This is a retrospective secondary analysis of the AQUAMARINE study. The study design and primary results of AQUAMARINE have been described elsewhere [11, 12]. In brief, 217 patients with AHF and renal dysfunction (estimated glomerular filtration rate, 15–60 mL/min/1.73 m²) were randomized within 6 h from hospitalization into two groups, either tolvaptan treatment or conventional treatment, to evaluate the efficacy of early treatment with tolvaptan. Fifty-three patients (48.6%) in the tolvaptan group received tolvaptan for more than 2 days, and no patient who was initially allocated to conventional group crossed over to tolvaptan during the first 48 h. The protocol of the study was approved by the ethics committees of all participating centers, and written informed consent was obtained from all the participants. This trial was registered at UMIN-CTR (Unique identifier: UMIN000007109).

In the AQUAMARINE study, data regarding blood pressure, heart rate, and improvement in dyspnea from baseline and blood samples were collected at 6, 12, 24, and 48 h from enrollment. Dyspnea was assessed according to the patient-reported seven-point Likert scale. Within 48 h, the amount of furosemide-equivalent loop diuretics, change in body weight from baseline, and urine output were noted down. Worsening renal function was defined as an increase of ≥0.3 mg/dL in the serum creatinine from the baseline at various pre-specified time points (6, 12, 24, and 48 h from randomization). The incidence of the combined endpoints for all-cause mortality and re-hospitalization for heart failure within 90 days was also evaluated.

We defined diuretic response as the change in body weight (kg) from baseline to 48 h per 40 mg intravenous furosemide administration. Oral furosemide was converted to half the dose of intravenous furosemide. The doses of oral loop diuretics that were considered equivalent to 40 mg intravenous furosemide were 10 mg torasemide and 60 mg azosemide [13, 14]. We also performed analyses using net fluid loss within 48 h as a measure of diuretic response. Diuretic response was measured according to body weight change in 189 cases after excluding 28 cases due to missing data on the total diuretic dose (n = 3) and body weight change (n = 25). Data on diuretic response based on net fluid loss were achieved in 171 cases and missing in 46 cases due to unavailability of information on water intake in 45 cases and on furosemide dose in 3 cases.

Data were expressed as mean ± standard deviation for normally distributed variables and as median with interquartile range (IQR) for non-normally distributed data. Categorical data were expressed as numbers and percentages. The relationships between baseline characteristics, outcomes and tertiles of diuretic response were compared by one-way analysis of variance, Kruskal–Wallis test, or χ² test, as appropriate. Correlation analysis was performed using Spearman’s rho. When necessary, variables were transformed for further analyses. Stepwise multiple linear regression analysis was performed using backward elimination method after including all variables with P values below 0.10 in the univariate analysis. Statistical analyses were performed using R version 3.1.2.

In spite of the lack of a universal definition, poor response to diuretic therapy has been shown to be one of the most powerful prognostic predictors in patients with heart failure [4, 13, 15, 16]. Initial studies used diuretic dose to define diuretic response, i.e., patients with persistent heart failure despite treatment with a certain dose of diuretics were defined to have a “poor diuretic response” [15, 17]. However, this definition used only amount of diuretics and hereby obviously ignored response to the diuretics and, therefore, assumed equal effectiveness. Recently, a novel definition of diuretic response based on urine/body weight response to a certain amount of diuretics was proposed [6]. In all studies that evaluated its prognostic potential, diuretic response consistently showed significant prognostic ability in patients with AHF when this modified definition was used [4, 5, 18, 19].

There has been no consensus on the parameter used to measure diuretic response to 40 mg furosemide or its equivalent, although recent studies have used either change in body weight, net fluid loss, or total urinary output. In the present study, we showed relatively poor correlation between the two measures of diuretic response. This result was in line with that of the DOSE trial and ASCEND-HF, which demonstrated a poor agreement between net fluid loss and weight loss [5, 20]. It is clear that we need better measures of diuretic response to encompass natriuretic response, change in volume distribution, and change in hemodynamic status. However, our consistent results on the improvement of diuretic response, by two different parameters, with tolvaptan supported our hypothesis.

According to this novel definition, the median diuretic response was −0.51 kg/40 mg/48 h furosemide in the conventional group in our study. This was greater than approximately 0.4 kg/40 mg of furosemide-equivalent diuretic response in the patients of the ASCEND-HF (weight change from admission to 48 h), RELAX-AHF (weight change from day 1 to 5), and PROTECT (weight change from day 1 to 4) studies [4, 5, 18]. This better diuretic response in this AQUAMARINE cohort did not match our expectations because our study included only AHF patients with renal dysfunction on admission and earlier studies suggest that renal dysfunction predisposes to worse diuretic response [4, 5, 18]. There are several conceivable speculations for this unexpected result. First, lower doses of loop diuretic were given in AQUAMARINE, compared to other studies and the additional effect of a drug usually decreases at higher doses. Second, although baseline creatinine values were higher in the AQUAMARINE cohort than in the ASCEND-HF cohort, levels of baseline blood urea nitrogen were not substantially different between these two studies. Given that blood urea nitrogen, but not creatinine, has been suggested by previous studies as the most powerful determinant of diuretic response [4, 21], this may be one of the reasons for discrepancy in our study. Third, median time till randomization from patient arrival was 2.1 h and 41.4% of all AQUAMARINE cohort was randomized before admission at the emergency department or clinic. This is surprisingly short given that mean time from admission to randomization was 15.5 h in ASCEND-HF and 7.9 h in RELAX-AHF [22]. This means AQUAMARINE randomized AHF patients much earlier, and we could, therefore, evaluate diuretic response in the very early phase which was not possible with previous diuretic response studies in AHF cohorts. This difference in the time window might be associated with the unexpected good diuretic response in our study cohort. Finally, our results lead to hypothesis that there may be a racial difference in diuretic response. All of the studies regarding diuretic response so far predominantly enrolled Western AHF patients and little is currently known about diuretic response in Asian AHF patients. This hypothesis is supported by the observation that the amount of intravenous loop diuretics used in the acute phase was very low (around or less than 100 mg/48 h) in Japanese AHF patients compared to Western patients [12, 23]. Therefore, influence of racial and/or genetic information on diuretic response needs to be elucidated in future studies.

For both diuretic response parameters, high blood pressure was associated with a good diuretic response. These findings were in accordance with the results of previous studies. In the PROTECT, RELAX-AHF, and ASCEND-HF cohorts, low diastolic blood pressure was an independent predictor of poor diuretic response [4, 5, 18]. Interestingly, intravenous unfractionated heparin was associated with good diuretic response measured with body weight. We have no clear explanation for this finding; however, hyperkalemia is known to be a rare but possible complication of heparin therapy [24], and hypokalemia was suggested as an independent predictor of poor diuretic response in PROTECT. Moreover, there is a case report that suggests a direct effect of heparin on diuresis in patients with AHF [25]. The association between intravenous heparin and diuretic response needs to be precisely elucidated in the future studies.

Poor diuretic response was significantly associated with a high incidence of worsening renal function and low rate of improvement in dyspnea at almost all time points. These results are in line with the findings of previous studies [4, 18]; however, it should be acknowledged that the number of events were very small and this study was obviously underpowered to evaluate prognostic significance of diuretic response.

Although some interventions to treat AHF patients with diuretic resistance have been investigated, there has been no proven therapy to improve diuretic resistance in this high risk population. In the ROSE-AHF study, neither low-dose nesiritide nor low-dose dopamine on top of standard of therapy was associated with a greater reduction in body weight within 72 h [26]. Given that the total amount of furosemide-equivalent diuretic used within 72 h was not significantly different, neither low-dose dopamine nor low-dose nesiritide was suggested to improve diuretic response. Likewise, in ASCEND-HF, nesiritide did not improve diuretic response [5]. In RELAX-AHF, serelaxin did not show a significant improvement in diuretic response of patients with AHF despite its potentially favorable effects on prognosis [18, 27, 28]. Rolofylline, an adenosine A1-receptor antagonist, on the other hand did improve diuretic response [4]. However, its clinical use was hampered by a neutral effect on prognosis and the concern for neurological adverse events. Ultrafiltration might be a promising decongestive strategy [29]; however, it has not been studied specifically in patients with a poor diuretic response.

In the present study, we showed that very early treatment with tolvaptan could improve diuretic response in AHF patients with renal impairment. The pathophysiological background of this favorable effect of tolvaptan on diuretic response remains to be elucidated; however, it may be attributed to certain differences in the mechanisms of action between loop diuretics and tolvaptan. First, time-dependent diuretic resistance was observed with loop diuretics. In patients who have been treated with diuretics for a long time, effectiveness is blunted gradually with time [30]. Second, loop diuretics have to be bound to plasma albumin and delivered to the proximal tubules to exert their effects. Therefore, hypoalbuminemia, which is common in patients with AHF, could contribute to poor diuretic response [31, 32]. Third, active secretion of loop diuretics into the lumen via an organic acid transporter is needed for them to act [33]. This transporter could be inhibited by endogenous organic anions [34]. However, compared with furosemide, tolvaptan has a different mechanism of action, i.e., inhibiting the activation of vasopressin-2 receptor by arginine–vasopressin and subsequent insertion of aquaporin-2 channels in the collecting tubules. This might be one of the reasons for the improvement in diuretic response in renal-impaired patients with AHF after intake of tolvaptan.

Contrary to our result, recent sub-analysis from EVEREST showed a lack of significant difference in prescription rate of tolvaptan between good/bad diuretic response groups [35]. There are some differences in patient backgrounds between EVEREST and AQUAMARINE that possibly explain this discordance (e.g., racial difference, baseline renal function). However, the most conceivable explanation for this discrepancy is time to treatment. In EVEREST, time from hospitalization to dyspnea assessment (the next calendar day after the first drug administration) was more than 36 h in 47.7%, and more than 60 h in 20.2% [36]. In AQUAMARINE about 40% of all patients were randomized before admission to the hospital ward and this early capture of AHF patients may lead to short time to randomization and better diuretic response. The association between time to therapy and diuretic response in AHF patient needs to be addressed in future studies.

Tolvaptan is expected to cause aquaresis but not natriuresis. As sodium retention plays a pivotal role in pathophysiology of AHF, aquaresis may have a different impact on prognosis from natriuresis in AHF patients. Although the pathophysiological background of the association between diuretic response and prognosis has yet to be elucidated, early successful decongestion and subsequent symptom relief are plausible mechanisms. Given that several studies, including AQUAMARINE, have consistently showed urine output with tolvaptan (i.e., aquaresis) could also lead to decongestion and subsequent symptom relief, improvement of diuretic response with early treatment with tolvaptan in AHF patients potentially improves outcome. From this perspective, EVEREST might not be suitable to evaluate this hypothesis as tolvaptan was used relatively late and did not improve diuretic response. As we showed improvement in diuretic response with very early treatment with tolvaptan for the first time, future studies on early use of tolvaptan for patients with AHF having poor diuretic response are warranted.

The most powerful limitation of this study which should be acknowledged is that this is a post hoc and non-pre-specified analysis. Moreover, several analyses were performed without adjusting for multiple testing. Given these points, our study result should be interpreted as an exploratory analysis and hypothesis generating.