2. PATIENTS AND METHODS 1 Patients
2.2 STUDY METHODS
The study methods used throughout this thesis are described below. Variations to these general methods are described in the relevant chapters
(a) ANGIOPLASTY PROCEDURE
Angioplasty was carried out according to methods described by Griintzig (Gruentzig et al, 1979) and subsequently modified throughout the series as new techniques and equipment were introduced. Initially, patients were treated with non-steerable balloon catheters until 1983 when steerable guidewire systems were used in all patients thereafter (Simpson et al, 1982). Throughout the study period, adaptations were made to accommodate technical advances. Several operators with varying degrees of experience contributed cases to the study population, and the dilatation strategy differed depending on the operator.
Prior to the procedure, written, informed consent to coronary angioplasty and emergency coronary bypass surgery was sought and gained. Patients were fasted overnight and received oral aspirin and vasodilators. Most of the procedures were performed via the femoral route whereby arterial access was gained using the Seldinger technique (Seldinger, 1953) followed by insertion of an arterial sheath. The brachial approach by antecubital cutdown and arteriotomy was used when femoral arterial puncture was contraindicated, usually as a result of peripheral vascular disease. Prior to the procedure, patients were premedicated with intravenous diazepam and at the start of the procedure, 10,000 units of intra-arterial heparin were administered. Subsequent boluses were given to maintain an activated clotting time of >300 seconds, or if the procedure was prolonged. Apparatus
produced by several different manufacturers was used, and the choice of equipment differed depending on operator preference. Balloon size was selected to approximate the diameter of the adjacent normal segment of the coronary artery.
Once the guiding catheter was engaged into the coronary ostium, single or multiple radiographic projections of the coronary arteriogram were obtained to delineate coronary morphology. A high quality video tape with freeze frame facilities, or a digital recording system allowed instant recall of the images recorded during the procedure and provided a 'road map' for operators to work with. The dilatation catheter was primed with radiographic contrast medium diluted 50% with heparinised saline solution. Inflations were administered until contrast injections showed wide dilatation of the stenosis or, in patients in whom the trans-stenotic gradient was measured, when the gradient had been effectively eliminated. After removal of the dilatation apparatus, a final coronary arteriogram was recorded. After angioplasty, the femoral sheath was left in place for 4 to 6 hours, and an intravenous infusion of heparin (1,000 units per hour) and nitrate (2-10 mg per hour) continued for 24 hours. Although drug therapy was not standardised, patients were usually discharged on oral nifedipine and a long acting nitrate for the first 4 to 6 weeks, and oral aspirin indefinitely.
In the event of an abrupt vessel closure occurring during or early after the angioplasty procedure, immediate redilatation was usually attempted. If this failed, or if evidence of severe coronary dissection was evident, patients with clinical or electrocardiographic evidence of myocardial ischaemia, or evolving myocardial infarction, were referred for emergency coronary artery bypass surgery. Recently, prolonged inflations using autoperfusion balloons, and coronary stents, have also been used successfully to treat abrupt vessel occlusions and coronary dissections.
Procedure-related myocardial infarction was diagnosed when new pathological Q- waves appeared on a predischzirge electrocardiogram.
(b) DATA COLLECTION
All angiographic details were assessed by 2 independent observers without knowledge of procedural outcome. If there was any disagreement, the opinion of a third cardiologist was obtained. Hand-held callipers were used for making quantitative measurements from the projected angiographic film using the guiding catheter for magnification scaling. A protractor was used for measuring lesion angulation.
Where relevant, lesion complexity was scored as type A, B, or C according to the guidelines published by the American College of Cardiology/American Heart Association Task Force (Ryan et al, 1988). Type B and C lesions were further subcategorised as suggested by Ellis et al (1990a) and Myler et al (1992) (Table
1.3).
Baseline and procedural data were recorded prospectively on a computerised database. Follow-up information was obtained at routine clinic visits, from the referring physicians, and by telephone interview with the patients. Clinical status, employment status, medication and the occurrence of new cardiac events (defined as death, myocardial infarction, coronary artery bypass grafting or coronary angioplasty) were recorded up to the date of census. Angina was graded using the Canadian Cardiovascular Society functional classification (Campeau, 1976) (Table 2.1). In all cases of death the cause was established and classified as cardiac or non cardiac. The cause of in-hospital deaths was established by review of the hospital records, death certificate, and post-mortem report when available. For out of
hospital deaths the cause of death was established by direct contact with the general practitioner, notes obtained from the local family practitioner committee, and by review of a post-mortem report if available. When the cause of death could not be established from these sources, copies of the death certificate were obtained from the office of Population Censuses and Surveys. A diagnosis of new myocardial infarction was made during follow-up when new pathological Q-waves appeared on a follow-up electrocardiogram, or when a documented clinical event was associated with other serial electrocardiographic changes and a 2-fold increase in the serum level of at least 1 cardiac enzyme (creatinine phosphokinase, aspartate transaminase, or lactate dehydrogenase).
Patients were followed closely after angioplasty by functional testing for the development of symptoms or signs of myocardial ischaemia. The indication for follow-up coronary angiography was strictly clinical: recurrence of angina pectoris or a positive symptom-limited treadmill exercise test performed routinely during out-patient visits.
(c) DEFINITIONS
The following definitions were used.
Complete revascularisation. Clinically successful angioplasty with no lesions greater than 50% diameter stenosis in any major epicardial vessel.
Epicardial vessel. Either the left anterior descending artery and its large diagonal branches, or the left circumflex artery and its large obtuse marginal branches, or a balanced or dominant right coronary artery.
Left ventricular dysfunction. Left ventricular function was categorised as abnormal if the ejection fraction was estimated to be <45% , assessed visually from a contrast ventriculogram in the right anterior oblique projection.
Significant stenosis. A 50% or greater diameter stenosis in at least 2 radiographic views, or 70% diameter stenosis in 1 view.
Single vessel disease. A significant stenosis or stenoses in a single major epicardial vessel or its large branches. Multivessel disease. Involvement of at least 2 major epicardial vessels or their large branches.
Unstable angina pectoris. Angina occurring at rest requiring intravenous medical therapy, and included patients with post-infarction angina pectoris.
(d) STATISTICAL METHODS
All baseline, angiographic, and procedural data were collected and entered into a microcomputer. The Microsoft Excel software was used for storing and organising data in the form of a spreadsheet. Data analysis was performed by exporting data for analysis by external programmes.
Simple Statistical Methods
The standard statistical descriptive terms of mean, median, standard deviation, standard error, and interquartile range are used throughout this thesis in describing most results. All categorical variables were analysed using the Chi-squared test (using Yates' correction for continuity), Fisher's exact test, or the Chi-squared test for linear trend. Continuous variables were compared using the unpaired t test, the
Kruskal-Wallis test, or the Mann-Whitney U-test. A p value of <0.05 was considered statistically significant.
Multiple Logistic Regression Analysis
The statistical technique of multiple regression was used to analyse data in chapters 3, 8 and 9. This technique allows the development of a model that uses the combination of the values of a group of explanatory variables to predict the probability of an individual having the outcome of interest. Because the outcome of interest is a binary variable, transformation of this probability is needed to avoid prediction of impossible probabilities outside the range 0 to 1. The transformation used is the logit transformation, written logit(p), and the approach used is called multiple logistic regression. If p is the probability of a patient having a particular outcome of interest, then 1-p is the probability that they do not have that outcome, and the ratio p /(l-p ) is the odds. The log odds is therefore:
logit(p) = Ioge(p/l-p).
If y = logit(p),
where y is a linear function based on the presence or absence of certain characteristics in a given patient, then
ey = p/l-p , and thus p = ey/(i+ey).
Single variable analyses and the backwards stepwise selection procedure were used to select variables that independently predict the outcome of interest significant at the 1 % level. First, all variables that were significantly related to the outcome by
univariate analysis were entered into the model. Unimportant variables were then removed 1 at a time until each of the remaining variables in the model contributed significantly and independently to the final model. At each step, the variable with the smallest contribution to the model was removed as long as that p value was greater than the chosen level. The value of y can simply be obtained by adding the regression coefficients corresponding to all those characteristics that are present in the patient and the regression coefficient of the constant.
Fisher's Stepwise Discriminant Analysis
The assignment technique of Fisher's stepwise discriminant analysis was used to analyse data in chapter 10 for producing classification functions to allow allocation of new patients to 1 of 4 possible outcome groups. The forwards stepwise selection procedure was used to select variables that independently predicted the outcome of interest significant at the 1% level. With this method, the simple relation between each potential explanatory variable and the outcome variable of interest is examined ignoring all the other variables. The single variable that has the strongest association with the dependent variable is found and entered into the model. The variable among those not in the model, when added to the model so far obtained explained the largest amount of the remaining variability, is then identified and entered into the model. This is repeated until the addition of an extra variable is not statistically significant at the chosen level. To predict which group a patient belongs to, one calculates the classification function for each of the 4 groups, and the largest of the 4 values will indicate the patient's most likely group.
Jacknife or "Leave-one-out" Method
The performance of the derived classification rules on the data were assessed by the jacknife method to compensate for the fact that the same data were being used to
test the prediction rule as had been used to derive it. Also called the 'leaving one out method', the classification rule is derived on the basis of (n-1) individuals and used to classify the individual not included, the whole process being repeated for each individual,
Kaplan-Meier Method and Logrank Test
Cumulative overall and event-free survival probabilities, and the corresponding standard errors and 95 % confidence intervals were analysed using the Kaplan-Meier method, computed from the time of coronary angioplasty (Kaplan and Meier, 1958; Peto et al, 1977; Machin and Gardner, 1988). Using this analysis, the proportion surviving a given length of time is calculated by multiplying the probabilities of surviving each successive time period up to that time. Only the times on which there is an event need to be considered. Hence, if Pj^ is the probability of surviving K days, is the number of subjects still at risk (ie still being followed up) immediately before the Kth day, and Fjr is the number of observed events on day K, then
= Pfc-1 ^ (^k - ^k/^k)
The number of patients still at risk, R]^, is changed both when patients suffer an event, and when patients reach the end of their follow-up period (ie when they are censored). The cardiac end points that were analysed included death, cardiac death, non-fatal myocardial infarction, coronary artery bypass surgery, and repeat coronary angioplasty. The survival times between different independent subgroups were compared with the logrank test. This is a non-parametric method for testing the null hypothesis that the groups being compared are samples from the same population as regards survival experience. The principle of the test is to divide the
survival time scale into intervals according to the distinct observed survival times, ignoring censored survival times. The logrank test produces for each group an observed (O) and an expected (E) number of events. These are then compared by obtaining the value of X^, calculated by the sum of
(0-E )2/E ,
and comparing this value to a x distribution to obtain the probability value.
Cox Multiple Proportional Hazards Regression Analysis
The survival analysis in chapter 7 used the Cox multiple proportional hazards regression analysis to identify which prognostic factors independently influenced long-term survival and the time to cardiac events during follow-up; prognostic factors were again selected using the backwards stepwise procedure (Cox, 1972, Christensen, 1987). This is a semi-parametric approach which allows the effects of several variables on survival to be investigated at the same time. Although no particular type of distribution is assumed for survival times, a strong assumption is made that the effects of the different variables on survival are constant over time and are additive in a particular scale. Although the overall hazard may change over time, changes in the hazard of any individual must be proportional to changes in the hazard of any other individual, and to changes in the underlying hazard (ho). For an individual with several independent variables of interest, say X \ to Xp, the instantaneous hazard of an event at time t, h(t), can be expressed as
where hg is the baseline hazard function (corresponding to the hazard when all the variables are zero), and b% to bp are regression coefficients. The instantaneous hazard gives the risk of an event at time t. Hence, the cumulative hazard, H(t), of an event between time 0 and time t can be obtained by adding up all the hazards up to time t, and is defined as
H(t) = Hg(t) X exp (biX i + b%X2 + . . . bpXp),
where Hg(t) is the cumulative underlying hazard function. The survival probability for any individual with specific values of the variables in the model can thus be estimated.
Software Packages
These statistical analyses were performed using commercially available statistical software programmes (SPSS for windows, BMDP programs IL and 2L, BMDP program LR).
Table 2.1. Angina Grade According to the Canadian Cardiovascular Society
Functional Classification.
Grade I. Angina occurs during strenuous or rapid or prolonged exertion at work or recreation. Ordinary physical activity, such as walking or climbing stairs, does not cause angina.
Grade
n.
Slight limitation of ordinary physical activities. Angina may be precipitated by walking or climbing stairs rapidly, walking uphill or stair climbing after meals, or in cold, or in wind, or other emotional stress, or only during the few hours after awakening.Grade IQ. Marked limitation of ordinary physical activity. Angina may be precipitated by walking on the level, or by climbing 1 flight of stairs at a normal pace, and in normal conditions.
Grade IV. Inability to carry out any physical activity without discomfort. Angina may occur when patient is at rest.
3. CORONARY ANGIOPLASTY OF CHRONIC TOTAL OCCLUSION