Minimal important differences of EORTC QLQ-C30 for metastatic breast cancer patients: Results from a randomized clinical trial

Treatment for metastatic breast cancer aims to prolong survival and palliate symptoms [1]. Considering that metastatic breast cancer remains incurable, maintaining quality of life (QOL) is an important therapeutic goal [2]. Instruments to measure QOL in patients with metastatic breast cancer include the European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire core 30 (EORTC QLQ-C30). Despite improvements in validation of the QLQ-C30, interpreting the numerical scores remains challenging.

A large clinical trial can show statistically significant differences for QOL scores, but the clinical relevance of these differences remains controversial. The concept of minimal important difference (MID) provides a measure of the smallest difference in QOL scores that patients can perceive as improvement or deterioration [3]. King has comprehensively reviewed the literature on this topic [4]. While both anchor- and distribution-based methods are available for estimating MIDs, the anchor-based method has the appeal of incorporating external instruments (criterion or anchor) that are relevant to patients [5].

Recommendations [6] or guidelines across cancer sites using meta-analysis [7, 8] have been provided to interpret QLQ-C30. These are useful for designing trials and interpretation, but MIDs can vary according to factors such as patient populations, cancer site, and disease stage. Research into MIDs for metastatic breast cancer remains limited. Musoro et al. recently used two published trials to estimate MIDs relying on clinical anchors from clinician examinations [9]. In contrast, the most widely used anchor is the rating of change in health status as obtained directly from patients [10].

The purpose of this study was to determine the MIDs of the QLQ-C30 in a population of metastatic breast cancer patients using data from the SELECT BC-CONFIRM trial. In that trial, transition items reported directly from patients and other clinical anchors were collected at multiple timepoints from the initiation of treatment; these trial data were thus considered desirable for the estimation of MID.

In total, 230 patients were randomized in the trial. To calculate the change in QOL from baseline, we included only the 154 patients who had completed a baseline QLQ-C30. Table 1 summarizes the demographic and treatment characteristics of the analyzed cohort. Median age was about 60 years. The majority of patients had a history of surgery (76.6%), and about two-thirds of patients had no history of chemotherapy.

Numbers of patients by anchor categories and corresponding mean change in QOL scores are shown in Supplementary Table 1. When using the transition item of global health, the proportions of “improvement”, “stable (no change)”, and “deterioration” were 17%, 38%, and 23%, respectively, with the remaining 22% responding two or more categories away from “stable (no change)”. When using the change in PS anchor, the proportions of “improvement”, “stable (no change)”, and “deterioration” were 6%, 78%, and 13%, respectively. Table 2 shows correlation coefficients between change in QLQ-C30 and anchors. For each anchor, correlated QOL scales (i.e., maximum correlation coefficient > 0.3) at one or more timepoints are listed. The performance status showed a correlation with change in QLQ-C30, mostly at 6 months. Supplementary Table 2 shows correlation coefficients for absolute or baseline scores with anchors. In most pairs of anchors and scales, correlations of transition items with change scores were greater than correlations of transition items with absolute scores. In most pairs of positively (negatively) correlated transition items and scales, transition items were negatively (positively) correlated with the score at baseline.

Figure 1 shows the range of MIDs from the mean change method for each QOL scale across multiple anchors. The mean change method identified MIDs for improvement and deterioration in 8 of 15 QLQ-C30 scales, and the GEE identified MIDs for at least one direction in 7 of the 15 QLQ-C30 scales. Estimated MIDs differed between scales of QOL, direction (improvement and deterioration), and methods of estimation. Estimates from GEE tended to be smaller than those from the mean change method. No MIDs were estimated from GEE on some scales. Supplementary Table 3 shows details of results from Fig. 1. Most MIDs estimated with GEE were p < 0.05, but this was not the case for MIDs estimated under the mean change method.

Table 3 summarizes MIDs from anchor-based methods, alongside results from the distribution-based method for comparison. Distribution-based estimates at 2 months are presented in Table 3 because results at other timepoints were similar, with at most a difference of 0.8 units for 0.2 SD. Most MIDs for interpretation of within-group differences were between 0.3 and 0.5 SD.

To the best of our knowledge, this study is the first to estimate MIDs of QLQ-C30 in a population of Japanese patients with metastatic breast cancer. The study identified MIDs for interpreting within-group differences in 8 scales and for interpreting between-group differences in 7 scales.

As far as we know, only one study has examined the MIDs of QLQ-C30 in patients with advanced breast cancer [9]. Musoro et al. used two published trials to identify MIDs [9], and most of the patients were from European countries. They identified at least one suitable MID in eight scales, and six of those eight scales (global quality of life, physical function, role function, social function, fatigue, and appetite loss) were in common with the present findings. In these scales, MIDs for interpreting between-group were similar. Our results thus support their conclusions, at least for these scales.

Conversely, MIDs for interpreting within-group deterioration tended to be larger in our study than in the previous study [9], such as for global quality of life (− 10 vs. − 8), role function (− 17 vs. − 6), social function (− 12 vs. − 7), and fatigue (− 10 vs. − 8). Feng et al. [23] investigated systematic differences between Japanese and European general populations in self-reported QOL questionnaires. They found that the percentage of Japanese respondents reporting no problems in morbidity, usual activities, pain, or anxiety was higher than that of European respondents [23]. Our study of Japanese patients revealed that minimally deteriorated patients described larger changes from baseline than the European patients reported by Musoro et al. [9]. These findings may indicate inter-country differences in the self-evaluation of health at a single point as well as changes between two timepoints, especially in terms of deterioration.

Consistent with recent findings on MIDs for other cancer sites [5, 19, 24, 25], the magnitudes of MIDs for deterioration and improvement differed, along with the scales of QLQ-C30. We observed no systematic differences between previous results and our findings. These estimates generally do not conflict with the previously proposed guidelines [7, 8], although we consider that MIDs for interpreting deteriorations and improvements should be distinguished. In addition, MIDs for interpreting within- and between-group changes differed, as shown in Table 3. These results strongly indicate that MIDs fitted for purpose are necessary.

The choice of an anchor is one of the critical components in the estimation of anchor-based MIDs [10]. Musoro et al. [9] estimated MIDs relying on clinician-examined anchors such as performance status and common terminology criteria for adverse events. However, patient-reported anchors appear preferable to clinician-reported anchors [10]. The present study used both types of anchor: transition items and performance status. We found that transition items were more sensitive to early changes in QOL than performance status, as shown in Table 2. This finding is consistent with findings from previous research that self-reports from patients are more sensitive than clinicians’ reports to underlying changes in health status, and tend to identify clinically meaningful symptoms earlier [26, 27]. MID estimates by mean change methods using performance status as the anchor lay within the range of MID estimates using transition items as the anchor (Supplementary Table 3). This finding may indicate that clinician-reported anchors provide similar results to patient-reported anchors for estimating MIDs. However, we recommend patient-reported anchors as more susceptible to changes in the underlying health status of patients than clinician-reported anchors. Note that one needs to be cautious about the use of transition items because they are often vulnerable to recall bias.

Distinguishing between unreliable and credible MIDs is essential for MIDs to help patients and clinicians in making decisions. One such method is referring to a recent guideline [10] and reporting the credibility level of the anchors used to estimate MID. We created Supplementary Table 2 for this purpose. We note here two points. First, the correlations of transition items with change score were greater than those with absolute score. Otherwise, the transition item seems to capture current health status more than the change in health status. Second, transition items showed some correlation not only to the change in QOL, but also to QOL at baseline. This correlation reflects the fact that patients appropriately respond to transition items while considering the health status at baseline. These points support the credibility of the transition items used in this study.

Several limitations need to be considered when interpreting the present results. First, due to the small or medium sample size for estimating MID, each estimate was not precise. We aggregated estimates into a single MID using correlation-weighted averages, so we consider our final estimates as reliable to some extent. Second, no change group or improved or deteriorated group showed statistically significant results in some scales. We do not believe, however, that this seriously influenced our conclusions, because all estimates of MIDs in this study exceeded effect size thresholds.

In conclusion, we provided the MIDs of QLQ-C30 using both patient- and clinician-reported anchors measured in a randomized trial of Japanese patients with metastatic breast cancer. Our results can aid patients and clinicians, as well as researchers, with interpreting QLQ-C30. Further studies should replicate this study in other populations or other anchors to validate the current MID estimates.