Predicting risk of fatal cardiovascular disease and sudden death in hypertension

In the current issue of the Journal, Le et al.[1] present a risk score for sudden death for patients with hypertension with or without cardiovascular disease. The risk score is based on randomized controlled trials of antihypertensive therapy included in the INDANA (INdividual Data ANalysis of Antihypertensive intervention trials) database. The INDANA project, initiated in 1990, was a challenging initiative by the Service de Pharmacologie Clinique, Lyon, France, with the intention to perform meta-analyses on individual patient data from available and published randomized controlled trials [2], preceding the initiative by the Blood Pressure Lowering Treatment Trialists’ Collaboration to conduct prospectively planned meta-analyses of the ongoing randomized trials of blood pressure (BP) lowering drugs [3]. The INDANA database includes individual patient data from randomized controlled trials assessing the efficacy of antihypertensive drug intervention versus placebo or no (or minimal) treatment on the basis of clinical outcomes for which data are available. Randomized controlled trials provide large cohorts with good follow-up, usually from several different countries. The main objective of the INDANA project was to identify the profile of hypertensive patients who may benefit most from treatment, and a number of secondary objectives were also formulated [2]. Assessing individual patient data offers several advantages over meta-analyses based on summary results, for example offering sufficient power to identify accurately the profile of patients who are responders to antihypertensive treatment. Main trials included in the original database were the Hypertension Detection and Follow-up Program [4], the Multiple Risk Factor Intervention Trial [5], the European Working Party on High Blood Pressure in the Elderly trial [6], the Medical Research Council (MRC) trial in mild hypertension [7], the trial by Coope and Warrender [8], the Systolic Hypertension in the Elderly Program [9,10], the Swedish Trial in Old Patients with Hypertension [11] and the MRC trial in older adults [12], published between 1979 and 1992; later on, the Systolic Hypertension in Europe trial from 1997 [13] was added. Before considering the contribution of the INDANA project to the prediction of risk, it may be of interest to reiterate a few of its original and clinically relevant achievements, including the benefit of antihypertensive treatment according to sex and in the very elderly, and a better insight into the J-curve phenomenon in hypertension. As it was not clear whether treatment benefit is similar in men and women, a subgroup meta-analysis was performed based on seven trials [4,6–8,10–12], including 20 802 women and 19 975 men [14]. The authors concluded that the odds ratios for all categories of events, that is fatal strokes, all strokes, fatal coronary events, all major coronary events, all major cardiovascular events, cardiovascular mortality and total mortality, did not differ significantly between men and women. In absolute terms, the benefit in women was seen primarily for strokes; in men, treatment prevented as many coronary events as strokes. At the time, it was not known whether treatment is beneficial in hypertensive patients aged 80 years and over. To tackle the issue, the INDANA investigators collected data from subgroups of randomized controlled trials to assess the evidence for and against antihypertensive treatment in the very elderly [15]. The meta-analysis of data from 1670 participants from six trials [6,8–11,13,16] suggested that treatment prevented 34% of strokes (P = 0.01). Rates of major cardiovascular events and heart failure were significantly decreased by 22 and 39%, respectively. There was no treatment benefit for cardiovascular death and a nonsignificant but potentially worrisome 6% relative excess of death from all causes. The authors concluded that an age threshold beyond which hypertension should not be treated cannot be justified, and that a large-scale specific trial is needed for definite conclusions that antihypertensive treatment is beneficial in very elderly hypertensive patients. It took 9 more years before the results of the dedicated Hypertension in the Very Elderly Trial [17] became available. The results confirmed the benefit of antihypertensive treatment in the very old and showed that concern for increased mortality was not justified; on the contrary, all-cause mortality decreased by 21% (P = 0.02), with amongst others significant reductions in the rate of stroke death (39%), any cardiovascular event (34%) and heart failure (64%). In addition, fewer serious adverse events were reported in the active-treatment group. In view of the controversy with regard to the J-shaped curve from individual studies, the INDANA investigators explored the J-shaped relationship between BP and mortality in 40 233 persons with hypertension from seven randomized controlled trials [4,6–8,10–12] included in the database [18]. The findings indicate an increased risk for death among both treated and untreated hypertensive patients with low BP, and this increased risk was apparent for cardiovascular deaths as well as for noncardiovascular deaths. The results supported the idea that the J-shaped curve is independent of treatment and explained by other factors, such as poor health associated with low BP. It is of note that the INDANA risk score for sudden death [1] was preceded by a score for predicting risk of death from cardiovascular disease, including stroke and coronary heart disease, based on individual data of 47 088 hypertensive men and women from eight randomized controlled trials [4–8,10–12] included in the original INDANA database [19]. A total of 1639 patients died from cardiovascular causes during a mean of 5.2 years of follow-up. Sixteen baseline risk factors were available in all datasets and had a priori plausibility as risk factors. A multivariable Cox proportional hazards model, stratified by trial, was fitted for all 16 risk factors simultaneously, with cardiovascular death as an outcome. A risk score was developed from 11 significant factors: age, sex, SBP, serum total cholesterol (TC) concentration, height, serum creatinine concentration, cigarette smoking, diabetes, left ventricular (LV) hypertrophy detected by ECG, history of stroke and history of myocardial infarction (MI). BMI, DBP, heart rate and serum uric acid were NS in the multivariable model and were not included in the risk score. The Cox model predictor was converted to an integer score directly related to an individual's probability of death, separately for men and women. An adult with very low risk, that is a women aged 35–39 years with the healthiest category of each risk factor, represented score 0; points were added for each factor according to its association with cardiovascular death, producing the total risk score for each individual. Participants in the two United Kingdom MRC trials [7,12] were chosen as the reference sample to give a country-specific probability of cardiovascular death linked to a person's risk score, and to provide age and sex-specific distributions of the risk score. A clear exponential relationship existed between risk score and dying from cardiovascular disease in 5 years. The risk score is meant to be useful for physicians when determining an individual's need for antihypertensive treatment and is available for users at the website www.riskscore.org.uk. It provides the probability of death from cardiovascular disease within 5 years, the individual's risk level compared with others of the same age and sex, and identifies the main modifiable risk factor(s) for the patient. Risk prediction models were also presented for fatal coronary heart disease, fatal stroke and all-cause mortality. In view of the paucity of data on the prediction of risk for sudden death and the absence of a risk score in hypertension, the objective of Le et al.[1] was to construct a risk score for sudden death based on individual patient date from six randomized controlled trials included in the INDANA database [5,6,8,10,11,13] with available and unbiased information regarding sudden death. A total of 345 sudden deaths occurred in 8044 women and 17 604 men during a mean follow-up of 5.16 years. The methodology was essentially similar to the methodology used by Pocock et al.[19]; in addition, the population was divided in derivation and validation subsets (2 : 1). Univariable analyses were conducted with 29 covariables, including the covariables used by Pocock et al.[19], except ECG LV hypertrophy, so that its possible contribution to sudden death was not assessed. Seven of these covariables were retained in the multivariable model: age, sex, smoking status, SBP, serum TC, diabetes and history of MI. It is not surprising that all of these are also included in the risk predictor of cardiovascular death. The final model was then converted into a single integer score for both sexes together, directly related to an individual's probability of sudden death within 5 years. Young investigators have been given the opportunity to use the INDANA database for the current analysis, which is a commendable initiative of the principal investigator to resolve an interesting scientific question. However, it would have been good to involve members of the original INDANA Steering Committee according to the rules outlined in the protocol [2] or at least acknowledge their contribution which was only done in the course of the review process. What is the relevance of the risk score for sudden death [1] in view of the risk score for overall cardiovascular death [19]? For clinical purposes, it appears to be more important to predict mortality from overall cardiovascular disease rather than just sudden death so that the latter risk score can be considered complementary to the one for cardiovascular death. It would be of interest to investigate if the outcome of the two risk scores would in fact differ in individual patients. Furthermore, the prediction of cardiovascular death appears to be more precise than the prediction of sudden death as it is based on nearly five times more events, considers 11 versus seven risk factors and provides separate risk scores for men and women versus one single sudden death score for both sexes. In addition, the risk score for sudden death is hampered by the different or even unknown definitions of sudden death in the included trials. Finally, the risk score for cardiovascular death is more practical as it is available online as mentioned above. With regard to practical consequences for antihypertensive treatment, it is of note that treatment was significant in the multivariable analysis on cardiovascular death but not in the analysis on sudden death. On the other hand, both risk scores have limitations. They are derived from randomized controlled trials of which the population is usually more healthy and not truly representative for the general hypertensive population, and data were collected in the last quarter of the previous century. Some variables such as HDL cholesterol were not available and changes may have occurred in risk factors and their impact on outcome over time. Finally, data were collected in Europe and North America and may not be applicable in other parts of the world. In principle, the ideal risk score should be constructed from data obtained in the population in which it is to be applied. Second best is external calibration of the risk score for specific populations or countries. Rates or cardiovascular death vary between communities and countries in ways not fully explained by known risk factors, which affects the generalizability of any risk score, but the relative importance of risk factors is likely to be similar in most developed countries. Pocock et al.[19] applied the risk score for cardiovascular death to the patients from the two United Kingdom MRC trials. Although the Systemic COronary Risk Evaluation (SCORE) model [20] has disadvantages of its own, for example being based on European cohort studies from the previous century, it is largely used, particularly by the European Society of Cardiology (ESC). Risk scores for dying from cardiovascular disease over 10 years have been constructed for low-risk and high-risk areas in Europe, and calibration of the SCORE risk charts for individual countries has been done for numerous European countries. In 1999, the WHO and the International Society of Hypertension accepted the concept that the initiation of antihypertensive treatment should be on the basis of the patient's total cardiovascular risk [21]. But rather than proposing a continuous risk score, patients were classified in one of three categories, that is low, medium and high risk, based on a number of risk factors, target organ damage and associated clinical conditions. This concept was further developed in the European Society of Hypertension (ESH)/ESC guidelines for the management of hypertension. In its last version [22], the classification in low, moderate, high and very high risk is considered to reflect the 10-year risk of cardiovascular mortality according to the SCORE model, that is less than 1%, from 1 to less than 5%, from 5 to less than 10% and at least 10%. In conclusion, many risk scores are available for the prediction of various cardiovascular diseases or death in different populations. All of them have advantages and limitations. They do offer a reasonable estimate of the patient's risk and have become indispensable for the initiation and guidance of antihypertensive treatment and overall management of the hypertensive patient. ESH/ESC guidelines for the management of hypertension have opted for a categorical system, but the continuous risk scores developed on the basis of the INDANA database may be a useful quantitative complement for the estimation of risk of the individual hypertensive patient, particularly the user-friendly version for cardiovascular death as it is available online. ACKNOWLEDGEMENTS Conflicts of interest There are no conflicts of interest.