APPENDICES: POSTED AS SUPPLIED BY AUTHOR

Appendix A

Definitions provided automatically by "help text" on electronic Case Record Form

For the purposes of recording ECG abnormalities:

ST depression = depression of the ST segment > 0.5 mm or > 0.05 mv below the isoelectric line

T wave inversion = inversion of the of the T wave of the QRS complex except in leads AVR, V1 and lead III.

Pathological Q waves =Q waves (except in AVR, V1 and III) which are >0.04 sec wide and 0.2mv deep or >20% of the accompanying R wave.

For clinical examination

Clinical signs of heart failure = any or all of the following: elevated venous pressure, hepatomegaly, dependent oedema, crepitations, gallop rhythm, or evidence of cardiac enlargement

For comorbid conditions and risk factors

Hepatic disease = Chronic hepatitis or cirrhosis, or other hepatic disease causing elevation of transaminases more than 3 times the upper limit of normal.

Peripheral Vascular Disease = Claudication either at rest or exertion, amputation for arterial vascular insufficiency, vascular surgery (reconstruction or bypass) or angioplasty to the extremities, documented aortic aneurysm, or noninvasive evidence of impaired arterial flow.

Chronic Renal Failure = Chronic dialysis or renal transplantation or serum creatinine greater than 200 �mol/l.

Chronic Respiratory Disease= Diagnosis previously made by physician, or patient receiving bronchodilators, or values of FEV1<75�%, arterial pO2<60%, or arterial pCO2>50% in prior studies.

Chronic Inflammatory Conditions = diagnosis of rheumatoid arthritis, systemic lupus erythematosis or other connective tissue diseases, polymyalgia rheumatica etc.

Peptic Ulcer Disease = Active peptic ulcer disease (PUD) or on treatment for PUD Malignancy = Diagnosis within 1 year or active malignancy

Hypertension = Treated hypertension, on anti-hypertensive therapy.

Hyperlipidaemia = Treated hyperlipidaemia, on lipid lowering medication, or specific dietary modification.

Diabetes = Treated diabetes, on insulin or oral diabetic medication, or specific dietary modification.

Family history of premature coronary artery disease (CAD) = A history of angina pectoris, myocardial infarction or sudden death among first-degree relatives under the age of 55 years for male relatives and 65 years for female relatives.

For clinical endpoints

Cardiac Mortality = Death due to myocardial infarction, or complications thereof, cardiac arrhythmia, sudden death, or congestive heart failure.

Cardiovascular mortality =Death due to cardiac causes (including myocardial infarction, cardiac arrhythmia, sudden death and congestive heart failure), cerebrovascular accidents or other vascular causes including bleeding and thrombo-embolic events

Non Fatal Myocardial infarction = Hospitalisation for myocardial infarction where myocardial infarction must involve two of the following:

Cardiac chest pain at rest lasting >20 minutes, or pulmonary oedema without significant valvular heart disease or known heart failure, or shock, without hypovolaemia or intoxication.

Transient elevation of CPK to twice the upper limit of normal for the laboratory, or CK MB to above the upper limit of normal for the laboratory, or elevation of troponin I or T above the 99^th centile of normal in the laboratory.

ECG series with the evolutionary changes of myocardial infarction. Development and disappearance of localised ST segment elevation, combined with the development of T wave inversion in at least two contiguous leads, and/or the development of pathological Q waves.

NB: If markers of myocardial damage are present the ECG changes could include ST depression, T wave inversion, Loss of R wave progression or new LBBB.

Unstable angina = Acute hospital admission for cardiac chest pain (CCS IV, either pain at rest, or rapidly progressive increase in symptoms) without biochemical marker elevation indicating myocardial infarction.

Hospitalisation for Heart Failure = Hospital admission involving a stay of at least 24 hours for increased severity of heart failure symptoms or signs necessitating either intravenous diuretics, an increase of existing oral diuretics by at least 50%, or commencement on inotropic support.

Cardiovascular hospitalisation = Admission to hospital involving a stay of at least 24 hours for worsening heart failure, acute coronary syndromes, stroke and thrombo-embolic events or other vascular causes

Diabetes = Treated diabetes, on insulin or oral diabetic medication, or specific dietary modification.

Family history of premature coronary artery disease (CAD) = A history of angina pectoris, myocardial infarction or sudden death among first-degree relatives under the age of 55 years for male relatives and 65 years for female relatives.

Appendix B

To develop the clinical risk score, stepwise regression was performed to select the variables most predictive of death or nonfatal MI when the results of investigation were coded as positive or negative rather than including "not performed" as a variable. Forward and backward selection were performed to assess the stability of the selection and the final model fitted to all patients without missing values for the selected variables. A positive (versus negative or inconclusive) result on non-invasive testing was not a significant independent predictor of adverse outcome and was not selected into the model.

A model was developed on a randomly selected 75% of the study patients (the `training set') and was then assessed in the remaining 25% (the `validation set'). The number of subjects included in the final model to create the score was 1896, with 66 events. The variables selected in the model were comorbidity, diabetes, symptom severity, symptom duration, abnormal LV function and ST changes on the resting ECG. The variables selected in the training set were identical to those selected when the same procedure was applied to the entire population.

The Cox regression coefficients of the selected model were then used to define a scoring system that can be used to predict the probability of death or non-fatal MI during the first year after presentation. The Cox proportional hazards model assumes that the event hazard (or rate) at time t, h_i(t), for a subject i with covariates x_1i , x_2i ,�, x_Ki , is:

h_i(t) = h_o(t) exp(β₁* x_1i + β₂* x_2i +�..+ β_K* x_Ki))

In applying these results to our data we have simplified the calculations of the scores by multiplying by 100 each "beta" in the equation above.The combination of all the predictors included in the model weighted by their estimated regression coefficients, (β₁, β₂, etc) is known as the score, i.e.

score_i = 100 (β₁* x_1i + β₂* x_2i + β₃* x_3i +�.. )

The probability that the event of interest occurs within time t (for example 1 year) for patient i can then be calculated as

Prob_i(event at time t)= 1 � Prob_i (surviving up to time t)= 1 � S_i(t)

Where S_i(t)= S₀(t) ^{exp(scorei/100)}

and S₀(t) is the survival probability at time t for somebody with score=0 which is estimated from Cox regression model (likewise the regression coefficients β₁, β₂, etc.)

The performance of the model on the on the validation set was assessed by computing the c statistic. The c statistic for this model applied to the remaining 25% of the population was 0.74 [95 % CI 0.64 to 0.83].