Our results indicate that early RRT initiation had no statistically significant survival benefit in ICU patients with ARDS and sepsis, with or without renal failure, when compared to no early RRT initiation or even no RRT initiation. Early RRT initiation strategies also did not significantly improve oxygenation or shorten mechanical ventilation time. This implies that although RRT is the standard treatment for AKI and there may be a theoretical basis for improving prognosis, an early RRT initiation strategy aimed at improving hydration status, oxygenation, and even clearing excessive inflammatory mediators in patients with ARDS combined with sepsis does not provide any clinical benefit, regardless of patients' renal function. "Primum non nocere," the adage seems to suggest that the RRT should be delayed or even reduced unless it is really needed . To our knowledge, this is the only clinical study to investigate the clinical benefits of an early RRT initiation strategy in a specific patient population of ARDS and sepsis with or without renal failure.
As mentioned above, patients with ARDS seem to benefit from early RRT initiation, which may promote rapid fluid negative balance. Han et al. compared oxygenation of continuous renal replacement therapy (CRRT) in a total of 53 ARDS patients with early initiation (within 12 hours after ARDS onset) and late initiation of CRRT (48 hours after ARDS onset) and found that oxygenation was significantly improved, and mechanical ventilation time was shortened in the early CRRT group . There were no hard-end points in this small sample study. Another single-center retrospective study compared the hard endpoints of early (≤ 48 hours postintubation) versus late (> 48 hours postintubation) RRT initiation in patients with ARDS, suggesting that the timing of RRT initiation was not associated with a survival benefit . But none of these studies compared hydration status. Our analysis of secondary outcomes showed that early RRT initiation with early onset may control fluid overload and achieve a negative balance 72 hours after admission, but this negative balance is insufficient to improve oxygenation index and shorten mechanical ventilation duration, or even improve survival. Gaudry et al. demonstrated in a post-hoc analysis of the AKIKI trial that it is preferable to use a diuretic rather than initiate RRT if the goal is only to control fluid balance., while patients without AKI may respond better to diuretic administration. Although the surge of pro-inflammatory mediators during sepsis can theoretically be removed by blood filtration , preclinical studies have also provided plenty of positive evidence [15–17]. Cole et al. found that in a total of 24 patients with early septic shock or infectious organ dysfunction, RRT did not reduce circulating levels of several cytokines and allergic toxins associated with organ dysfunction after septic shock or severe sepsis . Therefore, the effect of routine RRT on inflammatory molecules may be small, and higher flow RRT strategies may be required , but the impact on clinical outcomes remains uncertain [33–35]. The findings of our study are consistent with guidelines not to use renal replacement therapy in patients with sepsis or septic shock unless there is a clear indication . Furthermore, despite the increased severity of sepsis-related multiple organ failure, the proportion of early RRT use increased but did not improve survival outcomes (Fig. 4).
In the subgroup analysis, despite the small sample size, the survival benefit of early RRT was almost equal to that of delayed or even no RRT initiation in patients with eGFR ≤ 60 mL/min/1.73m2 (Fig. 4). Although patients delineated by baseline eGFR alone cannot completely replace the definition of AKI, this is broadly consistent with current results based on large multicenter randomized controlled trials (RCT) [19, 20]. In renal function, starting with the matched baseline serum creatinine level, although the early RRT initiation strategy resulted in statistically different serum creatinine levels compared to no early RRT at 48 hours, the difference disappeared at 72 hours. This suggests that the improvement in renal function brought about by the early RRT initiation strategy is delayed and transient. This finding does not support the faster recovery of renal function associated with the early initiation of RRT strategies. In fact, a growing body of evidence supports the hypothesis of renal injury induced by the artificial kidney , including RRT-related hemodynamic instability , biological incompatibility between blood and synthetic dialyzer membrane [38, 39], catheter-related complications (bleeding, infection), [19, 40] and other potential mechanisms involved in RRT-induced renal injury . It may even be an important factor in eliminating or even reversing the possible early RRT in promoting the improvement of prognosis.
The main advantage of this study is that the statistical method of PSM was used to construct a simulated RCT cohort. In fact, it is difficult to compare the advantages and disadvantages of the early RRT strategy to construct RCT in patients without considering renal function status. However, our study has the following limitations. First, these findings are based on an observational study that cannot determine a causal effect between early RRT and improved clinical outcome and still needs to be confirmed in randomized clinical trials, if at all; second, although we used PSM to minimize the bias, we could not rule out the effect of unmeasured confounders, but we assessed the robustness of the association with unmeasured or uncontrolled confounders by calculating an E-value  of 1.79 for early RRT and 1.67 for 30- and 90-day mortality, respectively. To our knowledge, unmeasured/unknown confounders would be less likely to have an association with both early RRT and study outcomes (30- and 90-day mortality) to an extent by having HRs exceeding 1.79 and 1.67, respectively; third, we could not use the Berlin definition  to define patients with ARDS due to lack of information (e.g., bilateral invasion on chest X-ray); fourth, there is a lack of circulatory and pulmonary fluid measurements measured by cardiac catheterization to directly evaluate whether early RRT alleviates pulmonary edema by reducing intravascular hydrostatic pressure, and serum inflammatory mediators and cytokines to directly evaluate whether early RRT alleviates overactivated inflammatory responses; fifth, due to the large heterogeneity of baseline characteristics of patients in intensive care Settings, the percentage of successful matches in the 1-to-1 matched cohort was low (49.4% of all patients with early RRT). However, the consistency of our results using multiple statistical methods and sensitivity analysis ensures the robustness of the results of the small sample cohort as much as possible.