SFW 118
SFW 118
SFW 118
3, 348356
doi: 10.1093/ckj/sfw118
Advance Access Publication Date: 12 January 2017
Original Article
ORIGINAL ARTICLE
Abstract
Background: The role of goal-directed therapy (GDT) in preventing creatinine rise following noncardiac surgery is unclear.
We performed a post-hoc analysis of a randomized controlled trial to assess the relationship between postoperative optimi-
zation of oxygen delivery and development of acute kidney injury (AKI)/creatinine rise following noncardiac surgery.
Methods: Patients were randomly assigned immediately postoperatively to receive either fluid and/or dobutamine therapy
to maintain/restore their preoperative oxygen delivery, or protocolized standard care (oxygen delivery only recorded).
Primary end point was serial changes in postoperative creatinine within 48 h postoperatively. Secondary outcomes were
development of AKI (KDIGO criteria) and minimal creatinine rise (MCR; no decline from preoperative creatinine), related to
all-cause morbidity and length of stay.
Results: Postoperative reductions in serum creatinine were similar (P 0.76) in patients randomized to GDT [10 mmol/L (95%
confidence interval, CI: 17 to 1); n 95] or protocolized care [8 mmol/L (95% CI: 17 to 6); n 92]. Postoperative haemody-
namic management was not associated with the development of MCR [78/187 (41.7%)] or AKI [13/187; (7.0%)]. Intraoperative
requirement for norepinephrine was more likely in patients who developed postoperative rises in creatinine [relative risk
(RR): 1.66 (95% CI: 1.042.67); P 0.04], despite similar volumes of intraoperative fluid being administered. Persistently higher
lactate during the intervention period was associated with AKI (mean difference: 1.15 mmol/L (95% CI: 0.481.81); P 0.01].
Prolonged hospital stay was associated with AKI but not MCR [RR: 2.71 (95% CI: 1.514.87); P 0.0008].
Conclusion: These data provide further insights into how perioperative haemodynamic alterations relate to postoperative
increases in creatinine once systemic inflammation is established.
Key words: acute kidney injury, cardiac output, noncardiac surgery, oxygen delivery
348
Acute kidney injury and oxygen delivery | 349
informed if haemodynamic management during the study target achieved, gelatin dose, systemic inflammation as
intervention period was involved directly. Postoperative reflected by CRP). Continuous variables are presented as mean
management adhered to enhanced recovery hospital protocols. (standard deviation) or median (quartiles), depending on nor-
Antibiotic use beyond prophylactic administration (i.e. after mality of distribution. Categorical variables are presented as n
24 h post-operatively) was a deviation from normal postopera- (%). Analyses were performed using NCSS 8 (Kaysville, UT, USA).
tive care. Significance was set at P 0.05 (two-tailed).
Outcomes Results
The primary outcome was creatinine rise within 48 h postopera- Patient population
tively stratified by postoperative haemodynamic intervention
arm. Creatinine rise refers to any creatinine value during the The study was conducted from 20 May 2010 until 12 February
first 48 h postoperatively that exceeded the preoperative value 2014, with 204 patients randomly assigned to receive postopera-
for an individual patient. Creatinine fall refers to creatinine val- tively either haemodynamic therapy designed to restore/main-
ues that were persistently lower throughout the first 48 h post- tain their individualized preoperative oxygen delivery (n 95) or
operatively, compared with preoperative values for each protocolized care (n 92; Figure 1). As detailed previously, 17
individual patient. Secondary end points were development of patients did not undergo the postoperative trial intervention as
AKI, MCR, haemodynamic characteristics, time to become mor- they developed exclusion criteria during the intraoperative
bidity free and hospital stay associated with AKI. We diagnosed period. Thus, we analysed data for 187 patients, with no further
AKI within the first 48 h postoperatively using changes in creati- loss to follow up. Demographic characteristics were similar
nine and/or oliguria, according to KDIGO criteria. MCR was between groups (Table 1).
defined similarly to other studies, whereby creatinine values
during the first 48 h postoperatively exceeded the preoperative Primary endpoint
value for an individual patient, but were below KDIGO-defined
Postoperative reductions in serum creatinine were similar
criteria for a significant creatinine rise. We calculated the Acute
(P 0.76) in patients randomized to GDT [10 mmol L1 (95% CI: 17
Physiology and Chronic Health Evaluation II (APACHE II) score
to 1)] and protocolized care [8 mmol/L (95% CI: 17 to 6); Figure
for each patient, which indicates the severity of acute physio-
2]. Postoperative haemodynamic management was not associ-
logical derangement following surgery. Lactate and C-reactive
ated with the development of MCR [78/187 (41.7%)] or AKI [13/
protein (CRP) were also measured as markers for tissue perfu-
187 (7.0%)]. Similar proportions of patients who developed MCR
sion and systemic inflammation, respectively.
were present in each haemodynamic therapy group (P 0.92;
Table 1). Thirteen patients developed AKI, of whom nine sus-
Statistical analyses tained stage 1 AKI. We found no association between AKI [RR:
0.74 (95% CI: 0.202.75); P 0.65] or MCR [RR: 1.43 (95% CI: 0.78
The primary objective of this trial was to establish whether indi-
2.63); P 0.24] and the trial-protocol-defined use of dobutamine
vidualized oxygen-delivery therapy reduced postoperative mor-
postoperatively.
bidity within the first 48 h postoperatively. On the basis of
previous studies, 60% of patients on postoperative day 2 sus-
tain significant morbidity [15, 16]. The sample size calculation Secondary clinical endpoints
(STATA version 10) has been detailed previously; briefly, at a Achievement of preoperative oxygen delivery was associated
statistical significance level of 5%, with a power of 80%, we esti- with a lower incidence of AKI [RR: 1.91 (95% CI: 1.183.09);
mated that at least 102 patients per treatment group would be P 0.03], but not MCR [RR: 1.33 (95% CI: 0.862.05); P 0.21],
required (allowing for 15% of patients to achieve their oxygen regardless of postoperative haemodynamic management. Six
delivery target regardless of intervention, and taking into out of 95 patients developed AKI in the haemodynamic therapy
account a further 20% dropout rate as a result of failure to group whereas 7/92 patients sustained AKI in the control group
adhere to the protocol and/or intraoperative withdrawals). [RR: 1.21 (95% CI: 0.423.45); P 0.77]. Markers of systemic
Analyses were performed according to an a priori statistical inflammation were similar between groups (Supplementary
analysis plan including all patients on an intention to treat Figure S2).
basis, regardless of protocol compliance. Repeated-measures
ANOVA was undertaken to compare haemodynamic changes
between patients who developed AKI, or maintained normal Length of hospital stay
renal function, at preoperative and postoperative (intervention) Early AKI was associated with delay in time to become morbid-
timepoints. Results of primary and secondary outcomes are ity free [unadjusted hazard ratio: 1.76 (95% CI: 1.122.76);
reported as relative risks (RRs) with 95% confidence intervals P 0.02] and prolonged hospital stay [hazard ratio: 1.91 (95% CI:
(CIs). The KaplanMeier method was used to summarize time to 1.232.94); P 0.02; Figure 3]. Prolonged hospital stay was associ-
become morbidity free and length of hospital stay, with the log- ated with AKI but not MCR [RR: 1.23 (95% CI: 0.911.68); P 0.16].
rank test used to analyse significant differences in time to Early AKI was not related to operation type (P 0.99) or chronic
event. Multiple logistic regression analysis was performed to kidney disease [RR: 1.37 (95% CI: 0.394.87); P 0.65].
assess associations between the primary outcome (creatinine
rise within 48 h postoperatively), preoperative factors (age, gen-
der, body mass index, cardiac comorbidity, preoperative creati-
Haemodynamic endpoints
nine, type of surgery) and perioperative management Both AKI [RR: 2.41 (95% CI: 1.244.67); P 0.02] and MCR [RR: 1.94
(requirement for norepinephrine, packed red cell transfusion, (95% CI: 1.043.62); P 0.03] were associated with intraoperative
lactate at end of operation, randomization arm, oxygen delivery requirement for norepinephrine to maintain mean arterial
Acute kidney injury and oxygen delivery | 351
Fig. 1. Trial enrollment and analysis plan. RCT, randomized controlled trial.
Creatinine fall (n 65) Creatinine rise (n 27) Creatinine fall (n 70) Creatinine rise (n 25)
Age (years) 68 6 10 69 6 6 68 6 10 69 6 7
Male 38 (58%) 18 (67%) 41 (59%) 17 (68%)
BMI (kg m2) 275 6 54.8 281 6 5.5 276 6 5.8 274 6 4.5
Malignancy 45 (69%) 19 (70%) 47 (67%) 17 (68%)
CKD stage 3 10 (15%) 2 (7%) 9 (13%) 6 (24%)
Diabetes mellitus 14 (22%) 6 (22%) 15 (21%) 4 (16%)
Hypertension 30 (46%) 10 (37%) 36 (51%) 15 (60%)
Albumin 43 6 5 42 6 5 42 6 5 42 6 4
CVD 46 (71%) 16 (59%) 55 (79%) 21 (84%)
Surgical procedure
UGI 16 (25%) 5 (19%) 14 (20%) 8 (32%)
Liver/hepatobiliary 25 (38%) 14 (52%) 28 (40%) 8 (32%)
Colorectal 13 (20%) 3 (11%) 17 (24%) 5 (20%)
Vascular 11 (17%) 5 (19%) 8 (11%) 4 (16%)
Data presented as mean6 standard deviation or n (%). BMI, body mass index; UGI, upper gastrointestinal surgery; CKD, chronic kidney disease; CVD, cardiovascular dis-
ease (stroke/ischaemic heart disease/peripheral vascular disease).
pressure (MAP), despite similar volumes of intraoperative fluid oxygen delivery (P 0.89) were similar between patients who
being administered (Table 2). By the end of the intraoperative developed, or avoided, MCR or AKI (Supplementary Figure S3).
period, venous lactate (P 0.001) was higher in patients who The difference in lactate between patients who developed AKI
went on to develop AKI (Supplementary Figure S3). At the end of and those that did not persisted throughout the intervention
the postoperative period, cardiac output (P 0.66) and absolute period (P 0.009), despite similar cardiac output and oxygen
352 | A. Patel et al.
Fig. 2. Oxygen delivery stratified by postoperative creatinine rise and allocation to postoperative haemodynamic intervention. (A) Oxygen delivery, indexed and
expressed as % individualized preoperative value. Mean values (95% CI) shown; numbers per group indicated within figure. Failure to reach preoperative oxygen deliv-
ery was associated with postoperative creatinine rise. Asterisk denotes P 0.008, for comparison between mean oxygen delivery during intervention period (standar-
dized to each patients preoperative value), by ANOVA. Post-hoc analysis showed a mean difference in standardized oxygen delivery between GDT and GDT-creatinine
rise was 22% [(95% CI: 145); P 0.05].
Fig. 3. KaplanMeier plot showing relationship between development of early AKI and subsequent length of hospital stay.
delivery throughout the same time period. Multiple logistic therapy failed to improve renal function after major abdominal
regression analysis confirmed that failure to achieve preopera- surgery compared with normal clinical care [17]. It is notewor-
tive oxygen delivery, use of packed red cells and/or intraopera- thy that postoperative interventional trials following noncar-
tive requirement for norepinephrine were significantly diac surgery have seldom reported the impact of GDT on renal
associated with increased postoperative creatinine over base- morbidity specifically [18]. Furthermore, our study afforded
line values by postoperative day 2 (Table 3). detailed, serial haemodynamic insight in a randomized con-
trolled setting, hence adding to this literature by providing
detailed haemodynamic profiles on patients randomized to
Discussion control carea feature notably lacking in preceding studies as
This re-analysis of a prospective randomized controlled, highlighted by a preceding systemic review [8]. This analysis
blinded study failed to demonstrate a benefit of the postopera- was therefore principally undertaken to contribute to this nota-
tive GDT on the primary endpoint, postoperative increases in ble lack of haemodynamic data reported in control groups, as
plasma creatinine. A similar prospective randomized trial also highlighted by a preceding systemic review [8]. We acknowledge
reported that algorithm-guided goal-directed haemodynamic that the original study was underpowered to explore
Acute kidney injury and oxygen delivery | 353
Intraoperative
Duration of surgery (min) 2706105 273 6 125 256 6 98 275 6 117
General anaesthetic only 29 (45%) 14 (52%) 27 (39%) 7 (28%)
Intravenous fluid (mL/kg/h) 13.4 6 9.2 13.9 6 7.4 13.3 6 5.8 12.3 6 5.0
Haemoglobin preoperative 12.9 6 1.7 13.1 6 1.5 12.2 6 1.7 12.9 6 1.3
Haemoglobin postoperative 10.9 6 1.5 10.6 6 1.7 10.6 6 1.5 11.0 6 1.7
Packed red cells [n (%)] 9 (14%) 7 (26%) 18 (26%) 7 (28%)
Vasopressor infusion [n (%)] 13 (20%) 8 (29%) 10 (14%) 8 (32%)
Lactate at end of surgery 1.9 6 1.1 2.2 6 1.2 2.1 6 1.4 2.3 6 1.2
Intervention period
APACHE II score 15 6 5 17 6 6 15 6 6 16 6 5
Gelatin (mL/kg/h) 1.5 6 1.3 2.1 6 2.0 2.9 6 1.7 2.7 6 1.7
Blood transfusion [n (%)] 7 (11%) 4 (15%) 15 (21%) 7 (28%)
Dobutamine infusion [n (%)] 0 0 22 (31%) 13 (52%)
Data presented as mean 6 standard deviation, median (interquartile range) or n (%). Excludes patients randomized but who met exclusion criteria by the end of their
operation.
Table 3. Multiple logistic regression analysis assessing perioperative factors associated with the development of creatinine rise above preoper-
ative baseline values 48 h after surgery
OR, odds ratio; PRC, packed red cells; DO2I, oxygen delivery; EndOp, end of operation.
mechanisms of AKI alone, but rather as part of the spectrum of increases in postoperative creatinine below the threshold for
postoperative morbidity that commonly develops in higher-risk AKI, as defined by KDIGO, do not appear to associate with worse
surgical patients. outcomes. These data are in contrast to cardiac surgery, where
Our detailed physiological data confirm expert consensus minimal increases in creatinine postoperatively are associated
that even relatively short periods of intraoperative hypotension with excess morbidity and mortality [6, 23].
requiring vasopressors may contribute to perioperative AKI [19]. Packed red blood cell transfusion is also an established peri-
These data show that the development of lactataemia and operative risk factor for AKI, at least in cardiac surgery [19]. Each
requirement for pressor support (norepinephrine) precedes the unit of perioperative blood that is transfused in cardiac surgery
subsequent development of AKI. However, AKI was not pre- is independently associated with a 1020% increase in the risk
vented by either GDT or standardized care after the intraopera- of AKI [24]. We cannot exclude an additional effect of transfu-
tive development of lactataemia and requirement for pressor sion on the development of AKI, since patients who developed
support. These data provide detailed haemodynamic data in AKI were more likely to receive erythrocyte transfusion postop-
accordance with several studies suggesting that intraoperative eratively. We can also not discount the possibility that the use
episodes of haemodynamic instability, characterized by relative of gelatin solutions may increase the risk of bleeding and renal
hypotension and lower perfusion pressure requiring the intrao- failure, as highlighted by a recent systematic review [25].
perative use of norepinephrine, are pathologically implicated in Notably, AKI patients did not differ in pre- or postoperative hae-
the development of AKI [2022]. In addition, we report that moglobin levels, compared with those who did not sustain renal
354 | A. Patel et al.
injury. Experimental models show that haeme derived from hae- hyperlactataemia as a result of adrenergic stimulation is also likely
moglobin imparts nephrotoxicity to vulnerable kidneys [26], par- to make a major contribution [43].
ticularly in older subjects subjected to major systemic Strengths of this study included the blinded, protocolized
inflammation [27]. We speculate that a 1-hit2-hit model may delivery of postoperative haemodynamic care. Serial analysis of
influence the development of AKI in this patient population, changes in creatinine was pre-specified in the analysis plan. In
where intraoperative hypotension generated by systemic inflam- contrast to preceding studies, haemodynamic variables were
mation that requires pressor support is accompanied by the fur- also measured in the control group. Significant limitations
ther injurious insult of haeme and nitric oxide consumption [28]. include the (predictably) low number of patients who sustained
The use of norepinephrine in ICU patients with hypotension AKI. This reflects that the original power calculation was based
has generated much controversy [29], particularly in the context on all-cause early morbidity (on postoperative day 2) rather
of acute renal injury [30]. In our study, the only pressor infusion than specifically the incidence of AKI, which is predictably far
used intraoperatively was norepinephrine. During systemic lower. However, a substantial number of patients who sus-
inflammation, restoration of blood pressure with noradrenaline tained MCR, a clinically relevant readout which has never fea-
may have a nephroprotective effect [31]. Norepinephrine infu- tured in non-cardiac surgical studies previously. A further
sion in experimental acute endotoxaemia reverses systemic limitation is a lack of specific biomarkers for renal injury, which
hypotension and improves renal blood flow, independent of may provide earlier information that guides haemodynamic
perfusion pressure. Norepinephrine increases renal vasodila- management.
tion through increased systemic blood pressure leading to vaso-
dilatation as a result of decreased renal sympathetic tone
through the baroreceptor response [32]. It remains unclear Conclusions
whether patients with significant comorbidity under general
The GDT protocol following major non cardiac surgery employed
anaesthesia require more pressor support to mount such a
in this study doubled the achievement rate of attaining individu-
response. The hypothesis that a subset of patients have a loss,
alized preoperative oxygen delivery values (from 33% to 60%),
or lack of, renal auto-regulatory reserve is supported by detailed
but failed to alter the trajectory of postoperative renal injury.
serial renal and cardiac output measurements in cardiac surgi-
Nevertheless, achievement of preoperative oxygen delivery
cal patients. A disconnect was observed between dose-
appears crucial in order to avoid postoperative kidney injury.
dependent increases in cardiac index with norepinephrine, yet
a MAP threshold at which pressure-dependent renal perfusion,
filtration and oxygenation improved [30]. However, very high
Supplementary data
doses of norepinephrine have been used to reverse AKI through
renal vasoconstriction in healthy animal models [33, 34]. In sep- Supplementary data are available online at http://ckj.oxford
sis, the development of AKI was not associated with changes in journals.org.
renal blood flow, oxygen delivery or histological appearance,
despite the use of norepinephrine to maintain arterial blood
pressure [35]. These data suggesting strongly that other mecha- Authors contributions
nisms must contribute to septic AKI.
G.L.A. designed study. A.P. analysed data. J.R.P. designed and
Parasympathetic autonomic dysfunction offers an additional
analysed data. POM-O Study Investigators collected trial data.
plausible mechanism that may contribute to perioperative AKI. We
have previously reported that GDT in the same trial is associated
with reduced cardiac (vagal) parasympathetic activity, as revealed
by changes in heart rate variability and despite similar heart rates
Conflict of interest statement
between groups [13]. A similar observation in reduced cardiac None declared.
(vagal) parasympathetic activity has been described during stress
echocardiography [36, 37]. Furthermore, we have also shown in
separate cohorts of patients that impaired baroreflex dys- Acknowledgements
functionwhich is in part characterized by reduced parasympa- This work was supported by Academy of Medical Sciences/
thetic toneis also associated with excess morbidity [38, 39]. Health Foundation clinician scientist award (to G.L.A.); and
These autonomic changes may impact on renal dysfunction, since
the Comprehensive Biomedical Research Centre, University
activation of vagal efferent outflow in a murine preclinical model
College London Hospitals NHS Trust/University College
of renal ischaemia-reperfusion minimizes injury, mediated by an
London (to G.L.A.). This work was undertaken in part at
anti-inflammatory mechanism requiring nicotinic a7 splenocytes
UCLH/UCL, who received a proportion of funding from the
[40].
Department of Healths NIHR Biomedical Research Centre
The intraoperative development of relative hyperlactataemia
associated with a requirement for pressor support in AKI patients
funding scheme. POM-O (PostOperative Morbidity-Oxygen
is not likely to be explained by tissue hypoxia. We found that lac- delivery) Study Group contributors: Sadaf Iqbal, Laura
tate levels at the end of the operation in patients who subsequently Gallego Paredes, Andrew Toner, Craig Lyness, Phoebe
developed AKI did not correlate with oxygen delivery and persisted Bodger, Anna Reyes, John Whittle, Steven Cone, Shamir
despite targeted haemodynamic therapy. As in septic shock, the Karmali, Gareth Ackland, Rumana Omar, Mervyn Singer,
presence of hyperlactataemia following resuscitation does not nec- Ana Gutierrez del Arroyo, Robert Stephens, Mark Hamilton,
essarily indicate oxygen debt but rather a metabolic change com- Susan Mallett, Massimo Malago, Charles Imber, Alastair
patible with elevated aerobic glycolysis [41]. Endotoxaemia, a likely Windsor, Alec Engledow, Robert Hinchcliffe, Muntzer
pathologic mediator in high risk surgery [42], is a potent driver of Mughal, Khalid Dawas, Tim Mould, Maurizio Cecconi,
increased lactate production that may stimulate aerobic glycolysis Nicholas Jenkins, Kirsty Everingham, Rupert Pearse, Martin
through stimulation of Na K ATPase activity [41]. Stress Lees, Robert Shulman and Majid Hashemi.
Acute kidney injury and oxygen delivery | 355
37. Sharma R, ODriscoll JM, Saha A et al. Differing autonomic through alpha7nAChR splenocytes. J Clin Invest 2016; 126:
responses to dobutamine stress in the presence and absence 19391952
of myocardial ischaemia. J Physiol 2015; 593: 21712184 41. Levy B, Gibot S, Franck P et al. Relation between muscle
38. Toner A, Jenkins N, Ackland GL et al. Baroreflex impairment NaK ATPase activity and raised lactate concentrations in
and morbidity after major surgery. Br J Anaesth 2016; 117: septic shock: a prospective study. Lancet 2005; 365: 871875
324331 42. Bennett-Guerrero E, Ayuso L, Hamilton-Davies C et al.
39. Ackland GL, Whittle J, Toner A et al. Molecular mechanisms Relationship of preoperative antiendotoxin core antibodies
linking autonomic dysfunction and impaired cardiac and adverse outcomes following cardiac surgery. JAMA 1997;
contractility in critical illness. Crit Care Med 2016; 44: 277: 646650
e614e624 43. Garcia-Alvarez M, Marik P, Bellomo R. Stress hyperlactatae-
40. Inoue T, Abe C, Sung SJ et al. Vagus nerve stimulation medi- mia: present understanding and controversy. Lancet Diabetes
ates protection from kidney ischemia-reperfusion injury Endocrinol 2014; 2: 339347