Untying the Gordian knot of sex and heart failure therapy

This article refers to ‘Dosing of losartan in men versus women with heart failure with reduced ejection fraction: the HEAAL trial’ by J.P. Ferreira et al., published in this issue on pages 1477–1484. It is well known that women have been under-represented in early clinical trials of heart failure (HF). Subgroup analysis of these studies, which failed to show sex-specific differences in efficacy of standard HF with reduced ejection fraction (HFrEF) therapies, led to a persistent belief that HF occurred similarly in men and women. Slowly developing literature has since emerged, demonstrating important sex-based differences in cardiac and clinical responses to many factors. This information has been largely based upon two types of data: the first, taken from observational studies using descriptive and regression techniques (to account for baseline characteristic differences) and the second (using similar techniques), from subgroup analyses of large, randomized clinical trials. Very few studies have prospectively assessed sex-based differences in cardiac and clinical response to injury and treatment. We have begun to recognize that females do not exhibit identical cardiac remodelling or clinical response characteristics as males do. While differential responses may appear relatively small in settings where left ventricular ejection fraction (LVEF) is below 40%, we note increasing male/female differences as LVEF approaches higher, or near ‘normal’ values. Indeed, these important differences have led to more rigorous evaluation of this phenomenon. Unfortunately, we are still in the early stages of carefully planned, properly powered and prospective studies involving both male and female patients with HF – the ‘gold standard approach’. In the meantime, newer analytic tools, such as machine learning tools, may help glean further insights that bridge this gap and directly inform the planning of these studies. Sex-based differences in HF epidemiology, aetiology, and outcomes [such as a higher representation of females with HF with preserved ejection fraction (HFpEF), and lower adjusted mortality on therapy] are now well recognized.1, 2 Several studies suggest that there are underlying differences in cardiac morphology and remodelling between male and female patients with HF. Gori et al.3 performed a secondary analysis of the Prospective Comparison of ARNI With ARB on Management of Heart Failure With Preserved Ejection Fraction (PARAMOUNT) study to investigate cardiac structural differences underlying the relative predisposition to HFpEF in women. They demonstrated that female patients had higher indexed left ventricular wall thickness and higher LVEF compared to male patients. However, global longitudinal strain was similar between sexes and mitral annular velocities by tissue Doppler imaging were lower in females.3 While different normal ranges have been established for females and males, myocardial remodelling patterns in response to stress likely also differ. For example, male elite endurance athletes have a greater increase in left and right ventricular volumes as well as increased left ventricular mass (eccentric hypertrophy) compared to female elite endurance athletes.4 Higher native T1 values in the female athletes suggests that the difference in remodelling patterns may be related to more cellular hypertrophy in male athletes.4 Similarly, in a cohort of patients with suspected coronary artery disease, female patients were more likely to have concentric remodelling and less likely to have eccentric hypertrophy defined using cardiac magnetic resonance imaging.5 Additionally, while concentric hypertrophy was associated with increased all-cause mortality overall, eccentric hypertrophy was independently associated with all-cause mortality only in female patients. These differences in cardiac morphology and remodelling patterns may partially explain differences in aetiology of HF,6 but also have important implications for patient diagnosis and management. In addition to body size and composition, other important differences must be considered.7 For example, testosterone decreases natriuretic peptide levels while oestrogens may increase them.7 Consequently, abnormal natriuretic peptide levels are associated with a lower relative risk for incident HF in female patients and elevated levels are less clearly associated with adverse events.7 Integrating sex-specific abnormal thresholds could help address these issues but are not typically applied in clinical practice. Additionally, there are significant differences in pharmacokinetics with female patients having smaller volumes of distribution for hydrophilic drugs and different activity levels for hepatic metabolic pathways, among many other differences.8 Several groups have investigated sex-specific responses in subgroup analyses of randomized trials. Ibrahim et al.9 demonstrated that female patients showed earlier and more consistent reverse remodelling with sacubitril/valsartan compared to male patients. Solomon et al.10 performed similarly critical work when they assessed for sex-specific differences in response to sacubitril/valsartan across a range of LVEFs. They demonstrated that all patients with lower LVEF derived a greater benefit with respect to the composite outcome of total HF hospitalizations or cardiovascular death. However, women derived benefit to a higher ejection compared to men with 95% confidence interval crossing at an LVEF ∼60% in women compared to ∼45% in men.10 Conversely, in a meta-analysis of randomized trials of implantable cardioverter-defibrillators (ICD) there was no benefit from ICD therapy in women.11 While these analyses are potentially informative, they are inherently limited by their retrospective nature and the under-representation of women in randomized trials. With all of these considerations in mind, a fundamental issue is to consider sex-based differences in response to medical therapy. In this issue of the Journal, Ferreira et al.12 perform a retrospective analysis to assess for possible sex-related differences in the Effects of High-Dose vs. Low-Dose Losartan on Clinical Outcomes in Patients with Heart Failure (HEAAL) study. The authors demonstrate, using simple subgroup analysis, that there was a differential female/male response to high- vs. low-dose losartan (interaction P = 0.018). While male patients appeared to benefit from high-dose losartan, female patients had no significant difference in response according to dose. However, the analysis did not stop there. The authors attempted to further address the issue of baseline confounders though a machine learning technique referred to as latent class analysis (LCA). LCA is an unsupervised machine learning technique, meaning it is not trained to predict a specific outcome or result. Instead, LCA is tasked with grouping similar patients without specific directions on which variables to use for grouping or the exact number of groups. While this may appear to de-emphasize clinical judgement, in reality the result allows objective visualization of relationships without influence from conscious or sub-conscious biases. This analysis demonstrated that groups with a higher likelihood of female sex clustered with other predictors of adverse outcomes including older age, more advanced symptoms, atrial fibrillation and worse renal function. Unlike the clusters including higher proportions of their male counterparts, patients in these clusters did not show improved outcomes when randomized to the 150 mg losartan vs. the lower 50 mg dose. The authors noted that clinical factors associated with female sex may have suggested more frail patients, who were possibly less able to tolerate high-dose medical therapy – a plausible underlying reason for the difference in dose–response relationship. For instance, worse renal function may predispose to hyperkalaemia and either treatment discontinuation or adverse outcomes with angiotensin receptor blockade.13 It is important to remember that randomized trials were developed in order to eliminate the confounding effect of baseline differences between study participants. Subgroup analysis, using simple regression techniques, do not fully achieve this objective. Subgroup analysis using machine learning methods may better control for multiple confounding factors that often exist in different subgroups and allow for a more fulsome understanding of response to therapy. This analysis by Ferreira et al.12 is not without the typical limitations which accompany retrospective analytic designs. While sex was a pre-specified subgroup, other sex-specific considerations were not made. The original study was not designed to incorporate a machine learning approach with inclusion of numerous variables, many of which may not have received full attention by study personnel. Many unmeasured variables may have confounded the results. The use of LCA without an external validation population potentially limit the external validity and/or generalizability. Additionally, the data were derived from participants in a clinical trial, further limiting generalizability. As such, the findings of the present analysis should be considered hypothesis-generating. Systematically incorporating known differences in left ventricular remodelling into patient selection criteria and sex-specific drug dosing regimens into clinical trial design seems like the most important next step. In other words, consideration of comorbid conditions which cluster with female sex should likely be prospectively considered in the enrolment, execution and the analytic planning of studies designed to elucidate differences related to sex. Ideally, large randomized trials for treatments of HF should be powered independently for both male and female participants. This will by necessity mandate inclusion of an adequate number of female participants. This type of objective information may help overcome subconscious bias leading to differences in utilization of HF therapies between female and male patients. Baumhäkel et al.14 demonstrated that male physicians were less likely to prescribe HF therapies to female patients and that lower doses were prescribed. As an extension, worse quality of life in women with HF15 may be a reflection of under-treatment analogous to how the association between hyperkalaemia and mortality may be mediated through HF therapy discontinuation.13 Addressing sex-based differences in response to medical therapies is critical to ensuring adequate care for all patients with HF. However, until these considerations are built into prospective clinical trial designs clinicians are left with more hypothesis-generating results. Conflict of interest: none declared.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.