Journal of Clinical Anesthesia 37 (2017) 116122

Contents lists available at ScienceDirect

Journal of Clinical Anesthesia

Original contribution

Dexamethasone and perioperative blood glucose in patients undergoing

total joint arthroplasty: A retrospective study
Michael Nurok a,, Jennifer Cheng a, Giulio R. Romeo b, Stephanie M. Vecino c, Kara G. Fields d, Jacques T. YaDeau a
a
Department of Anesthesiology, Hospital for Special Surgery, New York, NY, United States
b
Department of Molecular and Cellular Pharmacology, Joslin Diabetes Centre, Boston, MA, United States
c
Department of Anesthesiology, Weill Cornell Medical College, New York, NY, United States
d
Healthcare Research Institute, Hospital for Special Surgery, New York, NY, United States

a r t i c l e i n f o a b s t r a c t

Article history: Study objective: Perioperative dexamethasone is commonly used to prevent nausea. It can also increase blood glu-
Received 29 April 2016 cose levels, and recent concern about its blood glucose-elevating effect in humans has been raised. This study
Received in revised form 27 October 2016 aimed to demonstrate relationships between dexamethasone administration and elevated perioperative blood
Accepted 30 November 2016 glucose in patients undergoing total joint arthroplasty.
Available online xxxx
Design: Retrospective study.
Setting: Academic, orthopedic hospital.
Keywords:
Dexamethasone
Patients: A total of 625 patients (1899 years) who underwent total hip or total knee arthroplasty with an ASA 3
Blood glucose were included in the study.
Arthroplasty, replacement, hip Interventions: Patients who received dexamethasone perioperatively were compared to those who did not re-
Arthroplasty, replacement, knee ceive dexamethasone.
Diabetes mellitus Measurements: The primary outcome, which was any postoperative glucose N 200 mg/dl, was compared
between groups using multiple logistic regression. Demographic information, intraoperative information,
incidence of postoperative nausea and vomiting, white blood cell count, medication use, and length of stay
were also collected.
Main results: Perioperative dexamethasone (median [1st quartile, 3rd quartile] dose = 4 [4, 8] mg) was admin-
istered to 76% of patients. Only 5.6% (95% CI: 3.87.5) of patients had postoperative glucose levels N 200 mg/dl.
After covariate adjustment, there was no evidence of a difference in odds of experiencing postoperative glucose
levels N 200 mg/dl (odds ratio [95% CI]: 0.76 [0.282.07]; P = 0.594) and maximum glucose levels (P = 0.518)
between groups. Dexamethasone-treated patients had greater changes in white blood cell count between base-
line and postoperative days 01. There was no evidence of a difference in wound healing and length of stay be-
tween groups.
Conclusions: There was no evidence of an association between perioperative dexamethasone administration and
the odds of having postoperative glucose levels N200 mg/dl or higher maximum glucose levels. However, these
ndings may not be generalizable to patients having different baseline characteristics or procedures.
2016 Elsevier Inc. All rights reserved.

1. Introduction increase in blood glucose may impose in patients without diabetes

Dexamethasone is a corticosteroid that is commonly used to treat
postoperative nausea and vomiting (PONV) and pain in the orthopedic
setting [1,2]. However, dexamethasone has been shown to increase
blood glucose in animal models [3,4], and a similar effect has been sug-
gested in humans [5,6]. Concerns regarding the potential for periopera-
tive dexamethasone-induced hyperglycemia to negatively inuence
postoperative outcomes in patients and the harm that even a slight
Corresponding author at: Cedars-Sinai Heart Institute, 127 San Vicente Blvd., Suite
3100, Los Angeles, CA 90048, United States.
E-mail address: Michael.Nurok@cshs.org (M. Nurok).
have been raised. Thus, anesthetic regimens involving dexamethasone
may need to be re-evaluated to prevent postoperative hyperglycemia-
induced adverse effects [7], which can outweigh the benets of dexa-
methasone administration.
There are limited data demonstrating a clear relationship between
dexamethasone administration, elevated perioperative blood glucose,
and postoperative outcomes in patients undergoing orthopedic surgery.
This study aimed to evaluate the relationship between dexamethasone
administration and elevated perioperative blood glucose in patients
undergoing total joint arthroplasty. We hypothesized that dexametha-
sone administration would be associated with increased odds of
elevated perioperative blood glucose. The primary outcome was any

http://dx.doi.org/10.1016/j.jclinane.2016.11.012
0952-8180/ 2016 Elsevier Inc. All rights reserved.
 M. Nurok et al. / Journal of Clinical Anesthesia 37 (2017) 116122
postoperative blood glucose level N200 mg/dl. The secondary outcomes
included maximum blood glucose, the area under the curve (AUC) of
blood glucose measurements, changes in white blood cell (WBC)
count from baseline, average numerical rating scale (NRS) pain score,
wound healing, and length of stay.
2. Materials and methods
2.1. Ethics
This retrospective study of prospectively collected data was approved
by the Institutional Review Board at Hospital for Special Surgery (HSS) on
November 14, 2013 (#2013-103), and waiver of consent was obtained.
The study began in January 2014 and ended in July 2014. All procedures
followed were in accordance with the ethical standards of the responsible
committee on human experimentation (institutional or regional) and
with the Helsinki Declaration of 1975, as revised in 1983.
2.2. Patient enrollment
Patients aged 1899 years who underwent unilateral total hip
arthroplasty (THA) or total knee arthroplasty (TKA) were included.
The exclusion criteria were 1) intraoperative administration of stress-
dose steroids, 2) organ dysfunction, 3) American Society of Anesthesiol-
ogists Classication (ASA) N 3, 4) revision arthroplasty, and 5) additional
concurrent procedures.
2.3. Selection process
Patients were selected on a weekly basis. All patients who
underwent surgeries 2 weeks prior to the day of screening and met
the inclusion criteria were recorded. Anyone who t the exclusion
criteria or had documents missing from their charts, which would pre-
vent adequate screening and data collection, was excluded. Of the re-
maining patients, 625 THA and TKA patients were randomly selected
by a research assistant who was not otherwise involved in the study.
2.4. Study conduct details
Medications, blood glucose levels, WBC counts, and NRS pain scores
that were prospectively recorded at baseline (preoperative period), post-
operative day (POD) 0, POD 1, and POD 2 were collected from patients'
electronic charts. Blood glucose levels were measured via laboratory as-
sessments or point-of-care devices. Information about anesthesia and in-
traoperative medications was collected from the patients' anesthesia
records. Anesthesia start times and hospital discharge times were collect-
ed to calculate length of stay. All data were entered into Research Elec-
tronic Data Capture, which is a secure, web-based application for
building and managing online surveys and databases [8], hosted at HSS.
2.5. Blood glucose handling
For diabetic patients, point-of-care blood glucose measurements
were conducted upon arrival to the holding area. Glucose levels
b70 mg/dl or N200 mg/dl warranted treatments, as decided by an inter-
nist or anesthesiologist. Measurements of glucose were taken using a
point-of-care device every 6 h while on a soft diet and immediately be-
fore meals and at 10 PM while on a consistent carbohydrate diet. Serum
glucose levels were measured once daily. The inpatient glucose goal was
140180 mg/dl; if necessary, insulin was given to maintain these goals.
Blood glucose levels N400 mg/dl warranted the administration of 5
6 units of insulin and notication of the House Ofcer. For all other pa-
tients, serum glucose levels were measured once daily as part of the
basic metabolic panel (e.g., potassium, sodium, etc.), along with com-
plete blood count tests.
117
2.6. Statistical analysis
We estimated that 3364% of patients would be administered dexa-
methasone, and that 22% of patients not given dexamethasone would
experience postoperative blood glucose levels N 200 mg/dl. This was
based on the following estimates: (1) 10% of patients will have diabetes
[9]; (2) 20% of patients without diabetes who do not receive dexameth-
asone will have perioperative elevated blood glucose [10]; and (3) 40%
of patients with diabetes who do not receive dexamethasone will have
perioperative elevated blood glucose [11]. We calculated that a total of
625 patients would provide at least 80% power at a two-sided alpha
level of 0.05 to detect a 20% difference in the percentage of patients
that experience elevated postoperative blood glucose levels using mul-
tiple logistic regression, anticipating that model covariates would ex-
plain 70% of the variability in dexamethasone administration (i.e.,
model R [2] would be 0.7 if dexamethasone administration were
regressed on model covariates). The threshold of 200 mg/dl was chosen
to ensure good separation between groups and to prevent against the
likelihood of missing signals due to stress-induced hyperglycemia in pa-
tients who have not received dexamethasone. This narrow separation
has been documented in previous studies [12].
Continuous and ordinal variables are expressed as means with stan-
dard deviations or medians with 1st and 3rd quartiles, depending upon
the distribution of the data. Categorical variables are expressed as counts
and percentages. The association between dexamethasone administra-
tion and study variables was assessed using t-tests or Wilcoxon rank-
sum tests for continuous variables and chi-square or Fisher's exact tests
for categorical variables. Outcomes with a single measurement per pa-
tient were compared between groups using robust regression via Huber's
M-estimator [13] for continuous variables and multiple logistic regression
for categorical variables. For the analysis of postoperative glucose out-
comes, the last measurement to be included was the one closest to the
48-hour cut-off time; any measurements after that were eliminated to re-
duce large variations in time between rst and last glucose measure-
ments. Longitudinal measures were compared using regression based
on the generalized estimating equations (GEE) approach [14,15] with
an autoregressive [AR(1)] correlation structure. The GEE method accounts
for the correlation between repeated measurements on the same patient,
where the AR(1) correlation structure assumes a greater degree of corre-
lation among measurements recorded closer in time. Models for each out-
come initially included an interaction term between the treatment group
and time point. If no evidence of an interaction was found, the model was
ret without an interaction term, and results were reported as the overall
difference in means between groups or odds ratio with 95% condence in-
tervals (CI). In addition, a stratied analysis based on diabetes status was
performed to determine if dexamethasone was associated with the odds
of having postoperative glucose levels N200 mg/dl.
A propensity score analysis was performed to test the robustness of
regression results. Propensity scores were calculated via logistic regres-
sion using dexamethasone administration as the outcome and age, body
mass index (BMI), diabetes, daily insulin use, ASA status, white race,
procedure, preoperative antiepileptic drug use, preoperative oral
antiglycemic agent use, preoperative non-insulin injectable medication
use, preoperative antiemetic use, and baseline blood glucose level as
predictors. Patients who did not receive dexamethasone were matched
in a 1:1 ratio to patients who received perioperative dexamethasone.
Matching was performed exactly on diabetes status and by the nearest
neighbor method without replacement on propensity score. Two pa-
tients with type I diabetes who did not receive dexamethasone were ex-
cluded from analysis because they could not be matched exactly on
diabetes status to any patient in the dexamethasone group. Covariate
balance was assessed by calculating standardized differences on both
the original and matched samples and using the suggested cut-off of 0.1
to indicate a negligible difference between groups [16]. Outcomes with
a single measurement per patient were compared between groups in
the matched sample using paired t-tests for continuous variables and
118 M. Nurok et al. / Journal of Clinical Anesthesia 37 (2017) 116122

McNemar's test for categorical variables. Longitudinally measured out- Table 1

comes were compared between groups in the matched sample using Baseline and intraoperative characteristics.

mixed-effects modelling. All statistical hypothesis tests were two-sided, Dexamethasone

with P b 0.05 considered statistically signicant. R statistical software Yes No
(version 2.14.1; The R Foundation for Statistical Computing, Vienna, Aus- (N = 474) (N = 146) P value
tria) was used for propensity score calculation and matching via packages a
Age (years); mean SD 66 10 67 12 0.099
CBPS and MatchIt, respectively. SAS statistical software (version 9.3; SAS BMI (kg/m2); mean SD 29 6 30 7 0.521
Institute, Cary, NC) was used for all other statistical analyses. Women; N (%) 295 (62) 95 (65) 0.536
Caucasian; N (%) 402 (85) 121 (83) 0.766
ASA; N (%) 0.010
3. Results 1 16 (3) 0 (0)
2 373 (79) 109 (75)
3.1. Patient ow 3 85 (18) 37 (25)
Procedure; N (%) 0.815
Total knee arthroplasty 239 (50) 72 (49)
A total of 3245 patients who underwent unilateral THA or TKA from Total hip arthroplasty 235 (50) 74 (51)
December 2013 to July 2014 were assessed for eligibility. After eliminat- Diabetes; N (%) 32 (7) 27 (19) b0.001
ing charts with missing information and excluding for intraoperative Insulin use; N (%) 4 (1) 3 (2) 0.364
stress-dose steroids, ASA N 3, revision procedures, and additional concur- Oral antiglycemic agent use; N (%) 33 (7) 27 (19) b0.001
Blood glucose; mg/dl; median (Q1, Q3) 92 (84, 105) 93 (86, 111) 0.016
rent procedures, there were 2438 eligible patients. Of these, 625 patients
HbA1c; mmol/mol (%); mean SD 44 8.7 46 7.7 0.354
were randomly selected, and study data were collected for these patients. (6.2 0.8) (6.4 0.7)
Because detailed chart reviews revealed that 5 patients had intraopera- (n = 45) (n = 35)
tively received stress-dose steroids, only the remaining 620 patients Average NRS pain; median (Q1, Q3) 5 (4, 8) 6 (4, 8) 0.327
WBC Count; per nl (mean SD) 6.6 1.8 6.8 2.2 0.291
were included in the data analysis (309 for THA and 311 for TKA) (Fig. 1).
Intraoperative epinephrine use; N (%) 137 (29) 39 (27) 0.608
Periarticular injection use; N (%) 48 (10) 18 (12) 0.451
3.2. Demographics and dexamethasone administration Cement use; N (%) 282 (60) 91 (62) 0.541
Combined spinal-epidural anesthetic 407 (86) 117 (80) 0.094
use; N (%)
Perioperative dexamethasone (median [1st quartile, 3rd quartile] Spinal anesthetic use; N (%) 42 (9) 8 (5) 0.190
dose = 4 [4, 8] mg) was administered to 474 patients (76%; Table 1). Epidural anesthetic use; N (%) 12 (3) 20 (14) b0.001
Most patients received dexamethasone intraoperatively as an intrave- Peripheral nerve block use; N (%) 233 (49) 76 (52) 0.540
nous injection alone (87.1%), and several patients received it as a pe- General anesthetic use; N (%) 15 (3) 1 (0.6) 0.136
Postoperative PCA; N (%) 447 (94) 129 (88) 0.014
ripheral nerve block additive with (5.5%) or without (1.5%) the Epidural; N 393 123
intravenous injection. In some cases, dexamethasone was administered Intravenous; N 35 4
postoperatively via intravenous injection or oral pill; twelve patients Intra-articular; N 19 2
received it orally only (2.5%). Seven patients received dexamethasone a
Abbreviations - SD: standard deviation; BMI: body mass index; ASA: American Society
intravenously and orally (2.3%), and three patients received dexameth- of Anesthesiologists Classication; Q1: 1st quartile; Q3: 3rd quartile; WBC: white blood
asone intravenously, orally, and via a peripheral nerve block (0.6%). cell; HbA1c: glycated hemoglobin; PCA: patient-controlled analgesia.
p b 0.05.
There were no differences in age, BMI, sex, or race between patients
who were treated with or without dexamethasone. However, ASA sta-
tus was signicantly different between groups (P = 0.010); dexameth- in patients who did not receive dexamethasone, but the magnitude of
asone was administered to fewer ASA 3 patients (Table 1). the difference was small (median [1st quartile, 3rd quartile]; 93 mg/dl
[86, 111] vs. 92 mg/dl [84, 105]; P = 0.016). There were no differences
in baseline glycated hemoglobin (HbA1c) values (Table 1), although
3.3. Preoperative factors and intraoperative management data were available for only 80 patients.
Furthermore, no signicant differences in preoperative NRS pain
Fifty-nine patients (9%) had diabetes, and 32 of them received dexa- score and WBC count were observed between patients who were treat-
methasone. Preoperative blood glucose levels were signicantly higher ed with or without dexamethasone (Table 1). Intraoperatively, similar
percentages of dexamethasone- and non-dexamethasone-treated pa-
tients received epinephrine, periarticular injections, combined spinal-
epidural anesthetics, spinal anesthetics, peripheral nerve blocks, general
anesthetics, and orthopedic cement as part of the surgical procedure
(P = 0.541). A signicantly lower percentage of patients in the dexa-
methasone group received epidural anesthetics (P b 0.001), and a signif-
icantly higher percentage of patients in the dexamethasone group
received postoperative patient-controlled analgesia (P = 0.014)
(Table 1). Patient-controlled analgesia was administered via epidural
(bupivacaine/hydromorphone, bupivacaine/clonidine, or bupivacaine
alone), intravenous (hydromorphone), or intra-articular (ropivacaine)
lines. Throughout their perioperative stay, all patients were adminis-
tered Lactated Ringer's solution and/or normal saline. Only two patients
received intravenous dextrose 50%; both patients had diabetes.

3.4. Postoperative glucose levels and AUC of glucose

A relatively small number of patients (5.6%) had postoperative glu-

Fig. 1. Patient ow chart. The numbers of assessed, eligible, selected, and analyzed patients cose levels N200 mg/dl (Table 2). A higher percentage of patients who
are shown on the chart. did not receive dexamethasone had glucose levels N 200 mg/dl (11.6%
 M. Nurok et al. / Journal of Clinical Anesthesia 37 (2017) 116122 119

Table 2
Postoperative blood glucose outcomes adjusted analysesa.

Dexamethasone

Outcomes Yes (N = 474) No (N = 146) Odds ratio or difference in means (95% CI) P value

Blood glucose N 200 mg/dl (N; %) 18 (4) 17 (12) 0.76 (0.28, 2.07) 0.594
Maximum blood glucose; mg/dl; median (Q1, Q3) 127 (112, 149) 130 (114, 155) 1.6 (3.2, 6.3) 0.518
Maximum change in blood glucose; mg/dl; median (Q1, Q3) 34 (17.5, 54) 38 (22, 54) 1.0 (4.1, 6.2) 0.692
Positive incremental AUCb; median (Q1, Q3)c 748 (326, 1302) 726 (315, 1183) 0.98 (0.79, 1.22) 0.876

Diabetic patients
Blood glucose N 200 mg/dl (N; %) 5 (1) 4 (3) 0.32 (0.08, 1.24) 0.099

Non-diabetic patients
Blood glucose N 200 mg/dl (N; %) 13 (41) 13 (48) 1.70 (0.42, 6.86) 0.451
a
Blood glucose N 200 mg/dl and maximum blood glucose outcomes were adjusted for age, sex, body mass index, procedure (total hip arthroplasty vs. total knee arthroplasty), diabetes
(yes vs. no), ASA status (3 vs. 1 or 2), intraoperative epinephrine administration (yes vs. no), postoperative antiglycemic use (yes vs. no), postoperative steroid use (yes vs. no), postop-
erative antiemetic use (yes vs. no), and baseline glucose level. The maximum change in blood glucose was adjusted for all covariates except for baseline glucose level.
b
Results from unadjusted analysis only.
c
Abbreviations Q1: 1st quartile; Q3: 3rd quartile; AUC: area under the curve; CI: condence interval.

vs. 3.4%). After adjusting for age, sex, BMI, procedure, diabetes, ASA sta-
tus, intraoperative epinephrine, and postoperative use of antiglycemics,
steroids, and antiemetics, there was no evidence of an association be-
tween dexamethasone administration and the odds of experiencing
postoperative blood glucose levels N 200 mg/dl (odds ratio [95% CI]:
0.76 [0.28, 2.07]; P = 0.594). There was also no evidence of an interaction
between diabetes and dexamethasone (P = 0.351), so an interaction
term was not included in this nal logistic regression model. Furthermore,
there was no evidence of a difference in maximum postoperative glucose
levels between patients who were treated with or without dexametha-
sone (Table 2). When limiting the dexamethasone group to patients
who only received 4 mg IV dexamethasone (n = 299), there was no ev-
idence of an association between dexamethasone administration and
the odds of experiencing postoperative blood glucose levels N 200 mg/dl
(adjusted odds ratio [95% CI]: 0.60 [0.19, 1.87]; P = 0.379).
Additionally, the positive incremental AUC of blood glucose mea-
surements was calculated from baseline to 48 h post-anesthesia start
time. The positive incremental method gives the area under the curve
above the baseline value and does not factor in the area below baseline
[17]. The number of glucose measurements available to calculate the
AUC ranged from 2 to 30, with a median of 3. There were no differences
in glucose AUC between groups (Table 2).
Even though there was no evidence of an interaction between diabe-
tes and dexamethasone, a stratied analysis based on diabetes status was
also performed for the primary outcome. There was no evidence of an as-
sociation between dexamethasone administration and the odds of having
postoperative glucose levels N 200 mg/dl in diabetic patients or in non-di-
abetic patients (Table 2). In addition, there were seven prediabetic pa-
tients. In an exploratory analysis, prediabetes was not associated with
the odds of postoperative hyperglycemia (N200 mg/dl) (P = 0.196).
Further exploratory analyses were performed to test whether (1)
total dexamethasone dose, (2) preoperative pain, or (3) postoperative
pain was associated with the odds of postoperative blood glucose levels
N200 mg/dl. There was no evidence of an association between total
dexamethasone dose (P = 0.763), preoperative pain (P = 0.502), or
postoperative pain (0.406) and the odds of postoperative blood glucose
levels N 200 mg/dl.
3.5. Postoperative outcomes
Multiple regression analysis was used to measure the association be-
tween dexamethasone administration and postoperative outcomes.
NRS pain scores, which were on a scale of 010 and prospectively doc-
umented by nurses while patients were at rest, were collected and aver-
aged for POD 0, POD 1, and POD 2. On POD 0, patients who received
perioperative dexamethasone had lower pain scores (median [1st quar-
tile, 3rd quartile]; 1 [0, 2]) than those who did not receive dexametha-
sone (2 [0, 3]; P b 0.001). Pain scores for dexamethasone-treated

Table 3
Postoperative outcomes adjusted analysesa.

Dexamethasone

Outcomes Yes (N = 474) No (N = 146) Odds ratio or difference in means (95% CI) P value

Numerical rating scale painb; median (Q1, Q3)

POD 0 1 (0, 2) 2 (0, 3) 0.7 (1.0, 0.3) b0.001
POD 1 2 (1, 3) 2 (1, 3) 0.3 (0.5, 0.1) 0.015
POD 2 3 (2, 4) 3 (2, 4) 0.1 (0.2, 0.4) 0.363
Change in white blood cell count from baselinec; median (Q1, Q3)
POD 0 0.4 (0.8, 2.1) 0.6 (1.5, 0.5) 0.8 (0.3, 1.3) 0.002
POD 1 2.6 (1.2, 4.1) 1.5 (0, 3) 1.0 (0.6, 1.4) b0.001
POD 2 1.7 (0.5, 2.9) 1.9 (0.9, 3.6) 0.5 (1.0, 0) 0.031
Postoperative nausea and vomitingd; N (%)
POD 0 94 (19.8) 31 (21.2) 0.84 (0.60, 1.17) 0.305
POD 1 109 (23) 42 (28.8)
POD 2 46 (9.7) 13 (8.9)
Delayed wound healing; N (%) 0 (0.0) 1 (0.7) 1 (0.2) 0.236
Length of stay; days; median (Q1, Q3)e 3.1 (2.9, 4) 3.2 (3, 4.2) 0.12 (0.29, 0.04) 0.142
a
Delayed wound healing and length of stay were adjusted for age, sex, body mass index, procedure (total hip arthroplasty vs. total knee arthroplasty), diabetes (yes vs. no), ASA status (3 vs. 1
or 2), intraoperative epinephrine administration (yes vs. no), postoperative antiglycemic use (yes vs. no), postoperative steroid use (yes vs. no), and postoperative antiemetic use (yes vs. no).
b
All covariates plus preoperative numerical rating scale pain and timepoint dexamethasone interaction were included in the model.
c
All covariates plus timepoint dexamethasone interaction were included in the model.
d
All covariates except for postoperative antiemetic use were included in the model.
e
Abbreviations CI: condence interval; POD: postoperative day; Q1: 1st quartile; Q3: 3rd quartile.
p b 0.05.
120 M. Nurok et al. / Journal of Clinical Anesthesia 37 (2017) 116122

patients were also lower on POD 1 (P = 0.015). On POD 2, pain scores
were similar between groups (P = 0.363) (Table 3).
The association of dexamethasone administration with other
postoperative outcomes was assessed. Dexamethasone-treated
patients had greater changes in WBC counts between baseline (pre-
operative values) and POD 01 (P = 0.002 and b 0.001, respectively),
but a smaller change in WBC count between baseline and POD 2
(P = 0.031). There was no evidence of a difference in delayed
wound healing, PONV incidence, or length of stay between groups
(Table 3). Delayed wound healing was dened as the presence of
an infection in the wound area, which would then delay the healing
process. Only one patient had delayed wound healing and required
antibiotic treatments.
dexamethasone-treated patients had signicantly decreased lengths
of stay (P = 0.042) (Table 4).
3.7. Fasting status
An analysis was performed to compare fasting status values at the
time of the rst postoperative glucose measurement. Fifteen dexameth-
asone-treated patients and 15 non-dexamethasone-treated patients
were randomly selected. Patients fasted for similar lengths of time in
both groups (mean [standard deviation]; dexamethasone-treated
group: 18.7 [2.5] h; non-dexamethasone-treated group: 17.4 [2.8] h).
4. Discussion

This study was conducted to evaluate the relationship between

3.6. Propensity score matching
Propensity score matching was performed to test the robustness
of results and reduce selection bias [18]. Patients (n = 144) who
received dexamethasone were matched to patients who did not
receive dexamethasone. Except for ethnicity, preoperative insulin
use, preoperative use of antidepressants or clonidine, and anesthetic
approach, standardized differences between matched groups were
b10% for all covariates (Fig. 2). Analysis of these matched groups
showed results that were similar to the multiple regression models
for all outcomes, except for length of stay. After matching,
dexamethasone administration and perioperative blood glucose levels
in patients undergoing THA or TKA. Our main nding was that the inci-
dence of glucose levels N200 mg/dl was relatively low, whether or not
dexamethasone was given. There was no evidence of an association be-
tween perioperative dexamethasone administration and the odds of
having postoperative glucose levels N200 mg/dl. Dexamethasone-treat-
ed patients reported lower NRS pain scores on POD 01 compared to
non-dexamethasone-treated patients. There was also no evidence of a
difference in wound healing, PONV, and length of stay between groups.
Dexamethasone is used to treat various conditions [19,20]. The use
of dexamethasone to reduce nausea, vomiting, and inammation has

Fig. 2. Absolute standardized differences in patient and surgery characteristics between groups. Comparisons of patient and surgery characteristics between dexamethasone and non-
dexamethasone groups were quantied via absolute standardized differences before and after propensity score matching. Open circles: values before matching; open triangles: values
after matching.
 M. Nurok et al. / Journal of Clinical Anesthesia 37 (2017) 116122 121

Table 4
Postoperative outcomes after propensity score matching.

Dexamethasone

Outcomes Yes (n = 144) No (n = 144) Risk or mean difference (95% CI) P value

Blood glucose N 200 mg/dl (%) 9.0 10.4 1.4% (6.1, 3.3) 0.564
Maximum blood glucose; mg/dl; median (Q1, Q3)a 131.0 (116.5, 160.4) 129.0 (114.0, 153.0) 0.5 (6.3, 7.4) 0.879
Maximum change in blood glucose; mg/dl; median (Q1, Q3) 33.5 (18.5, 59) 37.0 (22.0, 53.5) 0.5 (8.0, 7.1) 0.902
Positive incremental AUC; median (Q1, Q3) 704.4 (245.3, 1302.2) 725.6 (314.7, 1182.8) 0.98 (0.79, 1.22) 0.876
Numerical rating scale pain; median (Q1, Q3)
POD 0 1 (0, 2) 2 (0, 3) 0.6 (0.9, 0.2) b0.001
POD 1 1.5 (1, 3) 2 (1, 3) 0.3 (0.6, 0.1) 0.151
POD 2 3 (2, 4) 3 (2, 4) 0.2 (0.2, 0.5) 0.369
Change in white blood cell count from baseline; median (Q1, Q3)
POD 0 0.3 (0.8, 2) 0.6 (1.5, 0.5) 0.8 (0.3, 1.4) 0.004
POD 1 2.5 (1, 3.9) 1.6 (0, 3) 1.0 (0.4, 1.5) 0.001
POD 2 1.6 (0.3, 2.7) 1.9 (0.9, 3.7) 0.7 (1.3, 0.2) 0.008
Postoperative nausea and vomiting; N (%)
POD 0 25 (17) 31 (22) 0.74 (0.48, 1.14) 0.171
POD 1 28 (19) 41 (28)
POD 2 15 (10) 13 (9)
Length of stay; days; median (Q1, Q3) 3.2 (3, 4.2) 3.1 (2.9, 4) 0.3 (0.6, 0) 0.042
a
Abbreviations CI: condence interval; POD: postoperative day; Q1: 1st quartile; Q3: 3rd quartile; AUC: area under the curve.
p b 0.05.

become more common in the perioperative setting, as factors such as
opioid administration are associated with PONV. However, dexametha-
sone has been reported to increase blood glucose levels in various ani-
mal models and humans [36]. This blood glucose-elevating effect has
been suggested to adversely affect postoperative outcomes [7]. In our
study, only 5.6% (95% CI; 3.8, 7.5) of all patients had postoperative glu-
cose levels N200 mg/dl. After adjusting for covariates, there was no ev-
idence of greater odds of experiencing postoperative blood glucose
levels N 200 mg/dl in the dexamethasone group. Baseline blood glucose
levels of non-dexamethasone-treated patients were signicantly higher
than those of dexamethasone-treated patients, although the magnitude
of the difference was small and may not be clinically meaningful. There
was no evidence of a difference in the maximum change in blood glu-
cose between groups. Maximum glucose levels postoperatively were
also similar between groups.
A larger percentage of non-dexamethasone-administered patients
had diabetes, and signicantly higher percentages of non-diabetic pa-
tients received dexamethasone. Dexamethasone itself can induce non-
insulin-dependent diabetes mellitus in rats [21]. In another study, pa-
tients with type 2 diabetes had greater maximum blood glucose concen-
trations, although dexamethasone-induced increases in these
concentrations were comparable between patients with and without
diabetes [22]. Recently, an editorial suggested that non-diabetic patients
may be more sensitive to dexamethasone-induced hyperglycemia than
diabetic patients [7]. There did not appear to be an interaction between
dexamethasone and diabetes in our cohort; however, this could be at-
tributable to a lack of statistical power.
There are controversies surrounding the effect of dexamethasone on
blood glucose levels, and concern has been raised over the potential ef-
fect of dexamethasone-induced hyperglycemia on postoperative out-
comes [7]. In our study, delayed wound healing was only observed in
one patient who did not receive dexamethasone. This study was likely
underpowered for a denitive conclusion about such rare events.
Length of stay was also not signicantly different between groups. How-
ever, dexamethasone-administered patients reported less pain on POD
0 and POD 1. The effect of dexamethasone on reducing postoperative
pain has been corroborated by other studies [1,2]. It is important to
note that the decrease in pain reported in the current study was rela-
tively small (median of 1) and may not be clinically relevant. A recent
meta-analysis showed that dexamethasone at doses exceeding
0.1 mg/kg can effectively reduce postoperative pain and opioid con-
sumption [23]. Moreover, the incidence of PONV in our study was
slightly lower in dexamethasone-treated patients, although this was
not statistically signicant. Many factors have been shown to contribute
to PONV [24]. Although most of the patients received regional anesthe-
sia (98%), there was a relatively high rate of PONV on POD 1 (24%), and
this may be due to the postoperative use of opioids. Interestingly, great-
er changes in WBC counts were observed between baseline and POD 0
1 in dexamethasone-treated patients. Increases in WBC counts were ob-
served in dexamethasone-treated patients, and decreases in WBC
counts were observed in non-dexamethasone-treated patients. This is
consistent with literature reporting dexamethasone-induced increases
in WBC in animal models [25]. However, smaller changes in WBC counts
were observed between baseline and POD 2, and WBC counts increased
more in non-dexamethasone-treated patients. Overall, these ndings
suggest that the incidence of adverse postoperative outcomes was sim-
ilar in the dexamethasone and non-dexamethasone groups.
It is important to note that multiple perioperative mechanisms may
account for elevations in blood glucose. It cannot be excluded that the
increases in blood glucose observed in this study may have been inde-
pendent of dexamethasone administration. However, the relationship
between dexamethasone administration and postoperative blood glu-
cose levels needs to be additionally investigated in other types of
procedures.
This study has several limitations. First, the lack of randomization
due to study design resulted in variability in our patient population, es-
pecially with regard to factors that can affect postoperative glucose
levels, such as the administration of anesthetics and other medications.
Dexamethasone was administered to more patients than expected in
making power calculations, thus giving us unequal group sizes. With
our original assumptions, we ultimately had 80% power at a two-sided
alpha level of 0.05 to detect 2.87 times the odds of experiencing postop-
erative elevated blood glucose in dexamethasone-administered pa-
tients versus non-dexamethasone-administered patients. However, it
is unlikely that increasing the sample size would allow for differences
to be detected, given our small effect size (adjusted odds ratio: 0.76).
In addition, fewer ASA 3 patients received dexamethasone, thus raising
the question of a patient-selection effect for exposure to the drug in
more complex subjects. With that being said, the current study design
allowed data to be examined from the entire eligible group without con-
cern about confounding factors that may be introduced via patient re-
fusals or stricter inclusion/exclusion criteria, which are inevitable in a
randomized controlled trial. Second, glucose samples were drawn at
various time points independent of fasting status or carbohydrate in-
take. With the exception of (1) ngerstick glucose measurements that
were taken from diabetic patients in the holding area and (2) the rst
postoperative glucose measurement, there was no way to determine
whether glucose samples were drawn during fasting, postprandial, or
122 M. Nurok et al. / Journal of Clinical Anesthesia 37 (2017) 116122
non-fasting conditions. Similarly, baseline preoperative measurements
were taken at various time points under different conditions. The num-
ber of glucose measurements ranged from 2 to 30, with diabetic and
prediabetic patients having more glucose measurements. Because dia-
betic patients tended to have more glucose measurements than nondi-
abetic patients, diabetes and number of glucose measurements were
collinear (i.e., correlated) variables. Therefore, we could not assess
whether number of glucose measurements was independently associat-
ed with the odds of postoperative hyperglycemia. This wide range in
number of glucose measurements, in addition to the usage of point-
of-care devices to measure ngerstick glucose levels, introduced anoth-
er level of variability. However, 81% of all glucose measurements were
performed on venous or arterial blood samples, and patients had a me-
dian of 3 glucose measurements. Third, our institution does not permit
elective surgery if an HbA1C is N 86 mmol/mol (10.0%) and therefore,
our population has relatively well-controlled blood glucose. This may
be different from practices in other institutions and may explain why
dexamethasone administration was not associated with adverse post-
operative outcomes in our population. Patients with poorly controlled
diabetes may react differently than those with well-controlled diabetes.
Finally, differences in times and doses of dexamethasone administration
may contribute to the variability in its effects on glucose levels. Unfortu-
nately, we did not account for the dosage of dexamethasone in the nal
analysis.
Findings from this study strongly suggest that perioperative dexa-
methasone administration in our population is not associated with the
increased odds of having postoperative glucose levels N200 mg/dl and
is also not associated with higher maximum levels of glucose postoper-
atively in a population with well-controlled blood glucose at baseline.
The use of perioperative dexamethasone appears to be a safe practice
in this setting. Future prospective studies will need to be performed to
closely examine the effects of perioperative dexamethasone administra-
tion on blood glucose levels in a more heterogeneous population.
Funding
This work was supported by the Research and Education Fund of the
Department of Anesthesiology at the Hospital for Special Surgery and
the National Center for Advancing Translational Science of the National
Institutes of Health (UL1TR000457).
Conicts of interest
The authors report no conicts of interest.
Acknowledgements
The authors thank Sean Ponzo for his help with data collection.
References
[1] Backes JR, Bentley JC, Politi JR, Chambers BT. Dexamethasone reduces length of
hospitalization and improves postoperative pain and nausea after total joint
arthroplasty: a prospective, randomized controlled trial. J Arthroplasty 2013;28(8
Suppl):117.
[2] Koh IJ, Chang CB, Lee JH, Jeon YT, Kim TK. Preemptive low-dose dexamethasone re-
duces postoperative emesis and pain after TKA: a randomized controlled study. Clin
Orthop Relat Res 2013;471(9):301020.
[3] Mokuda O, Sakamoto Y, Ikeda T, Mashiba H. Sensitivity and responsiveness of glu-
cose output to insulin in isolated perfused liver from dexamethasone-treated rats.
Horm Metab Res 1991;23(2):535.
[4] Gill AM, Leiter EH, Powell JG, Chapman HD, Yen TT. Dexamethasone-induced hyper-
glycemia in obese Avy/a (viable yellow) female mice entails preferential induction
of a hepatic estrogen sulfotransferase. Diabetes 1994;43(8):9991004.
[5] Nazar CE, Lacassie HJ, Lopez RA, Munoz HR. Dexamethasone for postoperative nau-
sea and vomiting prophylaxis: effect on glycaemia in obese patients with impaired
glucose tolerance. Eur J Anaesthesiol 2009;26(4):31821.
[6] Waldron NH, Jones CA, Gan TJ, Allen TK, Habib AS. Impact of perioperative dexa-
methasone on postoperative analgesia and side-effects: systematic review and
meta-analysis. Br J Anaesth 2013;110(2):191200.
[7] Dhatariya II K. Does dexamethasone-induced hyperglycaemia contribute to postop-
erative morbidity and mortality? Br J Anaesth 2013;110(5):6745.
[8] Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG. Research electronic
data capture (REDCap)a metadata-driven methodology and workow process
for providing translational research informatics support. J Biomed Inform 2009;
42(2):37781.
[9] Vulcano E, Lee YY, Yamany T, Lyman S, Valle AG. Obese patients undergoing total
knee arthroplasty have distinct preoperative characteristics: an institutional study
of 4718 patients. J Arthroplasty 2013;28(7):11259.
[10] Bersoux S, Cook CB, Kongable GL, Shu J, Zito DR. Benchmarking glycemic control in
U.S. hospitals. Endocr Pract 2014;20(9):87683.
[11] Desimone ME, Blank GE, Virji M, Donihi A, DiNardo M, Simak DM, et al. Effect of an
educational Inpatient Diabetes Management Program on medical resident knowl-
edge and measures of glycemic control: a randomized controlled trial. Endocr
Pract 2012;18(2):23849.
[12] Abdelmalak BB, Bonilla AM, Yang D, Chowdary HT, Gottlieb A, Lyden SP, et al. The
hyperglycemic response to major noncardiac surgery and the added effect of steroid
administration in patients with and without diabetes. Anesth Analg 2013;116(5):
111622.
[13] Huber PJ. Robust estimation of a location parameter; 1964 73101.
[14] Ma Y, Mazumdar M, Memtsoudis SG. Beyond repeated-measures analysis of vari-
ance: advanced statistical methods for the analysis of longitudinal data in anesthesia
research. Reg Anesth Pain Med 2012;37(1):99105.
[15] Zeger SL, Liang KY, Albert PS. Models for longitudinal data: a generalized estimating
equation approach. Biometrics 1988;44(4):104960.
[16] Austin PC. Balance diagnostics for comparing the distribution of baseline covariates
between treatment groups in propensity-score matched samples. Stat Med 2009;
28(25):3083107.
[17] Allison DB, Paultre F, Maggio C, Mezzitis N, Pi-Sunyer FX. The use of areas under
curves in diabetes research. Diabetes Care 1995;18(2):24550.
[18] Austin PC. An introduction to propensity score methods for reducing the effects of
confounding in observational studies. Multivar Behav Res 2011;46(3):399424.
[19] Borchorst S, Moller K. The role of dexamethasone in the treatment of bacterial men-
ingitis - a systematic review. Acta Anaesthesiol Scand 2012;56(10):121021.
[20] Fiel SB, Vincken W. Systemic corticosteroid therapy for acute asthma exacerbations. J
Asthma 2006;43(5):32131.
[21] Ohneda M, Johnson JH, Inman LR, Unger RH. GLUT-2 function in glucose-unrespon-
sive beta cells of dexamethasone-induced diabetes in rats. J Clin Invest 1993;92(4):
19506.
[22] Hans P, Vanthuyne A, Dewandre PY, Brichant JF, Bonhomme V. Blood glucose con-
centration prole after 10 mg dexamethasone in non-diabetic and type 2 diabetic
patients undergoing abdominal surgery. Br J Anaesth 2006;97(2):16470.
[23] De Oliveira Jr GS, Almeida MD, Benzon HT, McCarthy RJ. Perioperative single dose
systemic dexamethasone for postoperative pain: a meta-analysis of randomized
controlled trials. Anesthesiology 2011;115(3):57588.
[24] Gan TJ. Risk factors for postoperative nausea and vomiting. Anesth Analg 2006;
102(6):188498.
[25] Nakagawa M, Terashima T, D'Yachkova Y, Bondy GP, Hogg JC, van Eeden SF.
Glucocorticoid-induced granulocytosis: contribution of marrow release and
demargination of intravascular granulocytes. Circulation 1998;98(21):230713.