CARDIOVASCULAR JOURNAL OF AFRICA • Vol 21, No 2, March/April 2010
98
AFRICA
Demographic data associated with peri-operative cardiac risk,
5
and physiological
6
and surgical procedural data
7
associated with
in-hospital mortality were extracted from the hospital computer-
ised database. The following clinical risk factors are included in
the database: history of ischaemic heart disease (or pathological
Q waves on ECG), history of congestive heart failure, diabetes,
serum creatinine
>
180
µ
mol.l
-1
, age, gender, pre-operative and
postoperative haemoglobin and blood glucose levels, and a
history of hypertension. The vascular surgical procedural factors
included the type of surgery, duration of surgical operating time,
and whether surgery was undertaken out of elective surgical
working hours. Therefore this study includes both elective and
emergency surgical patients.
The physiological data included the mean daily heart rate
(HR), systolic blood pressure (SBP) and diastolic blood pressure
(DBP) calculated from recorded nurses’ observations during
the hospital admission. A history of pre-operative chronic beta-
blocker therapy and in-hospital drug administration was also
recorded in the database. It was therefore possible to identify
patients who did not have their beta-blocker therapy prescribed
during their hospital admission.
The cause of in-hospital deaths reported in the database were
determined using the stringent criteria of a previous study.
8
Cardiac deaths were defined as a postoperative cardiac event
being the primary event, which subsequently resulted in death.
These cardiac events included cardiac arrest, myocardial infarc-
tion or cardiac failure. A primary cardiac arrest was defined as a
witnessed cardiac arrest associated with ventricular fibrillation,
ventricular tachycardia or asystole in a patient who was previous-
ly considered stable. Myocardial infarction was defined as post-
operative clinical signs or symptoms consistent with myocardial
ischaemia and either an associated diagnostic rise in troponin T
or creatinine kinase-MB, or electrocardiographic (ECG) changes
consistent with acute myocardial infarction. Primary postopera-
tive cardiac failure was defined as clinical signs and symptoms
consistent with acute pulmonary oedema requiring inotropic
support without an obvious precipitant.
All patients in the database in whom chronic beta-blockade
was not administered on any of the first three postoperative
days were identified. Those patients who died during the hospi-
tal admission were identified as ‘cases’. These patients were
matched in two separate case–control studies: the first matched
‘cases’ with patients who were withdrawn from chronic beta-
blockade but survived, and the second matched ‘cases’ with
patients who were continued on chronic beta-blockade in the
peri-operative period and survived.
To identify matching patients, all the patients were strati-
fied according to age. The matched controls for the cases that
died were the two patients nearest in age to the case, who
were matched for the presence or absence of the following risk
factors (serum creatinine
>
180
µ
mol.l
-1
, surgery out of hours,
and a mean daily postoperative SBP of
>
179 or
<
90 mmHg).
These three risk factors were previously identified as independ-
ent predictors of all-cause in-hospital mortality in our vascular
surgical patients.
4
These three variables were chosen as two of
them were the strongest predictors of mortality in the postopera-
tive model (serum creatinine and surgery out of hours) and the
third variable was known to interact with postoperative heart rate
(SBP
>
179 or
<
90 mmHg).
4
The reason for withdrawal of chronic beta-blockade was iden-
tified by analysis of case notes. Attention to beta-blocker-associ-
ated side effects including bradycardia, hypotension, and periop-
erative complications requiring inotropic support were analysed.
If there was no obvious clinical indication for withdrawal, then
the reason for withdrawal was classified as ‘unknown’.
Statistical analysis
All demographic and possible risk factors were summarised by
group (case versus control) using descriptive statistics. Mean
[standard deviation (SD)] or median (interquartile range) was
used where appropriate for continuous variables and frequen-
cy (proportion) for categorical variables. Conditional logistic
regression was used to compare case and control groups for each
variable and to take into account matching in the comparison.
Categorical data of mortality associated with chronic beta-
blockade and withdrawal of beta-blockade were analysed using
the Fisher’s exact test.
Three risk factors possibly associated with mortality follow-
ing withdrawal of chronic beta-blockade were examined using
conditional logistic regression. Firstly, whether complete with-
drawal of beta-blockers for the first three days was associated
with mortality when compared with withdrawal for only a single
day. Secondly, whether the administration of postoperative
inotropes was associated with mortality. Thirdly, whether the
change in mean daily heart rate from the day of surgery to the
third postoperative day was associated with mortality. Where
patients died or were discharged before the third postoperative
day, mean daily heart rate for the last day in hospital was used.
If the change in mean daily heart was found to an important
univariate predictor, then a dichotomous heart rate variable was
to be derived from constructing a receiver operating characteris-
tic (ROC) curve and identifying the optimal cut-off point for the
change in mean daily heart rate from the day of surgery to the
‘last mean daily heart rate’.
The variables with the strongest association with postopera-
tive mortality following chronic beta-blocker withdrawal were
entered into a multivariate conditional logistic regression. A
p-
value less than the Bonferroni adjusted
α
(0.017
=
0.05/number
of possible risk factors) indicated statistical significance.
An independent samples
t
-test was conducted to compare
heart rates between cases and controls as the heart rate was
normally distributed.
P
-value, OR and CI are presented for all
comparisons. SPSS 15.0 for Windows (6 Sept 2006) was used
for data analysis.
Results
Out of 829 vascular surgical patients, 195 patients on chronic
beta-blockade were identified (23.5%). Both chronic beta-
blockade and withdrawal of chronic beta-blockade were associ-
ated with significantly higher in-hospital mortality rates (Tables
1 and 2). Fifty per cent of the patients who were withdrawn from
chronic beta-blockade were classified as having primary cardiac
deaths.
Of the 21 patients who were withdrawn from chronic beta-
blockade and died, seven were excluded from the matched case–
control studies, as four had missing haemodynamic data, and
for three cases it was impossible to identify adequate controls
for matching. The remaining fourteen cases were adequately
matched with two sets of 28 controls. For both the case–control
