Introduction

Fluid overload impairs organ function and increases mortality in critically ill patients [1,2,3,4]. A conservative fluid strategy improves outcomes in critically ill patients [5,6,7]. Fluid removal has thus become a key element of critical care, and diuretic therapy is one important mainstay treatment.

Although clinicians use intravenous loop diuretics in diverse ways [8], there are two main strategies, namely, continuous and intermittent administrations. Theoretically, the continuous infusion of loop diuretics is superior to intermittent administration in some aspects [9]: (1) consistent delivery of the drug to the nephron leads to more efficient diuresis by preventing rebound sodium retention and fluid reabsorption, that is, diuretic resistance [10,11,12]; (2) continuous diuretic infusion may decrease the fluctuation in intravascular volume [13]; and (3) continuous diuretic infusion allows the titration of diuretics depending on the patient’s condition.

Earlier studies that compared these two strategies included crossover trials [14,15,16,17,18], in which diuretic effect and resistance could have been carried over to the later phase due to the lack of adequate washout periods. The urine output measured in such trials was, therefore, affected by both the continuous and intermittent infusions of diuretics. Previous systematic reviews that compared two diuretic strategies included such crossover trials, which represented nearly half of the included studies [19,20,21]. Moreover, the populations of these studies were diverse, and the extent of diuretic resistance with respect to disease varied thereafter. Results from parallel-group trials in a homogenous disease population are reliable in precisely assessing the efficacy of a single diuretic strategy by eliminating the risk of diuretic effect and a variety of diuretic resistance.

Thus, we conducted a systematic review and meta-analysis of parallel-group randomized controlled trials that assessed two furosemide strategies. To eliminate the clinical heterogeneity, such as diuretic resistance, we evaluated these two strategies in adults with acute decompensated heart failure (ADHF).

Materials and methods

The conduct and reporting of this systematic review followed the Cochrane Collaboration methodology [22] and PRISMA statement [23], and the protocol is registered at PROSPERO (CRD42017083878).

Eligibility criteria

We considered parallel-group randomized controlled trials investigating adult patients with ADHF. The diagnosis of ADHF was based on the original study authors’ definition, and all etiologies and severities were considered.

The intervention and control groups had continuous and intermittent intravenous administrations of furosemide, respectively. We did not place restrictions on the dose of furosemide after randomization or whether patients received a loading dose before randomization. Given that the clinical management of heart failure may vary across settings, we allowed the concomitant use of other diuretics as long as they were administered to both groups.

We required that a trial used at least one of the following parameters as an outcome: mortality, length of stay, body weight loss, and urine output.

Search strategy

We searched Medline, EMBASE, and the Cochrane Central Register of Controlled Trials. We reviewed the references of the included trials. We also searched Google Scholar and Web of Science for relevant trials that prospectively cited eligible trials. No language or publication status restrictions were imposed. Our search strategy is outlined in Supplemental Table 1. We updated our search on May 26, 2018.

Study selection

Two authors (AK and SU) independently reviewed identified titles and abstracts and selected relevant articles after assessing the full text. Any disagreements were resolved through discussion.

Data extraction

The same authors extracted the data independently, using a pre-designed extraction form. We extracted the following information from each study: (1) patient demographics (age, sex, and left ventricular ejection fraction or New York Heart Association [NYHA] classification), (2) study characteristics (country), (3) information on interventions (dose and loading or furosemide, and duration of the interventions), and (4) outcomes of interest.

Risk of bias assessment

The same authors independently assessed the domain of bias with the Cochrane Risk of Bias assessment tool [24]. We assessed a trial to be at low risk of performance bias, when the study personnel were blinded to the type of interventions. We also examined industry sponsorship or conflicts of interest. Any inconsistency was resolved via consensus.

Statistical analysis

Our primary outcomes were (1) all-cause mortality, (2) length of hospital stay, and (3) reduction in body weight during the study period. Our secondary outcomes included (1) urine output, (2) hypokalemia, (3) hyponatremia, (4) increase in serum creatinine level, and (5) hypotension. For dichotomous and continuous outcomes, we calculated the risk ratio (RR) and weighted mean difference (WMD) with their corresponding 95% confidence intervals (CIs), respectively. When trials had zero events in either arm, continuity corrections were applied with the addition of 0.5 to each cell of the 2 × 2 tables from the trial [25]. Trial data available as median and interquartile range were converted to mean and standard deviation using the method proposed by Wan et al. [26]. Given that the studies were clinically diverse, we pooled data using the DerSimonian and Laird random effects model [27]. Statistical heterogeneity was assessed with I2 and Q statistics [28]. We tested for small-study effect or publication bias using Egger’s method [29] when there was a sufficient number of studies for an outcome.

We conducted a sensitivity analysis by excluding trials with a high or unclear risk of bias with regard to sequence generation, allocation concealment, blinding of personnel and outcome assessors, and sponsorship or conflicts of interest. We also conducted analyses limited to trials that used similar dosages between the groups. The threshold of statistical significance was set at p < 0.05. We conducted the analyses with Stata SE, version 15.1 (Stata Corp., College Station, TX).

Results

Overview of included studies

Our search produced 3221 articles (Supplemental Fig. 1). After applying the inclusion and exclusion criteria, we considered 12 parallel-group randomized controlled trials that compared the continuous and intermittent administrations of furosemide in 923 adults with ADHF (441, continuous; 482, intermittent) (Table 1) [30,31,32,33,34,35,36,37,38,39,40,41]. The reported mean age of participants ranged from 55.4 to 79.5 years, with the proportion of female patients ranging from 25 to 70.6%. The sample size ranged from 20 to 306. The daily amount of furosemide was similar between groups (range, 100 to 329 mg) in all trials, except for two trials. The first of these two trials compared two regimens of intermittent furosemide (20 mg every 6 and 8 h, respectively) with that of continuous furosemide infusion (10 mg per hour) [32]. The second trial used a protocol that allowed the titration of the furosemide dose in the group allocated to the continuous infusion of furosemide and administered furosemide of 62 and 157 mg in the continuous and intermittent infusion groups, respectively [36]. The daily dose of furosemide was fixed in five trials [32, 34, 37, 40, 41], three of which used the same accumulated dosages of furosemide per day between groups [34, 37, 40]; titration of the furosemide dose was allowed in the remaining seven trials. The durations of administrating furosemide varied across studies: 24 h (5 studies), 48 h (3 studies), 48 to 72 h (2 studies), and ≥ 100 h (2 studies). Primary outcomes varied across studies, and sample size calculation was performed in only four trials [31,32,33, 35]. All trials were published in full text. All trials were reported in English, except for one which was reported in Chinese [39]. Three trials were performed in India [34, 36, 37], two in the USA [30, 38], one in both Canada and the USA [31], and one each in China, Egypt, Israel, Italy, Spain, and Turkey [32, 33, 35, 39,40,41]. One trial author responded with data [34].

Table 1 Characteristics of included studies
Full size table

Risk of bias assessment

Overall, six trials (50%) had adequate sequence generation and four (33.3%) had adequately concealed allocations (Table 2). The study personnel and outcome assessors were judged to be adequately blinded in two (16.7%) and one (8.3%) trial(s), respectively. Seven (58.3%) and all studies were deemed to have a low possibility of incomplete outcome data and selective outcome reporting, respectively. Seven trials (58.3%) were free of industry sponsorship or conflicts of interest.

Table 2 Risk of bias in included studies
Full size table

Primary outcomes

Five trials with 499 patients provided data on all-cause mortality. Two trials reported data at 1 month, another two reported data at 2 and 3 months, and the remaining one reported in-hospital mortality, respectively. The continuous infusion of furosemide was not associated with an improvement in all-cause mortality (RR 1.19; 95% CI, 0.65 to 2.16; p = 0.58; Q = 0.97; df = 4; I2 = 0.0%), compared with the intermittent infusion of furosemide (Fig. 1).

Fig. 1
figure 1

All-cause mortality

Full size image

Eight trials involving 696 patients reported on the length of hospital stay. The continuous infusion of furosemide was not significantly associated with a reduction in length of hospital stay (WMD − 0.88 days; 95% CI, − 2.76 to 1.01; p = 0.36; Q = 54.04; df = 7; I2 = 87.0%), compared with the intermittent infusion of furosemide (Fig. 2). There was no evidence of publication bias (p = 0.91).

Fig. 2
figure 2

Length of hospital stay (days)

Full size image

Seven trials with 626 patients reported on the reduction of body weight loss. The continuous infusion of furosemide was associated with a greater body weight reduction (WMD 0.63 kg; 95% CI, 0.23 to 1.02; p = 0.002; Q = 1.66; df = 6; I2 = 0.0%), compared to the intermittent infusion of furosemide (Fig. 3). There was no evidence of publication bias (p = 0.17).

Fig. 3
figure 3

Body weight reduction (kg)

Full size image

Secondary outcomes

Nine (545 patients) and two (349 patients) trials reported on the amount of urinary output at 24 and 72 h, respectively. The continuous infusion of furosemide was not significantly associated with an increase in urine output at 24 h (WMD 489.17 mL; 95% CI, − 183.18 to 1161.51; p = 0.154; Q = 715.63; df = 8; I2 = 98.9%) (Supplemental Fig. 1) or at 72 h (WMD − 36.6 mL; 95% CI, − 335.9 to 386.9; p = 0.012; Q = 0.91; df = 1; I2 = 0.0%), compared to the intermittent infusion of furosemide. There was no evidence of publication bias for the outcome at 24 h (p = 0.41).

Seven trials screened for adverse effects, of which the most common were hypokalemia, hyponatremia, increased serum creatinine level, and hypotension. Compared with intermittent administration, the continuous infusion of furosemide was not associated with the incidence of hypokalemia (RR 1.41; 95% CI, 0.51 to 3.86; p = 0.51; Q = 6.12; df = 3; I2 = 51.0%), hyponatremia (RR 1.45; 95% CI, 0.75 to 2.80; p = 0.27; Q = 0.97; df = 2; I2 = 0.0%), increased serum creatinine level (RR 1.20; 95% CI, 0.85 to 1.69; p = 0.30; Q = 0.94; df = 3; I2 = 0.0%), or hypotension (RR 0.95; 95% CI, 0.48 to 1.88; p = 0.88; Q = 2.28; df = 1; I2 = 0.0%). Post-hoc analyses showed that the continuous infusion of furosemide was not associated with an increase in the change of serum creatinine level between before and after the intervention (Δ serum creatinine) (WMD 0.42 mg/dL; 95% CI, − 0.12 to 0.97; p = 0.129; Q = 372.30; df = 4; I2 = 98.9%) or in the absolute serum creatinine level at the end of the intervention (WMD, 0.18 mg/dL; 95% CI, − 0.06 to 0.41, p = 0.134; Q = 15.06; df = 3; I2 = 80.1%).

Sensitivity analyses

We conducted sensitivity analyses on all-cause mortality, length of hospital stay, body weight reduction, and 24-h urine output (Supplemental Table 2). A paucity of trials with adequate blinding of study personnel and outcome assessors precluded sensitivity analyses for these domains. Pre-planned sensitivity analyses otherwise yielded findings similar to the primary analyses.

Discussion

Our analysis suggested that the continuous administration of furosemide, compared to intermittent administration, was associated with a significant reduction in body weight and a tendency for increased 24-h urine output. There was no benefit in terms of all-cause mortality, length of hospital stay, and adverse events (hypokalemia, hyponatremia, increased serum creatinine level, and hypotension). Most sensitivity analyses were consistent with the primary analyses, thereby confirming the robustness of our findings.

There was no statistically significant difference in 24-h urine output between the two strategies; however, continuous administration was associated with a greater body weight reduction. There are some explanations for this inconsistency. First, three trials seemed to be outliers in terms of 24-h urine output. A post-hoc analysis excluding these trials tends to suggest that continuous infusion produced more 24-h urine output than intermittent administration, with reduced statistical heterogeneity (WMD 576.47 mL; 95% CI, 303.01 to 849.91; p < 0.001; Q = 8.09; df = 5; I2 = 38.2%). Second, trials with a higher risk of bias might have affected this outcome. Most sensitivity analyses with trials at lower risk of bias clearly showed that continuous infusion produced a significantly greater 24-h urine output than intermittent administration and reduced statistical heterogeneity. Thus, continuous infusion may generally produce an increased urine output compared to intermittent administration, which might lead to a great body weight reduction.

No significant benefits in all-cause mortality and length of hospital stay associated with continuous furosemide infusion were found in our analysis. Theoretically, a greater body weight reduction and urine output associated with continuous infusion of furosemide should help accelerate the reduction of congestive symptoms. However, ADHF is a multifactorial disorder [42]. While the continuous infusion of furosemide during the hospitalization may reduce the length of hospital stay and such a non-significant trend was confirmed in our analysis, it is reasonable that it does not directly affect mortality when the patients are discharged. Moreover, the sample size of each study was not calculated for these outcomes; thus, our meta-analysis may have been underpowered.

A theoretical merit associated with the continuous infusion of furosemide is the reduced possibility of electrolyte imbalance and hypotension. Our analysis, however, found no difference in hypokalemia, hyponatremia, and hypotension between the two strategies. This may have resulted from an underpowering due to the small number of trials included in the analyses, given that continuous infusion of furosemide was non-significantly associated with more frequent occurrence of adverse events.

An elevated serum creatinine level is a frequent and important adverse event associated with furosemide. Although statistically non-significant, our analyses suggest that the continuous infusion of furosemide can increase the incidence of increased serum creatinine level or the absolute serum creatinine level in comparison with the intermittent counterpart. Given that this tendency was consistent throughout the analyses, we may need to clinically recognize that the continuous infusion of furosemide is associated with an increased serum creatinine level in comparison with the intermittent furosemide administration.

Three previous systematic reviews have compared these two strategies of loop diuretics administration in heart failure: one focused on heart failure in general, including refractory chronic heart failure [21], another on hypervolemic status including ADHF [19], and the last on acute decompensated and chronic heart failure [20]. Nearly half of these included trials were small-sized crossover trials. As stated earlier, we focused on adults with ADHF in parallel-group randomized trials to investigate the exact efficacy of each furosemide strategy in a homogenous population in terms of diuretic resistance. A significant body weight reduction was consistently found in these reviews and ours. Unlike in our study, Salvador et al. found a significant reduction in the length of hospital stay and all-cause mortality with the continuous infusion of furosemide in refractory chronic heart failure or volume-overloaded status [20]. Two reviews reported a significant difference in 24-h urine output with the continuous infusion strategy; however, the participants were double-counted in some crossover trials [19, 20]. Such differences in patient and study designs might have resulted in clinical heterogeneity including diuretic resistance, thereby, leading to different findings.

Our study has some strengths. First, our search was comprehensive. We searched three large databases, supplemented by a manual search of two other platforms. This allowed us, compared to previous reviews that included both parallel-group and crossover trials, to review a larger number of participants as well as parallel-group trials. Second, we were able to examine clinically relevant, patient-oriented outcomes, such as mortality, length of hospital stay, and body weight reduction as our primary outcomes. Previous systematic reviews examined urine output as their primary outcomes, probably due to the limited number of parallel-group trials. Our study findings are more informative to clinicians who consider diuretic strategies. Third, we excluded crossover trials to eliminate the risk of a carryover diuretic effect and subsequent diuretic resistance. To our knowledge, this is the first systematic review of parallel-group randomized trials that purely compared two furosemide strategies in ADHF.

Our study also has limitations. First, the included trials differed in terms of diuretic protocols and the potential severity of heart failure. These might have led to high levels of statistical heterogeneity in the length of hospital stay and urine output. Lack of information on such variables precluded appropriate subgroup analysis. Second, most included studies were potentially at high risk of performance and attrition bias. Since six of our included studies allowed the titration of furosemide dosages at the treating physicians’ discretion, the study personnel should have been blinded from the type of interventions in order for a trial to be assessed as free of performance bias. A paucity of trials at low risk of such biases, however, precluded sensitivity analyses. Third, the durations of administrating furosemide in the included studies were relatively short and limited only to the acute phase of ADHF treatment; eight out of 12 studies used regimens with duration ≤ 72 h. Thus, our study failed to elucidate the impact on urine output and adverse events associated with either strategy of furosemide with a longer duration. Fourth, we did not discuss the adverse effects related to two furosemide strategies in detail. The CONSORT statement requires that trial investigators report “harms” associated with interventions [43]; however, randomized controlled trials in any fields generally under-report adverse events [44,45,46,47,48]. Only seven out of 12 trials partly reported on clinically important adverse events. Our study found no differences in hypokalemia, hyponatremia, serum creatinine level, and hypotension between the two strategies. Given that each of these outcomes was examined in a small number of studies, caution is still needed.

The continuous infusion of furosemide generally enhances body weight reduction, potentially increases urine output and reduces the length of stay in patients with ADHF. This evidence is derived from a relatively homogenous population and thus is reliable. Thus, clinicians need to select the continuous strategy, serially monitor patients, and try the intermittent one when there is inadequate urine response. Future studies need to select clinically homogeneous patients to reduce the variety of diuretic resistance, minimize the risk of performance and attrition bias, and report adverse events and patient-oriented outcomes.

Conclusions

Our study suggests that continuous infusion of furosemide leads to greater body weight reduction compared with intermittent administration. Limited available evidence suggests that there is no difference in adverse events between the two strategies. Thus, continuous infusion of furosemide should be considered for removing fluids in critically ill patients. However, the duration of the furosemide protocols examined in previous studies was mostly within 72 h, and thus their impact on urine output and adverse events in a longer duration remains to be elucidated.