Measures

The IW-SP, and two other patient-reported outcomes (PROs) instruments were used in this study to help determine the MIC of the IW-SP.

Impact of Weight on Self-perception Questionnaire (IW-SP)

The IW-SP contains three items [6]:

1.

How often do you feel unhappy with your appearance due to your weight?

2.

When going out in public, how often do you feel self-conscious due to your weight?

3.

When comparing yourself to others, how often do you feel unhappy, due to your weight?

Each item is rated on a five-point scale: 1 = “always,” 2 = “frequently,” 3 = “sometimes,” 4 = “rarely,” and 5 = “never.” The IW-SP total scores are derived by summing the item scores and dividing by the number of items. The score can be transformed to a range from 0 to 100. Higher IW-SP scores correspond to better self-perception, and the items apply to the present [6].

Ability to perform physical activities of daily living (APPADL)

The ability to perform physical activities of daily living (APPADL) contains seven items that assess how difficult it is for patients to engage in certain activities considered integral to normal daily life, such as walking, standing, and climbing stairs [10]. Items are scored on a five-point numeric rating scale, where 5 = “not at all difficult” and 1 = “unable to do.” A raw overall score is calculated by summing scores of the seven items and applying a linear transformation to yield a raw score ranging from 0 to 100. Higher scores indicate better ability to perform activities of daily living, and the items apply to the present.

Impact of weight on quality of life-lite clinical trials version (IWQOL-Lite-CT)

The impact of weight on quality of life-lite clinical trials version (IWQOL-Lite-CT) is a 20-item, obesity-specific PROM developed for use in obesity clinical trials. It assesses three primary composites of obesity-related health-related quality of life (HRQoL): physical (seven items), psychosocial (13 items), and a five-item subset of the physical composite—the physical function composite score. Items in the physical function composite describe physical impacts related to general and specific physical activities. All items are rated on either a five-point frequency scale (“never” to “always”) or a five-point truth scale (“not at all true” to “completely true”) [11, 12].

The IWQOL-Lite-CT composites and total scores are scored by averaging the responses of items within each scale, and then multiplying by the total number of items in the scale. Higher scores indicate better quality of life, and the recall period is 1 week.

Analyses

Anchor-based methods, recommended by the United States (US) Food and Drug Administration (FDA) to interpret PROM scores, were considered the primary approach, and distribution-based methods, recommended as supportive, were secondary [13, 14]. Anchor-based methods are those that explore the association between the targeted concept of a PROM (in this case, self-perception associated with one’s body weight, as measured by the IW-SP) and the same or a closely related concept measured by an independent instrument, the anchor. Anchors are typically single-item stems with a response scale that has descriptions accompanying each response option, so that meaningfulness to patients can be interpreted. Using this method, changes seen in the PROM are compared—or anchored—to changes on the anchoring instrument. An anchor must be conceptually or logically related to the concept measured by the PROM of interest and show an association with that PROM.

Analyses estimating an MIC for the IW-SP used SURPASS-2 clinical trial data [9]. The SURPASS-2 study was an open-label, 40-week phase 3 trial of patients with T2D randomly assigned to receive tirzepatide at a dose of 5 mg, 10 mg, or 15 mg or semaglutide at a dose of 1 mg. The SURPASS-2 trial did not include a global scale related to patient impressions of change or severity in self-perception of body weight. In the absence of a specific anchor, the current study used multiple exploratory anchors to gather a body of evidence supporting a meaningful change in the IW-SP. The exploratory anchors that were considered included weight loss and scales from conceptually related PROMs that were completed by participants in the trial, the APPADL and the three composites and total score of the IWQOL-Lite-CT. In measuring one’s perceived quality of life (QoL) associated with body weight, the IWQOL-Lite-CT was seen as logically related to self-perception of body weight. The APPADL, in measuring ADLs associated with obesity, is likely more distally related to self-perception and body weight, but a relationship may exist such that those losing weight improve in their ADLs as well as in their self-perception of body weight. Both PROMs have established MICs in people with either T2D or obesity. In addition, the IWQOL-Lite-CT also has two individual items that measure self-perceptions associated with body weight that were used as anchors: (Item 7) I feel less confident because of my weight (never, rarely, sometimes, usually, always) and (Item 20) I feel frustrated or upset with myself about my weight (not at all true, a little true, moderately true, mostly true, completely true).

Both anchor- and distribution-based approaches were used for the current analyses, first in 2/3 of the sample (estimation sample), then these initial results were confirmed in the other 1/3 of the sample (confirmation sample). The estimation and confirmation subsamples were created by random assignment. As most participants in the SURPASS-2 trial lost weight (approximately 73%), these analyses focused on change in terms of weight loss and improvement in PROM scores. Note that the process of deriving anchors was only performed during the estimation stage.

All analyses were conducted without consideration for treatment assignment in the trial, as the objective was to assess the meaningfulness of improvement in perception of body weight, irrespective of what drives weight loss. Change for all analyses was calculated using data from baseline and the trial endpoint (week 40), as these were the only visits where PROs were administered. All analyses were conducted using SAS Software (version 9.4).

Identifying potential anchors

The first step was defining prospective anchors, which could be used to group trial participants in terms of meaningful improvement. MIC values for weight loss from the literature show 5% and 10% change as frequently used indices of meaningful change in body weight [10, 15, 16], so both estimates were considered. The MIC for the APPADL was previously estimated as a range based on multiple approaches, from 6 to 14 points on a scale from 0 to 100 [10]. The midpoint of this range, 10 points, was used in the current analysis. The MICs for the IWQOL-Lite-CT were established by the developers as follows: physical composite (13.5 points), physical function composite (14.6), psychosocial composite (16.2 points), and total score (16.6 points); all scales range from 0 to 100 [17]. Change in Item 7 and Item 20 of the IWQOL-Lite-CT are discrete integers ranging from − 4 to 4, so the smallest change possible (one point) was used.

Finally, the percent of weight change that corresponds to the MICs of the APPADL and the IWQOL-Lite-CT Scales were estimated. These values were compared with the 5% and 10% values. These were estimated using regression, where change in the respective scale was the independent variable, percent change in body weight was the dependent variable, and the value of percent change in body weight predicted by each respective scale’s MIC was the value considered as a potential anchor.

Selection of anchors

First, only variables showing a minimum relationship of |0.30| (Spearman’s rho) in change [18] distributions (baseline to trial endpoint) were further included as anchors in the subsequent analyses. Second, redundant anchors for weight were eliminated. That is, potential anchors that grouped subjects by weight loss were eliminated if they converged on the same quantity of weight loss For example, seven anchors using percent change in weight—two from the literature [10, 15, 16] and five predicted by scale MICs—were explored as potential anchors for further analyses. A ≤ 5% point was identified a priori as the cut-off for redundant anchors, as 5% was the smallest MIC for weight change identified in our literature searches.

Application of anchors to form groups

The next step was to apply the anchors to form participant groups based on change in MIC. As the objective was to estimate the minimum meaningful change in the IW-SP, these analyses were performed using half MICs for all anchors, save for the IWQOL-Lite-CT items. Half MIC intervals were used in order to define a reasonable range within each change category for each of the anchor variables, as few (if any) people would have experienced exactly a 1 MIC change for each anchor variable. The intervals around the half MICs were shifted, such that groupings encompassed 1.0 times each anchors’ actual MIC (Table 1—middle column).

Table 1 Interval groupings of participants for IW-SP MIC estimation analyses

For the IWQOL-Lite-CT individual items, the “no change” group comprised participants indicating 0 points of change between baseline and trial endpoint, and the additional groups indicated improvements of 1–4 points between baseline and trial endpoint (five groups in all; Table 1—last column).

Assessing meaningful change and responsiveness

Once anchors were selected, group differences by anchor category were tested to establish responsiveness with the prediction that group differences on the IW-SP scores would be a function of differences in the conceptually related anchors. Responsiveness refers to the extent to which the instrument can detect true change in participants known to have changed on the concept of interest [19]. IW-SP responsiveness analyses compared change scores between participants with different degrees of anchor change using general linear models (analysis of covariance [ANCOVA]), controlling for age, gender, and baseline body mass index for each anchor.

These analyses also allowed assessment of the IW-SP’s ability to reliably discriminate groups defined by differing increments of change, adding evidence for the reliability of the estimated MIC.

Distribution-based approaches

Distribution-based methods for estimating the magnitude of meaningful change in PROM scores utilize statistical parameters from the clinical trial population, where a PROM change score is expressed relative to some measure of variability.

The distribution-based approach was a calculation of standard deviation (SD) units using the baseline score. The half SD has been shown to provide a reasonable approximation of a meaningful change in PROMs [20]. The standard error of measurement (SEM) [21] was not included, as the time interval between baseline and trial endpoint was too long to estimate test–retest reliability, which is required for calculation of the SEM.

Triangulation

A single estimate for an MIC for improvement in the IW-SP was achieved by triangulating findings of the anchor- and distribution-based analyses, supported by visualization of data on cumulative distribution function and probability density function plots [18, 22, 23]. Special attention was paid to the IW-SP change score that corresponded to a one-category improvement (0.75 to < 1.25 MIC improvement) when there was a statistically significant group difference between IW-SP scores corresponding to that anchor category and the adjacent “no change” anchor category. MIC estimates for the IW-SP from the confirmation and estimation stages are presented and discussed together, allowing the extent of the convergence of estimates to be evaluated. That is, comparison of the estimation and confirmation stage estimates is itself a form of triangulation.

The MIC for improvement in the IW-SP total score at this stage was expressed in terms of the transformed scale and the raw scale. The latter is more interpretable as it aligns directly with the actual IW-SP responses. Following this process, a final MIC estimate was produced and rounded to the nearest mathematically possible change value for the IW-SP total score, both in terms of raw and transformed scales.