Long-Term Survival of Very Elderly Patients with Heart Failure Managed in a Comprehensive Heart Failure Management Unit

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript provides a robust picture of the impact of comorbidities on the prognosis of very elderly heart failure patients. The study addresses a particular geriatric population segment (age ≥ 85 years), namely with a high prevalence of HF, many comorbidities and few data on long-term outcomes. The study covered a period of 12 years.

Major comments:

It is worth considering the contribution of modern therapies to improving survival. In this context, I believe it is useful to deepen some information.

- comments on coronary instrumentation (stent, bypass), valve prostheses, implantable cardiac devices (pacemaker, defibrillator, cardiac resynchronization therapy), as they are known to be important modifiers of prognosis.

- comments on modern pharmacological therapies. It can be noted that less than half of patients have medication that targets the RAAS, 10% have ARNI, and less than a quarter have SGLT2 inhibitors. An opinion on the role of these modern therapies (ARNI, SGLT2 inhibitors) in terms of outcome would be of great interest.

Minor comments:

Mortality was associated with ... prior HF. Authors are asked to consider whether “history of HF” is not more appropriate than “prior HF”.

The authors included only patients with HF and they monitored admissions for HF. Therefore, they are asked to provide clarifications on table 4: „Prior admission” and „ HF admission in the year prior to follow-up”. I recommend that it be clearly specified that it is prior admission for HF.

 

Thank you!

Author Response

Author's Reply to the Review Report (Reviewer 1)

Dear Editorial Team and Reviewer 1,

We would like to express our sincere appreciation for the careful and comprehensive evaluation of our manuscript. We are grateful for your thoughtful and constructive comments, which have provided valuable guidance for improving the quality, clarity, and scientific rigor of our work.

In accordance with your recommendations, we have thoroughly revised the manuscript and carefully addressed each of the points raised. A detailed, point-by-point response is provided below, and all corresponding modifications have been incorporated into the revised manuscript.

We trust that these revisions have adequately addressed your concerns and have substantially strengthened the manuscript. We sincerely appreciate your time and consideration throughout the review process.

Major comments

Comments 1: Comments on coronary instrumentation (stent, bypass), valve prostheses, implantable cardiac devices (pacemaker, defibrillator, cardiac resynchronization therapy), as they are known to be important modifiers of prognosis.

Response 1: We thank the reviewer for highlighting this important point. We agree that previous coronary revascularization, valve interventions or prostheses, and cardiac implantable electronic devices may influence prognosis in patients with heart failure. Unfortunately, this information was not systematically collected in our retrospective database and therefore could not be included in the analyses. We have now explicitly acknowledged this in the Limitations section, noting that the absence of these data represents an additional potential source of residual confounding and that the prognostic impact of these interventions and devices in patients aged ≥85 years with heart failure warrants further investigation.

 

Added text:

Limitations: Information on previous coronary revascularisation, valve interventions or prostheses, and cardiac implantable electronic devices was not systematically collected and could therefore not be included in the analyses. As these interventions may influence outcomes, the absence of these data represents a further potential source of residual confounding, and their prognostic impact in patients aged ≥85 years with heart failure warrants further investigation.

Comments 2: Comments on modern pharmacological therapies. It can be noted that less than half of patients have medication that targets the RAAS, 10% have ARNI, and less than a quarter have SGLT2 inhibitors. An opinion on the role of these modern therapies (ARNI, SGLT2 inhibitors) in terms of outcome would be of great interest.

 Response 2: We appreciate this thoughtful comment, which prompted us to address the role of contemporary disease-modifying pharmacological therapies more explicitly. As the reviewer notes, the uptake of these agents was limited in our cohort (SGLT2 inhibitors, 22.6%; ARNI, 10.3%). This may be explained, at least in part, by two main factors. First, recruitment spanned 2012–2023, whereas SGLT2 inhibitors were incorporated into guideline recommendations for heart failure with reduced ejection fraction only in 2021 and extended across the full spectrum of left ventricular ejection fraction only in the 2023 focused update, at the very end of our recruitment period. Second, only 13.6% of patients had reduced ejection fraction, limiting the proportion of patients for whom ARNI was guideline-recommended on the basis of heart failure phenotype during the study period.

We have added a paragraph to the Discussion addressing these points. We acknowledge that, despite the established prognostic benefit of SGLT2 inhibitors and ARNI in heart failure, their uptake in very elderly patients remains suboptimal, as has been reported in real-world cohorts of patients aged ≥80 years [Barry et al 2024]. This may reflect both the underrepresentation of this population in pivotal trials and concerns regarding tolerability, polypharmacy, and frailty. We further highlight that additional studies are needed to better define the implementation, tolerability, and net clinical benefit of these therapies specifically in patients of very advanced age.

Change to the manuscript: 

Discussion: The limited use of contemporary disease-modifying therapies in this cohort should be interpreted in the context of the long recruitment period and the evolving therapeutic landscape of heart failure. SGLT2 inhibitors were incorporated into guideline recommendations for heart failure with reduced ejection fraction only in 2021[15] and extended across the full spectrum of left ventricular ejection fraction in the 2023 focused update[39], at the very end of our recruitment period. In addition, only 13.6% of patients had reduced ejection fraction, limiting the proportion of patients for whom ARNI was guideline-recommended on the basis of heart failure phenotype during the study period. Evidence regarding the use of contemporary disease-modifying therapies in patients of very advanced age remains comparatively limited, and suboptimal uptake has also been reported in cohorts of patients aged ≥80 years[40]. Further studies specifically addressing the implementation, tolerability, and clinical impact of these therapies in patients of very advanced age are warranted.

References:

39. McDonagh TA, Metra M, Adamo M, Gardner RS, Baumbach A, Böhm M, et al. 2023 Focused Update of the 2021 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure. Eur Heart J. 2023;44(37):3627- 39. https://doi.org/10.1093/eurheartj/ehad195.
40. Barry AR, Grewal M, Blain L. Use of guideline-directed medical therapy in patients aged 80 years or older with heart failure with reduced ejection fraction. CJC Open. 2023;5(4):303-9. https://doi.org/10.1016/j.cjco.2023.01.002.

Are the conclusions supported by the results? Must be improved

In response to the reviewer’s assessment, we have revised the Conclusions to more closely reflect the observational nature of our findings and to avoid causal interpretation. We have also added a sentence acknowledging that the prognostic contribution of contemporary disease-modifying therapies and cardiovascular interventions and devices could not be assessed in this cohort and warrants further investigation.

Conclusions: In this cohort of patients aged ≥85 years with heart failure, long-term survival was poor, with approximately half alive at 2.5 years, and only one quarter at 5 years. Prognosis appeared to be more closely associated with clinical complexity than with chronological age. Comorbidity burden assessed using the Charlson Comorbidity Index, functional and cognitive status assessed using the Barthel Index and Pfeiffer test, NYHA functional class and NT-proBNP were independent predictors of mortality. Early HF readmission emerged as the strongest predictor when included in the model. These findings suggest that a comprehensive multidimensional assessment incorporating these domains may be useful in the management and prognostic evaluation of this population. The prognostic contribution of contemporary disease-modifying therapies and of cardiovascular interventions and devices could not be assessed in this cohort and warrants further investigation.

Minor comments:

Comments 3: Mortality was associated with ... prior HF. Authors are asked to consider whether “history of HF” is not more appropriate than “prior HF”.

Response 3: We agree with the reviewer and have replaced "prior HF" with "history of HF" at its only occurrence in the manuscript.

Comments 4: The authors included only patients with HF and they monitored admissions for HF. Therefore, they are asked to provide clarifications on table 4: „Prior admission” and „ HF admission in the year prior to follow-up”. I recommend that it be clearly specified that it is prior admission for HF.

Response 4: The requested clarification has been incorporated into Table 4. Specifically, "Prior admission" has been changed to "Prior HF admission" to explicitly indicate that it refers to hospital admission for heart failure.

 

Once again we would like to thank you for taking the time to read our article and for your constructive recommendations.

Regards.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Thank you for the opportunity to reveiw this interesting manuscript on the long-term survival of very elderly patients with heart failure managed in a specialised outpatient unit.

This is a relevant and clinically important study describing long-term outcomes and prognostic factors in a large cohort (n=514) of very elderly patients (>=85 years) with heart failure followed in a comprehensive HF management unit over more than 10 years. The focus on >=85-year-old patients, most with HFpEF and high multimorbidity, fills a clear evidence gap and will be of interest to cardio-geriatrics and HF clinicians. The main findings (poor 5-year survival; prognostic role of comorbidity burden, NT-proBNP and early readmission) are consistent and clinically plausible. 

Below are certain comments and suggestions to enhance the manuscript:

- Consider specifying "comprehensive heart failure management unit" or "UMIPIC programme" in the title, as this context is central to the intervention setting and may interest readers looking for models of care.
- In the abstract, please clarify whether survival percentages were derived from Kaplan-Meier estimates or crude proportions, and specify the median follow-up in the Methods part (currently only in Results).
- You report two multivariable models (with and without HF readmissions); in the abstract you mention Charlson index, NT-proBNP and readmissions as independent predictors, but in the model including readmissions Barthel and Pfeiffer fall out or become borderline. I suggest explicitly indicating that the first model (without readmissions) also identified cognitive and functional status as independent predictors, whereas in the second model, after including 1-year readmissions, Charlson index and NT-proBNP remained significant.

Introduction

- The background is generally adequate and appropriately referenced, but some paragraphs are quite dense and could be streamlined. For example, the description of cardiac and vascular ageing could be shortened to one concise sentence, focusing more on the evidence gap in >=85-year-old HF patients.
- Please sharpen the knowledge gap and aims at the end of the Introduction: e.g. specify that previous work on very elderly HF has largely focused on short-term outcomes after hospitalisation, whereas your study provides long-term survival and multivariable prognostic modelling in a structured outpatient programme.

Methods - design, setting and population

- Please state explicitly whether this is a single-centre study and confirm whether all consecutive eligible patients evaluated in the UMIPIC unit during 2012-2023 were included (consecutiveness affects risk of selection bias).
- Clarify whether patients were referred after an index HF hospitalisation, from primary care, or from other hospital services, and whether there were any criteria for referral to the UMIPIC unit that could limit generalisability.
- For exclusion criteria, you state that patients "lost to follow-up after the first visit" were excluded. Please define the minimum follow-up required (e.g. no further contact, or unknown vital status) and provide the number of patients excluded for this reason.

Methods - variables and definitions

- Indicate the calendar time at which Barthel, Pfeiffer, NT-proBNP and Charlson index were assessed (baseline visit) and clarify whether the Barthel and Pfeiffer scores were obtained systematically by trained staff or from routine clinical documentation.
- For Charlson index, specify whether age-adjusted or non-age-adjusted Charlson was used; for patients >=85, age-adjustment can substantially change the score and interpretation.
- Please define "new-onset HF" in more detail (first diagnosis within a specified time frame, or first hospitalisation, or first contact with cardiology/UMIPIC).
- For HF categories (HFrEF, HFmrEF, HFpEF), clarify whether LVEF was taken from the echocardiogram closest to the baseline visit and whether the same measurement was used for all patients.
- It would be useful to indicate whether frailty per se was assessed (e.g. Clinical Frailty Scale, gait speed, or similar) and to acknowledge explicitly in Methods that no formal frailty tool was used (currently only discussed as a limitation).

Methods - statistical analysis

- You mention that NT-proBNP was dichotomised using the Youden index at 2285 pg/mL and also log-transformed for Cox regression. Please clarify precisely how NT-proBNP entered each model (log-continuous versus dichotomous), and in Tables 5-6 label clearly whether "High NT-proBNP" refers to the cut-off determined by Youden.
- Provide some detail on how variables were selected for the multivariable Cox models: were all variables with p<0.05 in univariate analysis included, or did you apply any clinical pre-selection or stepwise procedure? Also, did you assess proportional hazards assumptions (e.g. Schoenfeld residuals, log-log plots)?
- Please report how missing data were handled. For several variables, denominators are <514 (e.g. Barthel, Pfeiffer, eGFR, HF readmission). Were these cases excluded listwise from the multivariable models? Clarifying this will help readers interpret any potential bias.
- Consider adding a brief statement on sample-size considerations for multivariable modelling (e.g. number of events per variable).

Results - baseline characteristics and treatments

- Baseline description is comprehensive. To improve readability, you might move some detailed percentages from text to tables and focus the text on key messages (e.g. "multimorbidity was marked, with median Charlson index 3 and >50% of patients having CKD, AF and anaemia").
- Given the age and comorbidity profile, it would be interesting to report the distribution of HF phenotype (HFrEF/HFmrEF/HFpEF) explicitly, not just "vast majority HFpEF" (e.g. HFpEF 73.9%, HFmrEF x%, HFrEF y%).
- For drug therapy, please indicate:
- In how many patients SGLT2i or ARNI were available (e.g. calendar-time effect);
- Whether prescription patterns differed by HF phenotype (not essential, but a short comment would situate the cohort against current guideline-directed therapy).

Results - outcomes and modelling

- You state that survival was 77% at 1 year and 25% at 5 years, while the abstract also mentions 50% at 2.5 years (median survival 30.5 months). It would be helpful to include the 2.5-year Kaplan-Meier survival estimate in the Results text and/or Figure 1 caption.
- Age is reported as not associated with survival in this >=85 cohort. It might be informative to provide HR per 5-year increase (e.g. 85-<90 vs >=90) to give readers a sense of effect size.
- In univariate analyses, several comorbidities (e.g. advanced CKD, diabetes, MI, dementia, cancer) were significantly associated with mortality, but only Charlson index remains in the final model. It would be worth briefly noting in Results that when Charlson index is entered, individual comorbidities lose significance, suggesting collinearity with overall comorbidity burden.
- Please ensure that the direction of HRs is always intuitive and consistent with how the variable is coded. For example, in Table 5 the HR for Barthel is 0.99 per point (protective), but in the text you sometimes refer to "functional dependence (lower Barthel)" as a risk factor. Clarifying that the HR is per unit increase in score will avoid confusion.
- In Table 6, consider showing the number of patients/events included in the model with HF readmissions (since n is smaller due to missing data on readmission status).

Discussion

- The Discussion is well-referenced and appropriately situates the cohort in the context of existing HF literature. To improve clarity, I suggest:
 - Condensing some of the comparisons with younger cohorts (several paragraphs) into a shorter synthesis, then expanding slightly on how your findings could influence risk stratification and follow-up in >=85-year-old patients.
  - More clearly separating observations derived from your baseline model (Charlson, NT-proBNP, Barthel, Pfeiffer, NYHA) from those of the model including HF readmissions, where readmission becomes a very strong predictor.
- When discussing the lack of association between age and survival, you might emphasise that the restricted age range likely limits statistical power to detect small differences and that markers of biological age (comorbidity, function, cognition) are more informative in this group.

Limitations

- The limitations are appropriately acknowledged (retrospective design, single-centre, missing formal frailty assessment). Two additional points could be mentioned:
  - Potential survivorship bias, since only patients referred and able to attend the outpatient unit were included; the sickest institutionalised or bed-bound patients may have been under-represented.
  - Possible unmeasured confounding by socio-economic status or caregiver support, which could be particularly important in this age group.

Language and style

- Overall, the English is understandable, but there are several minor grammatical issues and some long sentences that could be streamlined. Examples include: "included in an structured outpatient follow-up programme" (an structured -> a structured) or "multimorbidity HF" (multimorbid HF or HF with multimorbidity).

Overall, with clarification of methods, more precise statistical reporting, and language polishing, this manuscript provides valuable data on prognosis and multidimensional assessment in very elderly HF patients managed in a specialised outpatient unit.

Author Response

Author's Reply to the Review Report (Reviewer 2)

Dear Editorial Team and Reviewer 2,

We would like to express our sincere appreciation for the careful and comprehensive evaluation of our manuscript. We are grateful for your thoughtful and constructive comments, which have provided valuable guidance for improving the quality, clarity, and scientific rigor of our work.

In accordance with your recommendations, we have thoroughly revised the manuscript and carefully addressed each of the points raised. A detailed, point-by-point response is provided below, and all corresponding modifications have been incorporated into the revised manuscript.

We trust that these revisions have adequately addressed your concerns and have substantially strengthened the manuscript. We sincerely appreciate your time and consideration throughout the review process.

 

Title

Comments 1: Consider specifying "comprehensive heart failure management unit" or "UMIPIC programme" in the title, as this context is central to the intervention setting and may interest readers looking for models of care.

Response 1: We fully agree with the reviewer that specifying the management setting is of interest, as care delivered within a comprehensive heart failure management unit represents an important aspect of the study and may be relevant to readers interested in models of care. Accordingly, we have revised the title, which now reads:

"Long-term survival of very elderly patients with heart failure managed in a comprehensive heart failure management unit."

Abstract-Methods

Comments 2: In the abstract, please clarify whether survival percentages were derived from Kaplan-Meier estimates or crude proportions, and specify the median follow-up in the Methods part (currently only in Results).

Response 2: Both points have been addressed. To clarify the origin of the survival estimates, the Abstract now specifies that survival was analysed with the Kaplan–Meier method. In addition, the median follow-up has been incorporated into the Methods section, where it was previously absent.

Changes to the manuscript:

Abstract

Methods. […]Long-term survival was estimated using the Kaplan–Meier method, and predictors of mortality were analysed using Cox regression models.

Design and study population. […] The sample size was 514 patients. Patients were followed from the first visit until death or the last recorded contact; the median follow-up was 22.6 months (interquartile range [IQR] 10.6-40.3 months).

Comments 3: You report two multivariable models (with and without HF readmissions); in the abstract you mention Charlson index, NT-proBNP and readmissions as independent predictors, but in the model including readmissions Barthel and Pfeiffer fall out or become borderline. I suggest explicitly indicating that the first model (without readmissions) also identified cognitive and functional status as independent predictors, whereas in the second model, after including 1-year readmissions, Charlson index and NT-proBNP remained significant.

Response 3: We agree with the reviewer's observation that both multivariable models deserve to be reflected in the Abstract. Accordingly, the sentence has been revised to describe them separately.

Changes to the manuscript:In multivariable analysis, independent predictors of mortality were In the first multivariable model, worse cognitive and functional status, a higher Charlson comorbidity index, NYHA class III–IV, and higher NT-proBNP were independent predictors of mortality. When 1-year HF readmission was added to the model, the Charlson index (HR:1.101; 95%CI 1.01-1.19), elevated high NT-proBNP (HR:1.75; 95%CI 1.30-2.38) and HF readmissions within 1 year (HR:2.68; 95%CI 1.72-4.16) remained independently associated with mortality.

 

Introduction

Comments 4: The background is generally adequate and appropriately referenced, but some paragraphs are quite dense and could be streamlined. For example, the description of cardiac and vascular ageing could be shortened to one concise sentence, focusing more on the evidence gap in >=85-year-old HF patients.

Response 4: This passage was indeed overly dense. The description of cardiac and vascular ageing has been condensed into a single sentence, which allows the paragraph to focus more directly on the clinical profile of older patients. The evidence gap in patients aged ≥85 years is addressed in the subsequent paragraph, where we note that the clinical course of very elderly patients with HF remains poorly characterised despite representing the fastest-growing segment of the HF population; streamlining the preceding paragraph gives this point greater prominence. The original references have been retained.

Changes to the manuscript:

HF predominantly affects older adults, whose clinical, pathophysiological, and epidemiological characteristics differ substantially from those of younger individuals[4]. Aging leads to structural cardiac changes such as left ventricular hypertrophy, increased myocardial fibrosis, reduced ventricular compliance and diastolic dysfunction. Concomitantly, vascular ageing leads to arterial stiffening and endothelial dysfunction, while cardiac valves undergo fibrotic and calcific changes[5,6]. Ageing promotes structural cardiac, vascular, and valvular changes—including myocardial fibrosis, diastolic dysfunction, and arterial stiffening—that predispose to HF[5,6]. In addition, clinical manifestations may often be atypical or subtle and may remain unrecognised until severe decompensation occurs[4].

 

Comments 5: Please sharpen the knowledge gap and aims at the end of the Introduction: e.g. specify that previous work on very elderly HF has largely focused on short-term outcomes after hospitalisation, whereas your study provides long-term survival and multivariable prognostic modelling in a structured outpatient programme.

Response 5: The end of the Introduction has been sharpened as suggested. The knowledge gap now specifies that evidence in very elderly patients with HF has largely addressed short-term outcomes, frequently following hospitalisation for acute decompensation, and contrasts this with the present focus on long-term survival in a structured outpatient programme. The aim has been reworded accordingly to state that the study evaluates long-term survival and identifies independent prognostic factors in this population.

Changes to the manuscript:

The limited clinical evidence available for the management of these complex patients[8] together with the high healthcare burden associated with HF[9], has led to the development of care models based on structured and continuous management such as the Comprehensive Management Units for Patients with Heart Failure (UMIPIC programme), designed for patients with multimorbid HF and advanced age[10,11]. Although previous studies have demonstrated reductions in hospital admissions and emergency department visits among patients followed in these programmes[11], the clinical course of very elderly patients (≥85 years) with HF remains poorly characterised, despite representing the fastest growing segment of the HF population. Moreover, available evidence in this age group has largely focused on short-term outcomes, frequently following hospitalisation for acute decompensation, whereas long-term survival and its determinants in patients managed within structured outpatient programmes have been far less explored[12-14]. A better understanding of their prognosis and clinical profile may help optimise management and improve care strategies.

Therefore, the aim of this study was to evaluate long-term survival of patients aged ≥85 years with HF included in a structured outpatient follow-up programme, as well as to describe their clinical characteristics and to identify independent prognostic factors.

 

Methods - design, setting and population

Comments 6: Please state explicitly whether this is a single-centre study and confirm whether all consecutive eligible patients evaluated in the UMIPIC unit during 2012-2023 were included (consecutiveness affects risk of selection bias).

Response 6: The reviewer raises a valid methodological point. The Methods now state explicitly that this is a single-centre study and that recruitment was consecutive: all patients aged ≥85 years evaluated at the UMIPIC unit during the study period were included, with the sole exception of those meeting the predefined exclusion criteria, and no further selection was applied. This clarification addresses the potential risk of selection bias.

"We conducted a single-centre retrospective observational study that included all consecutive patients aged 85 years or older who were evaluated at the Comprehensive Management Unit for Patients with Heart Failure (UMIPIC) of a tertiary hospital from July 2012 to May 2023, except for those meeting the exclusion criteria described below."

 

Comments 7: Clarify whether patients were referred after an index HF hospitalisation, from primary care, or from other hospital services, and whether there were any criteria for referral to the UMIPIC unit that could limit generalisability.

Response 7: This is an important point, and addressing it has allowed us to characterise the cohort more accurately. The Methods now specify the referral pathways and clarify that, although there were no strict referral criteria, referral depended on clinical judgement and typically involved patients with more advanced HF whose functional status still permitted transfer to the hospital. We have expanded the Limitations accordingly.

Methods: “…The sample size was 514 patients. Patients were referred to the unit mainly after discharge from an HF hospitalisation or from the emergency department following a decompensation not requiring admission, and less frequently from other hospital services or from primary care. There were no strict referral criteria; however, referral was based on the responsible physician's judgement and typically involved patients with more advanced HF that was difficult to manage in primary care and/or with frequent decompensations or admissions, whose functional status still allowed transfer to the hospital.”

Limitations: “This study has several limitations inherent to its retrospective design, including potential information bias due to incomplete data recording and missing data for some clinical and laboratory measures. In addition, the single-centre nature of the study may limit the generalisability of the findings. Moreover, patients were referred on the basis of clinical judgement and had to be able to attend the outpatient unit. Consequently, the study population tended to include patients with more advanced heart failure but preserved functional status, which may limit the representativeness of the cohort.

.… Moreover, the structured and multidisciplinary nature of this care model may enhance the applicability and generalisability of the findings to comparable HF units, and could help inform clinical decision-making regarding the management and follow-up of very elderly patients with heart failure.”

 

Comments 8: For exclusion criteria, you state that patients "lost to follow-up after the first visit" were excluded. Please define the minimum follow-up required (e.g. no further contact, or unknown vital status) and provide the number of patients excluded for this reason.

Response 8: Patients lost to follow-up after the first visit were those with no subsequent clinical contact or recorded information after baseline. Of 1,519 patients assessed during the study period, 993 were excluded for not meeting the age criterion and 12 (2.3% of the 526 age-eligible patients) for loss to follow-up after the first visit, leaving 514 patients for analysis.

"Inclusion and exclusion criteria.

Patients were eligible if they had a diagnosis of HF according to the European Society of Cardiology heart failure guidelines[15] and were aged ≥85 years at the time of the first visit, and had their first visit between July 2012 and May 2023. Patients were excluded if they did not meet the age criteria or were lost to follow-up after the first visit, defined as the absence of any subsequent clinical contact or recorded information after baseline. Of 1,519 patients assessed at a first visit during the study period, 993 were excluded because they did not meet the age criterion and 12 because they had no contact after the first visit."

 

Methods - variables and definitions

Comments 9: Indicate the calendar time at which Barthel, Pfeiffer, NT-proBNP and Charlson index were assessed (baseline visit) and clarify whether the Barthel and Pfeiffer scores were obtained systematically by trained staff or from routine clinical documentation.

Response 9: The reviewer is right to ask for this precision. The Methods now specify that functional status (Barthel Index), cognitive status (Pfeiffer questionnaire), comorbidity burden (Charlson Comorbidity Index) and NT-proBNP were all assessed at the baseline visit. We have also clarified that the Barthel and Pfeiffer scales were administered by trained staff of the unit as part of its standardised assessment protocol, rather than extracted from routine clinical documentation.

Changes to the manuscript:

“Sociodemographic characteristics were collected, including sex, age, place of residence and cohabitants. The comorbidities included in the Charlson Comorbidity Index[16] and other conditions not included in the index, such as arterial hypertension, dyslipidaemia, and atrial fibrillation, as well as lifestyle factors (e.g., smoking, alcohol use), were also recorded at the baseline visit. The non-age-adjusted version of the Charlson Comorbidity Index was used. Functional status was assessed using the Barthel Index[17] and cognitive status using the Pfeiffer Short Portable Mental Status Questionnaire[18]. Both scales were administered at study entry by trained staff as part of the unit's standardised assessment protocol.”

 

Comments 10: For Charlson index, specify whether age-adjusted or non-age-adjusted Charlson was used; for patients >=85, age-adjustment can substantially change the score and interpretation.

Response 10: The non-age-adjusted Charlson Comorbidity Index was used; age was deliberately analysed as a separate independent variable to avoid incorporating it twice, which is particularly important in this age group. This is now specified in the Methods.

Changes to the manuscript:

“Sociodemographic characteristics were collected, including sex, age, place of residence and cohabitants. The comorbidities included in the Charlson Comorbidity Index[16] and other conditions not included in the index, such as arterial hypertension, dyslipidaemia, and atrial fibrillation, as well as lifestyle factors (e.g., smoking, alcohol use), were also recorded at the baseline visit. The non-age-adjusted version of the Charlson Comorbidity Index was used.”

 

Comments 11: Please define "new-onset HF" in more detail (first diagnosis within a specified time frame, or first hospitalisation, or first contact with cardiology/UMIPIC).

Response 11: A precise definition has been added to the Methods. New-onset HF was defined as HF first diagnosed at the time of referral to the unit in patients without a prior history of HF, irrespective of whether the presentation involved an episode of decompensation; chronic HF referred to a previously established diagnosis.

Change made:

“With regard to heart failure, […] HF was classified as new-onset when it was first diagnosed at the time of referral to the unit in patients with no prior history of HF – whether or not the presentation involved an episode of decompensation – as opposed to chronic HF, defined by a previously established diagnosis.”

 

 Comments 12: For HF categories (HFrEF, HFmrEF, HFpEF), clarify whether LVEF was taken from the echocardiogram closest to the baseline visit and whether the same measurement was used for all patients.

Response 12:  We thank the reviewer for this important methodological comment. We have clarified how LVEF was ascertained. In all patients, LVEF was obtained from the transthoracic echocardiogram closest to the baseline visit, applying the same approach throughout. A hierarchical method was used for quantification: the biplane Simpson's method whenever available, followed by the Teichholz method, and visual estimation when no quantitative measurement was feasible. This is now specified in the Methods.

"HF was categorised according to LVEF as reduced (HFrEF, LVEF ≤ 40%), mildly reduced (HFmrEF, 41-49%) or preserved (HFpEF, LVEF ≥50%)[15]. LVEF was obtained from the transthoracic echocardiogram closest to the baseline visit, applying the same approach to all patients. Whenever available, LVEF was quantified using the biplane Simpson's method; when this was not available, the Teichholz method was used; and when no quantitative measurement was feasible, LVEF was estimated visually."

 

Comments 13: It would be useful to indicate whether frailty per se was assessed (e.g. Clinical Frailty Scale, gait speed, or similar) and to acknowledge explicitly in Methods that no formal frailty tool was used (currently only discussed as a limitation).

Response 13: This is a helpful suggestion. The Methods now state explicitly that no formal frailty assessment instrument, such as the Clinical Frailty Scale or gait speed measurement, was used. The absence of a dedicated frailty assessment is also acknowledged in the Limitations.

Change made: "Functional status was assessed using the Barthel Index[17] and cognitive status using the Pfeiffer Short Portable Mental Status Questionnaire[18]. Both scales were administered at study entry by trained staff as part of the unit's standardised assessment protocol. No formal frailty assessment instrument was used."

 

Methods - statistical analysis

Comments 14: You mention that NT-proBNP was dichotomised using the Youden index at 2285 pg/mL and also log-transformed for Cox regression. Please clarify precisely how NT-proBNP entered each model (log-continuous versus dichotomous), and in Tables 5-6 label clearly whether "High NT-proBNP" refers to the cut-off determined by Youden.

Response 14:  We thank the reviewer for identifying this ambiguity, which we have clarified. NT-proBNP was analysed using two parameterisations. In the univariable analysis (Table 4), it was entered as a continuous log10-transformed variable; the corresponding hazard ratio (2.49) expresses the change in risk per one-unit increase in log10(NT-proBNP). For the multivariable models (Tables 5 and 6), NT-proBNP was dichotomised at the Youden-derived optimal cut-off of 2285 pg/mL (>2285 vs ≤2285 pg/mL). The Methods section has been revised to state both parameterisations explicitly, and Tables 5 and 6 now label this variable as "High NT-proBNP (>2285 pg/mL)" to distinguish it from the continuous term reported in Table 4. 

Methods: "The optimal NT-proBNP cut-off point for predicting mortality was calculated from the natriuretic peptide value (NT-proBNP) according to the Youden index, which is the value that maximises the sum of sensitivity and specificity[20], yielding a threshold of 2285 pg/mL. In addition, NT-proBNP was entered as a logarithmic transformation log10-transformed variable of NT-proBNP was in the univariable analyses, and as a dichotomised variable at the Youden-derived cut-off  (>2285 vs ≤2285 pg/mL) in the multivariable Cox models, to calculate Hazard Ratios (HR) and their confidence intervals using Cox regression analysis."

Results: "Similarly, elevated higher NT-proBNP (>2285pg/mL) considered as a continuous log10-transformed variable was associated with shorter survival (HR: 2.49; 95% CI 1.96–3.17). "

"In the multivariable Cox regression analysis (Table 5), after adjusting for age, sex, and variables significantly associated with mortality in the univariate analysis, independent predictors of lower survival …  and elevated high NT-proBNP (> 2285 pg/mL) (HR: 2.03; 95% CI 1.57–2.62)."

Tables 5 and 6:    "High NT-proBNP (>2285 pg/mL)". 

 

Comments 15: Provide some detail on how variables were selected for the multivariable Cox models: were all variables with p<0.05 in univariate analysis included, or did you apply any clinical pre-selection or stepwise procedure? Also, did you assess proportional hazards assumptions (e.g. Schoenfeld residuals, log-log plots)?

Response 15: Variables were entered using a forced-entry (enter) method, not a stepwise procedure. Candidate covariates were those significantly associated with mortality in the univariable analysis (p<0.05), with three pre-specified decisions: individual comorbidities were not entered, as overall comorbidity burden was already represented by the Charlson Comorbidity Index (avoiding redundancy and collinearity); infiltrative aetiology was excluded owing to its small sample size; and age and sex were forced into the model as clinically relevant adjustment variables. The proportional hazards assumption was assessed by testing time-dependent covariates; details have been added to the Statistical analysis section. These changes are reflected in the revised Methods. 

Changes to the manuscript:

Methods: "The variables that showed a significant association in the univariate analysis (p<0.05) were included in a multivariable Cox proportional hazards regression model using a forced-entry method. Individual comorbidities were not included, as overall comorbidity burden was represented by the Charlson Comorbidity Index; infiltrative aetiology was excluded owing to its small sample size; and age and sex were included as clinically relevant adjustment variables regardless of their univariable significance. The proportional hazards assumption was assessed by testing time-dependent covariates (interaction of each covariate with the logarithm of survival time); the main covariates showed time-dependent effects that attenuated during follow-up, so the reported hazard ratios represent average effects over the follow-up period."

Limitations: "Finally, the prognostic effects of the main covariates (comorbidity burden, functional class, functional dependence and NT-proBNP) were not constant over time, attenuating during a prolonged follow-up; the corresponding hazard ratios should therefore be interpreted as average effects over the study period rather than as time-invariant estimates."

 

Comments 16: Please report how missing data were handled. For several variables, denominators are <514 (e.g. Barthel, Pfeiffer, eGFR, HF readmission). Were these cases excluded listwise from the multivariable models? Clarifying this will help readers interpret any potential bias.

Response 16: Missing data were handled by complete-case or listwise analysis. Patients with a missing value for any variable included in a given multivariable model were excluded from that specific model. The baseline multivariable model (Table 5) included 425 of 514 patients (82.7%) with complete covariate data: 277 deaths and 148 censored, with 89 patients (17.3%) excluded. Readmission status was available in 377/514 patients; after complete-case selection, the model additionally including 1-year HF readmission (Table 6) included 318 patients (61.9%); 196 deaths and 122 censored; with 196 patients (38.1%) excluded. No imputation was performed. The number of cases and events included in each multivariable model has been added to the Statistical analysis section and table footnotes, and complete-case analysis together with the missingness of the readmission variable is now acknowledged as a limitation.

We have added:

 

Methods:  "Missing data were handled by complete-case (listwise) analysis: patients with a missing value in any variable included in a given multivariable model were excluded from that model, and no imputation was performed. The baseline model included 425 patients (277 deaths), and the model additionally including 1-year HF readmission included 318 patients (196 deaths)."

Results:

Table 5 (footnote): Model based on 425 patients with complete covariate data, including 277 deaths.

Table 6 (footnote): Model based on 318 patients with complete covariate data, including 196 deaths.

Limitations: Because the multivariable models were fitted using complete-case analysis, patients with missing covariate data—most notably 1-year HF readmission, missing in 38.1% of the sample—were excluded from the respective models, reducing the effective sample size and potentially introducing selection bias.

 

Comments 17: Consider adding a brief statement on sample-size considerations for multivariable modelling (e.g. number of events per variable).

Response 17: A statement on sample-size considerations has been added to the Statistical analysis. Events per variable (EPV) were calculated for each model from the events available after complete-case selection. Even the model incorporating HF readmissions, based on fewer patients owing to missing data (318 patients, 196 events; 10 covariates), maintained approximately 20 events per variable, above the commonly recommended minimum of ten events per variable for Cox regression. This supports the stability and reliability of the multivariable estimates.

 

Added text: "Missing data were handled by complete-case (listwise) analysis: patients with a missing value in any variable included in a given multivariable model were excluded from that model, and no imputation was performed. The baseline model included 425 patients (277 deaths), and the model additionally including 1-year HF readmission included 318 patients (196 deaths). The number of events per variable exceeded the recommended minimum of ten in both multivariable models (approximately 31 and 20 events per variable in the baseline and readmission models, respectively), calculated from the deaths available after complete-case selection. "

Results - baseline characteristics and treatments

Comments 18: Baseline description is comprehensive. To improve readability, you might move some detailed percentages from text to tables and focus the text on key messages (e.g. "multimorbidity was marked, with median Charlson index 3 and >50% of patients having CKD, AF and anaemia").

Response 18: Following this suggestion, the description of comorbidities has been streamlined: the detailed percentages are retained in Table 1, and the text now focuses on the key message, highlighting the marked multimorbidity.

 

Change made:

Results: The most frequent comorbidities were arterial hypertension (95.9%), dyslipidaemia (57.2%), diabetes mellitus (48.6%), chronic kidney disease (57.2%), atrial fibrillation (68.7%) and anaemia (55.3%) Multimorbidity was marked, with a median Charlson comorbidity index of 3; hypertension was almost universal (95.9%), and more than half of the patients had chronic kidney disease, atrial fibrillation and anaemia (Table 1).

 

Comments 19: Given the age and comorbidity profile, it would be interesting to report the distribution of HF phenotype (HFrEF/HFmrEF/HFpEF) explicitly, not just "vast majority HFpEF" (e.g. HFpEF 73.9%, HFmrEF x%, HFrEF y%).

Response 19: The distribution of HF phenotypes is now reported explicitly in the Results, replacing the previous general statement. HF with preserved ejection fraction was predominant (380 patients, 73.9%), followed by reduced (70, 13.6%) and mildly reduced ejection fraction (53, 10.3%); LVEF was not available in 11 patients (2.1%).

Regarding HF characteristics, HFpEF was the predominant phenotype (73.9%), followed by HFrEF (13.6%) and HFmrEF (10.3%); LVEF was unavailable in 2.1% of patients.the vast majority of patients (73.9%) had HF with preserved ejection fraction.

Table 4. Characteristics and evolution of heart failure in survivors versus deceased patients.

		Total N=514	Alive N=174 (33.9%)	Dead N=340 (66.1%)	p-value	HR (95% CI)
HF* characteristics
LVEF *  Median [IQR]  Mean (SD)		60 [47.25–66] 56.24 (14)	58 [45–67] 55.78 (14)	60 [48–66] 56.49 (14)	0.678	1.00 (0.99–1.01)
LVEF	Reduced	70 (13.6%)	28 (16.2%)	42 (12.7%)	0.071	1.35 (0.98–1.87)
	Mildly reduced	53 (10.3%)	17 (9.8%)	36 (10.9%)	0.72	1.12 (0.61-2.06)
	Preserved	380 (73.9%)	128 (74%)	252 (76.1%)	0.60	1.21 (0.73-1.71)
	Not available	11 (2.1%)	—	—	—	—

 

Comments 20: For drug therapy, please indicate: In how many patients SGLT2i or ARNI were available (e.g. calendar-time effect).

Response 20: The proportions of patients receiving these therapies are reported in Table 3 and are now also stated in the Results: SGLT2 inhibitors in 114 patients (22.6%) and ARNI in 52 (10.3%). As the reviewer notes, their relatively limited use reflects a calendar-time effect, since the study spanned 2012–2023 and these therapies were progressively incorporated into HF management during that period; patients recruited in earlier years could not receive treatments that were not yet available or indicated.

Change made: 

The most prescribed treatments (Table 3) were loop diuretics (89.7%), beta-blockers (61.1%), oral anticoagulants (59.9%) and statins (52.9%). SGLT2 inhibitors were prescribed in 114 patients (22.6%) and ARNI in 52 (10.3%).

 

Comments 21: For drug therapy, please indicate: Whether prescription patterns differed by HF phenotype (not essential, but a short comment would situate the cohort against current guideline-directed therapy).

Response 21: As suggested, we have now added a brief description of prescription patterns according to heart failure phenotype. ARNI and SGLT2 inhibitors were more frequently used in patients with reduced or mildly reduced ejection fraction than in those with preserved ejection fraction, whereas beta-blockers and mineralocorticoid receptor antagonists were used at similar frequencies across phenotypes. These patterns should be interpreted in the context of the long recruitment period (2012–2023), during which the evidence base and guideline recommendations for disease-modifying therapies evolved substantially. The widespread use of beta-blockers across phenotypes may also partly reflect the high prevalence of atrial fibrillation in this cohort.

Change made:

The most prescribed treatments … "By heart failure phenotype, ARNI and SGLT2 inhibitors were used more frequently in patients with reduced and mildly reduced than in preserved ejection fraction (ARNI: 40.0%, 24.5% and 2.6%; SGLT2 inhibitors: 38.6%, 32.1% and 18.4%, respectively), whereas beta-blockers (61.4%, 67.9% and 60.8%) and mineralocorticoid receptor antagonists (41.4%, 47.2% and 41.3%) were used similarly across phenotypes."

 

Results - outcomes and modelling

Comments 22: You state that survival was 77% at 1 year and 25% at 5 years, while the abstract also mentions 50% at 2.5 years (median survival 30.5 months). It would be helpful to include the 2.5-year Kaplan-Meier survival estimate in the Results text and/or Figure 1 caption.

Response 22: The 2.5-year estimate has been added to the Results, which now report survival of 77% at 1 year, 50% at 2.5 years and 25% at 5 years. This also aligns the Results with the figures given in the Abstract and Discussion.

Revised text:

The median survival was 30.5 months and the mean was 41.3 months. Survival rates were 77% at 1 year, 50% at 2.5 years, and 25% at 5 years.

 

Comments 23: Age is reported as not associated with survival in this >=85 cohort. It might be informative to provide HR per 5-year increase (e.g. 85-<90 vs >=90) to give readers a sense of effect size.

Response 23: As suggested, we have re-expressed the association between age and survival in more clinically interpretable terms. When modelled per 5-year increment, age was not significantly associated with mortality (HR 1.12; 95% CI 0.94–1.35; p=0.207). Consistently, comparing patients aged ≥90 years with those aged 85–89 years showed no significant difference in survival (HR 1.05; 95% CI 0.83–1.33; p=0.668; log-rank p=0.75, Figure 3). These findings reinforce our observation that, within this very elderly cohort, chronological age loses prognostic relevance relative to markers of biological age such as comorbidity burden, functional dependence and cognitive impairment.

Results: Age was not statistically significantly associated with survival, whether modelled per 5-year increment (HR 1.12; 95% CI 0.94–1.35) or comparing patients aged ≥90 versus 85–89 years (HR 1.05; 95% CI 0.83–1.33; log-rank p=0.75) (Table 1; Figure 3).

 

Comments 24: In univariate analyses, several comorbidities (e.g. advanced CKD, diabetes, MI, dementia, cancer) were significantly associated with mortality, but only Charlson index remains in the final model. It would be worth briefly noting in Results that when Charlson index is entered, individual comorbidities lose significance, suggesting collinearity with overall comorbidity burden.

Response 24: We thank the reviewer. As described in our response to Comment 15, overall comorbidity burden was represented by the Charlson comorbidity index rather than by the individual comorbidities; the two sets of variables were not entered simultaneously, precisely to avoid the collinearity and redundancy the reviewer describes, since the index already incorporates these conditions. We have added a brief clarifying sentence to the Results to make this design choice explicit.

Results: In multivariable models, overall comorbidity burden was represented by the Charlson index rather than by the individual comorbidities to avoid redundancy and potential collinearity. In the multivariable Cox regression analysis (Table 5)…

 

Comments 25: Please ensure that the direction of HRs is always intuitive and consistent with how the variable is coded. For example, in Table 5 the HR for Barthel is 0.99 per point (protective), but in the text you sometimes refer to "functional dependence (lower Barthel)" as a risk factor. Clarifying that the HR is per unit increase in score will avoid confusion.

Response 25: We thank the reviewer for this important observation. To keep the direction of each hazard ratio consistent with how the variable is coded, the Barthel index is now presented separately as a protective factor: higher functional status (higher Barthel index) was associated with better survival (HR 0.99), whereas higher Pfeiffer and Charlson scores were associated with increased mortality. This avoids the previous inconsistency between the label "functional dependence" and a hazard ratio below 1.

Change to the manuscript:

In the multivariable Cox regression analysis (Table 5), after adjusting for age, sex, and variables significantly associated with mortality in the univariate analysis, independent predictors of lower survival were cognitive impairment (higher Pfeiffer score) (HR: 1.07; 95% CI 1.00–1.14), functional dependence (lower Barthel index) (HR: 0.99; 95% CI 0.98–0.996), comorbidity burden (higher Charlson index) (HR: 1.10; 95% CI 1.03–1.18), worse NYHA functional class (HR: 1.30; 95% CI 1.01–1.68) and high NT-proBNP (HR: 2.03; 95% CI 1.57–2.62). Conversely, a better functional status (higher Barthel index) had an independent protective effect on survival (HR: 0.99; 95% CI 0.98–0.996).

 

Comments 26: In Table 6, consider showing the number of patients/events included in the model with HF readmissions (since n is smaller due to missing data on readmission status).

Response 26: As suggested, the number of patients and events included in the model with HF readmissions is now reported. Readmission status was available in 377 of the 514 patients; of these, 318 also had complete data for all the remaining covariates and were therefore included in the model, which comprised 196 deaths and 122 censored observations. This is now indicated in Table 6 and in the Results. See also our response to Comment 16.

Changes made:

Results: When HF readmission was included in the multivariable model (318 patients with complete covariate data, 196 deaths), it was strongly associated…

Table 6 (footnote): “Model based on 318 patients with complete covariate data, including 196 deaths.”

 

Discussion

Comments 27: Condensing some of the comparisons with younger cohorts (several paragraphs) into a shorter synthesis, then expanding slightly on how your findings could influence risk stratification and follow-up in >=85-year-old patients.

Response 27: Following this suggestion, the Discussion has been restructured. The comparisons with younger cohorts, previously spread across two detailed paragraphs, have been condensed into a shorter synthesis, while all the original references have been retained. In parallel, the clinical implications of our findings have been expanded: we now discuss how markers of biological age and disease severity—rather than chronological age—could inform risk stratification and the intensity of follow-up in patients aged ≥85 years with HF.

First part: condensing some of the comparisons with younger cohorts:

Marked up version: In this cohort of patients aged ≥85 years with heart failure (HF), long-term prognosis was poor, with an estimated survival of 50% at 2.5 years and 25% at 5 years. Although evidence on temporal trends in HF survival in the very elderly remains limited, These outcomes appear less favourable than those reported in younger incident HF cohorts, in which survival ranges from approximately 65–69% at 3–5 years in primary-care populations (mean age 76–77 years)[21,22] and reaches 56.7% at 5 years in a meta-analysis of community-dwelling patients, declining progressively with age[23].  managed in primary care populations. In a Mediterranean population-based study of ambulatory patients with incident chronic heart failure (mean age 77 years), survival at 1, 2, and 3 years was 90%, 80%, and 69%, respectively[18]. Likewise, a primary-care cohort showed a 5-year survival of approximately 65%[19]. These differences could be explained by a marked age gap, as the mean age in our cohort was 88 years, compared with 77 years[18] and 76 years[19] in the aforementioned studies. Furthermore, these differences may reflect the higher burden of multimorbidity and functional impairment typically observed in populations aged ≥85 years. A systematic review and meta-analysis of community-dwelling patients with chronic HF further supports the strong influence of age on prognosis, reporting survival rates of 72.6% at 2 years and 56.7% at 5 years, with a progressive decline in survival with advancing age[20]. Notably, in patients older than 75 years, survival at 5 years was 49.5%[20]. Similarly, A large European registry of ambulatory patients with chronic HF reported a 5-year survival of approximately 68% in a substantially younger population (mean age ~67 years), and identified age as an independent predictor of mortality[24]. The less favourable prognosis in our cohort likely reflects its advanced age (mean 88 years) and the higher burden of multimorbidity and functional impairment typical of patients aged ≥85 years.

Data focusing specifically on long-term outcomes in very elderly HF patients remain scarce. A Swedish population-based cohort of patients aged ≥ 85 years with newly diagnosed heart failure[12] reported a 1-year mortality of 30.5%, which is comparable to the 23% observed in the present cohort. Consistently, Baldasseroni et al. reported a 1-year mortality of 22.7% in a cohort of very elderly patients (mean age 89 years) managed in a multidisciplinary heart failure unit after acute decompensation[13].

 

Clean version: In this cohort of patients aged ≥85 years with heart failure (HF), long-term prognosis was poor, with an estimated survival of 50% at 2.5 years and 25% at 5 years. These outcomes appear less favourable than those reported in younger HF cohorts, in which survival ranges from approximately 65–69% at 3–5 years in primary-care populations (mean age 76–77 years)[21,22] and reaches 56.7% at 5 years in a meta-analysis of community-dwelling patients, declining progressively with age[23].  A large European registry of ambulatory patients with chronic HF reported a 5-year survival of approximately 68% in a substantially younger population (mean age ~67 years), and identified age as an independent predictor of mortality[24]. The less favourable prognosis in our cohort likely reflects its advanced age (mean 88 years) and the higher burden of multimorbidity and functional impairment typical of patients aged ≥85 years.

Data focusing specifically on long-term outcomes in very elderly HF patients remain scarce. A Swedish population-based cohort of patients aged ≥ 85 years with newly diagnosed heart failure[12] reported a 1-year mortality of 30.5%, comparable to the 23% observed in the present cohort. Consistently, Baldasseroni et al. reported a 1-year mortality of 22.7% in a cohort of very elderly patients (mean age 89 years) managed in a multidisciplinary heart failure unit after acute decompensation[13]. 

Second part: expand slightly on how our findings could influence risk stratification and follow-up in >=85 year-old patients.  

These findings highlight the importance of early identification of high-risk patients and the potential role of structured follow-up programmes in preventing adverse outcomes. In this cohort, the highest risk of death was identified by markers of biological age and disease severity, not by chronological age. These markers included comorbidity burden, functional and cognitive status, natriuretic peptide levels and early HF readmission. A multidimensional assessment based on these domains may therefore be more informative than age alone when estimating prognosis and tailoring the intensity of follow-up in this population.

 

Comments 28: More clearly separating observations derived from your baseline model (Charlson, NT-proBNP, Barthel, Pfeiffer, NYHA) from those of the model including HF readmissions, where readmission becomes a very strong predictor.

Response 28: This is an excellent point that has helped us improve the clarity of the Discussion. Following the reviewer's suggestion, the findings of the two multivariable models are now clearly separated, with each result explicitly attributed to the baseline model (Table 5) or to the model including HF readmissions (Table 6). We also state that, when HF readmission was added, it became the strongest predictor of mortality and attenuated the associations of NYHA class and functional and cognitive status, whereas the Charlson index and NT-proBNP remained independently associated with mortality in both models and were therefore the most consistent prognostic markers.

Changes made:

 

Markers of HF severity such as higher NT-proBNP levels and worse NYHA functional class (III–IV) were associated with worse prognosis as reported in previous studies[12,41–43]. In the baseline multivariable model (Table 5), both remained independently associated with mortality. This association remained significant in multivariable analysis. 

In addition, functional status (Barthel index) and cognitive status (Pfeiffer test) were independent predictors of survival in the baseline multivariable model (Table 5).

[…] Finally, the prognostic impact of HF readmissions was strong when included in the multivariable model (Table 6): readmission within the first year was the strongest predictor of mortality remained independently associated with mortality, while the associations of baseline functional class, cognitive status, and functional dependence were attenuated. In contrast, the Charlson index and high NT-proBNP remained independently associated with mortality, representing the most consistent prognostic markers across the two models.

 

Comments 29: When discussing the lack of association between age and survival, you might emphasise that the restricted age range likely limits statistical power to detect small differences and that markers of biological age (comorbidity, function, cognition) are more informative in this group.

Response 29: We agree with the reviewer. The Discussion continues to emphasise that, in this age group, markers of biological age—comorbidity burden, functional dependence and cognitive impairment—are more informative than chronological age. We have added a closing remark noting that the absence of an association with age may also partly reflect the narrow age range of the cohort (all patients ≥85 years, median 88), which limits the statistical power to detect small age-related differences in survival.

Added text:

Regarding factors related to survival, … Nonetheless, this may partly be due to the narrow age range of the cohort, in which all patients were aged ≥85 years (median 88 years, most between 86 and 90), which limits the statistical power to detect small age-related differences in survival.

 

Limitations

Comments 30: The limitations are appropriately acknowledged (retrospective design, single-centre, missing formal frailty assessment). Two additional points could be mentioned:

  - Potential survivorship bias, since only patients referred and able to attend the outpatient unit were included; the sickest institutionalised or bed-bound patients may have been under-represented.

  - Possible unmeasured confounding by socio-economic status or caregiver support, which could be particularly important in this age group.

Response 30: We thank the reviewer for these two suggestions, both of which have been incorporated into the Limitations. First, we now acknowledge a potential selection bias: because patients were referred on clinical judgement and had to be able to attend the outpatient unit, the study population tended to include patients with more advanced heart failure but preserved functional status, so the most severely ill or bed-bound patients may have been under-represented, which may limit the representativeness of the cohort. Institutionalised patients, however, were not excluded and were included whenever their functional status allowed them to attend the unit. Second, we acknowledge that data on socio-economic status and caregiver support were unavailable, and that unmeasured confounding by these factors—potentially relevant in this age group—cannot be excluded.

 

Text added in Limitations:  Moreover, patients were referred on the basis of clinical judgement and had to be able to attend the outpatient unit. Consequently, the study population tended to include patients with more advanced heart failure but preserved functional status, which may limit the representativeness of the cohort. Data on socio-economic status and caregiver support were also unavailable. Unmeasured confounding by these factors, potentially relevant in this age group, cannot be excluded.

 

Language and style

Comments 31: Overall, the English is understandable, but there are several minor grammatical issues and some long sentences that could be streamlined. Examples include: "included in an structured outpatient follow-up programme" (an structured -> a structured) or "multimorbidity HF" (multimorbid HF or HF with multimorbidity).

Response 31: We thank the reviewer for this helpful comment. The manuscript has been revised throughout for language, clarity and readability. The specific errors highlighted have been corrected, and, on re-reading, a small number of additional minor issues were amended. Spelling has been made consistent with British English conventions across the text and tables. In response to the comment on sentence length, several long sentences in the Methods and Discussion have been divided into shorter ones to improve readability, without altering their content.

Grammatical corrections:

Introduction: “in a structured outpatient”; “with multimorbid HF”

Results: “When HF readmission was included in the multivariable model, it was strongly associated…”

Limitations: biological age and vulnerability.  (full stop added)

 

Adapt everything to British English:

Results: “The aetiologies of heart disease associated with lower survival were ischaemic (HR: 1.49; 95% CI 1.18–1.88) and infiltrative cardiomyopathy (HR: 1.72; 1.06–2.78) (Table 4).”

Table 2 (title): “Functional status, cognitive status, and global comorbidity assessed by standardised scales”

 

Sentences shortened:

Methods (statistical analysis): “To assess the relationship between different variables and mortality, a univariate analysis was performed using the Chi-square test or Fisher's exact test for categorical variables and the Student's t-test or Mann-Whitney U test for continuous variables, depending on whether the distribution of the variables was normal or not.”

Simplified: “To assess the relationship between different variables and mortality, univariate analyses were performed. Categorical variables were compared using the Chi-square or Fisher's exact test, and continuous variables using the Student's t-test or Mann-Whitney U test, depending on their distribution.”

Discussion (comorbidity): “Patients exhibited a high burden of comorbidity, with hypertension being the most prevalent condition, affecting nearly all patients (95.9%), exceeding rates reported in most series (50%–90%)[10,22,26], likely reflecting the advanced age of the cohort.”

Simplified: “Patients exhibited a high burden of comorbidity. Hypertension was the most prevalent condition, affecting nearly all patients (95.9%); this exceeds the rates reported in most series (50%–90%)[10,12,26] and likely reflects the advanced age of the cohort.”

Discussion (prognostic framework): “These findings support a framework in which, among patients aged ≥85 years, prognosis is driven less by chronological age than by the interaction between heart failure severity, multimorbidity, functional dependence, and cognitive impairment, suggesting that prognosis in this population is better explained by clinical complexity than by age alone, reinforcing the need for a multidimensional geriatric approach to care.”

Simplified: “These findings support a framework in which, among patients aged ≥85 years, prognosis is driven less by chronological age than by the interaction between heart failure severity, multimorbidity, functional dependence, and cognitive impairment. This reinforces the need for a multidimensional geriatric approach to care.”

Discussion (ESC-EORP-HFA registry): “The ESC-EORP-HFA Heart Failure Long-Term Registry, carried out in 211 centres including 12,440 patients, corroborates the severity of HF once it requires admission, such that almost one third of patients die in the first year after admission and almost half are readmitted for heart failure[45].”

Simplified: “The ESC-EORP-HFA Heart Failure Long-Term Registry, carried out in 211 centres and including 12,440 patients, corroborates the severity of HF once admission is required. In that registry, almost one third of patients died in the first year after admission and almost half were readmitted for heart failure[45].”

Discussion (diabetes): “Diabetes mellitus was associated with worse survival, which is consistent with what has been described, both in observational studies[10] and in meta-analyses that have shown that diabetes is associated with a higher risk of all-cause death and cardiovascular death[7,33,34].”

Simplified: “Diabetes mellitus was associated with worse survival. This is consistent with previous observational studies[10] and with meta-analyses linking diabetes to a higher risk of all-cause death and cardiovascular death[7,33,34].”

 

Once again we would like to thank you for taking the time to read our article and for your constructive recommendations.

Regards.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The Authors have clarified the issues raised and introduced necessary information into the text, which has improved the quality of the manuscript.

Despite the limitations - honestly acknowledged by the authors - the manuscript provides the clinician with a useful overview of a specific population (the very elderly) with heart failure.

I have no further comments or recommendations.

Thank you!

Reviewer 2 Report

Comments and Suggestions for Authors

Dear Authors,

Thank you for the revised manuscript and the detailed response to the reviewers’ comments. The revised version has been substantially improved.

The study provides valuable long-term outcome data in a clinically important and underrepresented population of patients aged 85 years or older with heart failure managed in a structured multidisciplinary outpatient setting. The manuscript now presents the cohort selection, assessment methods, survival analyses, multivariable Cox models, proportional-hazards assessment, and handling of missing data with adequate clarity. The limitations of the retrospective single-centre design, referral-related selection, incomplete covariate data, and potential residual confounding are also appropriately acknowledged.

The results are clearly presented, the conclusions are consistent with the data, and the discussion appropriately places the findings in the context of the existing literature. I have no further substantive comments and consider the manuscript suitable for publication in its present form.