Three methods of estimating the SGRQ provide estimates that are very close to 4 units. No other instrument of this type has used three methods or has such consistency. The following is a review of MCID estimation for the SGRQ and a recommendation of how it is best used for treatment trials.
The identification of the MCID for the SGRQ has been reviewed extensively (1,2). Three basic methods have been used: patient judgment, clinician judgment and criterion referencing.
Recently an instrument has been published by Devji et al for assessing the credibility of anchor-based assessments (3) the following is an application of this to SGRQ MCID using their 5 core criteria using data from two studies (1, 4).
1.1. The anchor is rated by the patient
In both studies the anchor was reported by the patient and compared with change in SGRQ, with the patients being blind to their SGRQ scores.
1.2. The anchor is interpretable and relevant to the patient
The anchor was the patient’s overall perception of the treatment’s efficacy, which is highly relevant to clinical medicine.
1.3. The MID estimate is precise
In one of these studies the estimated MID was 3.9 and the other 4.1. An integer value of 4 is therefore appropriate, since a value to one decimal place (e.g. 4.0) would imply a false degree of precision.
1.4. The correlation between the anchor and the outcome measure reported by the patient is satisfactory.
Plots from both studies clearly show an ordered relationship between the SGRQ and the anchor.
1.5. The authors select a threshold on the anchor that reflects a small but important difference.
There is no agreed consensus on the description of “a small but important difference”. In terms of assessing the value of treatments they clearly must be “effective”, so the anchor meets that criterion. In terms of magnitude “slightly” is clearly detectably different from “no effect” and “moderately” is larger than “slightly”, so these anchors meet the criterion for “small”.
Full details of a complex analysis of clinician judgment in the creation of the SGRQ MCID is presented in reference (1). Since the SGRQ is designed to capture the effects of a wide range of effects of COPD, so physicians and nurses experienced in pulmonary care were asked to judge what would constitute a minimum clinically significant difference between two hypothesized populations using the following: frequency of cough; frequency of wheeze; level of dyspnoea in daily life; level of depression and 6-min walking distance. These differences, estimated individually, were then applied simultaneously in a multivariable model using SGRQ data. The calculated difference in SGRQ score between the two hypothesized populations was 3.9 units.
Criterion-based tests are often used in the field of PROMs validation. In a prospective study, patients discharged from hospital following an acute COPD exacerbation were followed for one year. The baseline SGRQ scores in those who were re-admitted or died was 4.8 units higher in the patients who did not have one of those major events (5).
Another criterion-based study could also be classified as patient-anchored. It compared SGRQ scores between patients who were in MRC Dyspnea Grade 4 (stop for breath after 100m) and Grade 5 (housebound). The difference in SGRQ score was 3.9 (6).
Scenarios have been published to illustrate what a 4-unit change in SGRQ may mean for a patient. For example, a 4-unit change would happen if a COPD patient no longer took a long time to wash or dress and became able to walk up stairs without stopping and could go out for entertainment (6). These are not trivial benefits.
5.1 A paper by Welling et al (7) proposed an MCID in the range 7-8. It used both a distribution-based estimate and physiological anchors using the MID for lung function and exercise capacity. The flaw with this approach is that the anchors were not patient-centred and fail to reflect the fact that the SGRQ is a global measure of impaired health, not just a measure related to loss of exercise capacity, for example.
5.2 A meta-analysis (8) that has clear evidence of selection bias because it did not include any of the papers concerning the SGRQ MCID discussed above. It also did not distinguish between anchor-based and distribution estimates. The latter can be dismissed because they neither measure the “minimum” nor identify what is “important”. Two distribution methods have been used most: the standard error of the estimate (SEE) and half the standard deviation. However, the SEE will be dependent on the number of subjects and an analysis of 11 published studies showed a 5-fold difference between the MCIDs calculated using these two distribution methods and even within one method there was a big range of estimates (2).
5.3 Of the three papers cited in that meta-analysis which used patient anchored methods, one has already been discussed (7). Another used a global rating of change (GRC) (9). Full details of that scale are not given, but it appears to have been a numerical 7-point category scale, to which descriptors were later applied to group the changes into none (0-1), minimal (2-3), moderate (3-4), major (6-7). Numerical response options are not meaningful by themselves, the meanings (i.e. descriptions) were added later, so the GRC does not meet Criterion 2 of Devji et al. A third paper calculated the MCID for the SGRQ from the MCID of one domain of another COPD instrument the CRQ (10) gave an estimated MCID of 3.1 (10).
Whilst many treatment studies compare treatments by estimating the mean difference, there is a significant disadvantage to this approach, not least because there is a risk that if the mean difference is <4 units, the treatment may be judged to be ineffective. However, for the mean difference to exceed 4.0, more than half of patients would need to improve by ≥4 units (if the data are normally distributed). This is a very high threshold for judging the efficacy of a treatment.
The alternative approach – use of a responder analysis, is much more informative since it reports a ratio of the proportion of patients who improved by ≥4 units with one treatment compared to compared to another. This is commonly reported as an Odds Ratio (OR). For example, an OR of 1.4 shows that one treatment had a 40% greater odds (or chance) of a clinically significant improvement compared to the other. An OR in this range is commonly seen when the mean different is 2-3 SGRQ units, so the advantage of the responder analysis is clear, because it provides a numerical estimate of the probability of benefit, which is a concept that clinicians are familiar with, rather than the superficial conclusion that, on average, there was no clinically significant benefit.
A further advantage if using a responder analysis is that it appears relatively immune to large differences in the threshold used for the analysis. In the case of the SGRQ, an analysis of therapeutic trials showed that the OR was consistent over the range 1.5 – 8.0 units (11).
- Jones PW. Interpreting thresholds for a clinically significant change in health status in asthma and COPD. European Respiratory Journal. 2002;19:398-404.
- Jones PW. St. George’s Respiratory Questionnaire: MCID. COPD: Journal of Chronic Obstructive Pulmonary Disease. 2005;2:75-79.
- Devji T, Carrasco-Labra A, Qasim A et al. Evaluating the credibility of anchor based estimates of minimal important differences for patient reported outcomes: instrument development and reliability study. BMJ. 2020;369:m1714.
- Jones PW, Bosh TK. Quality of life changes in COPD patients treated with salmeterol. Am J Respir Crit Care Med. 1997;155:1283-1289
- Osman IM, Godden DJ, Friend JA, Legge JS, Douglas JG. Quality of life and hospital re-admission in patients with chronic obstructive pulmonary disease. Thorax. 1997;52:67-71.
- Welling JBA, Hartman JE, Ten Hacken NHT, Klooster K, Slebos D-J. The minimal important difference for the St George’s Respiratory Questionnaire in patients with severe COPD. Eur Respir J. 2015;46:1598-1604.
- Alma H, de Jong C, Tsiligianni I, Sanderman R, Kocks J, van der Molen T. Clinically relevant differences in COPD health status: systematic review and triangulation. Eur Respir J. 2018;52:1800412.
- Alma H, de Jong C, Jelusic D et al. Health status instruments for patients with COPD in pulmonary rehabilitation: defining a minimal clinically important difference. NPJ Prim Care Respir Med. 2016;26:16041.
- Bestall JC, Paul EA, Garrod R, Garnham R, Jones PW, Wedzicha JA. Usefulness of the Medical Research Council (MRC) dyspnoea scale as a measure of disability in patients with chronic obstructive pulmonary disease. Thorax. 1999;54:581-586.
- Schünemann HJ, Griffith L, Jaeschke R, Goldstein R, Stubbing D, Guyatt GH. Evaluation of the minimal important difference for the feeling thermometer and the St. George’s Respiratory Questionnaire in patients with chronic airflow obstruction. J Clin Epidemiol. 2003;56:1170-1176.
- Jones PW, Gelhorn H, Wilson H et al. Responder Analyses for Treatment Effects in COPD Using the St George’s Respiratory Questionnaire. Chronic Obstructive Pulmonary Diseases: Journal of the COPD Foundation. 2017;4:120-127.