Discussion
In this study, we assessed the predictive value of ‘standard predictors’ routinely available for a general practitioner when he/she meets a patient with stable coronary heart disease whose visit is not prompted by renewed cardiac complaints. When no ‘standard predictors’ were included in the prediction, 63.2% of the model-based predictions of the composite outcome and 79.9% of all-cause death predictions were correct. Including all ‘standard predictors’ in the model increased the figures to 68.4% and 83.4%, respectively.
We have not been able to identify a study where death of all causes and cardiovascular insults have been assessed in patients with (1) stable coronary heart disease of a type like the above described, (2) with a 10-year follow-up, and (3) where the clinical examination was not prompted by renewed cardiovascular complaints. Taken together, most of the studies we identified were either analysing small study samples, for example,20 were developed in patients with acute coronary syndromes,21 or had short follow-up.22
In many respects, our patients are like those of the Prospective Observational Longitudinal Registry of Patients with Stable Coronary Heart Disease (CLARIFY) study23 which enrolled 20 291 patients of whom 20% had anginal symptoms corresponding to the 20% of our patients who took long-lasting nitrates. Patients with hospital admission for cardiovascular reasons (including revasculariation) in the past 3 months before enrolment or conditions interfering with life expectancy such as cancer and NYHA class 4 were excluded. So, in these respects, the CLARIFY patients are like our patients. However, the CLARIFY patients had been observed with a median of 24.1 months and enrolment took place 10 years later than in the CLARICOR trial. A total of 469 cardiovascular deaths or myocardial infarctions occurred in these patients (2.3%).23 By contrast, in our cohort the corresponding numbers found during the first 2 years of observation were 170 cardiovascular deaths and 7.7%, respectively, probably reflecting improved quality of treatment and more frequent statin treatment in the CLARIFY patients (84% vs only 41% in the CLARICOR patients).
Many of our ‘standard predictors’ showed that they contained statistically significant prognostic information. However, few of the ‘standard predictors’ contributed statistically significant prognostic information when all predictors were tested simultaneously. Moreover, the ‘standard predictors’ offered little prognostic discrimination.
Methodology
Regarding our methodology, the performance statistics reported here are minimal, but they suffice to show that the results are meagre, leaving room for improvement with advanced biomarkers.3 Prediction at 3, 6 and 9 years covers the follow-up as well as would a sophisticated integral over continuous time. Using 0.50 as a probability threshold is again a least arbitrary choice (the relative impact of false and true predictions being unknown), as is the ROC principle. Adding realism to the naked construct of a ‘completely effective preventive intervention’ (figure 2B) would again be subject to criticism in the absence of alternatives, with solid empirical backing.
Strengths
The strengths of the CLARICOR trial are the considerable size of the patient population, the long duration of follow-up, few losses to follow-up (0.5%), the ethnic homogeneity of the patient population, rarity of missing values, with focus on an operationally defined, homogeneous and relevant patient category. The design implies that the patients are sampled at random, presumably uneventful, time points during their stable state (as defined by the CLARICOR trial). Furthermore, the large biobank formed during the trial allows an extensive search for new advanced biomarkers to be made.
Limitations
Among those 7586 patients who declined our invitation to visit a cardiology centre, many must have been eligible for the CLARICOR trial, and we do not know how they looked and fared. With a response rate of about 50%, the cohort could represent a prognostic elite if responders were mostly mobile and health-conscious patients. So, selection bias cannot be excluded.
Furthermore, the patients recruited for the CLARICOR trial were diagnosed with coronary heart disease about 20 years ago. Because of the tremendous developments in treatment and rehabilitation, there has been a very significant and graded increase in prognosis of such patients,24 as shown in national data.25 Given these uncertainties, prognostic findings in the CLARICOR cohort should not directly be applied to present-day patient materials. However, the overall picture regarding the relative predictive effect of ‘standard predictors’ and advanced biomarkers should provide relevant and valuable information.
Potential weaknesses of the present cohort within the context of prognostication of patients with stable coronary heart disease as here defined include the lack of information about left ventricle function, body mass index and blood pressure, as well as the effects of changes in medications during follow-up. Information about postinfarction heart failure and postinfarction angina pectoris was not available to us. But information about the medication at entry into CLARICOR served as proxy information. The lack of left ventricular ejection fraction may be partially or completely compensated, as Solomon et al
26 found that age, sex, hypertension, prior acute myocardial infarction, creatinine, diuretics and digoxin were related to left ventricular ejection fraction, all quantities that we have included within the group referred to as ‘standard predictors’. The shortcomings of our data files are mitigated by the fact that by design the present study focuses on the situation where patients with stable coronary heart disease visit a physician for reasons unrelated to the coronary heart disease. Here it might be of value if simple clinical information readily available could be used to screen these patients with stable coronary heart disease to identify high-risk patients who might be referred to a more thorough cardiologic examination and follow-up.
It was our hope that the simple ‘standard predictors’ of table 1 might serve this purpose, but they proved too little informative. Behind this disappointing feature lie undoubtedly a meshwork of behavioural and statistical interconnections that are impossible to unravel. A reviewer has pointed out that ‘index event’ phenomena27–29 may also be involved. Briefly, to develop an index event (say, infarction), those with a given low-risk trait tend to have unrecorded high-risk traits, and vice versa, so, to the extent that risk traits are permanent, high-low follow-up comparisons will be biased when based on such patients. While this is true, there is no prognostic bias here as the index event is part of the definition of the population of interest.
Furthermore, patients, if any, who became eligible for the CLARICOR trial during the period 1993–1999 and then died before August 1999, are absent. Thus, our data do not represent patients as they enter a stable disease state (as delimited by CLARICOR exclusion criteria); instead, they may be regarded as community patients (subject to some self-selection) seen by their physician on a random date during their stable state.
Our model was tested using the same data that were used to derive it. Therefore, we do not feel it is advisable to use the results for predictive purposes without using some independent data to calibrate the model, and for this and the above-mentioned reasons we have elected not to present an explicit prediction model.