Background

Critically ill patients with sepsis develop acute kidney injury (AKI) in 20 to 50% of cases [1]. Despite its clinical importance, available methods for early detection of kidney damage have major limitations. One of the mechanisms implicated in the pathophysiology of this condition is hypoxia of renal tissue, particularly in the renal medulla [2]. Selective hypoxia in the renal medulla was observed in an ovine model of sepsis, despite the presence of global renal hyperemia [3]. In clinical practice, it is not feasible to measure renal medullary tissue oxygen tension in the intensive care unit (ICU). Nevertheless, it is possible to measure oxygen tension of bladder urine (PuO2) by the introduction of a fiber-optic probe in the bladder catheter [4, 5]. A number of experimental investigations have been performed to evaluate this technique and have documented a robust correlation between urinary PO2 and medullary PO2 [6].

One of the caveats for the understanding of relationship between urinary and medullary PO2 is that it may be cofounded by several factors. A concern arising from experimental and computational models is that systemic arterial oxygen tension (PaO2) may affect ureteric and bladder wall oxygenation which, in turn, may influence PuO2 [7,8,9,10]. Experimental observations in anesthetized rabbits suggested that oxygen exchange within the urinary tract is slow and is unlikely to be a major confounder of the relationship between renal medullary tissue PO2 and PuO2 [7]. Nevertheless, the potential for such confounding in human sepsis, where PuO2 measurement might be applied to guide management of risk of AKI and therapeutic interventions, remains unknown. Therefore, to assess whether systemic oxygenation has a potentially confounding impact on urinary oxygenation, we aimed to evaluate PuO2 in critically ill patients with sepsis during the periods before and after changes in fractional inspired oxygen (FiO2) instituted to manage PaO2 within clinically acceptable levels. Also, to support our clinical findings, we investigated the medullary and urinary oxygen tension levels in a sheep experiment within a similar range of FiO2 variation.

Methods

Observational study in septic patients

Study design

We performed a prospective observational cohort study in the ICU of a tertiary care hospital located in Melbourne, Australia, from January 2017 to March 2018. The protocol was approved by the Human Research Ethics Committee of the Austin Hospital (HREC/16/Austin/26). Written informed consent was obtained from all participants or their legal representatives.

Participants

A convenience sample of adult patients (18 years old or older) with suspected or confirmed septic shock was enrolled in the study. We excluded patients who were anuric, on chronic dialysis, pregnant, or who were kidney transplant recipients.

Measurement of PuO 2

For each patient, a fiberoptic luminescence optode (NX-LAS-1/O/E-5 m, Oxford Optronix, Abingdon, UK) was inserted into the lumen of the urinary catheter via a sterile procedure, as described in detail previously [11]. In brief, the sensing tip of the probe was advanced to the distal tip of the catheter so that the probe was placed inside the bladder. The fiberoptic probe was connected to a luminescence lifetime oximeter (Oxylite Pro, Oxford Optronix, UK) interfaced with a laptop computer running LabChart software (Version 8, ADInstruments, Bella Vista, NSW, Australia). PuO2 was recorded every minute from the time of the probe insertion until the removal of the urinary catheter by the treating medical team or at ICU discharge, whichever occurred first.

FiO 2 settings and measurement of PaO 2 and PuO 2

FiO2 was modified at the discretion of bedside clinicians to maintain a peripheral oxygen saturation level greater than 90%. Episodes of FiO2 change were documented in the laptop computer software by the bedside nurse at the moment the intervention took place and were verified against observation charts. To describe arterial oxygen levels, we identified arterial blood gases (ABG) that were collected before and after FiO2 modification (Fig. 1). To make before and after periods distinctive, we restricted the analysis to ABG samples that were obtained within a time difference from FiO2 change of 30 min or more. For each PaO2 measurement, we obtained the mean value of 30 PuO2 measurements centerd around the exact time when the ABG was collected (15 min before and 15 min after the blood was drawn). Blood gas analysis was performed with an ABL800 FLEX blood gas machine (Radiometer, Copenhagen, Denmark). No specific method was used to ascertain the ABG stability for the purposes of the study. However, the unit where the study was conducted is a world-class intensive care with a 1:1 nurse-to-patient ratio. The ABG analyzer is located inside the unit and the team is trained to obtain reliable measurements according to the institutional protocol.

Fig. 1
figure 1

Schematic representation of the procedure used to obtain PaO2 and PuO2 measurements before and after an episode of FiO2 change during the observation period. ABG: arterial blood gas; FiO2: fraction of inspired oxygen; PuO2: urinary oxygen tension

Full size image

Data extraction

Aside from the abovementioned variables, we collected information on age, gender, baseline creatinine, infection source, comorbidities and ICU severity scores. We also recorded data on duration of mechanical ventilation, number of hours in mandatory and spontaneous modes, mechanical ventilation parameters, end-tidal partial pressure of carbon dioxide (EtCO2), ICU and hospital length of stay, and hospital mortality.

Experimental study using healthy adult sheep

Animal preparation

We obtained data from 9 healthy adult Merino ewes included in an experimental study of sheep undergoing aseptic surgical procedures under general anesthesia [18]. The study was approved by the Animal Ethics Committee of the Florey Institute of Neuroscience and Mental Health under guidelines laid down by the National Health and Medical Research Council of Australia. A similar fiberoptic luminescence optode (Oxford Optronix, Abingdon, UK) used in human patients was inserted into the lumen of the urinary catheter and surgically inserted into the renal medulla. The tissue oxygen tension was continuously recorded at 100 Hz on a computer using a CED micro 1401 interface with Spike 2 software (Cambridge Electronic Design, Cambridge, United Kingdom).

Experimental protocol for the variation of FiO 2

The protocol had 4 components of 20-min duration: a 10-min period to allow oxygen levels to stabilize followed by a 10-min experimental period. Our primary experimental stabilization criteria included renal medullary PO2 and urinary PO2. This timing was determined by assessing the medullary and urinary PO2 values over time. A block randomization was used to set FiO2 at 0.21, 0.40, 0.60 and 1.0. The total gas flow on the mechanical ventilator was maintained at a constant rate of 1.5 L/min, whilst the ratio of the individual oxygen-to-air gas volumes was altered to achieve the target FiO2. For the current analysis, we obtained data from the periods when FiO2 levels were 0.21 and 0.40.

Statistical analyses

Continuous variables are reported as median (quartile 1, quartile 3) and categorical variables are reported as number (%). An aggregate measure was calculated per patient and the paired-sample Wilcoxon test was used to compare median values between the two time periods (before and after FiO2 change). Proportions were compared using Fisher’s exact test. To account for multiple episodes of FiO2 change per patient, we performed a mixed linear regression model to assess the relationship between the variation of PaO2 (ΔPaO2) and the variation of PuO2 (ΔPuO2).

In the sheep experiment, values for each FiO2 setting were calculated and a comparison between groups was performed by using Kruskal–Wallis test.

Statistical analysis was performed using R version 4.0.5. Two-tailed p ≤ 0.05 was considered statistically significant.

Results

Human septic patients

We studied 10 patients, whose clinical characteristics are reported in Table 1. During the study period, patients were mechanically ventilated for 733 h (88.9% of total duration), of which 459 h were in spontaneous mode (55.6%). The mechanical ventilation parameters during the study period are described in Table 2. Arterial blood gases were obtained in 233 occasions. Across the cohort of 10 patients there was a weak but statistically significant positive association between PaO2 and PuO2 (Fig. 2, r2 = 0.022, p = 0.004). This relationship was plotted for each patient (Additional file 1: Fig. S1).

Table 1 Baseline characteristics of study patients
Full size table
Table 2 Mechanical ventilation parameters of human patients
Full size table
Fig. 2
figure 2

Scatterplot of arterial and urinary oxygen tension measured in the ten patients enrolled in the study. PuO2: urinary oxygen tension; PaO2: arterial blood oxygen tension

Full size image

We observed 63 episodes of changes in the FiO2 setting: on 32 occasions FiO2 was decreased and on 31 FiO2 was increased. In the episodes where FiO2 decreased, the median [Q1, Q3] pre-intervention FiO2 was 0.36 [0.30, 0.39] and the median post-intervention FiO2 was 0.30 [0.23, 0.30] (p = 0.006). When FiO2 increased, the median FiO2 before the intervention was 0.30 [0.21, 0.30] and the median FiO2 after the intervention was 0.35 [0.30, 0.40], p = 0.008. There were 14 episodes of successive increase/decrease in the same patient. Such episodes were successive and conditional on the presence of a PaO2 measurement and the time difference between these episodes was 4.29 [2.49, 5.38] hours.

In the episodes when FiO2 was decreased, PaO2 fell from 83 [77, 94] mmHg to 72 [62, 80] mmHg (p = 0.009, Fig. 3). Nevertheless, PuO2 did not vary significantly across the two time points, being 23.2 [20.5, 29.0] mmHg before the intervention and 24.2 [20.6, 26.3] mmHg after the intervention (p = 0.557, Fig. 4). In such episodes, ΔPaO2 was − 14 [− 22.2, − 3.0] mmHg and ΔPuO2 was − 0.02 [− 4.3, 2.9] mmHg.

Fig. 3
figure 3

Box plot representation of arterial blood oxygen tension before and after changes in fractional inspired oxygen. FiO2: fractional inspired oxygen; PaO2: arterial blood oxygen tension

Full size image
Fig. 4
figure 4

Box plot representation of urinary oxygen tension before and after changes in fractional inspired oxygen. FiO2: fractional inspired oxygen; PuO2: urinary oxygen tension

Full size image

When FiO2 was increased, PaO2 increased from 64 [58, 72] mmHg to 71 [70, 100] mmHg (p = 0.038, Fig. 3). The corresponding PuO2 measurements were 25.0 [20.7, 26.8] mmHg before the intervention and 24.3 [20.7, 26.3] mmHg after the intervention (p = 0.652, Fig. 4). ΔPaO2 was 8 [− 5.5, 14.0] mmHg and ΔPuO2 was 0.5 [− 2.6, 4.1] mmHg. A mixed linear regression model showed a weak relationship between the change in PaO2 and the change in PuO2 (r2 = 0.003, p = 0.652, Fig. 5) Also, we obtained of an aggregated measure per patient and observed that ΔPuO2 was − 0.532 [− 1.410, 0.331] mmHg and ΔPaO2 was − 3.25 [− 8.880, − 0.125] mmHg, p = 0.1431.

Fig. 5
figure 5

Relationship between the changes in arterial oxygen tension (Δ PaO2) and urinary oxygen tension (Δ PuO2)

Full size image

Other laboratory parameters measured before and after FiO2 change were similar between the two time points (Table 3). The urine output before FiO2 change was 63.8 [32.5, 95.0] ml and 70.5 [30.6, 95.0] after FiO2 change, p = 0.9396 (Additional file 2: Fig. S2).

Table 3 Laboratory values before and after FiO2 change in human patients
Full size table

Experimental study

In the sheep experiment, we evaluated urinary and medullary tissue oxygen measurements in four FiO2 levels: 0.21, 0.40, 0.60 and 1.00 (Table 4). For each variable, a total of 36 measurements were obtained. The median PaO2 value at 0.21 FiO2 was 54.5 [51.3, 74.4] mmHg, 209 [181, 223] mmHg at 0.40 FiO2, p < 0.001. When comparing 0.21 and 0.40 FiO2, we found no statistically significant difference in oxygen tension measurements. The medullary oxygen tension was 25.3 [15.3, 30.5] mmHg at 0.21 FiO2 and 28.3 [15.9, 43.4] mmHg at 0.40 FiO2, p = 0.6766; and the urinary oxygen tension was 25.5 [21.6, 32.6] mmHg at 0.21 FiO2 and 30.0 [27.4, 33.6] mmHg at 0.40 FiO2, p = 0.3192.

Table 4 Measurements of PaO2, medullary oxygen tension and urinary oxygen tension under different FiO2 settings obtained in the experimental study using healthy adult sheep
Full size table

At higher FiO2 levels, PaO2 values increased (303 [282, 306] mmHg at 0.60 FiO2 and 510 [499, 525] mmHg at 1.00 FiO2. Furthermore, medullary oxygen tension tended to increase at these levels (33.4 [22.6, 45.0] mmHg at 0.60 FiO2 and 40.0 [34.0, 46.8] mmHg at 1.00 FiO2 (p = 0.087) while urinary oxygen tension values significantly increased (59.4 [36.5, 66.0] mmHg at 0.60 FiO2 and 87.9 [66.1, 99.8] mmHg at 1.00 FiO2 (p < 0.001).

For a larger change in FiO2 (from 0.21 to 0.60), medullary oxygen tension values were similar (25.3 [15.3, 30.5] mmHg vs. 33.4 [22.6, 45.0] mmHg, p = 0.2224) but urinary oxygen tension values increased (25.5 [21.6, 32.6] mmHg vs. 59.4 [36.5, 66.0] mmHg, p = 0.001).

Discussion

Key findings

We conducted an observational study in human septic patients to determine whether PuO2 is affected by changes in systemic oxygenation during routine care of patients with septic shock. As expected, changes of FiO2 resulted in significant changes in PaO2. However, we found no significant differences between PuO2 measured before and after the interventions occurred. We supported our clinical findings with data from an experimental sheep experiment showing that medullary and urinary oxygen tension measurements did not differ within a similar range of FiO2 variation.

Relationship to previous studies

To our knowledge, there have been no previous investigations of the relationship between systemic and urinary oxygenation in human patients with septic shock. Ngo et al. addressed this issue in a group of patients undergoing cardiac surgery, finding no significant relationship between PaO2 and PuO2 during cardiopulmonary bypass [6]. Importantly, however, these observations were obtained in a unique physiological state with non-pulsatile flow, an extracorporeal circuit and mild hypothermia. Our current observations are likely more generally applicable to patients in a critical care setting. They are also very consistent with the outcomes of a retrospective analysis of three 3 studies involving a total of 28 adult Merino ewes during experimental sepsis [4, 12, 13], in which only a weak linear relationship was found between PaO2 and PuO2, accounting for ≤ 6% of the variation of PuO2 [6].

The absence of detectable changes in PuO2 in response to modest but clinically significant changes in FiO2 and thus PaO2 indicate that renal medullary tissue PO2 was not markedly affected by these clinical maneuvers. Experimental evidence supports the concept that extreme variations in FiO2 and/or PaO2 lead to corresponding changes in the oxygen tension of renal tissue. For example, in anesthetized rats a reduction in FiO2 from 1.0 to 0.1 resulted in a decline in cortical and medullary microvascular PO2 as assessed by dual-wavelength phosphorimetry [14]. Likewise, studies using fluorescence optodes in anesthetized rabbits demonstrated variations in both cortical and medullary tissue PO2 with variations in FiO2 [7, 15,16,17]. These findings provide support to our experimental findings that greater FiO2 variation is associated with greater PuO2 response, in particular at 0.60 and 1.00 FiO2.

Our ovine study sample derived from a larger experiment comprising 18 healthy sheep undergoing abdominal surgery under total intravenous or volatile anesthesia. In this study, increasing FiO2 from 0.21 to 1.00 increased cortical and medullary tissue PO2 [18]. However, it is also well-established that renal medullary tissue PO2 is less responsive to changes in PaO2 than the renal cortical tissue PO2 [16, 18, 19]. This appears to be a consequence of counter-current diffusive shunting of oxygen between descending and ascending vasa recta, which acts to reduce delivery of oxygen to renal medullary tissue [20]. Consequently, small but physiologically (and clinically) significant changes in FiO2 and/or PaO2 may not appreciably alter renal medullary tissue PO2. In support of this concept, no appreciable difference was observed in medullary tissue PO2 in our group’s preceding experiment when FiO2 was varied from 0.4 to 0.6 [18]. Similarly, in anesthetized rats outer medullary microvascular PO2 did not vary significantly when FiO2 was varied from 0.21 and 0.30 [21]. We cannot directly measure renal medullary tissue PO2 in patients and can only draw indirect inferences from measurement of PuO2 and consideration of available experimental evidence. However, the most parsimonious interpretation of our current findings is that modest changes in FiO2 and thus PaO2 neither markedly alter renal medullary tissue PO2 in patients with sepsis nor confounded the relationship between medullary tissue PO2 and PuO2.

Study implications

Our findings suggest that commonly performed adjustments to FiO2 settings in patients with sepsis do not result in significant changes in PuO2. In consonance of these findings, we observed that variations of FiO2 between 0.21 and 0.40 did not alter either medullary or urinary oxygen tension measurements in a sheep experiment. Thus, variations of systemic oxygenation seem unlikely to confound or affect the utility of urinary oxygenation as a biomarker for risk of AKI. Nevertheless, at higher FiO2 (0.60 and 1.00), significantly increased PuO2 values were obtained. One possible explanation is that in our septic patients, the FiO2 gap was far smaller in comparison to the experimental study. Also, one could argue that a type 2 error was present in the observational study which may have been controlled for during the experimental protocol.

Additional investigation is needed to explore whether the lack of PuO2 variation in face of PaO2 changes derives from the presence of confounding factors affecting medullary oxygen values. Also, further studies in critically ill patients are needed to elucidate whether sustained differences in oxygen exposure [22, 23] influence renal related outcomes. Thus, changes in PaO2, as a consequence of altered FiO2 in routine care of patients with septic shock, is unlikely to be a major confounder of the relationship between renal medullary tissue PO2 and PuO2. In the current study, PaO2 was used as a measure of systemic oxygenation because it reflects the balance between oxygen delivery and consumption. Had SpO2 been used, the accuracy would have been affected by peripheral tissue perfusion, use of vasoactive agents and altered cardiac output. Finally, continuous measurement of PuO2 might be useful for monitoring the impact on renal medullary oxygenation.

Strengths and limitations

We evaluated systemic and urinary oxygenation in human septic patients and assisted our proposition with experimental data. Our findings are consistent with previous observations in sepsis [6] and provide additional evidence that the relationship between renal medullary tissue PO2 and PuO2 is unlikely to be confounded by changes in FiO2 or PaO2 in the range commonly encountered in the ICU. As such, they provide further support for the use of PuO2 as a clinical surrogate of renal medullary PO2.

Our study has several limitations. First, the clinical component was an observation designed to assess the effects on PuO2 where the observed intervention (change of FiO2) was not protocolized. Moreover, controlling for variables such as creatinine or urine output was not feasible due to technical limitations and the limited number of patients. However, we aimed to undertake an exploratory analysis to generate a preliminary hypothesis to guide advanced studies. Moreover, we added data from a sheep experiment where FiO2 variation was protocolized. Also, due to the lower number of measurements in the experimental study, greater heterogeneity was observed. Second, the inclusion of septic patients in our clinical study did not occur in the early stage of resuscitation. On the other hand, the instances of FiO2 change we captured took place in a stable state with lower propensity for PuO2 to be affected by additional confounding effects of interventions intended to optimize oxygen delivery to the tissues. Furthermore, changes in FiO2 performed under stable conditions might have reduced the likelihood of reverse causation or provided mitigation of any potential effect of other interventions. Third, the observational nature of the study may have led to confounding by indication. For instance, the reasons motivating the clinician to change FiO2 settings could have affected the relationship between FiO2 and PuO2. However, a larger degree of FiO2 change would be expected if optimization measures capable of affecting such relationship were in place. Our patients were enrolled in the stabilization phase of sepsis, a time when, in general, only limited interventions are performed to achieve physiologic parameters aiming to prevent organ dysfunction. Finally, we addressed only the variation of systemic oxygenation within the normoxemic range. However, such a normoxemic range is typical in the care of patients in the ICU.

Conclusions

Changes in FiO2 and PaO2 within the context of usual care did not appreciably affect PuO2. Our findings suggest that, within the values reported, PuO2 measured in a clinical and experimental setting is not confounded by changes in inspired oxygen fraction or arterial oxygen tension and that PuO2 can be used as biomarker of medullary oxygenation irrespective of FiO2.