A model for predicting angiographically normal coronary arteries in survivors of out-of-hospital cardiac arrest

Ethics statement

The study protocol was reviewed and approved by the ethics committee of Mito Kyodo
General Hospital, University of Tsukuba Hospital Mito Medical Center. The ethics committee
at our institution does not require its approval for observational studies using anonymous
data in existence such as this study. Also, informed consent from each patient was
waived for using anonymous data according to the informed consent guidelines in Japan.

A retrospective, observational, cohort study was conducted on consecutive adult patients
(age ?18 years) who were survivors of OHCA and who received immediate CAG. These patients
all presented to the emergency department (ED) of an urban teaching hospital in Japan,
between June 1, 2006 and March 31, 2011. The ED of St Luke’s International Hospital,
Tokyo, provides primary to tertiary care to a population of approximately 100,000.
The management of OHCA involves the Tokyo Fire Department (TFD) and the EDs of other
hospitals in Tokyo. Typically, the closest emergency medical technicians (EMTs) are
dispatched to the scene. Cardiopulmonary resuscitation (CPR) is initiated by EMTs
at arrival and continued according to the AHA standards. A 12-lead ECG is performed
in the ED immediately after return of spontaneous circulation (ROSC).

ROSC patients were brought directly from the ED to the cardiac catheterization laboratory.
CAG was performed according to a standard technique. Experienced cardiologists made
the decision to proceed to angioplasty only for critical lesions. Standard resuscitation
and stabilization were used during and after the procedure.

Data collection

The data were retrospectively obtained from computerized medical records and collected
in the Utstein style. Variables included age, gender, height, weight, risk factors
of ACS including hypertension (HT), hyperlipidemia (HL), diabetes mellitus (DM), history
of ACS, PCI, coronary artery bypass graft (CABG), heart failure, arrhythmia, chest
pain before arrest, witnessed collapsed patient, bystander initiated CPR, ventricular
tachycardia/ventricular fibrillation (VT/VF) on EMT arrival, estimated time of initiation
of CPR, and estimated time of cardiac arrest (interval until ROSC). The primary outcomes
were the CAG findings, (including normal coronary artery or not). Secondary outcomes
included PCI results intra-aortic balloon pumping (IABP), venoarterial-extracorporeal
membrane oxygenation (VA-ECMO), and therapeutic mild hypothermia. Normal coronary
arteries were defined as ‘no stenosis of any coronary arteries’ by experienced cardiologists’
reading.

Patients’ prognoses 1 month after admission were categorized by the Glasgow-Pittsburgh
cerebral performance category (GP-CPC) scale, from category 1 (good cerebral performance),
category 2 (moderate cerebral disability), category 3 (severe cerebral disability),
category 4 (coma or vegetative state), to category 5 (death). We also obtained data
on survival 1 month after resuscitation.

All ECG were recorded just after ROSC at the ED. ECGs were interpreted by two experienced
cardiologists who were unaware of the patients’ angiographic status. Disagreement
between the two experienced cardiologists was arbitrated by an independent third party.
ECG findings recorded included the following: heart rate (HR), axis deviation, atrial
fibrillation (AF) or atrial flutter (AFL), junctional rhythm, presence or absence
of P wave, abnormal shape of P wave, abnormal PQ interval, prolonged PQ interval,
any abnormal QRS shape, wide QRS, right bundle branch block (RBBB), left bundle branch
block (LBBB) including left anterior hemiblock (LAH) or left posterior hemiblock (LPH),
any bundle branch block (BBB), bifascicular block, presence or absence of Q wave,
any abnormal ST segment change, ST segment elevation, ST segment depression without
reciprocal change, any ST segment depression, prolonged QT interval, presence or absence
of inverted T wave (negative T wave, coronary T wave, or flat T wave), and abnormal
U wave. Abnormal ST depression is defined as ST depression of 1 mm (0.1 mV) measured
at 80 ms after the J point in at least two contiguous leads. Abnormal ST elevation
is defined as ST elevation at the J point in at least two contiguous leads of 2 mm
(0.2 mV) in men or 1.5 mm (0.15 mV) in women in leads V2–V3 and/or of 1 mm (0.1 mV)
in other contiguous chest leads or the limb leads. They were also asked to comment
on whether patients might have ACS based on the ECG findings.

Selection of participants

During the study period, there were 1390 patients with OHCA, of whom 472 patients
(34 %) were admitted to our hospital. Angiography was performed on 49 patients. Criteria
for CAG were as follows: probable favorable neurological outcome and no obvious non-cardiac
causes for the arrest, such as hyperkalemia, intoxication, or trauma. Two CAG patients
were excluded because of missing ECG data, leaving 47 patient cases analyzed. The
reason why we decided the patient would have probable favorable neurological outcome
was that most participants had witnessed cardiac arrest or collapse, bystander initiated
CPR, and a relatively short duration prior to ROSC.

Statistical analysis

The primary outcome was ‘normal coronary arteries’ according to CAG evaluation. Predictor
variables were assessed by univariate analysis with Fisher’s exact test for categorical
variables and t test for continuous variables. A two-tailed p value of less than 0.05 was considered statistically significant. We also analyzed
the inter-rater reliability of the cardiologist’s assessment of ECGs using the kappa
coefficient. Based on univariate analysis, these five variables were chosen as the
predictor (independent) variables: age, history of ACS, history of arrhythmia, history
of DM, or any abnormal ST segment change on the ECG. All patients 50 years old and
younger had normal coronary arteries. Although age was one of the strongest predictors
of normal coronaries, it is difficult to generalize this predictor to other populations
because Japanese under 50 are at relatively low risk of ACS compared with people in
other countries, so it was not used as a generalizable variable.

A multiple logistic regression model could not identify significant factors of the
ECG findings because all patients with no abnormal ST change on ECG had cardiac arrest
with normal coronary arteries. Thus, a recursive partitioning model was fit by analyzing
the relationship between cardiac arrest with normal coronary arteries and the chosen
four predictors. The recursive partitioning was conducted by maximizing the entropy
index. Validation is the process of using 90 % of this data set to estimate model
parameters and using the 10 % part to assess the predictive ability of the model using
k-fold cross-validation. Sensitivity analysis of the model was performed using the
area under the curve (AUC). Inter-rater reliability was analyzed by STATA version
11.2 (StataCorp, Texas). The other analyses were performed with JMP version 9.0.3
(SAS Institute, Cary, NC).