Clinical Investigation

Central venous pressure and the risk of

diuretic-associated acute kidney injury in
patients after cardiac surgery
Ian E. McCoy, MD, MS, Maria E. Montez-Rath, PhD, Glenn M. Chertow, MD, MPH, and Tara I. Chang, MD, MS
Palo Alto, CA

Background When prescribing diuretics in the postcardiac surgical intensive care unit (ICU), clinicians may use central
venous pressure (CVP) to assess volume status and the risk of acute kidney injury (AKI). In this study, we examined how the risk
of diuretic-associated AKI varied with CVP in patients undergoing cardiac surgery.
Methods We used the Medical Information Mart for Intensive Care database to study adults admitted to the postcardiac
surgical ICU at an urban, academic medical center between 2001 and 2012. We examined the odds of AKI per 1–mm Hg
increase in CVP among patients receiving intravenous loop diuretics using multivariable adjusted logistic regression. We
examined the risk of AKI among patients with diuretic use (vs nonuse) across tertiles of CVP using inverse probability treatment
weighting.
Results Among 4,164 patients receiving intravenous loop diuretics, the adjusted odds of subsequent AKI were 1.11 (95%
CI 1.08-1.13) times higher per mm Hg increase in mean CVP. This association was log-linear across the entire range of CVPs
observed. In the analysis of diuretic use (n = 5,396), the adjusted risk ratio for AKI with diuretic use (vs nonuse) was 1.33 (95%
CI 1.21-1.47) and did not materially differ across tertile of CVP.
Conclusions Higher rather than lower CVP is an independent marker of AKI risk. The risk of AKI associated with diuretic
use may not be influenced by CVP. Novel methods of assessing volume status and AKI risk are needed to guide patient
selection for diuretic therapy. (Am Heart J 2020;221:67-73.)

Diuretics are commonly used in the intensive care unit
(ICU) to treat and prevent fluid overload, but they can
also cause or contribute to acute kidney injury (AKI)
through excessive or overly rapid diuresis. When
weighing the risks and benefits of diuretic therapy,
clinicians may associate high central venous pressure
(CVP), one of the more objective measures of volume
status, with a lower risk of causing AKI with diuretic use. 1
Indeed, echocardiographic markers of right ventricular
dysfunction and volume overload have been associated
with a lower risk of worsening kidney function after
intravenous diuretic administration. 2
From the and Division of Nephrology, Stanford University School of Medicine,
Palo Alto, CA.
Roxana Mehran, MD, served as guest editor for this article.
Submitted July 24, 2019; accepted December 22, 2019.
Reprint requests: Ian E. McCoy, MD, MS, Division of Nephrology, Stanford University
School of Medicine, Palo Alto, CA 94304.
E-mail: imccoy@stanford.edu
0002-8703
© 2020 Elsevier Inc. All All rights reserved.
https://doi.org/10.1016/j.ahj.2019.12.013
In contrast, some studies have associated higher CVPs
with AKI in hospitalized patients, 3-5 although it is not known
whether the risk of AKI extends to critically ill patients being
treated with intravenous diuretics. Because prescription of
an intravenous diuretic implies volume overload in the
judgment of the treating clinician, the relation between AKI
and CVP in these patients may be distinct.
Herein we sought to examine the relation between CVP
and the odds of subsequent AKI in patients receiving
intravenous loop diuretics following cardiac surgery. We
hypothesized that higher CVP would be associated with
lower odds of AKI because the risk of excessive diuresis
would be lower in these patients. To specifically examine
AKI due to diuretic use, we estimated the risk of AKI with
intravenous loop diuretic use (vs nonuse) using methods
designed to mitigate confounding by indication and
hypothesized that the risk of AKI due to diuretic use
would be lower at higher CVP.
Methods
Study population
We analyzed deidentified data from the publicly
available Medical Information Mart for Intensive Care
 American Heart Journal
68 McCoy et al March 2020

Table I. Baseline patient characteristics, by tertiles of mean CVP

CVP b 9 mm Hg CVP 9-12 mm Hg CVP N 12 mm Hg
Characteristic (n = 1212) (n = 1511) (n = 1441)

Demographics
Age (y) 67 (12) 68 (12) 67 (12)
Female* 29% 32% 36%
White race 76% 76% 74%
Comorbidities
Diabetes* 28% 32% 35%
Hypertension 73% 73% 72%
Heart failure* 18% 25% 32%
CKD* 6% 7% 11%
Admit diagnosis
CAD* 56% 54% 44%
Aortic valve d/o* 15% 18% 20%
Mitral valve d/o* 9% 8% 6%
Other* 20% 20% 30%
Procedures/meds
CABG* 67% 68% 63%
CPB 92% 93% 93%
Left heart cath 28% 30% 31%
ACEi/ARB* 11% 10% 8%
NSAIDs* 35% 32% 27%
Vancomycin* 25% 23% 18%
Other AKI-associated meds* 0% 0% 1%
Severity of illness
ICU admission Cr (mg/dL)* 0.8 (0.3) 0.8 (0.3) 0.9 (0.4)
Mean MAP (mm Hg) 74 (6) 74 (6) 74 (6)
Mean CVP (mm Hg)* 7 (1) 10 (1) 15 (2)
Mechanically ventilated* 97% 99% 99%
Vasopressors* 76% 84% 87%
Minimum platelets (10 9/L) 145 (56) 143 (51) 143 (58)

Characteristics with * had statistically significant tests for trend at α ≤ .05. Continuous variables are expressed as mean (SD). All severity of illness variables and medication exposures
were measured for ICU day 1. Other AKI-associated medications include aminoglycosides, trimethoprim, amphotericin B, and calcineurin inhibitors. Cr, creatinine; d/o, disorder;
MAP, mean arterial pressure.

(MIMIC-III) database v1.4, which contains data on
N40,000 ICU patients at Beth Israel Deaconess Medical
Center between 2001 and 2012. 6 The database was
approved for research by the Institutional Review Boards
of the Massachusetts Institute of Technology and Beth
Israel Deaconess Medical Center, and studies of the
database are granted a waiver of informed consent. We
included the first ICU admission for each patient admitted
to the postcardiac surgical unit. We chose the postcardiac
surgical unit because almost all of these patients have
CVP measurements available (96%) and most (82%) of the
CVP data in the database come from this unit. Further-
more, the patients in the postcardiac surgical unit in the
MIMIC-III database are distinct from patients in other
ICUs, with higher rates of vasopressor use (70% vs 17%-
24%) and lower in-hospital mortality (4% vs 11%-15%).
Exclusion criteria were end-stage renal disease, lack of an
ICU day 1 CVP measurement, lack of either ICU day 1 or
ICU day 2-4 serum creatinine measurements, and missing
medication prescription information.
Because we hypothesized that the association of CVP
with AKI risk might be distinct among patients for whom
a clinician decided to order an intravenous diuretic, we
restricted our initial analyses to those patients who
received intravenous loop diuretics (furosemide, torse-
mide, bumetanide, and ethacrynic acid; 99.7% of
exposure was furosemide) during the first 24 hours of
ICU admission. In our analyses estimating the risk of AKI
with intravenous loop diuretic use (vs nonuse), we also
included patients who did not receive intravenous
loop diuretics.
Variable definitions
The mean of all ICU day 1 CVP measurements was our
exposure of interest. CVPs were directly measured about
once per hour, and measurements outside a reasonable
range (0-30 mm Hg) were considered to be spurious and
removed. We defined AKI by serum creatinine from the
Kidney Disease: Improving Global Outcomes (KDIGO)
criteria 7 and categorized as stage 1 if 1.5-b2× increase
from baseline or ≥0.3 mg/dL increase within 48 hours,
stage 2 if 2-b3× increase from baseline, and stage 3 if ≥3×
increase from baseline or increase to ≥4.0 mg/dL with
a ≥0.3-mg/dL increase from baseline. We used the first
serum creatinine measured on ICU day 1 as “baseline”
and compared to all creatinines measured 24-96 hours
American Heart Journal
Volume 221
Figure 1
Cohort assembly.
after ICU admission. We captured comorbidities from the
International Classification of Disease, Ninth Edition,
Clinical Modification (ICD-9-CM) codes according to
version 3.7 of the Elixhauser comorbidities defined by the
Agency for Healthcare Research and Quality. 8 Admission
diagnosis categorized the primary ICD-9 code for each
admission as coronary artery disease (CAD; codes 414.01
and 410.71), aortic valve disorder (424.1), mitral valve
disorder (424.0), or other. Cardiac procedures were
abstracted from ICD-9-CM procedure codes as cardiopul-
monary bypass (CPB; codes 3961 and 3966), coronary
artery bypass graft (CABG; codes 3610-3616), and left
heart catheterization (codes 3722-3 and 8853-7). 9
Statistical analysis
Continuous data were expressed as means with
standard deviations, and categorical data were reported
as proportions. We ranked CVP into approximate tertiles
(b9, 9-12, N12 mm Hg) and compared baseline charac-
teristics for trend using the Wilcoxon rank sum test and
the Cochran-Armitage trend test. We performed logistic
regression to compute odds ratios (ORs) for the
association of mean CVP on the odds of AKI. We
evaluated potential confounding with multivariable
McCoy et al 69
logistic regression. All covariates with univariate associ-
ations with AKI with P values ≤.2 were included in the
final multivariable model. AKI stage-specific ORs were
calculated from multinomial logistic regression.
To assess whether CVP modifies the risk of AKI with
intravenous loop diuretic use (vs nonuse), we performed
inverse probability treatment weighting (IPTW, Item S1)
to mitigate confounding by indication. 10 After dividing
the cohort into tertiles of CVP, we fit separate multivar-
iable logistic regressions for each tertile containing the
variables listed in Table I to estimate the propensity for
diuretic use. 11 We then computed stabilized, trimmed
weights, defined as the inverse of the propensity
multiplied by the marginal probability of treatment
received, truncated to 10 (0.1) if outside these bounds.
We fit generalized linear models (binomial distribution
with log link) with robust standard errors to report risk
ratios (RRs) for AKI together with 95% confidence
intervals (CI). Because all covariates were well balanced
after IPTW, no additional adjustments were made to the
models. Two-tailed P values b.05 were considered
statistically significant. We performed all statistical
analyses using SAS software, version 9.4 (SAS Institute,
Cary, NC).

70 McCoy et al
Figure 2
Incidence of subsequent AKI by mean CVP on ICU day 1.
Results
We identified 4,164 postcardiac surgical patients
treated with intravenous loop diuretics and with at least
1 CVP measurement in the first 24 hours after ICU
admission (Figure 1). These patients were almost
exclusively mechanically ventilated at some point on
ICU day 1, and most had exposure to cardiopulmonary
bypass. Patients with higher mean CVP were more likely
to have comorbidities of heart failure and chronic kidney
disease. They were also more likely to be treated with
vasopressors and less likely to be treated with
angiotensin-converting enzyme inhibitors (ACEi)/angio-
tensin receptor blockers (ARBs), nonsteroidal anti-
inflammatory drugs (NSAIDs), or vancomycin (Table I).
The incidence of AKI was higher among patients with
higher CVP (Figure 2, P for trend b .0001), ranging from
36% in patients with mean CVP b 9 mm Hg to 54% in
patients with mean CVP N 12 mm Hg. This trend was
consistent across all KDIGO AKI severity stages 1, 2, and
3. Fewer than 2% of patients received renal replacement
therapy. Logistic regression showed the odds of AKI were
1.11 (95% CI 1.09-1.13) times higher per 1–mm Hg
increase in mean CVP. The odds of AKI did not materially
change after sequential adjustment for potential con-
founders including demographics, comorbidities, hospi-
talization factors, and indicators of severity of illness (fully
adjusted OR 1.11, 95% CI 1.08-1.13, P b .0001) (Table SI).
Moreover, the odds per mm Hg of CVP were higher at
more severe AKI stages: adjusted odds ratios for KDIGO
stage 1 AKI: 1.09 (95% CI 1.06-1.11), KDIGO stage 2 AKI:
1.20 (95% CI 1.15-1.25), and KDIGO stage 3 AKI: 1.23
(95% CI 1.15-1.31).
American Heart Journal
March 2020
By graphing the log odds of AKI in 20 quantiles, we
found that the relationship between mean CVP and the
log odds of AKI was linear over the range of CVPs studied
(Figure 3). There was no sign of an upper or lower limit of
CVP beyond which the relationship changed.
In our analyses examining the risk of AKI with
intravenous loop diuretic use (vs nonuse), we included
1,232 patients not receiving intravenous loop diuretics on
ICU day 1 in addition to our original cohort of 4,164
patients. Compared to patients who did not receive
diuretics, patients who received diuretics were older,
with higher rates of CABG, CPB, and mechanical
ventilation. The groups were well balanced after IPTW
adjustment (Tables II and SII-SIV).
The RR for AKI with diuretic use (vs nonuse) from
IPTW analysis in the entire cohort was 1.33 (95% CI 1.21-
1.47). The RRs from the IPTW analyses in each CVP tertile
had overlapping CIs (Figure 4): CVP b 9 mm Hg 1.38
(95% CI 1.13-1.69), CVP 9-12 mm Hg 1.27 (95% CI 1.06-
1.51), and CVP N 12 mm Hg 1.40 (95% CI 1.20-1.64).
Thus, we found no evidence of significant difference in
the RR estimates across the range of CVP. 12
Discussion
In this cohort study, we found that higher mean CVP
was a marker of AKI risk in postcardiac surgical patients
receiving intravenous loop diuretics, but CVP did not
inform the risk of AKI due to diuretic use.
This association between CVP and AKI was log-linear
over the range of CVPs studied (Figure 3). Each mm Hg
increase in mean CVP increased the relative odds of AKI
by 11% even after adjustment for numerous covariates
American Heart Journal
Volume 221
Figure 3
Relationship between mean CVP on ICU day 1 and the log odds of subsequent AKI.
including mean arterial pressure and vasopressor use.
Surprisingly, there was no lower bound for CVP below
which the risk of AKI increased, although all of these
patients were prescribed intravenous diuretics that may
drop CVP even lower. Although most of the AKIs
observed in our study were KDIGO stage 1, the
association with CVP was significant for all AKI stages
and the OR was highest for KDIGO stage 3.
The association between CVP and AKI has been seen
before, 3 and higher CVP has been associated with poorer
kidney function in patients with and without heart
failure. 4, 5, 13 However, none of these previous analyses
were restricted to patients receiving intravenous loop
diuretics, so we hypothesized that the association
between CVP and AKI might be distinct in this
population. CVP may be a marker of other illness
variables not controlled for in our analysis, or it may
directly increase AKI risk through decreased renal
perfusion pressure and increased renal interstitial hyper-
tension from venous congestion. 14
Because diuretics decrease CVP and CVP is associated
with AKI risk, we hypothesized that the risk of AKI due to
diuretic use would be lower in patients with higher CVPs.
Although intravenous loop diuretics are often associated
with AKI presumably due to excessive diuresis, 15-18
sometimes diuretics reverse AKI caused by venous
congestion. 19, 20 Therefore, we expected that the risk
McCoy et al 71
of AKI with diuretic use would vary with CVP. However,
in our analysis, the magnitude of the association between
intravenous loop diuretic use (vs nonuse) and AKI did not
vary across tertiles of CVP (Figure 4).
The fact that the relative risk of AKI with diuretic use
was not different in different CVP tertiles (25%-40%
increase in all CVP tertiles) is difficult to explain.
Mechanical (positive pressure) ventilation does increase
CVP, but the rate of mechanical ventilation was similar in
all CVP tertiles (95%-99%) (Tables SII-SIV). Furthermore,
although the absolute CVP values may be artificially
increased in this cohort by mechanical ventilation, one
would still expect relative differences in CVP to indicate
volume status and meaningfully inform diuretic risk.
Indeed, many physicians use the CVP in critically ill,
mechanically ventilated patients to guide decisions on
diuretic and fluid management. Although CVP may
inform the chance of benefit from diuretic use (eg, to
pulmonary congestion or cardiac output), we conclude
that CVP does not appear to inform the risk of AKI from
diuretic use.
Our study has several limitations. First, although we
were able to adjust for numerous clinical variables and
used IPTW to mitigate confounding by indication, as with
any observational study, residual confounding cannot be
excluded. Second, preadmission serum creatinine deter-
minations were unavailable, so some patients may have
 American Heart Journal
72 McCoy et al March 2020

Table II. Baseline characteristics before and after IPTW adjustment

Full cohort IPTW cohort

Diuretics No diuretics
Characteristics (n = 4164) (n = 1232) Std diff Diuretics No diuretics Std diff

Demographics
Age (y) 67 (12) 65 (14) 0.16 67 (13) 67 (13) 0.00
Female 32% 31% 0.04 32% 31% 0.02
White race 75% 71% 0.09 75% 75% 0.00
Comorbidities
Diabetes 32% 26% 0.13 30% 30% 0.00
Hypertension 73% 63% 0.20 70% 70% 0.01
Heart failure 25% 24% 0.04 25% 25% 0.00
CKD 8% 7% 0.02 8% 7% 0.02
Admit diagnosis
CAD 51% 44% 0.15 50% 51% 0.03
Aortic valve d/o 18% 9% 0.25 16% 15% 0.02
Mitral valve d/o 8% 8% 0.00 8% 8% 0.00
Other 23% 39% 0.35 27% 26% 0.01
Procedures/meds
CABG 66% 55% 0.23 63% 65% 0.04
CPB 93% 75% 0.51 88% 88% 0.00
Left heart cath 30% 27% 0.07 29% 30% 0.02
ACEi/ARB 10% 4% 0.23 8% 6% 0.08
NSAIDs 31% 23% 0.18 29% 29% 0.00
Vancomycin 22% 19% 0.07 21% 21% 0.01
Other AKI-associated meds 0% 1% 0.05 1% 1% 0.01
Severity of illness
ICU admission Cr (mg/dL) 0.9 (0.4) 0.9 (0.6) 0.16 0.9 (0.4) 0.9 (0.5) 0.01
Mean MAP (mm Hg) 74 (6) 75 (7) 0.06 74 (6) 75 (7) 0.01
Mean CVP (mm Hg) 11 (3) 10 (4) 0.14 11 (3) 11 (4) 0.02
Mechanically ventilated 99% 94% 0.26 97% 97% 0.00
Vasopressors 83% 81% 0.06 82% 83% 0.02
Minimum platelets (10 9/L) 144 (55) 153 (69) 0.14 147 (62) 147 (60) 0.01

Continuous variables are expressed as mean (SD). All severity of illness variables and medication exposures were measured for ICU day 1. Other AKI-associated medications include
aminoglycosides, trimethoprim, amphotericin B, and calcineurin inhibitors.

Figure 4

Risk ratios of intravenous loop diuretic use (vs nonuse) for subsequent AKI across CVP tertiles.

already developed AKI on admission. Thus, the “baseline”
serum creatinine may have represented AKI “in evolu-
tion,” and thus, the incidence and severity of AKI might
have been underestimated. Third, although we used an
average of all CVPs measured on ICU day 1 as our
exposure of interest, we did not evaluate other related
and potentially informative exposures such as the change
in CVP during day 1 or beyond owing to increased
complexity and nonrandom missing data. Fourth, these
patients were almost exclusively mechanically ventilated,
American Heart Journal
Volume 221
which is known to increase CVPs. Thus, although we are
confident that the trend between AKI and CVP is a valid
one, the qualitative findings should be emphasized rather
than the risks associated with the absolute value of CVP,
which may differ by several mm Hg depending on
whether positive pressure ventilation is or is not being
delivered. Finally, we relied on administrative ICD-9-CM
codes to ascertain admission diagnoses and procedures.
In summary, higher rather than lower CVP is an
independent risk factor for AKI. However, CVP does
not inform the risk of AKI from diuretic use (vs nonuse).
Further research should aim to identify better tools to
assess volume status and to determine ICU patient groups
for whom diuretics can be most safely administered.
Acknowledgements
None.
Sources of funding
Dr McCoy was supported by a National Institutes of
Health T-32 grant (5T32DK007357), and Dr Chertow and
Dr Montez-Rath were supported by a National Institutes
of Health K-24 grant (2K24DK085446).
Disclosures
None.
Appendix. Supplementary data
Supplementary data to this article can be found online
at https://doi.org/10.1016/j.ahj.2019.12.013.
References
1. Heart Failure Society of America. Section 12: evaluation and
management of patients with acute decompensated heart failure. J
Card Fail 2010 Jun;16(6):e134-56.
2. Testani JM, Khera AV, St. John Sutton MG, et al. Effect of right
ventricular function and venous congestion on cardiorenal interac-
tions during the treatment of decompensated heart failure. Am J
Cardiol 2010 Feb;105(4):511-6.
3. Mullens W, Abrahams Z, Francis GS, et al. Importance of venous
congestion for worsening of renal function in advanced decompen-
sated heart failure. J Am Coll Cardiol 2009 Feb;53(7):589-96.
McCoy et al 73
4. Chen KP, Cavender S, Lee J, et al. Peripheral edema, central venous
pressure, and risk of AKI in critical illness. Clin J Am Soc Nephrol
2016;11(4):602-8.
5. Williams JB, Peterson ED, Wojdyla D, et al. Central venous pressure
after coronary artery bypass surgery: does it predict postoperative
mortality or renal failure? J Crit Care 2014 Dec;29(6):1006-10.
6. Johnson AEW, Pollard TJ, Shen L, et al. MIMIC-III, a freely accessible
critical care database. Sci Data 2016;3:1-9.
7. Kidney Disease. Improving Global Outcomes (KDIGO) Acute Kidney
Injury Working Group. KDIGO clinical practice guideline for acute
kidney injury. Kidney Int Suppl 2012;2(1):1-138.
8. Agency for Healthcare Research and Quality, Rockville M. HCUP
Elixhauser Comorbidity Software, Version 3.7. Healthcare Cost and
Utilization Project (HCUP). [Internet]. Available from: , www.hcup-us.
ahrq.gov/toolssoftware/comorbidity/comorbidity.jsp
9. Chu D, Bakaeen FG, Dao TK, et al. On-pump versus off-pump
coronary artery bypass grafting in a cohort of 63,000 patients. Ann
Thorac Surg 2009;87(6):1820-7.
10. Robins J, Hernan M. B B. Marginal structural models and causal
inference in epidemiology. Epidemiology 2000;11(55):550-60.
11. Brookhart MA, Schneeweiss S, Rothman KJ, et al. Variable selection
for propensity score models. Am J Epidemiol 2006 Jun 15;163(12):
1149-56.
12. Van Belle G. Statistical rules of thumb. Vol 699. John Wiley & Sons;
2011. 38–40 p.
13. Damman K, van Deursen VM, Navis G, et al. Increased central venous
pressure is associated with impaired renal function and mortality in a
broad spectrum of patients with cardiovascular disease. J Am Coll
Cardiol 2009 Feb;53(7):582-8.
14. Ronco C, McCullough P, Anker SD, et al. Cardio-renal syndromes:
report from the consensus conference of the acute dialysis quality
initiative. Eur Heart J 2010;31(6):703-11.
15. Felker GM, Lee KL, Bull DA, et al. Diuretic strategies in patients with
acute decompensated heart failure. N Engl J Med 2011 Mar 3;364
(9):797-805.
16. Grams ME, Estrella MM, Coresh J, et al. Fluid balance, diuretic use,
and mortality in acute kidney injury. Clin J Am Soc Nephrol 2011;6
(5):966-73.
17. Kane-Gill SL, Sileanu FE, Murugan R, et al. Risk factors for acute
kidney injury in older adults with critical Illness: a retrospective cohort
study. Am J Kidney Dis 2015;65(6):860-9.
18. McCoy IE, Montez-Rath ME, Chertow GM, et al. Estimated effects of
early diuretic use in critical illness. Crit Care Explor 2019 Jul;1.
19. Onuigbo M, Agbasi N, Sengodan M, et al. Acute kidney injury in
heart failure revisited—the ameliorating impact of “decongestive
diuresis” on renal dysfunction in type 1 acute cardiorenal syndrome:
accelerated rising pro B naturetic peptide is a predictor of good renal
prognosis. J Clin Med 2017;6(9):82.
20. Ross EA. Congestive renal failure: the pathophysiology and treatment
of renal venous hypertension. J Card Fail 2012;18(12):930-8.