood pressure, and calorie intake were compared as these variables, too, are often associated with a history of resistance to fat loss. Descriptive statistics across the treatment groups are presented in Table 1. Additionally, analyses were conducted to compare the three treatment groups on all baseline variables simultaneously (i.e., a multivariate analysis was conducted) as well as individually (i.e., a univariate analysis of variance was conducted on each baseline variable).