1 INTRODUCTION

Acute kidney injury (AKI) is a common and serious complication of cardiac surgery requiring cardiopulmonary bypass (CPB).¹ There is evolving evidence that current standard perfusion practice during CPB does not adequately protect the kidney,^2-4 particularly the renal medulla,^{5, 6} from hypoxia, thus increasing the risk of cardiac surgery-associated AKI. Experimental models of CPB in rats,⁷ pigs,⁸ and sheep^{5, 6} are all associated with renal medullary tissue hypoxia. Computational models of renal oxygenation also predict medullary hypoxia during CPB.^9-12 Furthermore, risk of AKI is associated with indices of intra-operative renal hypoxia, as assessed by urinary oxygen tension (PO₂)^13-15 or near infrared spectroscopy.^{16, 17} Thus, strategies aimed at improving renal oxygenation during CPB have the potential to mitigate post-operative AKI.

The clinical perfusionist could improve renal oxygen delivery (DO₂) during CPB by manipulating pump flow and/or blood haemoglobin concentration [Hb]. Findings from the Goal-Directed Perfusion Trial (GIFT) indicate that avoidance of a systemic DO₂ of less than 280 mL min⁻¹ m⁻² during CPB can mitigate the risk of AKI.¹⁸ The primary intervention in this trial was to increase pump flow. However, if this was not sufficient to achieve a systemic DO₂ of at least 280 mL min⁻¹ m⁻², or if the mixed venous oxygen saturation was <68% or systemic oxygen extraction was >40%, a unit of whole blood was given. Thus, the intervention was complex, involving variable combinations of increased pump flow and increased blood [Hb]. In an ovine model of CPB, we recently demonstrated that increasing pump flow or arterial pressure within a clinically achievable range improved renal medullary tissue oxygenation.⁵ However, the effects of varying blood [Hb] in this model remain to be determined. Therefore, in the current study, we tested the hypothesis that renal haemodynamics and oxygenation during experimental CPB are dependent on blood [Hb]. Using a cross-over design, blood [Hb] was varied between ~7 g dL⁻¹ and ~9 g dL⁻¹ during experimental CPB in sheep.

2 RESULTS

2.1 Variables in conscious sheep

In conscious sheep, all variables were similar to those we have observed previously and were within the normal physiological range (Figures 1-3, Tables 1-4).^{5, 6}

Table 1. Systemic variables

Variable	n	1. Conscious	2. Anaesthetized	P1v2	On bypass
					n	3. Low Hb	P2v3	4. High Hb	P2v4	P3v4
Body temperature (°C)	10	39.5 ± 0.4	39.4 ± 0.4	.63	10	35.9 ± 0.3	<.001	35.8 ± 0.4	<.001	.99
Central venous pressure (mmHg)	9	−0.2 ± 1.7	1.8 ± 2.5	<.001	9	2.3 ± 2.7	.02	2.3 ± 2.5	.07	>.99
Systemic vascular conductance (mL kg−1 mmHg−1)					10	1.02 ± 0.14		1.05 ± 0.14		.12
Systemic oxygen delivery (mL O2 kg−1 min−1)					10	7.45 ± 1.23		9.61 ± 2.32		.02
Systemic oxygen consumption (mL O2 kg−1 min−1)					10	2.48 ± 0.38		2.50 ± 0.53		>.99

Note

• P values ≤ 0.05 are shown in bold. Data are expressed as mean ± standard deviation. Systemic vascular conductance is pump flow divided by the product of body weight and mean arterial pressure. Systemic oxygen delivery is the product of pump flow and arterial blood oxygen content. Systemic oxygen consumption is the product of pump flow and the difference in oxygen content of arterial and mixed venous blood. P values are the outcomes of Student's paired t test, with a Dunn-Sidak correction to account for the four planned comparisons in the analysis.

Table 2. Renal haemodynamics, oxygenation and function

Variable	n	1. Conscious	2. Anaesthetized	P1v2	On bypass
					n	3. Low Hb	P2v3	4. High Hb	P2v4	P3v4
Renal vascular conductance (µL kg−1 min−1 mmHg−1)	10	70.8 ± 18.8	39.7 ± 13.2	<.001	10	28.7 ± 10.4	.08	27.4 ± 12.3	.03	.93
Cortical perfusion (units)	10	2250 ± 1398	1672 ± 958	.80	10	828 ± 610	.15	710 ± 313	.05	.90
Medullary perfusion (units)	10	674 ± 566	644 ± 447	>.99	10	272 ± 288	.25	368 ± 386	.36	.73
Urine flow (µL kg−1 min−1)	10	15.4 ± 7.2	18.7 ± 13.1	.97	10	41.6 ± 16.7	.02	18.9 ± 11.0	>.99	<.001
Sodium excretion (µmol kg−1 min−1)	10	1.82 ± 1.66	2.54 ± 3.39	.96	10	4.34 ± 3.07	.63	1.43 ± 1.66	.79	.003
CrCl (mL kg−1 min−1)	10	2.49 ± 1.05	1.88 ± 1.08	.70	10	1.12 ± 0.49	.25	0.69 ± 0.40	.01	.05
TNa+ (µmol kg−1 min−1)	10	350 ± 169	271 ± 155	.76	10	152 ± 70	.15	98 ± 60	.008	.08
Fraction Excretion of Sodium (%)	10	0.55 ± 0.44	0.79 ± 0.59	.74	10	2.62 ± 1.30	.01	1.27 ± 0.78	.40	.01
TNa+/renal VO2 (mol/mol)	8	80.6 ± 34.4	87.0 ± 46.8	>.99	7	83.7 ± 42.5	>.99	42.2 ± 23.1	.18	.13

Note

• P values ≤ 0.05 are shown in bold. Data are expressed as mean ± standard deviation. Renal vascular conductance is renal blood flow divided by the product of body weight and mean arterial pressure. Creatinine clearance (CrCl) is the product of urine flow and urinary creatinine concentration divided by plasma creatinine concentration. Sodium reabsorption (T_Na+) is the product of CrCl and plasma sodium concentration minus sodium excretion. P values are the outcomes of Student's paired t test, with a Dunn-Sidak correction to account for the four planned comparisons in the analysis.

Table 3. Arterial blood oximetry and chemistry

Variable	n	1. Conscious	2. Anaesthetized	P1v2	On bypass
					n	3. Low Hb	P2v3	4. High Hb	P2v4	P3v4
PO2 (mmHg)	10	97.3 ± 9.2	276.0 ± 24.8	<.001	10	307.9 ± 31.6	.08	317.1 ± 26.7	.05	.85
SO2 (%)	10	97.1 ± 1.7	99.5 ± 0.5	<.001	10	99.3 ± 0.5	.18	99.1 ± 0.4	.09	.12
Hb (g/dL)	10	11.3 ± 1.0	8.8 ± 1.2	<.001	10	6.9 ± 0.6	<.001	9.0 ± 1.5	>.99	.01
Oxygen content (mL O2/dL)	10	15.4 ± 1.2	12.8 ± 1.7	.002	10	10.4 ± 0.8	<.001	13.3 ± 0.7	>.99	.01
PCO2 (mmHg)	10	35.3 ± 3.1	29.2 ± 3.5	.002	10	45.4 ± 6.1	<.001	48.9 ± 7.1	<.001	.11
pH	10	7.52 ± 0.02	7.61 ± 0.04	<.001	10	7.43 ± 0.06	<.001	7.40 ± 0.07	<.001	.02
Lactate (mmol/L)	10	0.67 ± 0.39	1.01 ± 0.46	.24	10	2.27 ± 1.12	.04	2.41 ± 0.92	.02	>.99
Sodium (mmol/L)	10	137.6 ± 3.5	137.6 ± 3.5	>.99	10	139.8 ± 2.8	.54	141.0 ± 2.5	.14	.23
Potassium (mmol/L)	10	3.86 ± 0.30	3.53 ± 0.32	.01	10	3.09 ± 0.30	.001	2.99 ± 0.36	.002	.81
Chloride (mmol/L)	10	102.8 ± 4.0	103.8 ± 4.5	.86	10	101.5 ± 3.3	.29	102.6 ± 2.4	.76	.24
Calcium (mmol/L)	10	1.07 ± 0.10	1.03 ± 0.08	.77	10	0.99 ± 0.09	.84	1.05 ± 0.19	.99	.46
Bicarbonate (mmol/L)	10	28.7 ± 3.2	29.3 ± 2.5	.85	10	29.6 ± 1.7	>.99	29.5 ± 2.710	>.99	>.99
Base-excess (mmol/L)	10	5.84 ± 2.76	7.53 ± 2.10	.08	10	5.20 ± 2.16	.04	4.54 ± 3.21	.18	.88

Note

• P values ≤ 0.05 are shown in bold. Data are expressed as mean (standard deviation). Blood oxygen content was calculated as (0.0139 × [Hb] × SO₂) + (0.003 × PO₂). P values are the outcomes of Student's paired t test, with a Dunn-Sidak correction to account for the four planned comparisons in the analysis.
• Abbreviations: Hb, haemoglobin, PCO₂, partial pressure of carbon dioxide, PO₂, oxygen tension; SO₂, saturation of haemoglobin with oxygen.

Table 4. Venous blood oximetry and chemistry

Variable	n	1. Conscious	2. Anaesthetized	P1v2	On bypass				P2v4	P3v4
					n	3. Low Hb	P2v3	4. High Hb
Mixed venous blood
PO2 (mmHg)	10	39.6 ± 5.9	51.4 ± 11.1	.04	10	46.9 ± 5.6	.52	56.1 ± 12.8	.77	.07
SO2 (%)	10	64.5 ± 8.7	83.6 ± 8.1	.003	10	70.1 ± 9.9	.02	75.5 ± 11.7	.17	.23
P50 (mmHg)	10	31.4 ± 4.0	27.1 ± 4.8	.02	10	33.8 ± 5.8	.01	35.1 ± 6.0	.004	.25
Hb (g/dL)	10	11.1 ± 1.0	8.6 ± 1.3	<.001	10	6.8 ± 0.6	<.001	8.9 ± 1.6	.99	.01
Oxygen content (mL O2/dL)	10	10.2 ± 1.9	10.2 ± 1.6	>.99	10	6.8 ± 1.2	<.001	9.7 ± 2.7	.97	.03
PCO2 (mmHg)	10	40.5 ± 3.6	32.4 ± 3.6	<.001	10	52.4 ± 6.7	<.001	55.6 ± 8.5	<.001	.06
pH	10	7.46 ± 0.02	7.61 ± 0.13	.02	10	7.38 ± 0.06	.002	7.36 ± 0.07	.002	.36
Lactate (mM)	10	0.89 ± 0.42	1.10 ± 0.44	.67	10	2.35 ± 1.22	.06	2.54 ± 1.09	.02	.99
Renal venous blood
PO2 (mmHg)	8	57.5 ± 5.6	63.4 ± 7.2	.25	7	54.8 ± 8.0	.38	64.7 ± 12.0	>.99	.006
SO2 (%)	8	88.7 ± 2.3	90.9 ± 3.5	.22	7	80.8 ± 7.0	.10	83.3 ± 7.3	.25	.10
Hb (g/dL)	8	11.0 ± 1.2	8.5 ± 1.2	.006	7	6.8 ± 0.7	.01	9.1 ± 1.6	.96	.05
Oxygen content (mL O2/dL)	8	13.9 ± 1.3	11.0 ± 1.8	.006	7	7.8 ± 0.5	.01	10.7 ± 1.9	>.99	.03
PCO2 (mmHg)	8	36.2 ± 2.2	32.1 ± 3.9	.02	7	50.3 ± 5.3	.004	53.8 ± 8.8	.008	.23
pH	8	7.51 ± 0.03	7.58 ± 0.02	<.001	7	7.41 ± 0.06	.002	7.38 ± 0.08	.002	.19
Lactate (mM)	8	0.69 ± 0.44	0.96 ± 0.44	.47	7	2.36 ± 1.33	.16	2.14 ± 0.87	.10	.98

Note

• P values ≤ 0.05 are shown in bold. Data are expressed as mean (standard deviation). P₅₀ is the partial pressure of oxygen at which haemoglobin is 50% saturated with oxygen. It was estimated from PO₂ and SO₂, in mixed venous blood, by the method of Doyle.⁵⁰ Blood oxygen content was calculated as (0.0139 × [Hb] × SO₂) + (0.003 × PO₂). P values are the outcomes of Student's paired t test, with a Dunn-Sidak correction to account for the four planned comparisons in the analysis.
• Abbreviations: Hb, haemoglobin, PCO₂, partial pressure of carbon dioxide, PO₂, oxygen tension; SO₂, saturation of haemoglobin with oxygen.

3 DISCUSSION

In a clinically relevant ovine model of CPB, we were unable to detect changes in renal cortical or medullary tissue PO₂ as blood [Hb] was varied across a clinically relevant range, from an average of ~7 g dL⁻¹ to ~9 g dL⁻¹. This was despite the fact that increasing blood [Hb] increased both arterial blood oxygen content by 20% and systemic DO₂ by 29% and reduced renal fractional oxygen extraction by 17%. Thus, our findings do not support the proposition that blood transfusion during CPB is an effective intervention to improve renal tissue oxygenation. In contrast, renal medullary hypoxia in this model can be mitigated by increasing renal DO₂ through either increased pump flow, even in the absence of an associated increase in MAP, or by increasing MAP with the vasopressor, metaraminol.⁵ Thus, based on our findings using this model, manipulation of pump flow and arterial pressure appear to be more effective approaches to improve renal oxygenation during CPB than blood transfusion.

Pre-operative anaemia increases the risk of cardiac surgery associated AKI.¹⁹ This may at least partly be mediated by renal hypoxia, since experimental haemodilution can lead to renal tissue hypoxia,^20-23 including during experimental CPB in the rat.⁷ However, for the most part, renal hypoxia in experimental models has been seen with relatively severe anaemia, resulting in reductions in arterial blood oxygen content in the order of 50%. A more clinically relevant question is whether renal tissue PO₂ can be increased by blood transfusion, to increase blood [Hb] from a level which might trigger blood transfusion in the clinical setting during CPB (7-8 g dL⁻¹), to a clinically achievable target of 9-10 g dL⁻¹. Baseline blood [Hb] is slightly less in sheep than humans,^{24, 25} so we targeted blood [Hb] at the lower end of this range. We also applied a cross-over design rather than giving blood transfusions to all sheep after a period of haemodilution, to avoid potential time-dependent confounding. Our findings do not support the proposition that blood transfusion can improve renal tissue oxygenation during CPB. Thus, with regard to the GIFT trial,¹⁸ the critical component of the intervention may be increased pump flow rather than increased blood oxygen carrying capacity.

Our findings are also relevant to interpretation of the recent multicenter Transfusion Requirements in Cardiac Surgery III (TRICS III) Trial, which compared a restrictive (7.5 g dL⁻¹) versus a liberal (9.5 g dL⁻¹) threshold for blood transfusion during cardiac surgery.²⁶ TRICS III provided compelling evidence for lack of inferiority of a restrictive threshold compared with a liberal threshold, for the primary composite endpoint of death from any cause, myocardial infarction, stroke, or new onset renal failure with dialysis by hospital discharge or by day 28.²⁷ A similar conclusion was drawn after follow-up of these patients 6 months after their operation.²⁸ Importantly, in a pre-specified sub-study of TRICS III,²⁹ no difference in the incidence of post-operative AKI was found between the restrictive and liberal thresholds.³⁰ It is plausible that the lack of a reduction in the incidence of AKI when a more liberal transfusion threshold was used is at least partly attributable to the absence of a clinically significant impact of intra-operative blood transfusion on intraoperative renal medullary tissue hypoxia.

The failure of increased [Hb] to improve renal tissue oxygenation cannot be attributed to the potentially confounding effects of storage of donor blood, which increases the affinity of haemoglobin for oxygen.³¹ This effect can take between 5 days³² and two weeks³³ to reach an equilibrium, but can be appreciable within the first 24 hours of storage.³² It is due in part to loss of erythrocyte cytoplasmic 2,3-diphosphoglycerate (DPG).³¹ Additional factors may also contribute, since even autologous salvaged blood, which is returned to the patient within a few hours using modern cell-saving techniques, had significantly greater haemoglobin affinity for oxygen than fresh blood, despite no significant difference in DPG concentration.³⁴ However, calculated mixed venous blood P₅₀ was not less at high [Hb] than low [Hb], indicating no appreciable effect of the donor blood on the affinity of haemoglobin for oxygen in vivo. This may reflect the relatively short period of cold storage of donor blood, which was 18 hours or less in all cases. Thus, it remains possible that longer storage of donor blood, as may occur in a clinical setting,³⁵ could hinder tissue oxygenation.

The failure of increased [Hb] to improve renal tissue oxygenation is also unlikely to be due to increased local tissue oxygen consumption, since renal VO₂ did not increase with increased blood [Hb]. Indeed, creatinine clearance was significantly less, and total sodium reabsorption tended to be less (P = .08), at high blood [Hb] than low blood [Hb]. But the impact of this on renal VO₂ may have been offset by reduced efficiency of oxygen utilization for sodium reabsorption at higher blood [Hb], since there was at least a tendency for the ratio of sodium reabsorption to renal VO₂ to be less (by 41 ± 29 mol/mol). This apparent effect was statistically significant without application of the Dunn-Sidak correction (P = .03) but not after the correction was applied (P = .13). Thus, our failure to detect a difference may represent a type 2 error.

Our results are consistent with the proposition that increasing blood [Hb] increases oxygenation of blood in the kidney. Renal fractional oxygen extraction was less (and renal venous blood PO₂ was greater) at high than low [Hb], indicating improvement in the oxygen supply/demand relationship in the kidney. But this effect did not appear to lead to better tissue oxygenation. The failure of renal tissue PO₂ to increase with increased blood [Hb], even though renal venous PO₂ increased by ~10 mmHg, likely reflects reduced efficiency of oxygen extraction from blood to tissue. A similar phenomenon, of dissociation between changes in renal venous blood PO₂ and renal tissue PO₂, has been described in anaesthetized rabbits in response to changes in renal vascular resistance.^36-38 The mechanisms underlying these phenomena remain unknown. However, arterial-to-venous (and in the medulla descending to ascending vasa recta) oxygen shunting^{12, 39} and increased heterogeneity of tissue perfusion⁴⁰ and thus oxygenation⁴¹ could potentially lead to less efficient oxygen transport to tissue. Regardless, our current findings show that, during experimental CPB in sheep, renal venous blood PO₂ can be increased by increasing blood [Hb] without appreciable changes in cortical or medullary tissue PO₂.

CPB in humans is usually associated with the onset of a brisk diuresis.¹⁵ In our current experiment in sheep, low blood [Hb] during CPB was associated with approximately two-fold increases in creatinine clearance, fractional excretion of sodium, and urine flow, and a three-fold increase in sodium excretion. This phenomenon can likely be partly explained by the effects of reduced plasma oncotic pressure⁴² caused by haemodilution with crystalloid. A similar mechanism likely contributes to diuresis during CPB in humans, although other mechanisms, including the ability of mild hypothermia to inhibit renal metabolism,⁹ and of cold exposure to inhibit secretion of arginine vasopressin,⁴³ may contribute.

The haemodilution we observed under anaesthesia could be attributable to two potential mechanisms. Firstly, urine flow under anaesthesia (18.7 ± 13.1 µL kg⁻¹min⁻¹) was only about half the infusion rate of compound sodium lactate (33 µL kg⁻¹ min⁻¹), so an increase in whole body water after induction of anaesthesia is likely (~1 mL per kg of body weight over an hour). However, we believe this is probably only a minor effect since the blood sampling during the experimental period under anaesthesia occurred within 30 minutes of induction of anaesthesia and thus the commencement of the infusion. The other potential mechanism is auto-transfusion. In our current study MAP under isoflurane anaesthesia was similar to that in the conscious state. But, based on our previous observations with isoflurane anaesthesia,^{44, 45} cardiac output is less and so total peripheral resistance must be greater in the present study to sustain MAP. We would therefore expect capillary hydrostatic pressure to be less under general anaesthesia than in the conscious state, leading to bulk flow of extracellular fluid into the vascular compartment. The level of haemodilution appears to be similar with isoflurane anaesthesia⁵ and propofol/fentanyl anaesthesia,⁶ although this comparison is confounded by the fact that propofol anaesthesia is associated with an obligatory fluid load due to the vehicle for the intravenous anaesthetic administration.

Strengths of our current study include the use of a clinically relevant large animal model of CPB. We targeted blood [Hb] relevant to recent clinical trials of blood transfusion protocols (TRICS III)³⁰ and goal directed perfusion (GIFT)¹⁸ in cardiac surgery which have included the outcome of AKI. Nevertheless, the fact that we studied young and healthy sheep does limit our ability to generalize our findings to the population of patients who undergo cardiac surgery, who are often relatively old and suffer from multiple co-morbidities. We also cannot completely exclude the possibility that our failure to detect effects of altered blood [Hb] on cortical and medullary tissue PO₂ represents a type II error. But post-hoc analysis of the current data indicate that our study was adequately powered. The standard deviations of the differences in cortical and medullary PO₂ between the first and second 30 minutes period at each level of haemoglobin, reflecting within-subject variability, were both ~8 mmHg. Based on this, a sample size of 10 sheep would permit detection of a 12 mmHg change in cortical or medullary PO₂, even after application of the Dunn-Sidak correction to account for multiple comparisons. By comparison, in a previous study we found that increasing pump flow from 80 to 100 mL kg⁻¹ min⁻¹ increased medullary PO₂ by 9.5 mmHg, and increasing MAP with low dose metaraminol, from ~63 to ~90 mmHg, increased medullary PO₂ by 20.2 mmHg.

In conclusion, our findings provide evidence that blood transfusion, to increase blood [Hb] from ~7 g dL⁻¹ to ~9 g dL⁻¹, is an ineffective intervention to increase renal tissue oxygenation during experimental CPB in sheep. Our findings cannot be explained by the impact of donor blood on haemoglobin-oxygen affinity in vivo. They provide a potential explanation for the lack of efficacy of a liberal transfusion strategy to reduce the incidence of AKI in the TRICS III trial.³⁰ They also provide insight into the potential mechanisms underlying the success of goal directed perfusion, as demonstrated by the GIFT trial; namely that increasing pump flow and/or arterial pressure during CPB, rather than blood transfusion to increase blood oxygen carrying capacity, best protects the kidney from tissue hypoxia, thus mitigating the risk of post-operative AKI.

4 MATERIALS AND METHODS

CONFLICT OF INTEREST

None. The results presented in this paper have not been published previously in whole or part, except in abstract format.

AUTHORS’ CONTRIBUTIONS

CNM, YRL, RGE, ADC and RB contributed to study concept and design. YRL, SGH, ADC, CNM, BM, PRM, RGE and NO contributed to acquisition, analysis and interpretation of data. RGE contributed to drafting of the manuscript. YRL, CNM, RB, ADC, BM, PRM, SGH and NO contributed to critical revision of the manuscript for intellectual content. RGE, CNM, ADC and RB contributed to obtaining funding. RGE contributed to statistical analysis.

Funding information

This work was supported by the National Health and Medical Research Council of Australia (GNT1122455; GNT1185777), the Victorian Government Operational Infrastructure Support Grant, and the National Heart Foundation of Australia (101853).