Considering the Duration of Heart Failure: Using the Past to Predict the Future

This article refers to ‘Duration of chronic heart failure affects outcomes with preserved effects of heart rate reduction with ivabradine: findings from SHIFT’ by M. Böhm et al., published in this issue on pages 373–381. Despite recent therapeutic advances, heart failure with reduced ejection fraction (HFrEF) is generally considered a chronic and fatal condition with long-term survival comparable to many cancers.1, 2 However, although prognosis remains poor, heterogeneity within the HFrEF population is well documented and individual patients may vary widely. As such, a wealth of studies have elucidated several strong predictors of mortality and the use of validated multivariable risk scoring systems to estimate prognosis is endorsed by clinical guidelines.3-5 Despite an ever-growing body of literature centred on risk prediction in HFrEF, minimal attention has been paid to understanding the implications of the duration of heart failure (HF) diagnosis on prognosis. In the context of very limited data on this issue, two potential opposing effects seem plausible. First, HF is a progressive disease and it is conceivable that patients with longer HF duration may have higher subsequent mortality, reflective of prolonged exposure to attendant neurohormonal perturbations and a corresponding greater degree of maladaptive cardiac remodelling. Alternatively, one could hypothesize that select patients proven to be long-term survivors may represent a low-risk phenotype (i.e. ‘survivor bias’) and will continue to have favourable prognosis, given that they have already survived a lengthy period despite a diagnosis of HF. Confirming either one of these opposing associations could have significant implications across the spectrum of HF clinical care and research. To help address this question, in this issue of the Journal, Böhm et al. present a post hoc analysis from the SHIFT (Systolic Heart Failure Treatment with the If inhibitor Ivabradine) trial examining the influence of HF duration on clinical outcomes among ambulatory HFrEF patients.6 Notable selection criteria for SHIFT included a diagnosis of stable symptomatic chronic HF of ≥4 weeks duration, an ejection fraction (EF) of ≤35%, and a prior hospitalization for HF within the preceding 12 months.7 Among the 6505 patients for whom data were available on time of HF diagnosis, baseline clinical profile differed substantially by duration of HF, with those with longer duration tending to be older with worse functional class and a greater number of co-morbidities. After multivariable adjustment, the investigators found robust and statistically significant associations between increased duration of HF and heightened risk for adverse clinical outcomes.6 These findings were consistent across all study endpoints, including cardiovascular death, all-cause death and hospitalization for HF. The investigators also confirmed consistent therapeutic efficacy of ivabradine across the range of HF duration included in the trial, with a near 20% relative risk reduction for the primary endpoint irrespective of HF chronicity.6 To our knowledge, these authors present the first detailed analysis of the influence of HF duration on subsequent outcomes among ambulatory HFrEF patients. Although Böhm et al. are to be congratulated for execution of a timely and novel analysis, limitations of this work should be highlighted.6 First, by virtue of the SHIFT trial selection criteria, all patients in the present analysis were required to have a diagnosis of ≥4 weeks and thus results cannot be applied to ambulatory patients with de novo or more recently diagnosed HF. Second, although the directionality of the association between HF chronicity and clinical outcome is clear and consistent in this study, the SHIFT database is unable to distinguish the mechanism by which such a relationship might exist. For example, although patients with longer duration of HF may be more likely to have advanced cardiac remodelling, relevant information on factors beyond EF (such as ventricular volumes) is not available. Third, although multivariable models were utilized, the exact rationale for covariate inclusion is unclear and the possibility of significant residual confounding remains. Specifically, analyses of HF duration and study endpoints included only seven covariates, of which age was not one. In contrast, slightly different covariates were documented for models assessing the treatment effect of ivabradine, in which age was included and an additional sensitivity analysis including co-morbidities was performed. To date, prior work examining HF duration and clinical outcomes has mostly centred on hospitalized HF populations and differences between patients admitted with de novo vs. worsening chronic HF. Across three such studies, patients with de novo or recently diagnosed HF consistently had fewer co-morbidities, higher systolic blood pressure and better baseline renal function.8-10 This clinical profile translated to superior post-discharge outcomes, despite statistical adjustment, compared with patients hospitalized for worsening chronic HF. The most recent of these analyses was from the ASCEND-HF (Acute Study of Clinical Effectiveness of Nesiritide in Decompensated Heart Failure) trial, in which investigators used granular information on date of HF diagnosis to evaluate for a graded effect of HF duration on mortality and hospitalization outcomes (rather than the binary classification of de novo vs. acute on chronic HF used in prior studies).10 Results showed no suggestion of a continuous relationship between duration of chronic HF and short-term dyspnoea relief or post-discharge outcomes. Rather, the influence of HF duration on outcomes seemed only to manifest as a distinction between de novo or recently diagnosed (≤1 month) HF and any duration of chronic HF, with all duration groups >1 month independently associated with a near two-fold heightened risk for 180-day all-cause mortality. How can the lack of a continuous association with duration of chronic HF in ASCEND-HF be reconciled with the presence of such an association in SHIFT?6, 10 Although unproven, the answer may lie in differences between acute decompensated and stable HF populations. Hospitalization for acute HF has been traditionally viewed as a sentinel event in the natural history of chronic HF, an inflection point in the survival curve with a subsequent mortality risk up to three-fold higher than patients never hospitalized.11, 12 Thus, it is possible that following an HF hospitalization, prognosis is dominated by implications of the worsening HF event, overwhelming any potentially less powerful influence of HF chronicity. By comparison, in the absence of a worsening HF event, it may be possible to use differences in HF duration to risk stratify patients with chronic stable symptoms. Although this hypothesis warrants further confirmatory study, it may be the most obvious explanation for the differing results from the ASCEND-HF10 and SHIFT7 trials. What might explain a worse prognosis in stable HFrEF patients with increasing HF duration? One can imagine both biologic and epidemiologic forces being possible. From the biologic side, as discussed by the SHIFT authors, extended exposure to the neurohormonal and haemodynamic consequences of HFrEF would support a greater degree of cardiac and clinical deterioration, and hence worse outcomes in patients with longer HF duration.6 Indeed, prior data suggest continued cardiac injury is common among HFrEF patients despite clinical stability.13 From the epidemiologic standpoint, one might expect HF duration to systematically correlate with varying rates of cardiac recovery during follow-up. Specifically, patients with more recent HF diagnoses may have an increased likelihood of downstream recovery in comparison with patients with longer-standing persistent HFrEF despite therapy, and thus bias outcome results.14 Unfortunately, the lack of longitudinal data capture on recovery of ventricular function in SHIFT prevents examination of this hypothesis. Speculating on potential contributions of biology vs. epidemiology in the present data from SHIFT, it is unlikely that meaningful rates of cardiac recovery could have occurred within the two longest HF duration groups (mean durations of 2.6 and 7.1 years) to result in notable outcome differences between them. In contrast, invoking epidemiologic influences is more reasonable in ASCEND-HF, whereby meaningfully higher rates of cardiac recovery among patients with de novo diagnoses who are relatively naïve to therapy is more plausible. Notably, as compared to patients hospitalized for worsening of chronic HF, de novo and recently diagnosed patients in ASCEND-HF were markedly more likely to have a non-ischaemic HF aetiology, further suggesting potential for cardiac recovery in this group.10, 14, 15 Given the impact of HF chronicity on patient profile and event rates, it is conceivable that the efficacy of an investigational therapy might vary according to the duration of the HF diagnosis. The most common application of HF duration within trial selection criteria has been to require treatment with standard HF therapy for >30 days before enrolment, but many recent trials have encompassed all durations, including de novo.16-18 Moreover, with rare exceptions, prior large HF trials did not report subgroup analyses by duration of HF. In this respect, the present post hoc confirmation of ivabradine efficacy across a wide range of HF duration is reassuring and suggests it may never be ‘too late’ in the clinical course of stable symptomatic patients to derive benefit from this therapy (and with potentially broader implications for the introduction of other new therapies in this population). In contrast, although a possible chance finding, the significant interaction between HF chronicity and dyspnoea relief with nesiritide in ASCEND-HF further illustrates the potential for interplay between duration of HF and trial endpoints.10 Moving forward, the aggregate findings from ASCEND-HF and SHIFT suggest HF duration warrants consistent consideration within Phase III HF trials in at least two ways: (i) potential exclusion of de novo or recently diagnosed (within ≤30 days) patients, and (ii) prespecified subgroup testing by duration of HF. Each year, dozens of manuscripts are published in the area of HF risk prediction and describe a wide array of risk models, biomarkers, imaging parameters, and other measures. Although the importance of such work should not be discounted, the duration of HF as a marker of risk has thus far ‘flown under the radar.’ Prior data among acute HF populations, coupled with the current data from SHIFT, suggest that this simple variable may have powerful prognostic implications in both the hospitalized and stable ambulatory HF settings. Although further work is needed to verify and expand these findings, these studies target an underappreciated question: how do we define ‘chronic’ in chronic HF? Indeed, while the term ‘chronic’ is loosely applied in many conditions across medicine and is intrinsically reflective of a clinical course in the past, in the case of the HF patient, this label may also have important implications for the future. Conflict of interest: S.J.G. is supported by the National Heart, Lung and Blood Institute T32 post-doctoral training grant (T32HL069749-14) and a Heart Failure Society of America/Emergency Medicine Foundation Acute Heart Failure Young Investigator Award funded by Novartis.