Study population
The DOSE study was a prospective, randomized, double-blind, controlled trial that was designed to test continuous versus bolus administration of intravenous furosemide and high-dose versus low-dose furosemide therapy in AHF patients7. Patients were included if they had at least one symptom (dyspnea, orthopnea, or edema) and one sign (rales, peripheral edema, ascites, or pulmonary vascular congestion on chest radiography) of heart failure. A history of chronic heart failure and usage of oral loop diuretic for more than 1 month were also required. Patients were excluded if they had systolic blood pressure < 90 mmHg or a serum creatinine > 3 mg/dl, or if intravenous vasodilators or inotropic agents other than digoxin were required. There were no exclusion criteria pertaining to ejection fraction. The trial was approved by the ethics committee and an informed consent form was signed by each participant.
The trial used a 2-by-2 factorial design. Totally 308 patients were randomized in a 1:1:1:1 ratio to a low-dose or a high-dose strategy (daily intravenous furosemide dose equal to or 2.5 times their daily oral loop diuretic dose in furosemide equivalents) and to administration by continuous intravenous infusion or intravenous bolus every 12 hours. Physicians can adjust the treatment strategy based on patients’ response to therapy at 48 hours. They can maintain the treatment strategy, increase the dose by 50% while remaining blinded, or change to oral diuretic in preparation for discharge.
Heart failure subgroup
Subgroup of heart failure was determined by value of the last left ventricular ejection fraction (LVEF) measurement. Among 308 patients in the trial, 300 of them had available LVEF data and were included in this analysis. LVEF was measured by echocardiography in 287 patients, by radionuclide ventriculogram in 3 patients, by left ventriculogram in 6 patients, by MRI in 2 patients, and by other methods in 2 patients. One hundred and ninety-three patients with a LVEF < 40% were categorized as HFrEF, while the remaining 107 patients with a LVEF ≥ 40% were categorized as HFpEF.
The original data of the DOSE trial were obtained from the National Institutes of Heart, Lung, and Blood Institute's Biologic Specimen and Data Repository Information Coordinating Center.
Outcome of interest
This study aimed to evaluate the differential response to high-dose versus low-dose furosemide in HFrEF and HFpEF in terms of renal function, diuresis, and short-term clinical outcome.
For renal function, outcomes included change in creatinine and cystatin C at 72 hours measured by core lab, and development of cardiorenal syndrome (defined as an increase in creatinine of >0.3 mg/dl) within 72 hours. To capture the temporal characteristics, change of creatinine at 24 hours, 48 hours, 72 hours, 96 hours, and 7 days measured by local lab were also evaluated.
For diuresis, outcomes included weight change at 72 hours, net fluid loss at 72 hours, area under curve (AUC) of global and dyspnea Visual Analog Scale (VAS) at 72 hours, change in N-terminal pro-brain natriuretic peptide (NT-proBNP) measured by core lab at 72 hours, freedom from congestion at 72 hours, and worsening or persistent heart failure. Freedom from congestion was defined as jugular venous pressure < 8 cm, no orthopnea, and trace peripheral edema or less. Worsening or persistent heart failure was defined as need for rescue therapy over 72 hours. To capture the temporal characteristics, weight change at 24 hours, 48 hours, and 96 hours, and net fluid loss at 24 hours and 48 hours were also evaluated.
For short-term clinical outcome, we evaluated a composite outcome of death, total (first and recurrent) hospitalizations and unscheduled visits for heart failure. The original DOSE study only assessed the first clinical event and ignored recurrent events. However, the DOSE trial itself was underpowered to test the difference in clinical event, let alone the difference in subgroup analysis. It was shown that when treatment effect was consistent during follow-up (or treatment discontinuation rate was low), recurrent-event methods provided greater power than time-to-first methods10. Therefore, we decided to use a recurrent-event method.
Statistical analysis
Continuous variables were presented as mean ± standard deviation or median (25th-75th percentile) and were compared using Student's t test or rank sum test, depending on their normality. Categorical variables were presented as number (percentage) and were compared using chi-squared test or Fisher exact test.
Continuous outcomes were evaluated by linear regression model and binary outcomes were evaluated by logistic regression model. For outcomes that had a relevant baseline value, such as change in creatinine, the baseline value was also adjusted. To evaluate the composite clinical outcome, incidence rates were compared. The effect of furosemide dose was also visualized by Nelson-Aalen cumulative hazard curves11 and quantified by a marginal risk set model proposed by Wei, Lin, and Weissfeld12. Mode of furosemide administration was adjusted in above models. Because most of the baseline characteristics were comparable between high-dose and low-dose treatment arms in both heart failure subgroups (except for the modest difference in oxygen saturation in HFpEF) (Additional file 1), no additional variable was adjusted.
To test the interaction between heart failure type and treatment strategy, heart failure type and heart failure type-treatment interaction term were added in above models. There is possibility that the detected heart failure type-treatment interaction was confounded by interaction between other baseline characteristics and treatment. Therefore, sensitivity analyses were performed to adjust for the effect of baseline characteristics that differed significantly between HFrEF and HFpEF. For example, gender proportion differed between HFrEF and HFpEF. In the sensitivity analysis, heart failure type, gender, treatment, heart failure type-gender interaction, heart failure type-treatment interaction, and gender-treatment interaction were all included in the model.
Results were reported with 95% confidence interval (CI). A P value < 0.05 was regarded as statistical significance.