Use of loop diuretics in chronic heart failure: do we adhere to the Hippocratian principle ‘do no harm’?

Although approaches to optimize volume status, such as modifying fluid and salt intake, are important,1 use of loop diuretics remains the mainstay of treatment for congestion in patients with both acute (AHF) and chronic heart failure (CHF), regardless of the underlying left ventricular ejection fraction.2 Characteristically, more than 80% of patients with CHF receive regular treatment with some kind of oral loop diuretic.3, 4 The 2016 European guidelines on heart failure (HF) recommend the use of diuretics for patients with signs and symptoms of congestion (recommendation class I, level of evidence B).5 However, loop diuretics may exert a range of adverse effects including electrolyte depletion, which predispose to life-threatening ventricular arrhythmias, hyperglycaemia, hyperuricaemia, orthostatic hypotension, vestibular symptoms and renal function deterioration.2 Furthermore, they seem to activate the neurohormonal system and hamper the up-titration of guideline-recommended HF treatment.6, 7 Thus, current guidelines recommend use of diuretics at the lowest dose needed to achieve and maintain euvolaemia, meaning that reduced loop diuretic dose may often be indicated in stable CHF patients.5 Recently, the Heart Failure Association (HFA) of the European Society of Cardiology (ESC) put together a comprehensive consensus document on the use of diuretics in HF,1 providing, among others, guidance on the challenging task of assessing HF patients' fluid status. Herein we summarize the existing literature on loop diuretic dose changes in CHF and call for randomized trials. Loop diuretic strategies in AHF are not addressed. Contemporary, double-blind, randomized controlled trials (RCTs) of loop diuretic use vs. placebo in HF patients are lacking. Given that loop diuretics are administered for alleviation of bothersome, congestion-related symptoms and signs, willingness of physicians to randomize patients in this setting is doubtful, while the absence of a clear dose–response relationship complicates the design of such studies. Moreover, lack of interest from industry in these generic and inexpensive drugs further reduces the likelihood of future RCTs, although registry-based RCTs may be a suitable design,8 possibly in combination with home weight and symptom monitoring and portable devices for assessing serum electrolyte levels. However, a few randomized studies specifically addressing loop diuretic withdrawal or dose decrease are available in the literature. The only meta-analysis of randomized studies on loop diuretic withdrawal was published 15 years ago, while an attempt to update it in 2012 was judged fruitless due to the paucity of new studies.9 It showed lower HF hospitalization rates among patients who continued loop diuretics compared with patients in whom diuretics were discontinued.9 It is notable that this single meta-analysis is cited to support the recommendation on use of loop diuretics in the most recent European HF guidelines.4 Nonetheless, this study has received significant criticism, as the only four withdrawal studies that it included were performed more than two decades ago and the populations that they included have little resemblance to the contemporary CHF population.10 A recent multicentre double-blind study randomly assigned patients with CHF New York Heart Association (NYHA) functional classes I and II and reduced left ventricular ejection fraction (≤45%), who were receiving furosemide at a maximum stable dose of 80 mg daily, either to loop diuretic continuation n = 93, median N-terminal pro B-type natriuretic peptide (NT-proBNP) 696 (interquartile range, IQR 282–1776) or withdrawal n = 95, median NT-proBNP 644 (IQR 287–1527) (Table 1).11 In this study, loop diuretic withdrawal did not affect patients' self-perceived dyspnoea severity and was not associated with increased needs for additional diuretics during 90-day follow-up. The authors concluded that furosemide withdrawal may be feasible and safe in a sub-group of stable CHF patients, given that they are closely monitored during follow-up. However, these findings warrant confirmation in larger studies and different CHF populations. Only one randomized study of loop diuretic dose decrease in patients with CHF has recently been performed.12 Researchers randomized 40 patients with stable CHF under treatment with stable, high doses of oral furosemide (≥120 mg daily) either to maintenance of dose or reduction of furosemide dose to approximately one third. During follow-up (838 ± 262 days), only one patient (5%) in the dose decrease group required subsequent up-titration of loop diuretics (Table 1). The decrease of loop diuretic dose was not associated with deterioration of NYHA class, increase of body weight or worsening of exercise capacity. On the contrary, maintenance (vs. reduction) of high-dose loop diuretics was associated with significant increase in serum creatinine and decrease in plasma haemoglobin. However, serum creatinine has serious limitations as a surrogate marker of renal function, especially if used to evaluate acute changes. Current literature is abundant with non-randomized studies that examined the association of (i) loop diuretics and their doses, and (ii) changes in loop diuretic dose, with various endpoints in patients with CHF. A recent meta-analysis compared the use (vs. non-use) of loop diuretics and the use of high-dose (vs. low-dose) loop diuretics relative to the incidence of all-cause death and HF-related hospitalizations in adult patients with CHF.13 The meta-analysis included 24 non-randomized studies with a total of 96 959 patients and demonstrated an association between use of loop diuretics and higher all-cause mortality and HF hospitalizations. An association between use of high-dose loop diuretics (median dose 80 mg daily) and increased all-cause mortality and HF hospitalizations was also shown. Interestingly, a recent sub-analysis of the Medicare-linked Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients with Heart Failure (OPTIMIZE-HF) registry reported that among the elderly, loop diuretic-naïve patients who received loop diuretics at discharge after a HF hospitalization had significantly better outcomes at 30 days post-discharge.14 Several putative explanations can be postulated for the conflicting results of the latter study with older age, loop diuretic naivety, lower use of HF pharmacotherapy and shorter follow-up being the most obvious. Some prospective, non-randomized studies have demonstrated that dose decrease or even withdrawal of loop diuretics is feasible in selected stable CHF patients with no recent HF hospitalization or HF pharmacotherapy change, with uneventful reduction occurring in 58–83% of patients (Table 1).15-18 Two of these studies also reported that loop diuretic dose decrease was accompanied by improvements in renal function,16, 17 especially in patients with impaired renal function at baseline.17 Importantly, in the latter study, renal function was assessed by measuring glomerular filtration rate (GFR) of iothalamate and not by estimating GFR with use of equations based on serum creatinine,17 in this way likely reflecting true changes in renal function. Few studies to date have evaluated the association of loop diuretic dose change with clinical outcomes in patients with CHF. A sub-analysis of the CardioMEMS Heart Sensor Allows Monitoring of Pressure to Improve Outcomes in Class III Heart Failure (CHAMPION) trial demonstrated that the intervention that patients of the study's active arm more often underwent and that likely drove the positive results of the trial was changes in loop diuretics (both increases and decreases).19 Patients randomized to active monitoring had a 33% lower risk of HF hospitalization compared with the control group,20 which was even higher among patients with diuretic changes (38%).19 A recent analysis of the ESC EURObservational Research Programme Long-Term Heart Failure (ESC-EORPHF-LT) Registry in 8130 patients with CHF demonstrated that increased loop diuretic dose at baseline visit was independently associated with worse outcomes, while decreased loop diuretic dose was associated with better outcomes compared with patients who kept their loop diuretic dose unchanged over a median follow-up of 372 (IQR 363–419) days21 (Table 1). Loop diuretic dose decrease was successful in half of patients it was performed. Importantly, independent predictors of successful loop diuretic dose decrease, namely higher systolic arterial pressure, absence of sleep apnoea, absence of peripheral congestion and absence of moderate to severe mitral valve regurgitation, were identified for the first time, providing guidance for the potential selection of patients in whom dose reduction could be attempted. In a recent study from nationwide administrative registers in Denmark, all patients (n = 74 990) who were diagnosed with HF over a 15-year period (2001–2016) and received angiotensin-converting enzyme inhibitors/angiotensin receptor blockers and beta-blockers within 120 days of diagnosis were identified.22 Diuretic intensification, defined as new addition or doubling of loop diuretics or addition of thiazide to loop diuretics, was associated with a 75% increased risk of 1-year all-cause mortality after adjusting for various confounders, including age, sex and diabetes mellitus,22 consistent with more severe HF requiring greater diuretic use. More than 80% of contemporary patients with CHF receive a loop diuretic.21 Nonetheless, contrary to current guideline recommendations, in <10% of patients who visited a specialized HF clinic across 33 European and ESC-affiliated countries was loop diuretic dose decreased, a proportion which did not differ considerably between clinically stable and unstable patients.21 Recent data suggest that appropriate adjustment of loop diuretic therapy, tailored to patients' haemodynamic status, may reduce the risk of HF hospitalization.19 At the same time, a growing body of literature also demonstrates that (inappropriate?) use of loop diuretics in real-life settings, especially at high doses, is associated with poor prognosis in patients with stable CHF.13 Despite extensive multivariable adjustment in these studies, variables such as diuretic resistance, ascites, right-sided HF and refractory oedema, which are at the same time associated with both HF severity and greater need for loop diuretics, are often not available for use in multivariable models. Thus, it remains difficult to discern whether loop diuretic use is a risk factor for adverse events or merely a risk marker of severe disease, representing unmeasured confounding. However, it is a reasonable course of action to attempt decrease or even withdrawal of loop diuretics in selected hypovolaemic or euvolaemic patients with stable CHF (Figure 1), especially in the absence of sleep apnoea, peripheral congestion and moderate to severe mitral valve regurgitation, and in the presence of higher systolic arterial pressure. On the other hand, it is well established that patients with AHF are often discharged with residual congestion, which is a strong driver of early readmissions and unfavourable outcomes.23, 24 This may be in part explained by the fact that fluid overload is often in the interstitial rather than the intravascular space, thus complicating both absorption and effects of loop diuretics. Congestion is itself a major contributor to kidney dysfunction, and discharge prior to achieving full decongestion should be avoided. An escalated approach with combination of treatments (medication of different categories with diuretic effects and/or ultrafiltration) should be implemented to this end. Continuing an outpatient parenteral administration scheme under close monitoring for a short period post-hospitalization may also be warranted. The latter could also apply for HF outpatients who have symptoms and signs of congestion despite increasing doses of oral diuretics, as venous congestion is a key contributor to renal function deterioration and diuretic resistance in patients with CHF.25 Acknowledging that a ‘one size does not fit all’ approach is inappropriate in the case of loop diuretics, it is of paramount significance to phenotype patient profiles and clinical scenarios where loop diuretic dose changes may be feasible and beneficial. In this direction, we are in need of studies to provide data on appropriate dosing, route of administration, potential combination therapies and dynamic changes. A recent study demonstrated that sacubitril/valsartan use resulted in more loop diuretic dose reductions and fewer dose increases compared with enalapril.26 Furthermore, decongestion through glycosuria and osmotic diuresis has been proposed as one of the multiple putative mechanisms of the beneficial effects noted with use of the novel sodium–glucose co-transporter 2 (SGLT2) inhibitors .27-29 The potential diuretic effect of SGLT2 inhibitors was corroborated by the results of the recent Effect of Sotagliflozin on Cardiovascular Events in Patients with Type 2 Diabetes Post Worsening Heart Failure (SOLOIST-WHF) trial.30 Patients who received treatment with the SGLT1/2 inhibitor, sotagliflozin, during or shortly after hospitalization for AHF showed a significantly lower risk for HF readmissions or urgent emergency department visits compared to patients receiving placebo.30 Considering that (i) a significant proportion of patients hospitalized for HF are discharged with residual symptoms and signs of congestion,23, 24 (ii) residual congestion is associated with a high risk for 60-day rehospitalization,31 and (iii) the beneficial effects of sotagliflozin are evident early (within 90 days) after initiation of treatment,30 the effects of sotagliflozin are likely mediated via diuretic/haemodynamic mechanisms. Surprisingly, however, randomization to dapagliflozin in the Dapagliflozin And Prevention of Adverse-outcomes in Heart Failure trial (DAPA-HF) was not associated with lower doses of loop diuretics during follow-up compared to placebo.27 Nevertheless, these hypotheses may only be confirmed in randomized trials. Conflict of interest: none declared.