Weight loss required by the severely obese to achieve clinically important differences in health-related quality of life: two-year prospective cohort study

Class II (body mass index (BMI) 35 to 39.9 kg/m²) and Class III (BMI ≥40 kg/m²) obesity (hereafter collectively referred to as `severe’ obesity) have increased by 400% over two decades and lead to substantial morbidity, mortality and reduced health-related quality of life (HRQL) [1]-[4]. Contemporary guidelines define 5% to 10% weight reductions as clinically important, citing expert opinion and statistically significant (albeit modest) improvements in cardio-metabolic risk as evidence for this contention [5]-[8]. Although many studies have examined HRQL (perceptions of physical, mental and social functioning) changes following weight loss [9],[10], to our knowledge, none have attempted to calculate the amount of weight loss required to achieve minimal clinically important differences (MCIDs) in HRQL or verify that 5% to 10% weight reductions result in clinically important HRQL improvements.

An MCID is the smallest difference in score in the domain of interest which patients perceive as beneficial and which would mandate a change in the patient’s management [11],[12]. HRQL, unlike various cardio-metabolic parameters or weight, is not a surrogate or intermediate measure; rather, it is a patient-reported outcome of tremendous clinical importance in its own right [8]. With the expanded use of HRQL endpoints and the increasing number of HRQL instruments (each with its own scoring structure and scale), interpreting HRQL in the context of MCID improvements is imperative. Thus, empirically determining the weight reduction thresholds corresponding to these MCIDs is needed. These instrument-specific weight loss thresholds could then be used to assess whether new or existing treatments are producing clinically important HRQL improvements.

The objective of this study was to examine treatment-related HRQL change and define clinically important weight loss as it relates to HRQL (that is, to determine the weight reductions required to achieve HRQL MCIDs). Specifically, we used two-year longitudinal data from 500 severely obese patients enrolled in a population-representative bariatric program to determine: (1) the two-year changes in weight and HRQL with medical and surgical treatment; and (2) the amount of weight loss required to attain MCIDs for three validated HRQL instruments.

A detailed study protocol for the Alberta Population-based Prospective Evaluation of the Quality of Life Outcomes and Economic Impact of Bariatric Surgery (APPLES) study, a prospective two-year observational evaluation of surgically treated, medically managed and wait-listed severely obese patients has been previously published [13]. The University of Alberta Health Research Ethics Board approved the study and all patients provided written informed consent.

Two major findings are noteworthy from this analysis of 500 patients enrolled in a publicly funded Canadian bariatric care program. First, compared to wait-listed patients who lost little weight over two years, HRQL improved following both medical and surgical treatment, with the most clinically important improvements found with surgery. Second, for most HRQL instruments, the percent weight reductions required to achieve HRQL MCIDs are substantially higher than currently promoted thresholds of 5% to 10% and are more in the order of 20% or greater.

Minimum weight loss thresholds of 5% of initial body weight are commonly cited as sufficient to improve health [5],[7]. Regulatory agencies also use a 5% placebo-subtracted weight loss threshold as one requirement for approval of new anti-obesity drugs [26]. Our findings suggest that this 5% threshold is not associated with clinically important improvements in HRQL in most patients. Even 10% weight loss was insufficient for most of the HRQL instruments examined, while 20% weight reductions appeared a more appropriate threshold to achieve clinically important HRQL improvement. A recent paper reported that a 1 kg decrease in weight following a modestly successful (5% weight loss on average) two-year behavioral intervention was associated with statistically significant improvements of 0.25 points in the SF-12 PCS, 0.09 points in the SF-12 MCS, 0.54 points in EQ-VAS and 0.002 in EQ-Index score [25],[27]. Except for the MCS, these results are similar to those reported in the present study.

The two-year HRQL changes we observed are comparable to those reported in the Utah Obesity Study, a prospective cohort study that enrolled 308 surgical patients, 253 patients who sought to undergo surgery but did not and 272 population-based controls [28]. After two years of follow up, surgically treated patients reported clinically important improvements in the IWQOL-Lite, PCS and MCS scores [29]. In APPLES, surgical-treatment was associated with the greatest improvements in HRQL (compared to the medically managed and wait-listed groups). HRQL improvements in medically managed patients were surprisingly high given that relatively modest weight losses were observed. We speculate that provision of behavior counselling, relatively frequent contact, and/or greater physical activity may have played a role in improving HRQL independent of weight loss, and further research into weight-independent and -dependent effects is needed [30]. Wait-listed patients experienced no change or small improvements in HRQL over two years despite minimal intervention. This confirms that no substantial deterioration in HRQL over this time occurs in patients awaiting bariatric care, which is relevant because patients trying to access publicly-funded bariatric care often face protracted wait times [31], and we can be assured that at least in terms of HRQL there are no overt harms associated with being wait-listed. In fact, there may be small improvements associated with wait listing or, perhaps, the HRQL changes in the wait-listed group are a product of temporal variation.

The relatively large sample size; inclusion of surgical, medical and wait-listed patients; long follow-up; simultaneous use of three validated HRQL measures; and population-representativeness of the study sample are major strengths of this study. However, there are several limitations. First, the interpretation of our results relies entirely upon the accuracy and validity of the HRQL instruments and established HRQL MCIDs - misidentification of an MCID would result in misspecification of a weight loss threshold. This may account for the discrepancy between the relatively low weight losses needed to attain EQ-5D MCIDs and those needed to attain MCIDs with all other instruments. The 0.03 MCID for the EQ-Index score was not derived from a population of obese patients or patients with chronic disease and, thus, may not appropriately generalize to our study population. In addition, the ceiling effects commonly seen in the three level version of the EQ-5D may make it ill-suited to assess HRQL change with weight loss [30]. While there is much debate over the appropriate ways in which to determine HRQL MCIDs [12], the MCIDs for the instruments we used are well-established and widely accepted, and we conservatively predefined our MCIDs using the higher end of the plausible ranges. Second, censoring was high in the wait-listed and medical groups because of the naturalistic study design in which patients were allowed to sequentially cross over to their next treatment phase. We handled this by using an intent-to-treat framework and LOCF imputation for missing data, as is routinely done in randomized trials of obesity management [32]. We note that study attrition was fairly low (<20%) and LOCF imputation was mostly needed for cross-overs into more intensive treatment (wait-listed transitioned to medical management and medical to surgical treatment). LOCF assumes that any observed changes occurring early in follow-up are maintained over two years. This might result in an overly optimistic assessment of treatment effectiveness. Third, baseline between-group imbalances in weight and HRQL were present (likely because of the sequential nature of the data), and while we adjusted for observed differences some residual confounding is possible. Because of the sequential nature of the data, propensity score matching was not used. Fourth, we did not collect data on the extent to which health behaviors changed and, thus, could not assess whether these contributed to weight-independent improvements in HRQL with medical therapy. Last, our study population was predominately white and female, and all were severely obese, and, therefore, our results may not generalize to a more heterogeneous population or necessarily to those with lesser degrees of obesity; however, our recent meta-analysis of weight loss interventions, including participants with initial BMIs ranging between 25 kg/m² and 55 kg/m², showed little to no improvement in HRQL with modest weight loss [32]. As well, Weight Wise is a publicly funded bariatric program in one region in Canada where all patients have universal healthcare coverage, and it may be that we enrolled more treatment-resistant severely obese patients than those typically seen in other studies conducted in other settings.

In a severely obese population, only bariatric surgery consistently led to statistically significant and clinically important weight reductions. Medical-management may have led to weight-independent HRQL improvements and this requires further study. HRQL increments per percent of weight loss were small and, for most severely obese patients and most instruments, a 20% weight loss over two years is required to achieve clinically important HRQL improvements predictably. If replicated, these findings also indicate that, from the HRQL perspective, future non-surgical obesity treatments will need to be more efficacious than current ones if clinically meaningful HRQL improvements are to be achieved.

All authors had full access to all of the data and RSP takes responsibility for the integrity of the data and accuracy of the data analysis. RSP and LMW developed the initial study concept, which was then refined by SMR, JAJ, CBA, CFR, AMS, SWK, SK and DWB. RSP, LMW and CBA were responsible for statistical analysis. LMW wrote the initial draft of the manuscript, with RSP, SMR, JAJ, CBA, CFR, AMS, SWK, SK and DWB contributing to the editing of the manuscript. All authors have read and approved the final manuscript.