Introduction

Acute respiratory distress syndrome (ARDS) is defined1 as a bilateral lung injury inducing hypoxaemia (PaO2/FiO2 ratio < 300 mmHg) occurring less than a week after a pulmonary injury (pneumonia, trauma, pancreatitis, etc.) associated with bilateral opacities (not explained by effusions, atelectasis or nodules) and not fully explained by left heart failure. Another condition is a positive end-expiratory pressure (PEEP) greater than 5 cm H2O during invasive mechanical ventilation (IMV).

Patients with severe ARDS often present acute kidney injury (AKI) and it is well established that AKI is a factor of complications and mortality in ARDS2.

Taking the example of the COVID 19 epidemic, a recent study3 showed an incidence of AKI of 26% at KDIGO stage 3, 15% at stage 2 and 17% at stage 1, giving a total of 58% for all stages combined.

A similar result was described by McNicholas et al.4 in a large cohort where they depicted that AKI could be associated with difficulties in weaning from IMV.

Thus, Leite et al.5, showed a causal association between high PEEP levels, impaired respiratory compliances and the severity of AKI.

Ottolina et al.6 confirmed these results for patients with high PEEP (> 13cmH20) while Panitchote et al.7 investigated the factors associated with non recovery.

Rocha et al.8 showed in an interventional animal study that gradually decreasing PEEP combined with restrictive fluid management strategy decreased renal damage and AKI. This restrictive strategy was also confirmed by Upadhyaya et al.9. Finally, Panitchote et al.10, investigated long-term recovery of AKI during ARDS.

While all these studies focused on prognostic factors and therapeutic strategies in the onset of AKI, to our knowledge, there is no investigation concerning specifically the early phase of renal recovery during the weaning from positive expiratory pressure.

The onset of AKI has been suggested to be related to the inflammation associated with ARDS, but also to venous congestion11 that can be increased when using high PEEP. This venous congestion could also be related to right ventricular failure which detection remains challenging12.

Indeed, the venous congestion associated with the positive expiratory pressure is responsible for higher pressure in the inferior vena cava and therefore higher renal downstream pressure in the renal vein. This may cause a lower renal flow rate11.

It is therefore interesting to investigate the renal function immediately after PEEP weaning.

The main objective of our study was to evaluate the mean variation of the urine output (UO) after PEEP weaning.

The secondary objective was to investigate the significant factors associated with a strong variation of UO.

Methods

Study design

We conducted a monocentric retrospective study in the adult medical ICU of Brest university hospital concerning the period from January 2015 to December 2023.

All research was performed in accordance with relevant guidelines, and informed consent was obtained from all participants or their legal guardians.

Data analysis was performed within the ReaSTOC data-warehousing project including all patients admitted to our ICU. It was approved by the ethical committee of Brest University Hospital and registered in Clinical trial RE2SDRA : 29BRC24.0031–RE2SDRA, NCT06754787.

Patients

Patients were included if they were intubated for more than 48 h, exposed to a PEEP ≥ 8 cmH20 followed by a successful PEEP weaning defined as ≤ 6 cmH20 for more than 72 h. The weaning protocol of our ICU follows the protocol of the EXPRESS study13. It consists in weaning the PEEP to 5cmH20 when FiO2 ≤ 60% and paO2/Fio2 > 150mmHg.

Moreover, we selected patients with impaired ourly urine output (UO) (≤ 1 ml/kg/h) the day prior to PEEP weaning.

Patients were excluded if they had chronic end-stage renal disease on admission, refusal to participate, whose non-opposition could not be obtained from themselves or their trusted support person in the event of incapacity or if they were under legal protection.

Data collection

Clinical data were collected manually within the computerized medical file stored in the ICCA software (IntelliSpace Critical Care and Anesthesia, Phillips Healthcare, Amsterdam, Netherlands) and included : demographic data (age, sex), physical data (weight, height), medical history on admission, daily body weight, fluid balance, routine biological results (renal balance, blood gas), medical treatments (vasopressors, diuretics, antibiotics, neuromuscular blockers, initiation of RRT), and mechanical ventilation data (PEEP and FiO2).

Severity of patients condition was assessed at ICU admission using the SAPS 2 score.

Physiological and mechanical ventilation parameters were collected daily for each patient.

The PEEP weaning day (D0) was defined when the PEEP level was decreased to 6 cmH20 or below, without any increase during the following 72-hrs. For each day, the value of PEEP, FiO2 and norepinephrine was measured at 6:00 am. They were consistent with the mean values for the last 24 h.

For each patient we calculated the variation of UO defined as the difference between the mean UO 48 h after PEEP weaning (D2) and 24 h before the weaning (D−1) : ΔUO = UOD2-UOD−1.

The baseline serum creatinine was calculated using the MDRD formula using an estimated glomerular filtration rate (eGFR) of 90/mL/min/1,73m2.

In our unit, we use a strategy of late initiation of RRT based on the classical metabolic parameters: hyperkalemia > 6 mmol/L, metabolic acidosis with pH < 7,2 urea excess > 40 mmol/L or oliguria for more than 72 h.

Data definitions

Two groups were defined depending of the variation of UO. The high responders (HR) group (n = 55) was defined with a variation of UO > 0.35 mL/kg/h while the poor responders (PR) group (n = 65) was defined with a variation of UO ≤ 0.35 mL/kg/h.

Statistical analysis

Statistical analysis was performed with R (Rmcdr for Linux Ubuntu).

Normality of continuous data were assessed with Shapiro-Wilk test.

Outcomes were compared with unpaired Student t-test.

Discrete outcomes were compared with chi-squared. The alpha risk was set to 5%.

We summarized descriptive data as average values for numerical data and as counts (percentages) for categorical data.

Multivariate analysis was performed with a logistic regression with a backward and forward stepwise approach to find the most significantly independent and relevant variables selected from the univariate analysis with p < 0.1.

Cumulative event curves were realised using Kaplan-Meier plots. Log-rank tests were realised to compare the sub-groups.

For all tests, p value ≤ 0.05 were considered statistically significant.

All analyses included complete cases except for few patients transferred from other ICU at the initial phase. When missing, daily weight was calculated from the last known weight.

Results

Trial population

From January 2015 to December 2023, 338 patients required IMV for ARDS within our MICU (Fig. 1). Among them, 128 were excluded because they died before PEEP weaning. 210 were successfully weaned from PEEP (i.e. no increase of PEEP during 72 h) and among them 90 were excluded because they did not have an impaired UO (i.e. was greater than 1 ml/kg/h) or had end-stage chronic kidney disease.

Fig. 1
figure 1

Flowchart showing the selection of patients for study population. UO = Urine Output (mL/kg/h).

Full size image

Finally, 120 patients were included for a primary analysis in our study (65 in poor responder group and 55 in the high responder group with a UO cut-off of 0.35 mL/kg/h).

Table 1 presents the baseline characteristics of our population and unadjusted outcomes according to responding status to PEEP weaning.

Table 1 Baseline characteristics and outcomes. Values are given as average values for numerical data and as counts (percentages) for categorical data. PR = Poor responder, hr = high responders.
Full size table

The mean age of our population was 59 year, 21% of them were women, the mean BMI was 28.

28% had an ARDS due to COVID-19 and their mean SAPS 2 was 50.

There was no significant difference in the baseline characteristics of the two groups in univariate analysis except for the BMI (29,4 vs. 26,6, p < 0.001) and chronic liver failure (9 vs. 2, p = 0.04) which were both increased in the PR group.

Outcomes

Figure 2a shows the time variation of the mean PEEP for the total population while Fig. 2b shows an important increase of UO between D-1 and D2 (around 0.4 mL/Kg/h) before reaching a steady state value.

Fig. 2
figure 2

Variation of PEEP (a) and urine output (b) as a function of time for the total population.

Full size image

While there seem to be no major difference in the PEEP values between the sub-groups (Fig. 3a), UO is increased in the HR group (Fig. 3b).

Fig. 3
figure 3

Variation of PEEP (a) and urine output (b) as a function of time for the sub-groups. PR = Poor responder, HR = High responders.

Full size image

More data related to the weaning period can be found in supplementary materials such as the decrease of Fio2 (figure S1) and norepinephrine (figure S2) while a slight increase of creatinine for the PR group can be noticed in the serum creatinine evolution (figure S3).

As shown in Table 1, we find that 77.5% of our patients had an increase of UO with a mean value of 0.46 mL/Kg/h (p < 0.001).

By definition, UO variation is significantly decreased in the PR group vs. HR group (0.038 vs. 0.976, p < 0.001). Note that there is no significant difference between urine output the day before the weaning.

Moreover, 84% patients had acute kidney injury on the 72 h before PEEP weaning. 50% had at least KDIGO 2 or 3 stages with no significant difference between the sub-groups.

A poor response was associated with a higher total dose of furosemide (186 vs. 81 mg, p = 0.03) as compared with the high responders with a non-significant trend towards more use of diuretics in the PR group at the PEEP weaning period.

There is also a significant difference in the net fluid balance in the PR group (−1.5 vs. −22 mL/kg, p < 0.001).

Moreover, total days of oliguria (11.6 vs. 4.8, p < 0.001) or AKI KDIGO 3 (7.9 vs. 2.4, p < 0.001) were increased in the PR group. They were recorded during the entire stay at the ICU.

The number of sessions of RRT with fluid loss are also increased in the PR group (4.2 vs. 1.6, p = 0.003).

The IMV duration (28 vs. 17, p > 0.001) and PEEP duration (defined as the number of days between the intubation and PEEP weaning) (11.1 vs. 7.5 p = 0.017) are increased in the PR group. This is as well observed for the prone position duration (2.8 vs. 1.5 sessions, p = 0.024).

ICU length of stay is also longer in the PR group (34 vs. 23 days, p < 0.001).

The maximum onset of PEEP, Fio2 or norepinephrine before PEEP weaning has no significant effect on the response to PEEP weaning.

In this univariate analysis, there is also no significant incidence on mortality between the two groups.

Note that concerning the different causes of the ARDS, we observed a predominance of COVID-19 (28,3%), followed by aspiration pneumonia (11,6%), Streptococcus pneumoniae (8,3%), Influenza (7,5%) and Pneumocystis jirovecci (7,5%).

Table 2 shows the results of the multivariate analysis.

Table 2 Multivariate analysis. Odds ratio are given with 95% confidence interval. The analysis was performed with a logistic regression with a backward and forward Stepwise approach.
Full size table

The high response to PEEP weaning was significantly associated with a lower number of days with KDIGO 3 AKI (OR 0.89 [0.82–0.94]) and with a lower number of days before PEEP weaning (OR 0.93 [0.87–0.99]).

The initiation of RRT was significantly increased for patients who had higher doses of norepinephrine (0R = 1.77, p = 0.003), pre-existing chronic kidney disease (OR = 28, p < 0.001), a higher number of days with AKI KDIGO 3 (OR = 1.43, p < 0.001), an increased PEEP duration (OR = 1.09, p = 0.049) while it was inversely associated with an ARDS due to COVID-19 (OR 0.07, p = 0.01).

Finally, ICU mortality was significantly higher in patients who needed RRT (OR 3.14 p = 0.02) and for older patients (OR 1.06 p = 0.01).

We represented in Fig. 4 the cumulative events for a high response to PEEP weaning for two groups depending whether the number of days with AKI KDIGO 3 was lower or greater than 5.

Fig. 4
figure 4

Kaplan-Meier analysis of likelihood of for a high response for two KDIGO 3 durations (Red > = 5 days, Blue < 5 days). Time-to-event was set equal to PEEP duration + 2 days. Log-rank tests were realised to compare the sub-groups. For all tests, p value ≤ 0.05 were considered statistically significant.

Full size image

As found by the multivariate analysis, we notice a significant difference between the two groups with a higher proportion of high responders when the total number of days with AKI KDIGO 3 is lower than 5 days.

Discussion

The objective of our study was to investigate the variation of the urine output after PEEP weaning for patients with ARDS and invasive mechanical ventilation.

We showed that 77,5% of the patients had an improvement in diuresis with a significant mean increase (0.46 mL/Kg/h, p < 0.001) after PEEP weaning. It could be related to the effect of high PEEP on the venous congestion possibly responsible for renal failure14.

We separated the patients in two groups (high responders n = 55 and poor responders n = 65).

While there was no significative difference on the AKI stages, 72 h before PEEP weaning, there were significantly more RRT sessions in the PR group before and after PEEP weaning. This could suggest that, even tough not well described by the AKI KDIGO stages, the PR group would have more severe renal damage due to increased PEEP durations which could influence the urine output evolution.

There are many studies that investigated the factors associated with the development of an acute kidney injury in ARDS3,7,10. Indeed, higher levels of PEEP, higher tidal ventilation volumes, older age, the severity of ARDS, the presence of diabetes and acidosis contributed to the development of AKI7.

Higher BMI, history of heart failure and peak pressures were associated with the severity of the AKI7.

While all these results concerned the development of AKI, few studies investigated the factors associated with the renal recovery.

In a retrospective study of 244 patients, Panitchote et al.10 studied the renal recovery following the definition of Kellum et al.15 based on the variation of their serum creatinine levels and urine output. They defined five patterns of renal recovery: early-sustained reversal (7 days from the start of AKI), early reversal with possible relapse, relapse without final recovery, late reversal (after 7 days) and no reversal.

In fact, they studied all the patients with ARDS including those who never recovered from the respiratory or initial hemodynamic failure making difficult to conclude specifically on the effect of PEEP weaning.

Although in our work, we focused on the increase of the urine output 48 h after PEEP weaning, comparable results were found in the multivariate analysis.

Moreover, we found that both the poor response and the use of RRT were significantly associated with a higher number of days with AKI KDIGO 3 while in their study10, non-recovery was associated with a higher stage of AKI and a delayed onset of AKI.

They also found that non-recovery was more associated with septic shocks while we found that the initiation of RRT was increased for higher doses of norepinephrine.

At the exception of chronic renal failure which was significantly associated with the initiation of RRT3,16, we did not find significant association to patient characteristics anterior to the admission to ICU while Panitchote et al.10 found that a history of malignancy was associated with non-recovery. Although not strictly comparable, it was not the case for a history of immunosuppression or hematological malignancy in our study.

We also found that COVID-19 seems to be a protective factor for the initiation of RRT. Although some studies3,17, covering principally the period 2019–2020, found worse initial AKI for patients with multiple critical failures, especially when comparing to influenza17, among the survivors of the initial phase, there was a significant lower risk for long-term kidney function decline and all-cause mortality compared with non-COVID-1918.

Note however that the initial severity in our cohort measured by the SAPS 2 score for COVID-19 patients was significantly lower (41 vs. 54 p = 0.001) compared to other ARDS causes even though SAPS 2 was not significant in our multivariate model. This could be related to the fact that our investigation covers a later period (until December 2023) with a generalisation of the use of Dexamethasone and Tocilizumab in our ICU making COVID-19 infections less severe.

Finally, the main contribution of our study is to focus on the early days following PEEP weaning. We showed that a longer exposition to high PEEP was negatively associated with an early increase in diuresis independently from the previous maximum onset of PEEP or norepinephrine. Moreover, the mean urine output before the weaning was not a predictive factor for the weaning outcome.

While our study found interesting outcomes using an original approach, it has several limits.

First, it was monocentric and therefore the weaning protocols with a slower PEEP weaning19 could be different in other ICU limiting the generalisation of our results as we had a strategy of fast decrease of PEEP during the weaning process13.

In addition, even though many results were significant, there could be a lack of significance for some of the variables related to the small number of patients.

Moreover, we did not focus on creatinine serum levels as we usually lack the daily serum creatinine value and because the kinetics of variation of creatinine is slower than the variation of urine output.

We also did not classify AKI in term of KDIGO stage 1 or 2 after PEEP weaning because serum creatinine and precise urine output quantification was usually missing for 6 h periods.

Finally, the renal response to the PEEP weaning focused specifically on 72 h of the stay in ICU. This could be compared to an early reversal of AKI (i.e. in the first 7 days based on the definition of Kellum et al.15. Indeed, we did not focus on the delayed renal recovery (after 7 days) or relapse, which could be multifactorial.

Moreover, some of our high responders also needed an initiation of RRT (mean delay of initiation of 17 days for a PEEP duration of 7.5 days) because of multiple possible factors (late relapse of ARDS, new septic shock, etc…). This prevents us to conclude on the long-term evolution of the renal function from our primary outcome.

Conclusion

48 h after PEEP weaning in ARDS patients, the mean urine output was significantly increased. The factors associated with an increased urine output response to PEEP weaning were a decreased duration of IMV before the PEEP weaning, and a decreased number of days with AKI KDIGO 3.

The initiation of RRT was associated with higher doses of norepinephrine, pre-existing chronic kidney disease, higher number of days with AKI KDIGO 3 and increased PEEP duration while it was inversely associated with an ARDS due to COVID-19.