Clinical status in heart failure is conventionally assessed by the physician’s evaluation, patients’ own perception of their symptoms, quality of life (QoL) tools, and a measure of functional capacity. These aspects can be measured with tools such as the New York Heart Association functional class, QoL tools such as the EuropeanQoL-5 dimension, the Kansas City Cardiomyopathy Questionnaire, patient global assessment (PGA), and by 6-minute walk test (6MWT), respectively. The ferric carboxymaltose in patients with heart failure and iron deficiency (FAIR-HF) trial demonstrated that treatment with intravenous ferric carboxymaltose in iron-deficient patients with symptomatic heart failure with reduced left ventricular function, significantly improved all 5 outcome measures. This analysis assessed the correlations between the longitudinal changes in the measures of clinical status, as measured by QoL tools and the changes in the measures of functional capacity as measured by the 6MWT. This analysis used the database from the FAIR-HF trial, which randomized 459 patients with chronic heart failure (reduced left ventricular ejection fraction) and iron deficiency, with or without anemia to ferrous carboxymaltose or placebo. The degree of correlation between QoL tools and the 6MWT was assessed at 4, 12, and 24 weeks. The data demonstrate highly significant correlations between QoL and functional capacity, as measured by the 6MWT, at all time points (p <0.001). Changes in PGA, Kansas City Cardiomyopathy Questionnaire, and EuroQoL-5D correlated increasingly over time with changes in 6MWT performance. Interestingly, the strongest correlation at 24 weeks is for the PGA, which is a simple numerical scale ( r = −0.57, p <0.001). This analysis provides evidence that QoL assessment show a significant correlation with functional capacity, as measured by the 6MWT. The strength of these correlations increased over time.
There are few studies assessing correlation between quality of life (QoL) tools and exercise capacity in patients with heart failure. Previously, weak but statistically significant correlations have been found between the Minnesota Living with Heart Failure Questionarie and exercise capacity, but a substudy of the Heart Failure: A Controlled Trial Investigating Outcomes of Exercise Training (HF-ACTION) population did not find any strong correlations between QoL tools and functional capacity. One small study assessed correlations between Kansas City Cardiomyopathy Questionnaire (KCCQ) and functional capacity measured by the 6-minute walk test (6MWT) in a small study of patients with heart failure. They found a strong correlation in patients recovering from decompensated heart failure but did not find any significant correlation in stable patients. As a result, further assessment of clinical implication and internal validation of these end points are warranted. The FAIR-HF trial collected data on symptoms, QoL, and 6MWT during the follow-up period of 24 weeks. These substantial and sustained changes after the intervention in the FAIR-HF study provide a unique opportunity to quantify the degree of concordance between clinical status and the 6MWT. We specifically investigated the correlations between patient functional capacity, as measured by the 6MWT and different health-related QoL tools. In addition, we quantified the strength between these clinically meaningful assessments during longitudinal follow-up in a well-defined heart failure population.
The analyses presented in this report are based on the data collected in the FAIR-HF study which has been previously reported. Briefly, FAIR-HF was a randomized, multicenter, double-blind, placebo-controlled study investigating the effects of ferric carboxymaltose (FCM) in iron-deficient patients with symptomatic heart failure with reduced left ventricular function, with or without anemia. The main study inclusion criteria were New York Heart Association (NYHA) class II/III, left ventricular ejection fraction ≤40% (NYHA II) or ≤45% (NYHA III), hemoglobin levels between 95 and 135 g/L, and iron deficiency. Iron deficiency was defined as either serum ferritin <100 μg/L, or serum ferritin <300 μg/L if transferrin saturation was ≤20%.
The substantial treatment effect in the FAIR-HF enabled us to assess correlations between the efficacy measures. The FAIR-HF study showed a positive treatment effect for FCM over placebo. In particular, significant treatment differences were observed for the coprimary efficacy end points, patient global assessment (PGA) at week 24, and change in NYHA class at week 24 (p <0.001). In addition, statistical significance in favor of the FCM group was seen for the 6MWT, QoL (when evaluated using the EuroQoL-5 dimension (EQ-5D) visual assessment score (VAS) and also for the KCCQ overall summary score [OSS]) at weeks 4, 12, and 24 (p <0.001 for all comparisons). Further details on the results of the efficacy and safety end points are described in Anker et al. Summary analyses presented in this report are based on the 6MWT, self-reported patient PGA, EQ-5D VAS, and KCCQ.
The 6MWT was performed in accordance to the protocol developed by Guyatt et al. Subjects were allowed to rest on the chairs during the test but were encouraged to resume walking as soon as they felt physically capable to do so. The distance walked in 6 minutes, to the nearest meter, was recorded in meters. Patients performed the test at baseline, 4, 12, and 24 weeks after the start of treatment. Missing 6MWT values in subjects who were known to be alive and not hospitalized at the time of the test were not imputed (i.e., test was considered as missing). For subjects who were hospitalized and unable to exercise, the worst nonnull test across the study was used. For subjects who had died, the value was set to zero.
The analysis of treatment difference on the 6MWT was compared using the least squares means of the respective visit based on a model for repeated measures including terms for treatment, baseline, time, and treatment-by-time with an unstructured covariance matrix to model the within-patient variability. Summary of the 6MWT distances, change from baseline, and treatment difference over the study time points are presented in Table 1 in the result section.
|Study time point|
|Baseline||Week 4||Week 12||Week 24|
|Ferric Carboxymaltose (N=304)|
|Distance covered (meters) ∗||274±6||294±7||313±7||313±7|
|Change from baseline (meters) †||22±3||39±4||39±5|
|Distance covered (meters) ∗||269±9||269±10||272±10||277±10|
|Change from baseline (meters) †||1±5||2±5||5±6|
|Change from baseline (meters) †||21±6||37±6||35±8|
|p -value †||<0.001||<0.001||<0.001|
The EQ-5D is a standardized non–disease-specific instrument measuring health outcome. It has been used widely in clinical trials for assessing QoL and has been established as a valid QoL assessment in heart failure. The EQ-5D VAS is on a vertical, graduated (0 to 100 points) scale, with a score of 100 representing “best imaginable health state” and 0 representing “worst imaginable health state”. Missing values in subjects who were known to be alive and not hospitalized were imputed from the previous follow-up visit using last observation carried forward (LOCF). If a subject was hospitalized and unable to complete the questionnaire, or if the subject had died, the worst reply was used for the imputation (i.e., 0 for the EQ-5D VAS).
The KCCQ is a self-administered instrument for measuring health-related QoL that is specific to heart disease and has been validated for patients with heart failure. It is a 23-item questionnaire that quantifies physical function, symptoms (frequency, severity, and recent change), social function, self-efficacy and knowledge, and QoL. The OSS is a weighted average of the different domains and has a continuous scale that ranges from 0 to 100, with a higher score indicating a better QoL. Missing values in subjects who were known to be alive and not hospitalized were imputed using LOCF. If a subject was hospitalized and unable to complete the questionnaire, or if the subject had died, the worst reply was used for the imputation (i.e., 1 for the KCCQ subquestions).
The self-reported PGA tool is a numeric scale (1 to 8) where the subject subjectively reports functioning or response to an intervention. PGA has been used extensively in clinical trials and correlates well with other QoL tools in patients with heart failure.
The FAIR-HF subjects completed the PGA in a quiet place and before any other assessment, interview, or drug administration, so that their assessment was not influenced by study procedures. Taking into account subjects who were hospitalized at that time or had died, the PGA was ranked as follows:
A little improved
A little worse
Missing PGA values were imputed from the previous follow-up visit using LOCF for subjects who were known to be alive and not hospitalized at the time of the assessment. For hospitalized subjects, the missing PGA value was ranked as “much worse” and for subjects who had died; the value was ranked as “died.” All QoL tools were self-administered at baseline and 4, 12, and 24 weeks after start of treatment.
Consistent with Anker et al, the analyses presented are based on the full analysis set population. Baseline characteristic summary was described using count and percentages for categorical variables and mean and SDs for continuous variables. The treatment comparison analyses presented for 6MWT, EQ-5D VAS, KCCQ OSS, and PGA are taken from Anker et al and consistent with the FAIR-HF clinical study report. In summary, for the continuous variables 6MWT, EQ-5D VAS, and KCCQ OSS change from baseline values at weeks 4, 12, and 24 were analyzed using a model for repeated measures with treatment (FCM vs placebo) as a factor and baseline score as a covariate. Statistical presentation includes means and standard errors split by treatment group and least squares means, standard errors and associated p values. For the categorical variable, PGA, treatment comparison between FCM and placebo was tested using ordered polytomous regression. Presentation for the PGA analysis includes the count and percentages of the reported scores split by treatment group and odds ratios, 95% CIs, and associated p values. Correlations between 6MWT and PGA, EQ-5D VAS, and KCCQ OSS were based on Pearson product-moment correlation coefficient ( r ). These correlations were further explored graphically. Consistent with the methods presented in Anker et al, all data used in the correlation analyses except PGA was adjusted for the respective baseline score. Statistical significance was assumed to be at p <0.05. For the 6MWT, normal missing values were not imputed, and for the other end points, the rate of imputation was low, being less than 2%. Statistical analyses were performed using SAS version 9.1 and R.
Of the 459 subjects who were randomized and started treatment, 304 were assigned to FCM and 155 to placebo (intention-to-treat/full analysis set population). Subjects were matched for baseline clinical and laboratory characteristics and the use of cardiovascular medications at the time of enrollment. Summary baseline characteristics split by treatment group are presented in Table 2 .
( N =304)
( N =155)
( N =459)
|Female||159 (52.3%)||85 (54.8%)||244 (53.2%)|
|II||53 (17.4%)||29 (18.7%)||82 (17.9%)|
|III||251 (82.6%)||126 (81.3%)||377 (82.1%)|
|Left ventricular ejection fraction||31.9±5.5||33.0±6.1||32.3±5.7|
|6-minute walk distance (meters)||274±105||269±109||272±106|
|Ischaemic cause of HF||245 (80.6%)||123 (79.4%)||368 (80.2%)|
|Atrial fibrillation||94 (30.9%)||44 (28.4%)||138 (30.1%)|
|Hypertension||243 (79.9%)||128 (82.6%)||371 (80.8%)|
|Coronary revascularisation||64 (21.1%)||31 (20.0%)||95 (20.7%)|
|Serrum ferritin (μg/L)||52.5±54.5||60.1±66.5||55.0±58.9|
|Transferrin saturation (%)||17.7±12.6||16.7±8.4||17.3±11.4|
|Diuretic||283 (93.1%)||141 (91.0%)||424 (92.4%)|
|ACE inhibitor or ARB||282 (92.8%)||141 (91.0%)||423 (92.2%)|
|Beta-blocker||263 (86.5%)||129 (83.2%)||392 (85.4%)|