Association studies between the angiotensin-converting enzyme insertion/deletion polymorphism and hypertension: still interesting?

In this issue of the journal, Matsubara et al. [1] report on a population-based association study that investigates the relationship between the insertion/deletion (I/D) polymorphism in the angiotensin I-converting enzyme (ACE, dipeptidylcarboxypeptidase 1, DCP1) gene and blood pressure in two cohorts of Japanese subjects. The authors found no evidence for an association of the ACE I/D polymorphism and blood pressure or prevalence of hypertension, respectively. In view of the large number of similar studies that have been published before with largely negative results, it may be a timely question to ask whether association studies with the ACE I/D polymorphism and blood pressure or hypertension should still be performed and published. This question is part of a more general discussion regarding the future of population-based association studies, which has been addressed by a number of recent editorial comments accompanying association studies in several journals [2–7]. The authors of these editorials all agree that, after a decade of association studies, future investigations will have to meet more stringent requirements in order to be considered as useful contributions to our understanding of the genetic basis of complex diseases. A glance back When Rigat et al., in 1990, reported that the I/D polymorphism in the ACE gene accounts for approximately one-half of the variance of serum enzyme levels [8] and Cambien et al., in 1992, proposed that this polymorphism is a potent risk factor for myocardial infarction [9], a tide of association studies between this polymorphism and several cardiovascular phenotypes, including blood pressure, hypertension, coronary heart disease, myocardial infarction, stroke, diabetic nephropathy and others, avalanched down on the readers of genetic and cardiovascular journals, resulting in a rather confusing patchwork of positive and negative results that often directly contradicted each other. The results of many of these studies have been put together and reviewed by several meta-analyses [10–13]. While some of these meta-analyses found that possession of the D allele was associated with an increased risk for myocardial infarction [10,11], stroke [10,12] or diabetic nephropathy [10], others did not confirm these results [13]. An association of the ACE I/D polymorphism with blood pressure or prevalence of hypertension was not reported in these meta-analyses. Not surprisingly, all meta-analyses found evidence for publication bias. In general, small studies tended to show larger effects than large studies. Thus, Agerholm-Larsen et al. [13] showed that the risk of myocardial infarction and coronary heart disease was increased in carriers of the DD genotype in small but not in large studies. This suggests selective non-publication of small, imprecise studies with negative results, while studies of similar quality but with positive results tend to be published. Thus, not only is there a need for more stringent criteria for high-quality association studies, but also for good opportunities to publish high-quality negative studies. In this issue of the journal, Matsubara et al. [1] report a negative study showing no association between the ACE I/D polymorphism and blood pressure or prevalence of hypertension, respectively. In their meta-analysis of 23 studies comprising 6923 subjects, Staessen et al. [10] also found no evidence for an association of the ACE I/D polymorphism and hypertension. Similarly, Agerholm-Larsen et al. [13] found no evidence for an association of the ACE I/D polymorphism with blood pressure in their meta-analysis of 19 studies comprising 15 942 subjects. Should the study by Matsubara et al. therefore be dismissed as the twentieth-or-so-fold replication of a negative result? Certainly not. Association studies must be interpreted within their respective ethnic context. Staessen et al. [10] only included two reports from Japan in their meta-analysis; Agerholm-Larsen et al. [13] included none. In the studies included in the Staessen meta-analysis, the D allele was associated with an excess risk for hypertension. More recently, Higaki et al. [14] found an association between the D allele and hypertension in male but not in female Japanese. In contrast to these studies, Zaman et al. [15] found no association between the ACE I/D polymorphism and hypertension in a Japanese population. Thus it is currently unclear whether there is an association between the ACE I/D polymorphism and blood pressure or prevalence of hypertension in the Japanese population. Criteria for high-quality association studies Recently, a number of criteria for high-quality association studies have been suggested [2,3,6,7]. These criteria include (i) a well defined phenotype, (ii) a clear a priori hypothesis, (iii) a large sample size and an a priori estimation of the statistical power of the analysis, (iv) a small P- value and a high odds ratio and/or attributable risk, (v) biological plausibility of the hypothesis and functional significance of the polymorphism under investigation, (vi) independent replication of the results in different populations, and (viii) confirmation in family based studies. In order to rule out that a certain locus contributes to a given phenotype, several (at least three) common polymorphisms at this locus should be tested and haplotype analyses should be performed. Obviously, not all criteria need to be fulfilled by each individual study; however, these criteria provide a useful orientation when an association study is to be evaluated for its scientific merits. Does the study by Matsubara et al. [1] fulfil these criteria? Phenotype The major phenotype in this study is hypertension, which was defined according to standard definitions. The analysis was extended on actual blood pressure data from home blood pressure measurements and ambulatory 24-h blood pressure measurements. When interpreting these data, it should be considered that approximately one-third of the study population was under antihypertensive medication. Regarding the history of cardiovascular disease, the authors rely on information from questionnaires and medical records, which is not specified further. Hypothesis The study was designed to determine the association between the ACE I/D polymorphism and hypertension in two Japanese populations. To this end, the hypothesis is clear. No clear hypothesis is stated with respect to the actual blood pressure data, which were analysed separately for systolic and diastolic blood pressure, as well as for pulse pressure and circadian blood pressure variation. Extreme caution should be exerted when interpreting the results of some exploratory post-hoc analyses in subgroups of the population which are not driven by a clear a priori hypothesis. Sample size With more than 1200 subjects included, the home blood pressure part of the study is comparatively large. Among the 19 studies that were included in the meta-analysis by Agerholm-Larsen et al. [13] on the association between the ACE I/D polymorphism and blood pressure, only three contained a greater number of subjects. Nevertheless, a statistical power calculation should have been performed, particularly since negative results are presented. To illustrate the problem: in the meta-analysis by Agerholm-Larsen et al. [13], the mean difference in systolic blood pressure between DD and II subjects was 0.5 mmHg. In order to demonstrate a difference in systolic blood pressure of 1 mmHg in a similar setting on an α = 0.05 test significance level with 80% power would have required more than 3000 subjects per group. Functional significance and biological plausibility The D allele is associated with increased plasma ACE activity [8]. It is well accepted that the 287 bp Alu repeat element in intron 16, the presence or absence of which constitutes the I/D polymorphism, is not itself the functional variant directly affecting plasma ACE activity [16]. As the true functional variant is currently unknown, an association study may well be performed with the I/D polymorphism which obviously is in linkage disequilibrium with the functional variant, at least in Caucasians. A further question is whether the criterion of biological plausibility can also be considered as being fulfilled? In other words, it remains to be determined how an increased ACE activity would lead to chronic elevations in blood pressure and to hypertension. While various scenarios have been proposed, much work needs to be done to show that one of them is really in effect in hypertension [4]. The negative results obtained by Matsubara et al. [1] and others [10,13] on the relationship between the ACE I/D polymorphism and hypertension may make these questions somewhat less urgent. Independent replication While there is plenty of evidence indicating no association between the ACE I/D polymorphism and hypertension in other ethnic groups [10,13], much less information is available for Japanese populations. Thus, there is the need for independent replication of the data in this ethnicity. Simultaneous testing of multiple polymorphisms: haplotype analysis The lack of association between a single diallelic polymorphism, such as the ACE I/D polymorphism, and a phenotype such as hypertension does not rule out the gene under investigation as a potential candidate gene. In order to allow for more reliable conclusions as to the role of a candidate gene for a complex phenotype, it has been suggested that at least three common polymorphisms within the gene should be considered and their combination should be analysed (haplotype analysis) [3]. Similar to many other regions of the human genome, there exist an enormous degree of genetic diversity in the ACE gene. Rieder et al. [17] have sequenced the ACE gene in 11 subjects of European-American and African-American descent. They identified at least 78 varying sites in coding and non-coding regions, which resolved in 13 distinct haplotypes. Evaluating haplotypes makes the statistical power problem even more critical because it increases the number of genotypes to be compared with each other, and thus requires larger sample sizes [5]. Conclusions The last decade has seen a huge number of genetic association studies with often contradictory results. Positive findings in small studies were often not replicated in larger, better-designed studies. This has led to considerable scepticism against association studies. Meanwhile, stringent criteria for high-quality association studies have been proposed that, when adhered to, will greatly enhance the quality of future studies. The future will see large-scale approaches with new diagnostic methods, such as multiplex polymerase chain reactions and DNA microarrays or gene chips, which allows simultaneous testing of many single nucleotide polymorphisms in many candidate genes as functional groups. Only approaches such as these will make analyses of functionally interrelated networks or physiological pathways technically feasible [7]. As the number of subjects required for testing increases with each gene variant being investigated, these approaches will also necessitate more large-scale collaborations, leaving a doubtful fate for small classical association studies such as the one by Matsubara et al. [1].