The evaluating self-management and educational support in severely obese patients awaiting multidisciplinary bariatric care (EVOLUTION) trial: rationale and design

Intensive lifestyle modification (diet, exercise ± behavioural modification) with a goal to reduce weight by 5-10% over 6–12 months is recommended as preferred initial therapy and should be delivered in a multidisciplinary setting [10]. Behavioural modification involves specific techniques to increase self-monitoring and self-efficacy. Examples include food journaling, goal setting, recognizing and addressing weight loss barriers (such as maladaptive eating behaviours), managing stress, increasing motivation and reframing negative emotions into positive ones [11]. Multidisciplinary teams are typically comprised of nurses, dietitians, physicians, mental health experts, physiotherapists and occupational therapists. One-year weight reductions in clinical trials examining intensive lifestyle modification average 6% of initial body weight (with most weight loss occurring within the first 3–6 months) and this degree of weight loss is associated with clinically significant improvements in cardiovascular risk factors (blood pressure, lipids, glycemic control) and reductions in diabetes incidence [10, 12]. In Canada, publicly funded care for severely obese individuals is often delivered within multidisciplinary bariatric specialty clinics because many primary care settings lack the required support of the allied health team and infrastructure [13].

Adding orlistat, a pancreatic lipase inhibitor and the only antiobesity drug approved in Canada, to lifestyle modification results in an additional ~3 kg of weight loss on average [14]. However, orlistat frequently leads to gastrointestinal adverse effects, has poor long-term persistence rates, and is rarely used [15].

In Canada, bariatric surgery is currently indicated in patients with BMI levels of ≥ 40 kg/m² or BMI levels of ≥ 35 kg/m² with a major medical complication (e.g., hypertension, diabetes, sleep apnea) who are refractory to non-surgical therapy [16]. Bariatric procedures either involve stomach restriction alone or combined restriction plus intestinal diversion [17]. Surgery reduces weight by 33% after 2–3 years; [18] mortality rates by 29% after 15 years; [19] and leads to resolution or improvement of type 2 diabetes, hypertension, dyslipidemia and sleep apnea in 70-86% of cases [18]. However, surgery can lead to serious complications including rates of perioperative death of 0.1-0.5%, a 10% rate of serious complication and long-term nutrient deficiencies such as symptomatic B12 or iron deficiency [20]. Thus, the benefit-risk ratio must be carefully assessed and patients require multidisciplinary care to prepare for surgery and lifelong multidisciplinary follow-up and monitoring thereafter [21].

Assessment of clinical effectiveness will include examination of changes in dichotomous outcomes (proportion of patients achieving 5% and 10% weight loss, prevalence of diabetes, hypertension, dyslipidemia, and sleep apnea as defined by lab parameters, patient report and/or use of active treatment) and continuous outcomes (anthropometric indices, blood pressure, lipid profile, glycemic control).

First, variables will be examined descriptively and graphically, including assessments of temporal trends and tests of normality. Second, the 3-month and 9-month mean change from baseline measures in each outcome will be calculated and compared between cases and controls using unpaired t-tests for continuous outcomes and chi-squared tests for dichotomous ones. Third, multivariable predictors of the 9-month change in a given outcome will be identified using appropriately constructed and calibrated linear regression models for continuous outcomes or logistic regression models for dichotomous ones. Herein, we present a detailed example of our modelling strategies for body weight. First, the proportion of patients achieving 5% weight loss (5% responders, a value widely considered as a clinically worthwhile weight change) [10] and the mean 9-month change in weight will be calculated in each study arm. Second, chi-square tests will examine for significant differences between cases and controls in 5% response rates and unpaired t-tests will examine for significant differences in mean body weight change scores. Third, a multivariable logistic regression model for 5% responders and a linear regression model for mean body weight will be constructed to examine the independent effect of the WWCM program on the 9-month change in 5% response rate or mean body weight, adjusting for any statistically significant (P<0.1) or clinically important between group differences. Missing data for weight change will be addressed using a conservative ‘baseline-value carried forward’ approach and examined in a sensitivity analysis using group mean-imputation methods to confirm the robustness of results [37]. All analyses will be conducted according to the “intention to treat principle”. No interim analyses, for either efficacy or futility purposes, have been planned.

Our sample size estimation is based upon the desire to detect a 15% difference between the WWCM interventions and controls in the proportion of 5% weight responders with an alpha level of 0.05 and a power of 0.90. We assume that the enhanced control arm will result in a 5% weight loss in 20% of subjects (i.e., the control event rate = 20% at one year) [38]. Given these parameters, the minimal required sample size will be ~180 patients per arm or 540 patients total (http://​statpages.​org/​proppowr.​html). As losses to follow-up have reached 30% in previous obesity related trials; [39] we will also adjust for a 30% potential drop out rate. Therefore, our final estimated sample size will be 220 patients per arm, for a total sample size of 660 patients.

We chose a 5% weight loss threshold as the primary outcome measure of individual patient success (rather than using mean absolute body weight or a similar continuous parameter) based upon the recommendations of current consensus guidelines, which endorse this degree of weight loss as the minimum target to achieve clinically meaningful health benefits after 6–12 months of intervention [10]. We chose a 15% absolute difference between study arms as the minimum clinically worthwhile (important) difference (MCID) by consensus of clinical experts and trials methodologists and have used this type of consensus-based threshold definition in prior studies [40]. We felt that if the WWCM program produced less than a 15% improvement in weight loss [5% response] compared with controls it would not be worth continued funding - unless very compelling benefits on humanistic outcomes or cost-effectiveness were also demonstrated. We partly based our rationale for choosing this threshold on the opinions of prior consensus-based expert panels, that have proposed a 10% MCID for 1-year weight change patients with type 2 diabetes (while recognizing that weight reductions in type 2 diabetics are generally lower than non-diabetics) and a 15% MCID for statin uptake in patients with coronary disease [41, 42].

The instruments to be used to assess health related quality of life (SF-12, EQ-5D and PHQ-8), self-efficacy (WEL), depression (PHQ-8) and patient satisfaction (PSQ) are described above. The mental and physical domains of each instrument (where applicable) will be examined in subgroup analyses and (where available) compared with population-based normative data available for the Alberta population.

Descriptive analyses will be performed and data will be examined graphically and serial changes described. Because these are self-reported data collected every 3 months, we will use a “last-value carried forward” approach to handle missing data as our primary analytic strategy. For example, temporal trends in the SF-12 will be assessed graphically and using analysis of variance (ANOVA) as a test of trends over time. Multivariable, adjusted 9-month changes in continuous measures will be assessed using linear regression in a manner similar to that detailed for the clinical outcome analysis. Patient satisfaction will be categorized as a dichotomous variable [31] and multivariable predictors of deterioration in satisfaction will be determined in serially constructed models.

The sample size of 660 subjects has more than adequate power. For example, for a nine month change in SF-36 domains such as physical function, with baseline score of 60, SD 19 [43], a deterioration of 10 points, two-sided alpha=0.05, beta=0.90, and a 10% attrition rate, only 85 total patients would be required. Similarly, for an outcome such as satisfaction, with baseline measures of 70% satisfied with care, a 30% deterioration in satisfaction over 1 year, and otherwise similar power-related assumptions, 136 total patients are required.

The total and incremental cost of WWCM (web-based and in-person) vs. controls (by program and per patient) will be determined and the incremental cost per QALY gained (cost-utility), and cost per incremental difference in BMI (cost-effectiveness) will be calculated.

A validated economic model of interventions in bariatric patients will be adapted to compare each version of the WWCM program vs. controls while adhering to recommended best practices for conduct of economic evaluation [18, 44]. The model will be informed by primary data, including costing data, utility scores (EQ-5D) and changes in BMI in the 3 groups at 9-months. A 9-month time horizon and a public health care payer perspective will be used. No discounting of costs and benefits will be performed in the reference case (given the short time frame). Uncertainty and variability will be explored through sensitivity analysis, including one-way and probabilistic sensitivity analysis to determine if results are modified over a plausible range of model inputs (for example, 95% confidence intervals of differences in EQ-5D, or variability in cost per patient of the WWCM program). If differences in BMI are found between the treatment groups at 9-months, we will perform an exploratory lifetime analysis linking BMI at 9-months to longer term health outcomes (including probability of developing complications such as diabetes, sleep apnea, long-term quality of life, and risk of death) as conducted in our previous work [18, 19, 44, 45].