Loading
A trial can report a statistically significant change in a quality of life score that no person would notice. The minimally important difference is the attempt to say how much a score has to move to correspond to something a patient would call a change, and it differs by questionnaire, by scale, by cancer and by direction.
The founding study asked patients directly. Osoba and colleagues gave people on chemotherapy for breast cancer or small-cell lung cancer a subjective significance questionnaire alongside the QLQ-C30, asking them to rate perceived change since they last completed it on a seven-category scale from much worse through no change to much better. The result: "For patients who indicated 'no change' in the SSQ, the mean change in scores in the corresponding QLQ-C30 domains was not significantly different from 0. For patients who indicated 'a little' change either for better or for worse, the mean change in scores was about 5 to 10; for 'moderate' change, about 10 to 20; and for 'very much' change, greater than 20." That five-to-ten-point rule of thumb is still the most quoted figure in the field.
The EORTC then did it the other way round, from the literature and from expert judgement. Cocks and colleagues had 34 experts, blinded to the actual results, predict differences across 2,217 contrasts from 152 articles, then combined the real data meta-analytically. Their conclusion is scale-dependent: "the recommended minimum to detect medium differences ranges from 9 (cognitive functioning) to 19 points (role functioning)". Role functioning needs the largest change to be meaningful, which is consistent with it being the least reliable scale on the instrument.
The modern approach is anchor-based and disease-specific, and the EORTC has published separate guidance for several cancers. In advanced breast cancer, using clinical anchors such as performance status, "MIDs for within-group change ranged from 5 to 14 points (improvement) and -14 to -4 points (deterioration), and MIDs for between-group change over time ranged from 4 to 11 points and from -18 to -4 points", with correlation-weighted values for most scales between 4 and 10 points. Equivalent guidance exists for lung cancer and malignant pleural mesothelioma and for prostate cancer.
Five things follow that a reader should hold on to.
The number is not symmetric. It takes a different amount of change to count as a deterioration than as an improvement, which the advanced breast cancer estimates show directly.
The number is not transferable. A threshold derived in advanced breast cancer does not apply to early prostate cancer, and the published sets differ.
Group differences and individual change are different quantities. The amount by which two trial arms must differ on average is not the amount by which one person's score must move for them to notice, and the second is usually larger. The FACT-Cog work makes the distinction explicit, separating clinically important differences for groups from meaningful change thresholds for individuals.
Stating the threshold in advance is what makes it work. The PRO-TECT trial declared its own and reported honestly against them: physical function with a stated minimum clinically important difference of 2 to 7 points, and "no MCID defined for symptom control or HRQOL". The mean differences it found were about 2.5 points on a 0 to 100 scale, and the authors reported them as such rather than calling them large.
A threshold can be gamed. If the pre-specified difference is chosen after seeing the data, or the largest published value is quoted for a negative result and the smallest for a positive one, the device stops working. Pre-specification is the only defence and is not universal.
Anchor-based methods estimate how much a score changes in people who report a small but definite change on an external criterion; distribution-based methods use a fraction of the standard deviation or the standard error of measurement. Anchor-based estimates are preferred because they are tied to something a person said, and distribution-based estimates are used to check them.
Query for this technology: (TITLE:"What a difference has to be before a person would notice it" OR ABSTRACT:"What a difference has to be before a person would notice it") AND (cancer OR tumor OR tumour OR oncology OR carcinoma OR lymphoma OR leukemia OR leukaemia OR myeloma OR sarcoma OR melanoma OR glioma). Results are unfiltered search hits about What a difference has to be before a person would notice it, not a curated reading list.
Shares The questionnaires that were never returned, and the people never asked, EORTC QLQ-C30: the questionnaire most cancer trials use, Asking people how they are, every week, as a treatment in its own right, How recovery is measured: the questionnaires behind the numbers and the tags rejuvenation, survivorship, measurement, instruments.
Shares FACT-G and the FACIT family: the other main questionnaire, EORTC QLQ-C30: the questionnaire most cancer trials use, How recovery is measured: the questionnaires behind the numbers, Patients lack understanding, navigation and agency and the tags rejuvenation, survivorship, measurement, instruments.
Shares How recovery is measured: the questionnaires behind the numbers, Patients lack understanding, navigation and agency, Quality of life, HR-positive / HER2-negative breast cancer and the tags rejuvenation, survivorship, measurement, instruments.
Shares The questionnaires that were never returned, and the people never asked, Asking people how they are, every week, as a treatment in its own right, Quality of life, HR-positive / HER2-negative breast cancer and the tags rejuvenation, survivorship, measurement, instruments.
Shares The questionnaires that were never returned, and the people never asked, Asking people how they are, every week, as a treatment in its own right, Patients lack understanding, navigation and agency, Quality of life and the tags rejuvenation, survivorship, measurement.
Shares EQ-5D: health reduced to one number, and what that number is for, Quality of life, HR-positive / HER2-negative breast cancer, Prostate cancer and the tags rejuvenation, survivorship, measurement.
Shares FACT-G and the FACIT family: the other main questionnaire, Quality of life, HR-positive / HER2-negative breast cancer, Prostate cancer and the tags rejuvenation, survivorship.
Shares The questionnaires that were never returned, and the people never asked, How recovery is measured: the questionnaires behind the numbers, Quality of life, HR-positive / HER2-negative breast cancer and the tags rejuvenation, survivorship, measurement.