Changes in frailty based on minimally important difference and the impact of spironolactone on frailty in heart failure with preserved ejection fraction: insights from the TOPCAT trial

Abstract

Background

The minimally important difference (MID) for frailty variation associated with adverse outcomes remains unknown in patients with heart failure and preserved ejection fraction (HFpEF), and whether spironolactone can ameliorate frailty progression in this population remains unclear.

Methods

We analysed data from 1767 participants in the Treatment of Preserved Cardiac Function Heart Failure with an Aldosterone Antagonist trial. The MID for frailty was calculated using an anchor-based approach, with the EuroQol-Visual Analogue Scale (EQ-VAS) as the anchor. Frailty index (FI), defined as a 35-item cumulative deficit score, and EQ-VAS were assessed at baseline and 1-year follow-up. The primary composite outcome (cardiovascular death, aborted cardiac arrest or heart failure hospitalisation) was assessed from the 1-year follow-up visit. Adjusted Cox proportional hazards models evaluated the link between FI changes (ΔFI≥MID) and the primary outcome. Longitudinal FI changes were analysed using linear mixed-effects models to evaluate spironolactone’s effect.

Results

The MID for the FI was 0.03 points. An FI reduction ≥MID was associated with a lower risk of the primary composite outcome (aHR, 0.63; 95% CI 0.48 to 0.82), all-cause mortality (aHR, 0.57; 95% CI 0.42 to 0.76) and heart failure hospitalisation (aHR, 0.55; 95% CI 0.40 to 0.75) after adjusting for baseline FI, age, sex, New York Heart Association class, smoking status and treatment assignment. No between-group difference in FI change was observed with spironolactone versus placebo (aOR, 0.85; 95% CI 0.67 to 1.09).

Conclusions

Frailty improvement exceeding the 0.03 FI threshold predicts better prognosis in HFpEF, underscoring the value of routine assessment. Spironolactone use was associated with neutral effects on frailty progression in our analysis, suggesting potential safety in this vulnerable population.

Trial registration number

NCT00094302.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Frailty is a well-documented comorbidity in patients with heart failure with preserved ejection fraction (HFpEF), associated with a substantially elevated risk of adverse clinical outcomes. The frailty index, a validated and widely adopted tool, serves as a key metric for assessing frailty severity in both clinical practice and research contexts.

WHAT THIS STUDY ADDS

• In patients with HFpEF, a 0.03-unit change in frailty index, representing the established minimally important difference (MID), was significantly associated with increased risk of adverse clinical outcomes.

• Spironolactone use was associated with neutral effects on frailty progression in our analysis, suggesting potential safety in this vulnerable population.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• This study suggests that the established MID in the frailty index may serve as a practical threshold for monitoring clinically meaningful changes in frailty status among HFpEF patients. Additionally, spironolactone appears to be a safe therapeutic option without adversely affecting frailty progression in this population.

Introduction

Frailty is a complex, age-related clinical syndrome characterised by reduced physiological reserve across multiple organ systems, leading to increased vulnerability to stressors, higher hospitalisation rates and mortality.^{1 2} In patients with heart failure with preserved ejection fraction (HFpEF)—a population particularly susceptible to frailty due to multisystem involvement—the prevalence of frailty ranges from 35% to 94% across studies.^{2 3} This striking variation reflects both the heterogeneity of frailty phenotypes and methodological differences in assessment tools.1,3 Identifying strategies for early intervention to ameliorate frailty and improve clinical outcomes remains a critical unmet need. While several trials have examined frailty-targeted interventions in older adults,4,6 optimal approaches for HFpEF patients remain undefined.

The frailty index (FI), developed using the Rockwood method, is widely used to assess frailty in both clinical and research settings.⁷ In patients with heart failure (HF), higher FI values—indicating greater frailty severity—are associated with worse quality of life, increased cardiovascular mortality and higher HF hospitalisation rates.8,10 A post-hoc analysis of the PARAGON-HF trial categorised FI into three strata but found no significant differences in primary outcomes across frailty levels. However, when analysed as a continuous variable, FI demonstrated a significant interaction with treatment efficacy, with the most frail patients deriving greater benefit.¹¹ This discrepancy may reflect limitations of categorical frailty classification in detecting treatment effects, whereas continuous FI captures graded associations. Given these findings, frailty changes may be better quantified using the minimally important difference (MID),¹² defined as the smallest change in a health outcome perceived as clinically meaningful.¹³ Unlike categorical classifications, MID provides a standardised threshold for defining clinically relevant changes in frailty. Previous studies proposed an MID of 0.02 for FI in community-dwelling older adults¹⁴ and 0.03 in hospitalised geriatric patients,¹⁵ but whether these thresholds apply to HFpEF remains uncertain.

Data from the Treatment of Preserved Cardiac Function Heart Failure with an Aldosterone Antagonist (TOPCAT) trial¹⁶ demonstrated that spironolactone reduced HF hospitalisations in HFpEF patients, with no attenuation of benefit by baseline frailty status.² However, it remains unknown whether longitudinal frailty progression independently predicts clinical outcomes in HFpEF and whether spironolactone can modify frailty progression trajectories. Therefore, in this study, we will analyse spironolactone’s effect on frailty changes using MID-based thresholds. Our objectives are: (1) to establish the MID for FI changes in TOPCAT’s Americas cohort; (2) to assess the association between 1-year FI changes and adverse outcomes; and (3) to determine if spironolactone affects frailty severity versus placebo.

Methods

Data of the TOPCAT study

This secondary analysis used data from the TOPCAT trial, a multicentre, randomised, double-blind, placebo-controlled study evaluating spironolactone (15–45 mg daily) in 3445 symptomatic HF patients with left ventricular ejection fraction ≥45%. The original trial design and primary outcomes have been previously reported.^{17 18} While the primary composite endpoint was neutral overall, post-hoc analysis demonstrated significant treatment benefits in the Americas cohort (n=1767 from the USA, Canada, Brazil and Argentina), with spironolactone reducing cardiovascular mortality by 26% and HF hospitalisation by 18%.¹⁶ Data were obtained from the Biologic Specimen and Data Repository Information Coordinating Center of the US National Heart, Lung, and Blood Institute, following approval of our proposal. Our analysis included 1767 participants enrolled in the Americas regions (USA, Canada, Brazil and Argentina) due to potential differences in HFpEF phenotyping among patients enrolled in Russia and Georgia.

Frailty index

We constructed a 35-item FI following Rockwood’s cumulative deficit model,¹⁹ with variables selected based on clinical relevance to frailty, availability in TOPCAT and prior validation in cardiovascular cohorts (online supplemental table S1). Each patient’s FI was calculated as the proportion of deficits present (range 0–1), with FI>0.21 defining frailty based on established cut-offs from population studies.^{20 21}

Minimally important difference

Previous studies have used the EuroQol-5D utility score as an anchor to determine the MID via anchor-based methods¹⁴ and EuroQol-Visual Analogue Scale (EQ-VAS) has demonstrated consistency with health state severity across multiple EQ-5D-defined health states.²² In this study, we employed an anchor-based MID approach following COSMIN guidelines,¹³ using EQ-VAS scores (0–100; higher=better health) assessed at baseline and 12 months as the anchor variable.²³ EQ-VAS scores were categorised into five levels: very good (80–100), good (60–79), fair (40–59), poor (20–39) and very poor (0–19), with changes from baseline to 12 months classified into seven levels: improvement (+3 (great), +2 (moderate), +1 (mild)), no change (0) and worsening (−1 (mild), −2 (moderate), −3 (great)).^{14 15} For each EQ-VAS change category, we computed mean ΔFI with 95% CIs and defined the MID as the absolute value of the mean ΔFI difference between both the ‘mild improvement’ (+1 level) and ‘mild worsening’ (−1 level) groups relative to the ‘no change’ group (0), since these differences showed comparable magnitudes. The validity of this MID required both a Pearson correlation |r|≥0.3 between ΔFI and ΔEQ-VAS and the expected negative correlation direction (where FI reduction corresponded with QoL improvement).¹³

Statistical analysis

Normally distributed continuous variables were presented as mean±SD, and non-normally distributed variables as median (IQR), assessed by Shapiro-Wilk tests. Categorical variables were summarised as frequency (%). Group comparisons used Student’s t-tests or Mann-Whitney U tests for continuous variables, and χ² or Fisher’s exact tests for categorical variables, as appropriate.

The MID was estimated by calculating the mean FI differences between the ‘mildly improved’ and ‘unchanged’ groups, and between the ‘mildly worse’ and ‘unchanged’ group. The primary outcome was a composite of death from cardiovascular causes, cardiac arrest and hospitalisation for HF, assessed from 1-year follow-up onward. Cox proportional hazards models were used to assess the relationship between FI changes and primary outcome, adjusted for baseline FI (continuous), age, sex, New York Heart Association (NYHA) class, smoking status and treatment assignment (spironolactone or placebo). Variables that were part of the FI were not adjusted for, as the FI categorisation was conditioned on these factors. A repeated measures analysis was conducted to investigate the effects of spironolactone on changes in FI over time, incorporating all follow-up time points. Time was treated as a categorical variable (months 4, 12, 24, 36, 48 and 60), and the interaction between treatment and time was tested to determine if the treatment effect on changes in each outcome varied by the time point. Furthermore, we tested for interaction between treatment group (spironolactone vs placebo) and 1-year frailty index change (ΔFI) to evaluate whether the treatment effect on the outcomes varied by longitudinal frailty progression.

Patients with >20% missing FI items or death within 1 year were excluded from the primary analysis. In sensitivity analyses, we addressed these exclusions by applying multiple imputation (assuming missing at random) for missing FI items in all patients. For surviving patients with missing EQ-VAS values, the last observation carried forward method was applied separately. A two-tailed significance level <0.05 was applied to all assessments. All analyses were performed using RStudio, version V.4.2.

Results

Baseline characteristics

Among the 1767 enrolled patients, 1355 (76.7%) had FI data with ≤20% missing items at 1 year, and EQ-VAS data were available for 1759 (99.6%). Overall, 1346 (76.2%) patients had complete EQ-VAS and FI data at follow-up. Baseline characteristics for these patients are presented in table 1, stratified by study arm. No significant between-group differences were observed. The spironolactone group showed a marginally lower baseline FI than the placebo group (0.38 vs 0.39, p=0.15), though this difference was non-significant.

Table 1. Baseline characteristics of patients with both EQ-VAS and FI data at 1 year.

	Placebo N=658	Spironolactone N=688	P value
Age, years	72±9	72±10	0.55
Female sex	326 (49.5)	340 (49.4)	1
Race
White	530 (80.5)	546 (79.4)	0.74
Black	101 (15.3)	108 (15.7)
Other	27 (4.1)	34 (4.9)
Current smoker	32 (4.9)	37 (5.4)	0.76
History of diseases
HF hospitalisation	372 (56.5)	405 (58.9)	0.42
Myocardial infarction	147 (22.3)	139 (20.2)	0.37
Atrial fibrillation	288 (43.8)	298 (43.3)	0.91
Hypertension	593 (90.1)	620 (90.1)	1
Diabetes	292 (44.4)	302 (43.9)	0.90
Ejection fraction (%)	58.2±7.9	58.6±7.8	0.40
Systolic BP (mm Hg)	127.8±15.7	127.6±15.0	0.82
Heart rate (bpm)	68.8±10.7	68.5±11.7	0.68
NYHA class 3 and 4	214 (32.5)	234 (34.1)	0.59
BMI (kg/m2)	33.8±7.8	33.9±7.9	0.79
Laboratory
Serum potassium  (mmol/L)	4.20±0.41	4.18±0.42	0.22
Serum creatinine  (µmol/L)	1.18±0.33	1.15±0.33	0.10
Albumin (g/dL)	3.67±2.80	3.45±2.39	0.11
Medications
Diuretics	586 (89.2)	609 (88.5)	0.76
ACEI/ARB	528 (80.4)	542 (78.8)	0.51
Beta-blocker	523 (79.6)	531 (77.2)	0.31
FI at baseline	0.39±0.11	0.38±0.11	0.15

Values are mean ± SD or n (%).

ACEI/ARB, angiotensin converting enzyme inhibitor/angiotensin receptor blocker; BMI, body mass index; BP, blood pressure; EQ-VAS, EuroQol-Visual Analogue Scales; FI, frailty index; HF, heart failure; NYHA, New York Heart Association.

Minimally important difference

FI changes correlated significantly with the EQ-VAS anchor (r=−0.35, p<0.001). At 1 year, FI changes stratified by clinical trajectory were: mild improvement (mean Δ −0.086 (SD 0.085), n=303), no change (mean Δ −0.056 (0.088), n=583) and mild worsening (mean Δ −0.026 (0.100), n=240) (figure 1). Using an anchor-based approach, the MID for ‘little improvement’ (transition from no change to mild improvement) was −0.030 (95% CI −0.042 to −0.018, p<0.001), while the MID for ‘little deterioration’ (transition from no change to mild worsening) was 0.030 (0.016 to 0.044, p<0.001). Based on these findings, we defined an MID threshold of |0.030| for frailty change assessments.

Figure 1. Distribution of patients into EQ-VAS categories. The figure shows the number of patients for each of the EQ-VAS categories. EQ-VAS, EuroQol-Visual Analogue Scales; FI, frailty index.

Relationship between FI change and adverse outcomes

Using the predefined MID threshold of |0.03|, patients were categorised as either: (1) improved frailty (FI decrease >0.03) or (2) deteriorated (increased) or stable frailty (FI change ≤0.03 or increase >0.03; combined due to small sample sizes). Baseline characteristics by FI change category are shown in online supplemental table S2.

During a mean 2-year follow-up period (from year 1 onward) with 259 (19.2%) primary outcomes, adjusted Kaplan-Meier analysis (figure 2) revealed that compared with patients with stable or worsened frailty, those showing frailty improvement (FI decrease >0.03) had significantly lower risks of the primary composite outcome (CV mortality, aborted cardiac arrest and HF hospitalisation; adjusted HR 0.63, 95% CI 0.48 to 0.82), all-cause mortality (adjusted HR 0.57, 0.42 to 0.76) and HF hospitalisation (adjusted HR 0.55, 0.40 to 0.75) after adjustment for baseline FI, age, sex, NYHA class, smoking status and treatment assignment (spironolactone or placebo).

Figure 2. Kaplan-Meier estimation of adverse outcomes by FI change groups. The primary outcome was a composite of cardiovascular mortality, aborted cardiac arrest and heart failure hospitalisation. Models adjusted for baseline FI, age, sex, New York Heart Association and smoking. FI, frailty index; HF, heart failure.

Effects of spironolactone on frailty

Of 1346 randomised patients, 688 (51.11%) received spironolactone and 658 (48.89%) received placebo. Frailty trajectories, categorised into three classes based on FI changes from baseline to 60 months, did not differ significantly between groups (online supplemental figure S1). Although the spironolactone group maintained lower FI levels than placebo throughout follow-up (figure 3), neither group exhibited significant within-group FI changes over time. Multivariable repeated-measures analysis confirmed no between-group difference in frailty progression (adjusted OR 0.85, 95% CI 0.67 to 1.09; p=0.20). No significant interaction was observed between treatment allocation (spironolactone vs placebo) and FI changes in affecting clinical outcomes (P_interaction=0.88 for the primary composite outcome; see online supplemental table S3 for details).

Figure 3. The frailty index at every visit in each group. FI, frailty index.

Sensitivity analyses

When including patients who died or had >20% missing FI items, the MID for ‘little improvement’ in FI was −0.018 (−0.046 to 0.011, p=0.23), while MID for ‘little deterioration’ in FI was 0.036 (0.004 to 0.068, p=0.27).

Discussion

This study provides novel evidence on two clinically important aspects of frailty in HFpEF: first, we established 0.03 as the MID threshold for FI changes, and second, we demonstrated that frailty improvement exceeding this threshold was associated with significantly reduced cardiovascular risk. Of clinical relevance, spironolactone use appeared to have neutral associations with frailty progression, which may indicate an acceptable safety profile in this population.

Frailty is highly prevalent in HF patients and portends poor clinical outcomes. Analyses from the PARADIGM-HF and ATMOSPHERE trials demonstrated a 63% prevalence of frailty in HFrEF patients, with significant associations with increased morbidity and mortality.⁸ The GUIDE-IT trial further revealed suboptimal utilisation of guideline-directed medical therapy among frail HFrEF patients,²⁴ while the HF-ACTION trial established that aerobic exercise training reduces all-cause hospitalisation risk in this population.²⁵ In HFpEF, TOPCAT post-hoc analyses showed that greater baseline frailty predicted elevated cardiovascular risk without modifying spironolactone’s treatment effects.²

In this exploratory analysis of HFpEF patients, spironolactone treatment was not associated with significant modifications in frailty trajectories. While these neutral effects suggest a plausible safety profile, the observational nature of this post hoc analysis precludes definitive conclusions. Further investigation in prospective studies designed specifically to assess frailty endpoints is warranted. Emerging evidence suggests that SGLT2 inhibitors may benefit frail HFpEF patients,²⁶ raising important questions about potential synergies with spironolactone and other HFpEF-directed therapies. Prospective, randomised trials are urgently needed to validate frailty assessment tools in HFpEF, evaluate targeted interventions for frailty and determine optimal combination therapies in this high-risk population.

Our observed mean baseline FI (0.37) was higher than previously reported values (0.21–0.26) in HF populations.^{11 26 27} This discrepancy likely reflects fundamental differences in FI construction: whereas prior studies relied primarily on binary-coded medical history data, our FI incorporated multi-category health variables from the TOPCAT dataset, resulting in fewer zero responses. This methodological approach is supported by another TOPCAT-based analysis reporting a similar FI (0.38),² validating our measure’s reliability in HFpEF. Importantly, our FI’s components are clinically accessible, enabling reproducible frailty assessment in routine practice.

Frailty represents a dynamic clinical state, with FI-derived MID serving as a key metric for assessing its progression.²⁸ While prior studies reported MID values of 0.08–0.11 for the FI in community-dwelling older adults and 0.037 in nursing home residents,^{12 14 29} our study identifies a significantly lower MID (0.03) in HFpEF patients. This finding likely reflects the heightened clinical vulnerability characteristic of HF patients, where even small FI changes (|0.03|) may carry substantial clinical implications. Notably, patients achieving frailty improvement had lower baseline FI scores than non-responders, suggesting greater treatment responsiveness in those with more advanced baseline frailty—an observation aligning with existing literature.

Frailty progression substantially elevates vulnerability to adverse clinical outcomes. A pooled analysis of three prospective cohorts demonstrated that transitions from robust to prefrail or frail states over 2 years significantly increased cardiovascular disease risk.³⁰ Mechanistically, frailty and HF share pathophysiological underpinnings including multimorbidity burden, chronic inflammation and sarcopenia.³¹ In HF patients, these interlinked processes create a vicious cycle: frailty exacerbation reduces physical capacity and amplifies symptom burden, while simultaneously increasing fall risk and hospitalisation rates—all of which may further deteriorate cardiac function.^{31 32} Our findings underscore the imperative for routine frailty assessment in HFpEF management protocols to disrupt this deleterious cycle and improve outcomes.

Study limitations

Several limitations warrant consideration. First, the FI was constructed retrospectively using cardiovascular risk factors, comorbidities and quality-of-life measures, which may not capture all relevant frailty domains. Nevertheless, its consistency with previously validated FIs in HFpEF populations supports its applicability. Second, the MID determination relied exclusively on EQ-VAS changes over 12 months, an approach potentially biased by substantial mortality during this period. Third, using EQ-VAS as an anchor may introduce bias since its changes can reflect factors beyond true frailty progression. Although prior studies support EQ-5D as an anchor and demonstrate EQ-VAS/EQ-5D correlations, future studies should validate alternative measures (eg, KCCQ or objective frailty indicators) for more specific frailty assessment. Fourth, while the primary composite endpoint analysis appropriately accounted for competing events by design, our secondary analysis of HF hospitalisation alone did not formally address competing risks (particularly all-cause death). These exploratory findings should therefore be interpreted as hypothesis-generating, and future studies should employ competing-risk models for such analyses. At last, as a post-hoc analysis of the TOPCAT trial, our findings may be limited by the original study’s sample size and design. Prospective studies with larger cohorts and extended follow-up are needed to confirm these observations.

Conclusion

In conclusion, we established 0.03 as the MID for FI changes in HFpEF patients, a threshold associated with clinically meaningful improvements in health status, quality of life and adverse outcomes. Notably, patients achieving FI improvements exceeding this MID demonstrated significantly reduced risks of all-cause mortality and HF hospitalisation compared with those with stable/worsening frailty. These findings underscore the prognostic value of frailty monitoring and the urgent need for targeted interventions to modify frailty trajectories in HFpEF management.

Data availability statement

Data may be obtained from a third party and are not publicly available.