Author + information
- ↵*Reprint requests and correspondence: Dr. David J. Cohen, Cardiovascular Division, Beth Israel Deaconess Medical Center, 330 Brookline Ave., Boston, Massachusetts 02215
Despite recent advances in catheter technology, acute complications, including death, delayed abrupt closure and periprocedural myocardial infarction, continue to occur in 10% to 15% of patients undergoing percutaneous coronary intervention (PCI) (1,2). Because numerous observational studies have now confirmed a close association between ischemic complications of PCI and late mortality (3), prevention of such complications remains a central goal of the practicing interventional cardiologist. In recent years, it has become increasingly clear that the coagulation cascade plays a critical role in the pathogenesis of such complications (1,2,4,5). All forms of catheter-based revascularization produce local endothelial injury, thus exposing underlying tissue factor, which binds to and activates circulating Factor VII (FVII). This activated form of FVII is then free to activate other clotting factors (including Factors IX and X), thus leading to the generation of thrombin, platelet activation and, ultimately, to the formation of thrombus at the site of the catheter-induced arterial injury.
Following from this model, one might postulate that the risk of subacute ischemic complications after catheter-induced endothelial injury should be influenced by circulating levels of FVII. In theory, higher levels of circulating FVII should increase the risk of an ischemic complication, while lower levels of circulating FVII would be protective. Recently, several genetic polymorphisms that influence FVII levels have been identified, including an A10,976 G point mutation on exon 8 of the FVII gene that leads to a substitution of glutamine for arginine at codon 253 (Arg253 Gln polymorphism) (6). In the Framingham Heart Study, this mutation was shown to lead to a dose-dependent decrease in FVII activity (7). It is conceivable, therefore, that persons who have one or two copies of this allele might be at lower risk of ischemic complications after PCI.
In this issue of the Journal, Mrozikiewicz et al. (8) describe the results of a preliminary study to test this hypothesis. In a cohort of 666 consecutive patients undergoing PCI at a single center, ischemic complications, including death, myocardial infarction, or repeat target vessel revascularization within 30 days, occurred in 6.4% of patients. Although most traditional clinical and angiographic factors failed to predict the occurrence of these complications, they found that patients who were either heterozygous or homozygous for the Arg253Gln polymorphism had a substantially reduced risk of ischemic complications after PCI, compared with patients without this polymorphism. This effect seemed to be fairly consistent regardless of the type of PCI employed (balloon angioplasty, directional atherectomy or stenting) and persisted after adjustment for a variety of clinical and angiographic factors previously shown to predict such complications. The authors conclude, therefore, that activated FVII is important in the pathogenesis of ischemic complications after PCI and suggest that pharmacologic therapy directed at inhibiting activated FVII during PCI may be a useful approach for further reducing the incidence of such complications.
Although these findings are interesting and of potential pathophysiologic and therapeutic relevance, the main value of this study is in highlighting many of the principles of genomic epidemiology that are becoming increasingly important techniques for biomedical research. Complex genomic association analysis, of which this study is an example, is dedicated to discovering alleles and genotypes that are associated with various common clinical outcomes. Identifying and understanding these associations is likely to fundamentally change our basic understanding of cardiovascular disease, and this study illustrates several ways in which genomics is likely to influence cardiovascular medicine.
The study by Mrozikiewicz et al. (8) clearly illustrates that, unlike simple monogenic or Mendelian disorders, even genotypes that have an apparently profound effect on a common complex disease actually have only a very small absolute effect. When expressed as the reciprocal of the absolute risk difference, the magnitude of a complex genetic association can be thought of as the number needed for genetic effect (NNGE) (9). The NNGE is the number of persons with a particular genotype who are likely to experience the associated disease as the result of the specific genetic effect. For example, in this study, individuals with the Arg253/Arg253 genotype had an approximately three times greater risk for ischemic complications following PCI relative to individuals with at least one Arg253Gln allele (relative risk = 7.7%/2.5% = 3.1). By contrast, when expressed as the reciprocal of the absolute risk difference, the NNGE is 19.2 (i.e., 1/[0.077 − 0.025]). These same data thus indicate that despite a substantially increased relative risk, only one of every 19 individuals with the Arg253/Arg253 genotype will experience an ischemic complication of PCI as the result of a genetic effect.
This concept is important because it illustrates that, contrary to popular expectation, most genotypes associated with common complex cardiovascular diseases are not likely to be determinative, or even particularly predictive. Only a small fraction of individuals who inherit a genotype associated with a common cardiovascular disease are likely to ultimately develop the disease as the result of a genetic effect. Instead, complex genomic associations are much more likely to help cardiologists identify subgroups of patients who are at various levels of risk based on their genotype, and to help physicians choose treatment strategies that are most appropriate to a particular patient’s genomic profile.
Furthermore, elucidating the various genetic pathways that may contribute to a common clinical outcome is likely to reveal that most common cardiovascular events, such as ischemic complications of PCI, are actually the common manifestation of diverse genetic influences. Appreciation of the different genetic etiologies of common clinical events will reduce the heterogeneity inherent in any complex genetic disease and allow physicians to choose more rational individualized primary and secondary prevention treatment strategies based on genotype. For example, in the example provided by this study, perhaps only persons with the Arg253/Arg253 genotype will benefit from activated FVII inhibition during PCI, or perhaps optimal dosing may depend on FVII genotype. Elucidation of these genetic pathways will not only improve our understanding of how and why disease develops but it will also ultimately suggest novel treatment strategies to interrupt the disease process.
Clearly, genomics has the potential to profoundly impact cardiovascular medicine. With the completion of the Human Genome Project imminent, the number of clinical genomic epidemiologic studies is likely to increase substantially. Because its potential influence is so great, the cardiovascular community must demand that these studies be conducted with the most rigorous methodologies and be subjected to the most critical scrutiny prior to being published. Unfortunately, the study by Mrozikiewicz et al. (8) in this issue of the Journal, not unlike many genomic epidemiologic studies, fails to adhere to several basic principles of epidemiologic science. Thus, the results of this study must be interpreted with considerable caution.
The most obvious deficiency in this study is the fact that the authors do not mention, nor did they specifically attempt to control for population stratification bias, the single most important bias peculiar to genomic epidemiology (10). Comparing the genotypes of unrelated persons with and without a particular clinical outcome, such as ischemic complications following PCI, is a deceptively simple study design. Without adequate sampling restrictions, however, this design can easily lead to spurious genetic associations. Population stratification bias consists of two independent components, confounding and sampling bias, either or both of which can occur in a genetic case-control study. The overall impact of this bias is equal to the effect of the confounding bias multiplied by the effect of the sampling bias, and either component bias alone, or the combined effect of both biases, can cause a spurious genetic association.
Both components of population stratification bias are caused by the presence of distinct genetic subgroups within the assembled study sample. These genetic subgroups are sometimes difficult to detect, but they are usually present within most unrestricted sampling frames. Therefore, the most important feature of any genomic epidemiologic study is that it must specifically attempt to control for the possible effect of population stratification bias by restricting the sampling frame.
The sampling bias component of population stratification is caused by sampling affected and unaffected persons from different genetically admixed populations. It can be eliminated by sampling all affected and unaffected persons from the same enumerated population of eligible subjects. This sampling strategy ensures that the case and control subjects will have comparable genetic backgrounds. Fortunately, the cohort design employed by Mrozikiewicz et al. (8) accomplishes this goal because all affected and unaffected persons are derived from the same cohort of the 666 consecutive patients undergoing PCI. Therefore, sampling bias due to population stratification is not likely to have affected the results of this study.
The confounding bias component of population stratification is caused by the presence of distinct genetic subgroups within the study population, even when both the case and control subjects are derived from the same source population. If these genetic subgroups have both a different frequency of the genotype under study and a different rate of the outcome event in interest, then an epidemiologic confounding bias will be present. This bias can be largely eliminated by further restricting the study to a few self-identified subgroups within a single main ethnic group. The potential for confounding bias due to population stratification can then be eliminated by stratifying the genetic analysis by subgroup. These calculations can be efficiently performed using stratified logistic regression analysis, in which outcome is the dependent variable, genotype is the independent variable, and ethnic subgroup is the stratifying variable. Unfortunately, the authors of the current study did not attempt to explicitly take into account the genetic origin of the persons enrolled in their study. Therefore, confounding due to population stratification bias cannot be excluded as a possible explanation for their results.
In addition to the potential for bias, it is also important to consider the possibility that chance alone explains the results of this study. In general, relatively small studies such as this (with only 46 outcome events) should be viewed as hypothesis-generating, at best, because the temptation to test hundreds of genetic polymorphisms in the same cohort is too great. Using a nominal significance level of p < 0.05, one can expect that many of these polymorphisms will be regarded as “statistically significantly” associated with a common disease merely by chance. In addition, small sample sizes do not permit the investigators to evaluate the increasingly clinically pertinent topic of gene by gene, and gene by environment, interactions. It is curious that in the study by Mrozikiewicz et al. (8), the level of FVII activity in those persons with an Arg253Gln allele who suffered an ischemic complication was the same as the FVII activity among persons with the Arg253/Arg253 genotype who experienced such an event. This observation strongly suggests the possibility that some environmental factor may have been interacting with the Arg253Gln allele to modify its effect on FVII activity in exposed persons and, therefore, modified their risk of ischemic complications. Unfortunately, this possibility cannot be evaluated, because the study was not large enough to detect and characterize such a potentially clinically pertinent gene by environment interaction.
Considering these important limitations, what have we learned from this study? We have learned that aspects of the coagulation cascade, in addition to the platelet, may play an important role in the pathogenesis of ischemic complications of PCI. In the future, novel anticoagulant strategies directed specifically against activated FVII or the FVII-tissue factor interaction may be an important adjunct to our pharmacologic armamentarium for PCI patients. Moreover, we have identified a relatively common genetic polymorphism (the Arg253Gln polymorphism) that may be protective against ischemic complications in PCI. If this association is confirmed in other studies and a simple genetic screen is developed for this polymorphism, it might be possible to tailor a patient’s treatment to his or her specific risks of complications. For now, however, these conclusions are highly preliminary and await confirmation by larger, prospective studies.
The caveats outlined above notwithstanding, Mrozikiewicz et al. (8) should be congratulated for performing an intriguing study that illustrates both the potential and the limitations of genomic medicine. Perhaps the most important contribution of this study, however, is to point out that there is an urgent need for a closer collaboration among geneticists, epidemiologists and clinical scientists in order to improve the quality of genomic epidemiologic research. Editors, reviewers and readers of cardiovascular journals must demand that elegant molecular genetic methods be accompanied by rigorous epidemiologic methods. Because the potential impact of genomics is so great, the cardiovascular medicine community should demand more, not fewer, such studies. We should, however, also demand that such studies be large enough to draw firm conclusions and rigorous enough to persuade us that genomic information can make important contributions to the practice of cardiovascular medicine.
☆ Dr. Cohen was supported in part by a Clinician-Scientist Award from the American Heart Association.
↵∗ Editorials published in the Journal of the American College of Cardiology reflect the views of the authors and do not necessarily represent the views of JACC or the American College of Cardiology.
- American College of Cardiology
- Califf R.M.,
- Abdelmeguid A.E.,
- Kuntz R.E.,
- et al.
- RESTORE Investigators
- Green F.,
- Kelleher C.,
- Wilkes H.,
- et al.
- Feng D.,
- Tofler G.H.,
- Karson M.G.,
- et al.
- Mrozikiewicz P.M.,
- Cascorbi I.,
- Zeimer S.,
- et al.
- Ference B.A.,
- Horwitz R.I.,
- Feinstein A.R.
- Witte J.S.,
- Gauderman W.J.,
- Thomas D.C.