Impact of furosemide on mortality and the requirement for renal replacement therapy in acute kidney injury: a systematic review and meta-analysis of randomised trials

Acute kidney injury (AKI) constitutes a serious clinical hazard in critically ill patients. There are multiple acknowledged risk factors for AKI which can be found in the intensive care unit (ICU) setting, including sepsis, circulatory shock, trauma, and use of nephrotoxic drugs [1]. Therefore, it is not surprising that AKI develops quite frequently, affecting one in five of all hospitalised patients [2], and more than half of all patients admitted to the ICU [3]. AKI has serious sequelae: progression to chronic renal failure; increased risk and progression of multiple organ failure; increased risk of cardiovascular disease; and, finally, increased in-hospital mortality [4].

Pharmacological and non-pharmacological interventions are made for AKI treatment, including metabolic and haemodynamic stabilisation, growth factor and adenosine receptor antagonist’s administration. Unfortunately, some of these were found to have unsatisfactory results. According to KDIGO guidelines [1], diuretics should not be used for the treatment of AKI, except for fluid overload. However, this is based on low evidence (II C recommendation). RRT is the only treatment option for severe AKI.

Previously, several investigators were concerned with increased mortality in patients with AKI treated with furosemide [5, 6]. These findings were rejected by many others [7, 8]. Due to the discrepancy mentioned earlier, the impact of furosemide on mortality in AKI patients has become a subject of several systematic reviews and meta-analyses [9‐13]. As all of these meta-analyses included papers published over 20 years ago, their results should be cautiously extrapolated to current clinical practice due to evident changes in the treatment of the patients at risk of AKI (e.g. those with cardio-renal syndrome, sepsis, in the perioperative period, in the ICU setting). Therefore, we attempted to verify this interesting relationship taking into account the most recent data. We also sought to investigate the possible impact of furosemide on the requirement for renal replacement therapy (RRT).

In this systematic review, we conducted a comprehensive literature search for studies published in English from 1998 up to October 2018. We searched the electronic databases PubMed (Medline) and Embase using a pre-specified strategy (Additional file 1: Table S1). The MeSH (Medical Subject Headings) terms and key words were: furosemide, mortality, renal replacement therapy, acute kidney injury, acute renal failure and renal insufficiency. We searched for all related randomised controlled trials (RCTs) that compared prevention or treatment with furosemide, to a placebo, standard of care or RRT in adult patients with AKI, or at risk of AKI (or acute renal failure). Moreover, additional relevant studies were searched manually by checking the reference lists of identified studies or reviews. We excluded unpublished reports and conference abstracts. Two independent investigators (LJK and PFCz) screened the abstracts and/or manuscripts and extracted appropriate data. Short-term mortality was considered a primary outcome while the requirement for RRT was a secondary outcome. If the outcomes (i.e. OR; 95% CI) for individual RCTs were not revealed in the index publications, we calculated them based on raw data. Subgroup analyses defined by intervention strategy (prevention/treatment), AKI origin (cardio-renal syndrome, contrast-induced nephropathy, post-cardiopulmonary bypass, critical illness), control arm comparator (RRT, saline/placebo/standard of care) and furosemide dose (< 160/≥ 160 mg) were also performed. PRISMA guidelines were followed for appropriate data preparation and reporting [14]. A review protocol regarding this study has not been published before. Flow diagram of the study selection process is presented in Fig. 1. The quality of the trials was verified using the RoB 2.0 tool for assessing the risk of bias in randomised trials (Table 1) [15]. No disagreements occurred between the investigators on the quality of data extracted.

We identified 20 relevant studies (2608 patients: 1330 in the treatment arm and 1278 in the control arm) [7, 8, 16‐33] (Table 2, Additional file 2: Table S2). There were 7 studies on AKI prevention and 13 studies describing AKI treatment with furosemide. Eight trials investigated AKI secondary to heart failure (i.e. cardio-renal syndrome), 4 described contrast-induced AKI (i.e. contrast-induced nephropathy), and 4 studies were conducted in the ICU setting (i.e. AKI of heterogeneous origin). Eighteen trials revealed data on mortality, while 8 trials presented data on the need for RRT. The quality of the papers was deemed suboptimal with the risk of bias noticeable across them (Table 1). Nineteen were included in the meta-analysis (i.e. except from the study of Lim [29]). It was possible to calculate the daily furosemide dose in 17 studies (i.e. apart from the studies of Barbanti [18], Hanna [28] and van der Voort [33]), based on mean or median values given by the authors, or based on the infusion rate and body weight of the participants. This ranged from 20 to 2500 mg, with a median value of 160 mg. A detailed description of all reviewed and included trials is presented in Supplemental Digital Content.

Furosemide had no impact on mortality (random effect model OR = 1.051, 95% CI 0.825–1.339, p = 0.4; fixed effect model OR = 1.051, 95% CI 0.827–1.334, p = 0.4) (Fig. 2). Heterogeneity between studies was low (I² = 0%, 95% CI 0–10.4%, p = 0.9), and publication bias was low (Additional file 3: Figure S1A). Furosemide had no impact on the need for RRT (random effect model OR = 0.947, 95% CI 0.521–1.721, p = 0.2; fixed effect model OR = 0.878, 95% CI 0.602–1.281, p = 0.7) (Fig. 3). Heterogeneity was acceptable (I² = 28.62%, 95% CI 0–69.29%, p = 0.2), and publication bias was low (Additional file 4: Figure S1B).

The results of additional analyses are shown in Table 3. Irrespective of intervention strategy (prevention/treatment), AKI origin (cardio-renal syndrome, post-cardiopulmonary bypass, critical illness), control arm comparator (RRT, saline/placebo/standard of care) and furosemide dose (< 160/≥ 160 mg), there was no effect of furosemide either on mortality or the need for RRT. However, subjects who received furosemide with matched hydration in the prevention of CIN had a less frequent need for RRT (by 79%, with borderline significance).

In this systematic review of 20 RCTs covering over 2600 patients with AKI, or at risk of AKI, we found that furosemide had an impact neither on mortality nor on the requirement for RRT. Furosemide had no effect on the outcomes regardless of intervention strategy, AKI origin, control arm comparator and furosemide dose. Only patients at risk of CIN may benefit from furosemide administration.

These observations are comparable with the available data. In the most recent meta-analysis, Bove et al. [13] revealed no association between furosemide and mortality, the need for RRT, length of hospital stay and worsening of AKI. The authors found significant improvement in survival of patients receiving furosemide as a preventive measure (OR = 0.62; 95% CI 0.41–0.94). It needs to be underlined that they included 8 studies published over 20 years ago and compared intermittent furosemide administration (only) with any comparator (including continuous furosemide infusion) in subjects with or at risk of AKI. In their meta-analysis published in 2006, Ho and Sheridan [9] also found no impact of furosemide, either on mortality or on the need for RRT, regardless of the fact whether it was used to prevent acute deterioration of renal function or treat acute renal failure. However, they included only 9 RCTs describing ARF patients, while only 3 of them investigated the requirement dialysis. Moreover, Bagshaw et al. [10] in 2007 described no effect of furosemide on mortality and renal recovery. In their study, loop diuretics were associated with a shorter duration of RRT (by 1.4 days), a shorter time to spontaneous decline in serum creatinine level (by 2.1 days) and a greater increase in urine output from the baseline (OR = 2.56; 95% CI 1.35–4.85). However, they analysed 5 RCTs only, 4 of which were published over 20 years ago, and all of which were of very low quality. Last but not least, Cheng et al. [34] in their meta-analysis of 7 RCTs, revealed comparable mortality between patients with decompensated heart failure receiving heamofiltration or diuretics (OR = 0.95; 95% CI 0.58–1.55).

Our findings are also in line with the results of Putzu et al. [35] who evaluated whether the administration of furosemide with matched hydration using the RenalGuard System might have reduced the incidence of CIN. Based on 4 trials, they found the beneficial impact of furosemide use on CIN occurrence (OR = 0.31; 95% CI 0.19–0.50), as well as the need for RRT (OR = 0.19; 95% CI 0.05–0.79).

From a clinical point of view, our observations are of particular importance to practitioners as, based on experimental data, furosemide may have some detrimental effects in patients with or at risk of AKI [5, 6, 36]. This observation plays also a key role in determining the necessity of early initiation of RRT in critically ill subjects with deterioration of renal function and/or fluid overload. This is in line with the above-mentioned observations of Cheng et al. [34].

Fluid overload is frequently found in AKI patients in the ICU setting. Fluid excess of at least 10% is associated with increased mortality and morbidity, including pulmonary oedema, cardiac failure, delayed wound healing, tissue breakdown and impaired bowel function [37, 38]. Although there are several methods of evaluating fluid status, most of them are fairly inaccurate [39, 40]. Moreover, the optimisation and management of fluid therapy remains difficult [41‐43]. Diuretics are considered a reasonable first-line treatment but only in responsive subjects [1]. When selecting a diuretic, clinicians should consider evidence-based indications, possible adverse effects, compatibility, pharmacokinetics and other issues of particular diuretics [36]. Diuretics are not recommended to treat AKI per se but are suggested to treat volume overload [37]. According to current data, it is still difficult to predict which patients will respond to diuretics and which patients will benefit from early initiation of RRT.

The severity of AKI alters both the pharmacokinetics and pharmacodynamics of furosemide [40], making appropriate dosing of this agent difficult. The requirement for increased dosing of loop diuretics is associated with worse AKI prognosis in patients with congestive heart failure [44] and in other groups of critically ill subjects [45]. The use of high-dose furosemide to convert oliguric to non-oliguric AKI may exacerbate injury to the kidneys (increased markers of oxidative stress), especially in hypovolaemic patients with reduced renal perfusion [46].

Based on our results, implementation of RRT should be considered at an early stage of critical illness in patients with fluid overload, as well as those with AKI and concomitant indications for immediate extracorporeal treatment [47]. The effect of RRT modality seems to be of lower importance [48, 49].

Based on our meta-analysis of RCTs, furosemide administration has no impact either on mortality or on the requirement for RRT, regardless of intervention strategy, AKI origin, control arm comparator and furosemide dose. However, patients at risk of CIN may benefit from furosemide treatment, if proper hydration is guaranteed beforehand. Further well-designed RCTs are needed in order to verify these findings.