In the present study, we found that early loop diuretics were associated with reduced 28-day mortality in ARDS patients after adjustment of both time-fixed and time-varying confounders. LCA identified three phenotypes and patients in subtype3 who were characterized by worse renal function and higher CVP, might benefit from diuretics. Subgroup analysis indicated that the association between diuretics and reduced 28-day mortality was more remarkable among female, sepsis induced ARDS, and those with lower PaO2/FiO2 ( < = 150mmHg), higher MAP ( > = 65mmHg).
Fluid therapy is the fundamental treatment for ARDS, while volume overload is rather common and related to increased risk of death[19]. Diuretics are frequently prescribed in the critically ill to facilitate liquid removal and have become a pharmacologic adjuvant therapy in ARDS[20]. Studies found that compared with liberal fluid strategy or standard care, conservative fluid management achieved by restricting fluid intake, use of diuretics or hemofiltration, was associated with improved oxygenation, increased ventilation free days, and decreased mortality[21–24]. What’s more, it has been proposed that correction of fluid retention may rely on diuretics or renal replacement therapy once the hemodynamic status is stabilized [25]. Of note, early diuretics use was independently associated with lower mortality, which had been found in a less rigorous study, using logistics regression based on the time-fixed baseline variables[7]. We explored the effect of diuretics on 28-mortality using MSCM to adjust time-dependent confounders, and the results further support its use in ARDS patients.
There are evident distinctions in the etiology, physiology, and biology of ARDS patients, leading to different responses to the same therapy[10]. We identified three subtypes by LCA and MSCM indicated that diuretics correlated with reduced 28-day mortality in subtype3, who were distinguished by elevated serum creatinine, higher CVP, more complications including diabetes mellitus, hypertension and heart disease. Previous study has found that in ARDS patients especially with concomitant acute kidney injury, positive fluid balance was associated with higher mortality[26]. However, when used appropriately, frusemide may prevent and even resolve acute kidney injury as well as improving survival[27, 28]. Further research revealed that diuretics were significantly associated with lower mortality in the positive fluid balance subgroup while insignificant in the negative fluid balance subgroup[29]. Moreover, diuretics have been recommended in patients with hypertension and heart failure to promote water and sodium excretion and reduce volume load[30, 31]. The effect of diuretics on mortality might be attributed to the improvement of renal function and reduction of fluid retention.
Fluid resuscitation is highly recommended in sepsis management[32] while persistent positive fluid balance was an independent risk factor for death[33]. Actually, in ARDS patients complicated by septic shock, achieving both early goal-directed cardiovascular resuscitation and late conservative fluid therapy was related to the lowest mortality[34]. So far, the conservative strategy has been recommended for sepsis-induced ARDS who do not have evidence of tissue hypoperfusion[32]. The results that the diuretics correlated with decreased morality in sepsis-induce ARDS patients and those with higher MAP are consistent with the current clinical practice. Additionally, we found that diuretics might be favorable to patients with PaO2/FiO2 < = 150mmHg, possibly due to the reduction of EVLW. Since EVLW estimates the fluid in pulmonary interstitial and alveolar spaces, and is strongly associated with deterioration of PaO2/FiO2, severer lung injury and higher mortality[3, 35], and decrease in EVLW may be associated with improved survival[36]. We postulated that diuretics may obviously alleviate pulmonary edema in the worse oxygenation subgroup and contribute to better survival.
The present study is the first to explore the effect of loop diuretics use on 28-day mortality of ARDS patients, using MSCM to account for both time-fixed and time-dependent confounders. The phenotypes derived based on variables accessible from medical history and routine laboratory tests may inspire clinicians to more precise treatment. Also we have to acknowledge that several limitations exist. Firstly, we were unable to adjust confounders after the first seven days, since the original trial was planned for a maximum of one week. However, when evaluating seven-day mortality, the benefit of diuretics was also significant (Table S5). The survival advantage of diuretics treatment might persist for a long time. Secondly, we did not include inflammatory biomarkers in LCA, due to limited access to data extraction. Nevertheless, we divided the patients into three categories of mild, moderate and severe disease severity, which was in accordance with clinical practice. In addition, the retrospective secondary analysis was insufficient to explain the causality, well designed randomized controlled clinical trials are required.