KCCQ

KCCQ in Heart Failure Trials and Economic Evaluation

The Kansas City Cardiomyopathy Questionnaire, usually shortened to KCCQ, records how people with heart failure have felt and functioned over the previous two weeks. For a health economist the KCCQ sits at an awkward junction. Cardiology trials use the questionnaire widely and the US Food and Drug Administration (FDA) has qualified the KCCQ as a clinical outcome assessment, yet its scores carry no preference weights, so a cost-utility analysis can use them only after mapping to a utility measure such as the EQ-5D or by defining model health states on the score. In the empagliflozin appraisal TA773 a NICE committee accepted an economic model whose health states were quartiles of a KCCQ summary score, which means the questionnaire now shapes QALY estimates in heart failure appraisals as well as the clinical evidence behind them.

Structure, recall period and scoring of the questionnaire

Green, Porter, Bresnahan and Spertus developed the KCCQ as a self-administered, disease-specific health status measure for people with congestive heart failure and published its development and first evaluation in 2000. The original version has 23 items. Its domains cover physical limitation, symptoms, self-efficacy, social interference and quality of life.

The developers chose a two-week time frame so that patients could integrate a representative period of recent functioning into their answers. Each response option receives an ordinal value starting at 1 for the lowest level of functioning, and the item values are summed within each domain. Missing answers within a domain are given the average of the answered items in that domain. Each domain total is then rescaled.

$$S = 100 \times \frac{x-x_{\min}}{x_{\max}-x_{\min}}$$

where $S$ is the domain score on the 0 to 100 scale, $x$ is the patient's summed raw score for the domain, $x_{\min}$ is the lowest possible raw score and $x_{\max}$ is the highest possible raw score. Higher scores mean better health status.

Trials and appraisals draw on several derived scores, which are easy to confuse. Spertus and Jones report an overall summary score, and the FDA qualification for drug development names the total symptom score, the physical limitations score and the clinical summary score.

Green and colleagues tested the instrument in two cohorts, 70 stable and 59 decompensated patients with ejection fractions below 40 per cent, who also completed the Minnesota Living with Heart Failure Questionnaire and the SF-36 at baseline and three months later. KCCQ domains correlated with the criterion measures at 0.46 to 0.74. Among patients with stable heart failure who remained stable the mean domain change was 0.8 to 4.0 points, while patients whose decompensated heart failure improved changed by 15.4 to 40.4 points, and the authors judged the KCCQ substantially more sensitive than either comparison instrument. Sensitivity to change of this kind, known as responsiveness, was the stated aim of building a disease-specific measure for heart failure.

The 12-item short form

Length limited uptake of the full questionnaire. Spertus and Jones derived and validated a 12-item version, the KCCQ-12, using data from three clinical studies covering 4,168 patients, and published it in 2015.

The KCCQ-12 keeps scores for symptom frequency, physical limitation, social limitation and quality of life, together with an overall summary score. Spertus and Jones reported correlations with the original scales above 0.93 for all scales in all clinical settings, test-retest reliability above 0.76 for all domains and prognostic value comparable to the full instrument. A trial that records the KCCQ-12 can therefore report an overall summary score that tracks the 23-item version closely.

Interpreting score changes in heart failure trials

A mean difference between trial arms on a 0 to 100 scale says little on its own about whether patients noticed the change. Developer-led studies have tied score changes to clinical judgements, which gives the instrument an empirical minimal clinically important difference.

Spertus and colleagues followed 476 outpatients at 14 centres in a prospective cohort, assessed at baseline and again at 6 ± 2 weeks. Cardiologists, blinded to every measure except weight and New York Heart Association class, graded each patient's clinical change on a scale running through small, moderate and large deterioration or improvement. The KCCQ fell by a mean of 5.3, 17 and 25 points for small, moderate and large deteriorations, and rose by 5.7, 10.5 and 22.3 points for small, moderate and large improvements. In that 2005 study the KCCQ had the highest c statistic for monitoring individual patients, ahead of the NYHA functional classification and the 6-minute walk test, and neither of those two measures separated small deterioration from no change. Because the clinical judgement serves as the external anchor, these figures are an anchor-based MID and not a statistical rule of thumb.

In their 2020 review Spertus, Jones, Sandhu and Arnold summarised this evidence as changes of 5, 10 and 20 points marking small, moderate to large and large to very large clinical change. The same review argued that trials should report the proportions of patients whose scores improve or worsen by clinically important amounts, because mean differences between arms are hard to translate into clinical meaning. Responder proportions also suit an economic model better than a mean, since a model needs the share of patients in each state of health and not an average score.

FDA qualification as a clinical outcome assessment

Two FDA centres have qualified the KCCQ as a clinical outcome assessment, one for medical devices and one for drugs. Under the device programme, qualification means that FDA staff can rely on the measure in its qualified context of use without having to reconfirm that the tool is suitable.

The Center for Devices and Radiological Health qualified the KCCQ through its Medical Device Development Tools programme in October 2017, for the paper, self-administered 23-item version in feasibility studies and in the main approval studies of devices for patients with symptomatic heart failure. The Center for Drug Evaluation and Research (CDER) qualified the KCCQ-23 on 9 April 2020 as DDT COA #000084, for patients with heart failure, with a concept of interest of heart failure symptoms and their impact on physical limitations. CDER's statement covers the total symptom score, the physical limitations score and the clinical summary score, and notes that further evaluation is warranted of the ability to detect change and of how to interpret clinically meaningful within-patient change. Sponsors planning to use the KCCQ-23 as a confirmatory endpoint are asked to discuss the plan with the relevant review division.

Turning KCCQ scores into utilities for QALYs

The KCCQ is a patient-reported outcome measure without preference weights. Its 0 to 100 scale has no anchor at full health or at death, and no value set converts its scores into utilities, so a KCCQ score cannot be used directly in a QALY calculation. Some heart failure trials collected the KCCQ without the EQ-5D, which leaves an analyst with two routes, a mapping function that predicts utility from KCCQ responses, or a model whose health states are defined by KCCQ score bands with a utility estimated for each band.

Two published mappings show how the first route works. Hunger and colleagues used data from the PARADIGM-HF trial, splitting the sample into a derivation set of 4,465 patients and a validation set of 1,892, and mapped the KCCQ onto the EQ-5D-3L with the Japanese and UK value sets. They compared ordinary least squares, two-part, Tobit and three-part models. All of their models tended to underpredict utilities above 0.75 and overpredict utilities below 0.5, although they performed well at the population level, and the authors concluded that EQ-5D-3L scores can be estimated from KCCQ responses for cost-utility analysis of trials that collected only the KCCQ.

Thomas and colleagues developed models from two United States cohort studies that predict EQ-5D-3L and EQ-5D-5L utilities from both the 23-item and 12-item KCCQ. The 3L models came from 1,000 assessments in 547 patients with heart failure with reduced ejection fraction and reached an internally validated R² of 48.4 to 50.5 per cent, falling to 33.7 to 45.6 per cent in external validation on two clinical trials. The 5L model, from 3,925 patients, reached 57.7 per cent. The authors presented the mappings for cost-utility analysis when EQ-5D data are not available.

KCCQ health states in NICE heart failure appraisals

NICE committees have met the KCCQ as a source of model structure as well as of clinical evidence. In the two appraisals below the company models defined their health states by KCCQ score bands.

In the appraisal of empagliflozin for chronic heart failure with reduced ejection fraction (TA773), the company's model moved people between quartiles of the KCCQ clinical summary score, with a specific utility and cost associated with each state. The guidance records the score as running 0 to 100, with high scores indicating lower symptom burden. The committee considered whether the KCCQ is an appropriate measure of quality-of-life outcomes, heard clinical experts describe it as well validated for heart failure, and noted that there may be some inconsistencies in how the size of effect translates from trial data into modelled quality-of-life effects. It concluded that the company's model structure was appropriate for decision making.

In the appraisal of dapagliflozin for chronic heart failure with preserved or mildly reduced ejection fraction (TA902), the company's health states were quartiles of the KCCQ total symptom score. The external assessment group noted that including the KCCQ total symptom score as a predictor of all-cause and cardiovascular mortality generates an indirect survival benefit for dapagliflozin, because people having dapagliflozin are more likely to remain in better KCCQ states. The committee questioned the appropriateness of health states defined by the KCCQ total symptom score, noting that older cost-effectiveness models of chronic heart failure had used health states defined by New York Heart Association class, but observed that the structure was similar to those of previous NICE technology appraisals. It accepted that dapagliflozin plausibly lowers mortality while judging the benefit uncertain from the DELIVER data, whose mortality hazard ratios had confidence intervals crossing 1. A model of this kind turns a quality-of-life effect into a survival effect, and the assumption needs separate justification.

Illustrative worked example with KCCQ quartile health states

All numbers in this example are illustrative and do not come from any appraisal. A model defines four health states by quartile of the KCCQ clinical summary score, Q1 being the worst, and attaches an EQ-5D utility to each state of 0.55, 0.68, 0.77 and 0.85. Over one year with no difference in survival, standard care leaves 25 per cent of the cohort in each quartile, while a new treatment leaves 20, 23, 27 and 30 per cent in Q1 to Q4. The treatment adds GBP 420 per patient per year.

1. Mean utility in each arm.

$$\bar{u} = \sum_{k=1}^{4} p_k u_k$$

where $\bar{u}$ is the mean utility of the arm, $p_k$ is the proportion of the cohort in quartile $k$ and $u_k$ is the utility attached to quartile $k$. Standard care and the new treatment then give the following.

$$\bar{u}_C = 0.25 \times 0.55 + 0.25 \times 0.68 + 0.25 \times 0.77 + 0.25 \times 0.85 = 0.7125$$

where $\bar{u}_C$ is the mean utility under standard care, with each quartile holding 25 per cent of the cohort.

$$\bar{u}_T = 0.20 \times 0.55 + 0.23 \times 0.68 + 0.27 \times 0.77 + 0.30 \times 0.85 = 0.7293$$

where $\bar{u}_T$ is the mean utility under the new treatment.

2. QALY gain. Over one year the gain is 0.7293 − 0.7125 = 0.0168 QALYs per patient.

3. Cost per QALY.

$$\text{ICER} = \frac{420}{0.0168} = 25000$$

where 420 is the incremental cost in GBP and 0.0168 is the incremental QALY gain, so the ICER is GBP 25,000 per QALY.

At an illustrative threshold of GBP 30,000 per QALY the treatment would be accepted. The QALY gain rests entirely on two inputs, the shift between quartiles and the utility gaps between them. If the Q4 utility were 0.79 instead of 0.85, leaving Q3 and Q4 only 0.02 apart, the same shift between quartiles would give the following.

$$\Delta Q = (0.20-0.25) \times 0.55 + (0.23-0.25) \times 0.68 + (0.27-0.25) \times 0.77 + (0.30-0.25) \times 0.79 = 0.0138$$

where $\Delta Q$ is the one-year QALY gain, each bracket is the change in the proportion of the cohort in a quartile and each multiplier is that quartile's utility. The ICER becomes 420 ÷ 0.0138, about GBP 30,435 per QALY, just above the illustrative threshold, without any change in the trial's KCCQ result. A model that also linked mortality to the quartile would add a survival gain on top, which is the indirect benefit the external assessment group questioned in TA902.

How the KCCQ differs from neighbouring measures

The KCCQ belongs to several overlapping classes, and each comparison clarifies a different use. Against a generic instrument such as the EQ-5D, the KCCQ trades comparability across diseases for sensitivity within heart failure, and Green and colleagues found it more responsive than the generic SF-36 in their first cohorts. The EQ-5D has value sets and feeds a QALY directly, while the KCCQ needs a bridge.

Against the NYHA functional classification, the KCCQ is patient-reported and continuous, while the NYHA class is a clinician's judgement on four levels. Spertus and colleagues found that NYHA class did not distinguish small deterioration from no change, while the KCCQ did. As a patient-reported measure the KCCQ is one kind of clinical outcome assessment. The Minnesota Living with Heart Failure Questionnaire is an older heart failure measure, and Green and colleagues used it as one of the criterion measures when they developed the KCCQ.

Limitations and common misreadings

Several errors recur when KCCQ results reach an economic evaluation. Most come from treating a descriptive score as if it were a utility or a survival measure.

• Reading scores as utilities. A score of 75 does not mean a utility of 0.75. The scale has no anchor at death, and equal score gaps need not carry equal value to patients.
• Treating mapped values as observed values. The models of Thomas and colleagues explained roughly half of the variation in EQ-5D utilities, and Hunger and colleagues found that predictions were too low at the top of the range and too high at the bottom, so mapped utilities suit group estimates more than individual patients.
• Mixing summary scores. The overall summary, clinical summary and total symptom scores are different quantities. TA773 used clinical summary score quartiles and TA902 used total symptom score quartiles, so results based on one cannot be carried across to the other without checking.
• Confusing the versions. The KCCQ-12 reproduces the overall summary score of the 23-item version closely, but the FDA drug qualification refers to the KCCQ-23 and its named scores.
• Attributing a survival gain the trial did not demonstrate. Spertus and Jones reported prognostic value for both versions of the KCCQ, so a model that lets the score drive mortality as well as utility can credit a treatment with a survival gain that its trial did not show, the concern raised about DELIVER in TA902.

Sources

• Green CP, Porter CB, Bresnahan DR, Spertus JA. Development and evaluation of the Kansas City Cardiomyopathy Questionnaire: a new health status measure for heart failure. Journal of the American College of Cardiology. 2000;35(5):1245-1255. doi:10.1016/S0735-1097(00)00531-3
• Hunger M, Eriksson J, Regnier SA, Mori K, Spertus JA, Cristino J. Mapping the Kansas City Cardiomyopathy Questionnaire (KCCQ) onto EQ-5D-3L in heart failure patients: results for the Japanese and UK value sets. MDM Policy & Practice. 2020;5(2):2381468320971606. doi:10.1177/2381468320971606
• National Institute for Health and Care Excellence. Empagliflozin for treating chronic heart failure with reduced ejection fraction (TA773). 2022. https://www.nice.org.uk/guidance/ta773/chapter/3-Committee-discussion
• National Institute for Health and Care Excellence. Dapagliflozin for treating chronic heart failure with preserved or mildly reduced ejection fraction (TA902). Published 21 June 2023. https://www.nice.org.uk/guidance/ta902/chapter/3-Committee-discussion
• Spertus JA, Jones PG. Development and validation of a short version of the Kansas City Cardiomyopathy Questionnaire. Circulation: Cardiovascular Quality and Outcomes. 2015;8:469-476. doi:10.1161/CIRCOUTCOMES.115.001958
• Spertus JA, Jones PG, Sandhu AT, Arnold SV. Interpreting the Kansas City Cardiomyopathy Questionnaire in clinical trials and clinical care: JACC state-of-the-art review. Journal of the American College of Cardiology. 2020;76(20):2379-2390. doi:10.1016/j.jacc.2020.09.542
• Spertus J, Peterson E, Conard MW, et al. Monitoring clinical changes in patients with heart failure: a comparison of methods. American Heart Journal. 2005;150(4):707-715. PMID 16209970
• Thomas M, Jones PG, Cohen DJ, et al. Predicting the EQ-5D utilities from the Kansas City Cardiomyopathy Questionnaire in patients with heart failure. European Heart Journal Quality of Care and Clinical Outcomes. 2021;7(4):388-396. doi:10.1093/ehjqcco/qcab014
• US Food and Drug Administration, Center for Drug Evaluation and Research. DDT COA #000084: Kansas City Cardiomyopathy Questionnaire (KCCQ), qualified 9 April 2020. https://www.fda.gov/drugs/clinical-outcome-assessment-coa-qualification-program/ddt-coa-000084-kansas-city-cardiomyopathy-questionnaire-kccq
• US Food and Drug Administration, Center for Devices and Radiological Health. Medical Device Development Tool (MDDT) qualification decision summary for the Kansas City Cardiomyopathy Questionnaire (MDDT020), date qualified 19 October 2017. https://www.fda.gov/media/108301/download