SGRQ

SGRQ in Respiratory Trials and Cost-Utility Models

The St George's Respiratory Questionnaire, published by Paul Jones and colleagues at St George's in London in 1992, measures health impairment in chronic obstructive pulmonary disease (COPD) and asthma and is used as an outcome in respiratory trials. Each answer carries an empirically derived weight, and the weights are combined into three component scores and a total, each expressed on a scale where 0 is the best possible score and 100 the worst. For a health economist the SGRQ matters in two ways. The FDA and the developer read a change of 4 units in the total score as the smallest important change, and mapping algorithms turn SGRQ totals into utilities for cost-utility models when EQ-5D data are thin. In the NICE appraisal of dupilumab for COPD (TA1142), for example, the EQ-5D-5L had been recorded at infrequent timepoints, so the company mapped from the SGRQ.

Structure of the questionnaire and its two parts

Jones and colleagues described the SGRQ in 1992 as a fixed-format self-complete questionnaire yielding symptoms, activity and impacts scores and a total. They tested its repeatability in 40 patients with asthma and 20 with COPD, its validity in 141 patients and its change over a year in 133. The 1992 paper describes a 76-item questionnaire, while the SGRQ manual counts 50 items with 76 weighted responses and describes the questionnaire as designed to measure health impairment in patients with asthma and COPD.

The questionnaire has two parts. Part 1 asks how often respiratory symptoms have occurred over a recall period, and Part 2 asks about the patient's current state. The US Food and Drug Administration (FDA) guidance on the instrument lists versions of the 50-item questionnaire with 1-year, 3-month and 4-week recall periods, notes that as of March 2016 only the 3-month and 4-week versions were available, and notes that not all versions have equivalent validation information.

The items feed three components, named Symptoms, Activity and Impacts, and a total score summarises all three.

How the weighted scores are calculated

Each questionnaire response has an empirically derived weight. The SGRQ manual states that the lowest possible weight is zero and the highest is 100, and that each component is scored separately. When a patient gives more than one positive response to an item, the weights of those responses are averaged before being added to the sum.

$$S_c = 100 \times \frac{\sum_{j \in c} w_j}{W_c}$$

where $S_c$ is the score for component $c$ (or for the total), $w_j$ is the weight of each positive response $j$ in that component, and $W_c$ is the maximum possible weight for the component after the weights of any missed items have been deducted.

The deduction for missed items keeps the score on the 0 to 100 scale when a patient skips a question, but only up to a limit. The manual tolerates at most 2 missed items in Symptoms, 4 in Activity and 6 in Impacts, and beyond that no component or total score is calculated, which turns heavy item non-response into missing data for the analysis. Because the weights differ between items, two patients with the same number of positive answers can have different scores.

The SGRQ-C and other versions

The SGRQ-C is a shorter version for COPD, derived from the original after detailed analysis of data from large studies in COPD. Its manual describes 40 items, against the original 50, and states that its scores need a small arithmetic adjustment to make them directly comparable to those obtained with the SGRQ. For the Symptoms component, for example, the SGRQ-equivalent score is the SGRQ-C score multiplied by 0.99 plus 0.94 units.

$$S_{\text{SGRQ}} = 0.99 \times S_{\text{SGRQ-C}} + 0.94$$

where $S_{\text{SGRQ-C}}$ is the Symptoms score from the SGRQ-C and $S_{\text{SGRQ}}$ is the equivalent Symptoms score on the original SGRQ.

The COPD-specific version was the comparator when the COPD Assessment Test was developed, and Jones and colleagues reported a correlation of r = 0.80 between the two in their US sample. The questionnaires are copyright material held by St George's, University of London (now City St George's). Its permissions page states that clinicians may use both questionnaires without charge, that commercial organisations pay a licence fee charged per patient per study, and that investigator-led studies and student projects do not attract a fee.

The 4-unit threshold in trials and regulation

The SGRQ threshold for a minimal clinically important difference rests on several lines of evidence. Jones reviewed the evidence in 2005 and reported that estimates from earlier studies, made with different methods including anchor-based approaches and judgements by experts and patients, clustered around 4 units. He also noted that distribution-based estimates varied and permitted no consistent estimate, and that MCIDs should be used as indicative values. The SGRQ manual states that a minimum change in score of 4 units was established as clinically relevant after patient and clinician testing.

The FDA guidance for industry on use of the SGRQ as a patient-reported outcome assessment tool, issued in March 2018, sets the regulatory reading. It states that the minimum clinically important difference for the total score, both between patients and within a patient, is at least 4 units, and that there is no evidence to support other values. Sponsors should use only the total score and not the component scores, and responder analysis is the FDA's preferred primary method for reporting results.

In the dupilumab appraisal (TA1142) the committee concluded that dupilumab improved SGRQ score compared with placebo. One clinical expert at the committee meeting noted that the incremental benefit over placebo was less than 4 points while the improvement from baseline in the dupilumab arm was 9.9 points, a distinction between change within an arm and the difference between arms that a model has to keep.

SGRQ scores in cost-utility analysis

An SGRQ score is not a health state utility value. Its weights describe how much each response contributes to impairment on a 0 to 100 scale, while a utility for a QALY must be anchored at 1 for full health and 0 for dead and must come from preferences over health states. A cost-utility analysis that has only SGRQ data needs a mapping function, estimated in a mapping study where both instruments were completed by the same patients.

Starkie, Briggs, Chambers and Jones developed such an algorithm in 2011 from the TORCH trial, in which patients completed the SGRQ and the EQ-5D at baseline and every 24 weeks for three years. They compared ordinary least squares, generalised linear models and two-part models, selected on root mean squared error, and found that a simple ordinary least squares model performed about as well as the more complex options.

$$U = 0.9617-0.0013 \times T-0.0001 \times T^2+0.0231 \times M$$

where $U$ is the predicted EQ-5D utility, $T$ is the SGRQ total score and $M$ is 1 for a male patient and 0 otherwise. The root mean squared error of the model was 0.1723, and the coefficient 0.0001 on the squared term is rounded as published, so predictions made with it carry a small rounding error.

Starkie and colleagues found that the ranking of treatments by QALY gain depended on the method used to estimate utility. They judged the mapping usable in some settings but recommended that health technology assessment submissions, where precision matters, derive utility scores directly from the trial population.

The dupilumab appraisal shows the same route taken with newer data. The company collected the SGRQ and the EQ-5D-5L in the NOTUS and BOREAS trials, but the EQ-5D-5L was recorded at infrequent timepoints, so it built a mapping algorithm from visits where both were collected and converted the predicted EQ-5D-5L values to EQ-5D-3L values with the UK crosswalk tariffs. The external assessment group noted that in the utility regression only the coefficients for SGRQ at baseline, severity of airflow obstruction and exacerbation risk were statistically significant.

Illustrative worked example with weighted scoring and the Starkie mapping

The skipped-item weight, the patient's summed weights and the cost in this example are illustrative. The mapping coefficients are those Starkie and colleagues published, used here only to show the arithmetic.

1. Scoring one component with a missed item. The manual gives the Activity component a maximum possible weight of 1209.1. Suppose a patient skips one Activity item, within the limit of 4, and that item's maximum weight is an illustrative 100.0. The adjusted maximum is 1209.1 − 100.0 = 1109.1. If the weights of the patient's positive responses sum to an illustrative 554.55, the Activity score is 100 × 554.55 / 1109.1 = 50.

2. Mapping a 4-unit improvement at a total of 50. For a male patient with an SGRQ total of 50, the linear term is 0.0013 × 50 = 0.065, the squared term is 0.0001 × 2500 = 0.25, and the predicted utility is 0.9617 − 0.065 − 0.25 + 0.0231 = 0.6698. At a total of 46 the terms are 0.0013 × 46 = 0.0598 and 0.0001 × 2116 = 0.2116, and the predicted utility is 0.9617 − 0.0598 − 0.2116 + 0.0231 = 0.7134. The 4-unit improvement is worth 0.7134 − 0.6698 = 0.0436 in utility.

3. The same improvement at a total of 20. At 20 the terms are 0.0013 × 20 = 0.026 and 0.0001 × 400 = 0.04, giving 0.9617 − 0.026 − 0.04 + 0.0231 = 0.9188. At 16 they are 0.0013 × 16 = 0.0208 and 0.0001 × 256 = 0.0256, giving 0.9617 − 0.0208 − 0.0256 + 0.0231 = 0.9384. The same 4 units are worth 0.9384 − 0.9188 = 0.0196, less than half the gain at a total of 50.

4. Cost per QALY. If the gain at a total of 50 lasts one year, the QALY gain is 0.0436 × 1 = 0.0436. With an illustrative incremental cost of GBP 872 per patient,

$$\text{ICER} = \frac{872}{0.0436} = 20000$$

where 872 is the incremental cost in GBP and 0.0436 is the mapped QALY gain.

The ICER is GBP 20,000 per QALY for patients starting at 50, but because the curve is quadratic, the same 4-unit improvement in milder patients near 20 maps to 0.0196 QALYs and the ratio rises to 872 / 0.0196, about GBP 44,490 per QALY. Two treatments with identical mean SGRQ effects can therefore have different mapped cost-effectiveness if their trial populations differ in severity. Starkie and colleagues reported a related sensitivity at the level of the analysis, since the ranking of treatments by QALY gain changed with the method of estimating utility.

How the SGRQ relates to the CAT and generic measures

GOLD anchors its symptom cut-point on the SGRQ. Its 2025 report describes the SGRQ as the most widely documented comprehensive measure, notes that scores below 25 are uncommon in diagnosed COPD and that scores of 25 or more are very uncommon in healthy persons, and gives a CAT score of 10 as the equivalent cut-point. The CAT has eight unweighted items and is quicker to complete, while the SGRQ gives component scores and is the subject of the FDA guidance. Kon and colleagues found that changes in the CAT after pulmonary rehabilitation correlated with changes in the SGRQ, so the two tend to move together without being interchangeable.

The SGRQ is a patient-reported outcome measure, a disease-specific instrument and a non-preference-based measure of health-related quality of life in respiratory disease. The EQ-5D is generic, has no respiratory content and yields utilities through a value set, so the two answer different questions in an evaluation. An evaluation in respiratory disease economics, the wider application of economic evaluation to lung conditions, can therefore report the SGRQ as clinical evidence of benefit and take its QALYs from the EQ-5D.

Limitations and common misreadings of the SGRQ

Most errors with the SGRQ in economic work come from mixing versions or scales. Others come from reading the 4-unit figure more widely than its sources support.

• Treating weights as utilities. The weights are fixed for each response and place a score on a 0 to 100 impairment scale. A total of 30 does not mean a utility of 0.70.
• Pooling SGRQ and SGRQ-C scores unadjusted. The SGRQ-C manual gives conversion formulas because the two versions do not produce identical scores.
• Confusing within-arm and between-arm change. As the TA1142 discussion shows, a 9.9-point improvement from baseline can sit alongside a difference from placebo of less than 4 points, and a model needs the difference between arms.
• Reporting component scores as primary results. The FDA guidance asks for the total score only, so a component result is supporting evidence.
• Treating mapped utilities as measured. Starkie and colleagues recommended direct utility data where precision matters, and the quadratic form of their model means that the utility value of a fixed SGRQ change depends on where a patient starts.

Sources

• City St George's, University of London. Permission to use SGRQ, SGRQ-C and SGRQ-I and licensing of copyrighted material. https://www.staff.sgul.ac.uk/about/research-operations/research-administration/st-georges-respiratory-questionnaire/permissions (accessed 17 October 2026)
• Global Initiative for Chronic Obstructive Lung Disease. Global Strategy for the Diagnosis, Management, and Prevention of Chronic Obstructive Pulmonary Disease, 2025 Report. https://goldcopd.org/wp-content/uploads/2024/11/GOLD-2025-Report-v1.0-15Nov2024_WMV.pdf
• Jones PW. St. George's Respiratory Questionnaire: MCID. COPD: Journal of Chronic Obstructive Pulmonary Disease. 2005;2(1):75-79. doi:10.1081/COPD-200050513
• Jones PW. St George's Respiratory Questionnaire Manual. City St George's, University of London; March 2022. https://www.citystgeorges.ac.uk/__data/assets/pdf_file/0009/899631/SGRQ-Manual-March-2022.pdf
• Jones PW. St George's Respiratory Questionnaire for COPD Patients (SGRQ-C) Manual. City St George's, University of London; January 2023. https://www.citystgeorges.ac.uk/__data/assets/pdf_file/0004/899653/SGRQ-C-Manual-January-2023.pdf
• Jones PW, Harding G, Berry P, Wiklund I, Chen WH, Kline Leidy N. Development and first validation of the COPD Assessment Test. European Respiratory Journal. 2009;34(3):648-654. doi:10.1183/09031936.00102509
• Jones PW, Quirk FH, Baveystock CM, Littlejohns P. A self-complete measure of health status for chronic airflow limitation. The St. George's Respiratory Questionnaire. American Review of Respiratory Disease. 1992;145(6):1321-1327. doi:10.1164/ajrccm/145.6.1321
• Kon SSC, Canavan JL, Jones SE, et al. Minimum clinically important difference for the COPD Assessment Test: a prospective analysis. Lancet Respiratory Medicine. 2014;2(3):195-203. doi:10.1016/S2213-2600(14)70001-3
• National Institute for Health and Care Excellence. Dupilumab for maintenance treatment of uncontrolled chronic obstructive pulmonary disease with raised blood eosinophils (TA1142). https://www.nice.org.uk/guidance/TA1142/chapter/3-committee-discussion (accessed 17 October 2026)
• Starkie HJ, Briggs AH, Chambers MG, Jones P. Predicting EQ-5D values using the SGRQ. Value in Health. 2011;14(2):354-360. https://pubmed.ncbi.nlm.nih.gov/21402304/
• US Food and Drug Administration. Chronic Obstructive Pulmonary Disease: Use of the St. George's Respiratory Questionnaire as a PRO Assessment Tool. Guidance for Industry. Center for Drug Evaluation and Research; March 2018. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/chronic-obstructive-pulmonary-disease-use-st-georges-respiratory-questionnaire-pro-assessment-tool