Our analyses were part of an EORTC project [12] on MID development for the QLQ-C30 scales in various cancer entities and adds prostate-specific MIDs to the EORTC MID portfolio.
The main results of the study are anchor-based MIDs for deterioration for seven QLQ-C30 scales (physical functioning, role functioning, social functioning, pain, fatigue, global quality of life, diarrhoea), and for improvement for three QLQ-C30 scales (role functioning, social functioning, diarrhoea) both for within-group and between-group differences. MIDs varied by scale and direction (between 5 and 13 points for deterioration and 4 and 10 points for improvement), whereby the direction was always in accordance with the anchor change category (i.e. anchor scores indicating a low health status were associated with lower HRQOL scores). This compares well to MIDs already developed in this EORTC project for head and neck cancer [22], advanced breast cancer [21], malignant melanoma [20], colorectal [23] and ovarian [24] as well as to other similar research [26,27,28]. With two exceptions (global quality of life, diarrhoea), these MIDs were larger for deterioration compared to improvements. This aligns with existing findings even beyond the QLQ-C30 [17, 20, 21], suggesting that patients may have a higher sensitivity to favourable differences [26, 29, 30]. However this effect is not universal as other studies have reported no systematic differences in the magnitude of change between deteriorating and improving scores [15, 19, 22].
Overall, our MID estimates, with few exceptions, lie between 5 and 10 points, which corresponds to the thresholds suggested by Osoba et al. in 1998 [9] where patients’ reports on subjective change were used as clinical anchors. While these thresholds had been developed in breast and small-cell lung cancer patients, they have also been observed in various other cancer sites [21,22,23, 26,27,28, 31].
There seems to be a certain universality of an MID of 5–10 points on QLQ-C30 scales, but smaller and larger MIDs have been repeatedly found, especially for role functioning [20, 23] including the present study. This highlights that the scales and different sites are not to be tarred with the same brush.
Limitations
There are though some limitations to be considered when interpreting the presented results.
Most importantly, after careful evaluation of 14 potential clinical anchors, only the CTCA diarrhoea scale and the WHO PS were suitable for MID estimation as the others showed low correlations with HRQOL scales. A reason may be found in certain insensitivity of these rating systems to HRQOL differences due to a low interrater reliability in toxicity identification with CTCAE [32] or somewhat wide WHO PS categories (e.g. between 0- fully active and 1-able to carry out light work). Ideally, multiple anchors including patient self-reports which might be able to shed some light on the issue of subjectively perceived change on different scales would be considered. Furthermore, it has to be noted that in the present statistical approach ordinal scales are treated as interval scales, disregarding the fact that a difference between “not at all” and “a little” might be different from the difference between “quite a bit” and “very much”. This is where item-response-theory based methods can provide valuable information in future research. Finally, only two trials could be included, none of which was covering metastasised disease. Hence, the application of the here developed anchor-based MIDs to a prostate cancer population with stage IV disease needs to be done with caution. A further limitation is that, based on the available data, no anchor-based MIDs for improvement could be developed for some scales. This needs to be covered by future research along with the investigation of additional anchors to further approach the concept of minimal change. Meanwhile, the presented distribution-based MIDs may provide some guidance.
It is a strength of the present study, though, that MIDs did not vary across the different data sources, i.e. a trial in locally advanced prostate cancer on the effect of androgen suppression and a trial on effectiveness of radiation therapy with or without bicalutamide and goserelin in localized prostate cancer, indicating a certain stability of the estimated values. Our results may therefore support sound hypothesis for HRQOL in clinical trials targeting similar patient groups.