1. Top of page
  2. Abstract
  3. Disclosure of Conflict of Interests
  4. References

See also Mannucci PM, Asselta R, Duga S, Guella I, Spreafico M, Lotta L, Merlini PA, Peyvandi F, Kathiresan S, Ardissino D. The association of factor V Leiden with myocardial infarction is replicated in 1880 patients with premature disease. This issue, pp 2116–21.

A causal association between abnormalities of the hemostatic system and arterial disease (i.e. either cardio- or cerebro-vascular disease) has always been purported. This contention was further strengthened by the findings of the Northwick Park Heart Study in the early 1980s [1,2], which suggested an important association between elevated plasma fibrinogen and factor (F)VII, and myocardial infarction. However, subsequent studies tempered the enthusiasm that surrounded these data, mostly because of the aspecific effects on all-cause mortalities that were observed for these abnormalities, and of the probable presence of residual confounding variables [3]. The interest in this area returned with the discovery of common genetic polymorphisms that modulate plasma levels of clotting factors [4,5]. Genotyping has the obvious advantage of not being influenced by other factors, which dramatically reduces the influence of confounding variables in association studies. Moreover, genetic analysis was felt to be analytically more reliable than phenotyping (at least for simple bi-allelic loci). However, despite this the disadvantage remains that selective pressure is expected to reduce the prevalence of those polymorphisms that are associated with an increased risk of arterial disease during the reproductive age. Therefore, it might be expected that common polymorphisms do not carry a particularly increased risk of vascular disease. For this reason, the value of adding these polymorphisms to everyday clinical practice is limited [6]. This has also been demonstrated by a recent meta-analysis by Ye et al. on 66 155 cases and 91 307 controls, in which factor V (FV) Leiden and the prothrombin G20210A variant (two common polymorphisms associated with venous thromboembolism) were found to be associated with cardiovascular disease, although the degree of association was low (OR = 1.17 and 1.31, respectively) [7].

In this issue of the Journal of Thrombosis and Haemostasis, Mannucci and co-workers were able to confirm the results of the larger meta-analysis by Ye et al. in a case–control study performed in a cohort of 1880 young patients with acute myocardial infarction (AMI) and 1880 controls [8]. The study is largely a confirmatory one, but still deserves attention.

First, the study by Mannucci et al. indicates that genotyping may be less accurate than originally accepted, and that this may severely impact upon the study results. The study included a subset of 1210 AMI patients and 1210 controls previously enrolled in a similar study, and in whom no association was originally found between FV Leiden and AMI [9]. In that earlier study, both FV Leiden and prothrombin G20210A variant status were assessed using electrophoresis after enzymatic restriction, a technique that requires visualization of two bands in heterozygous carriers and just one band in wild-type carriers. Hence, the fragment-restriction technique used in the ‘negative’ study is expected to have a sub-optimal sensitivity, but a very high (around 100%) specificity. Indeed, in a proficiency survey among Italian laboratories mainly using fragment-restriction as the genotyping technique, the sensitivity and specificity for heterozygous FV Leiden and G20210A mutations was 95% and 100%, respectively [10]. Upon re-examination of the same cohort of 2420 cases and controls with a newer technique, the authors found an increase in the allele frequency (about 6%) of both mutations in carriers but a decrease in the allele frequency of FV Leiden in controls. This means that some controls had been previously classified as carriers when in fact they were not (i.e. the specificity of genotyping in the previous study was below 100%).

Sub-optimal sensitivity and specificity of measured variables in epidemiologic studies has long been known as a cause of bias, to form the information bias and the diagnostic bias (i.e. the error that arises when case and controls are incorrectly diagnosed) [11]. Incomplete (i.e. below 100%) specificity has the largest impact on bias. Incomplete sensitivity results in under-estimation of the prevalence of the exposure in both cases and controls. This necessitates greater sample sizes to maintain statistical power, although it does not significantly distort the degree of association if one assumes a non-differential sensitivity in both cases and controls. In contrast, incomplete specificity always reduces the magnitude of any association, i.e. reduces the observed relative risk and therefore may increase the probability of finding no association, even if an association exists. This issue can be appreciated in Fig. 1, which depicts the consequences of incomplete specificity on a hypothetical case–control study. Using a computer simulation, a population of 1 million cases and 1 million controls was generated, and the prevalence of heterozygous FV Leiden was fixed at 5% in control subjects, and at 8%, 10% and 12% in cases. Odds ratios were computed under a misclassification error because of a specificity ranging from 90 to 100%. From this, it could be readily appreciated that even a small decrease of specificity (i.e. from 100% to 98%) could induce a significant underestimation of the relative risk. How then, might the effect of incomplete sensitivity and specificity be corrected? Replication studies (at least in a subsample of the whole study) are always valuable in demonstrating the analytical reliability of the method used. Indeed, this is now a requirement of the recently released STREGA (STrengthening the REporting of Genetic Association studies) statement [12]. Once the error rates are known, one can then apply published formulae that account for incomplete specificity, sensitivity and prevalence to estimate the ‘true’ odds ratios from the observed ones [13].


Figure 1.  Changes in the observed odds ratio in a simulated case–control study on 106 cases and 106controls. Prevalence of heterozygous carriers of a mutation (e.g. FV Leiden) was fixed at 5% in controls and at 8% (diamonds), 10% (squares) and 12% (triangles) in cases. Under a non-differential misclassification model, for both cases and controls, the sensitivity of the method was fixed at 100%, whereas specificity ranged from 90% to 100%.

Download figure to PowerPoint

In addition to incomplete specificity, another possible bias that may explain the discrepancies disclosed by the retesting process is a differential error in genotyping between cases and controls. A differential misclassification error exists when sensitivity and specificity are not uniform, but rather vary between cases and controls [11]. For instance, DNA storage, collection or processing protocols may be different in cases, in whom blood processing is required during an emergency admission, than in controls, in whom blood sampling could be more ‘relaxed’. At variance with errors as a result of low specificity that always reduces relative risk estimates, differential misclassification errors lead to unpredictable under- or over-estimation of the true relative risk estimate, depending on the group of subjects having the highest error.

The decision to re-evaluate a previously investigated sample together with a new sample of additional 670 cases and controls was certainly bold. By re-analyzing the expanded sample, the authors found an increase in the relative risk of AMI for both FV Leiden and the prothrombin G20210A variant that was very similar to that observed in the much larger meta-analysis of Ye et al., thus further reinforcing a causal link between hemostasis and arterial vascular disease. The pathophysiological mechanisms responsible for this increase of risk are not entirely clear, however, a possible explanation might be offered by a secondary analysis of the present study. At variance with the results of meta-analysis of Ye et al., the authors were able to observe that the increase of AMI risk associated with FV Leiden was confined to those subjects with hypercholesterolemia, suggesting an interaction between these two vascular risk factors. This finding supports the concept that whereas identification single genetic polymorphisms by themselves has little practical value per se, they may be clinically relevant when looking at selected subgroups of patients. Clearly, these findings from secondary analyzes warrant further confirmation, possibly by meta-analyzes pooling individual subjects databases (a technique that allows a complete analysis of interactions between risk factors [14]). Should the hypothesis of an interaction between carriership of a hemostatic abnormality and other risk factors (e.g. hypercholesterolemia) be confirmed, an entirely new perspective for primary and secondary prophylaxis could arise, at least for carriers of polymorphisms that so far have only been associated with venous thrombophilia.

In conclusion, the study of Mannucci et al. reopens the case of the association between thrombophilic mutations and arterial disease, particularly stressing the fallacies of genotyping errors and opening some new horizons for research, and calls for additional studies on prospective cohorts.

Disclosure of Conflict of Interests

  1. Top of page
  2. Abstract
  3. Disclosure of Conflict of Interests
  4. References

The author states that he has no conflict of interest.


  1. Top of page
  2. Abstract
  3. Disclosure of Conflict of Interests
  4. References
  • 1
    Meade TW, Brozovic MM, Chakrabarti RR, Haines AP, Imeson JD, Mellows S, Miller GJ, North WR, Stirling Y, Thompson SG. Haemostatic function and ischaemic heart disease: principal results of the Northwick Park Heart Study. Lancet 1986; 328: 5337.
  • 2
    Meade TW, Ruddock V, Stirling Y, Chakrabarti R, Miller GJ. Fibrinolytic activity, clotting factors, and long-term incidence of ischaemic heart disease in the Northwick Park Heart Study. Lancet 1993; 342: 10769.
  • 3
    Danesh J, Lewington S, Thompson SG, Lowe GD, Collins R, Kostis JB, Wilson AC, Folsom AR, Wu K, Benderly M, Goldbourt U, Willeit J, Kiechl S, Yarnell JW, Sweetnam PM, Elwood PC, Cushman M, Psaty BM. Plasma fibrinogen level and the risk of major cardiovascular diseases and nonvascular mortality: an individual participant meta-analysis. JAMA 2005; 294: 1799809.
  • 4
    Fowkes FG, Connor JM, Smith FB, Wood J, Donnan PT, Lowe GD. Fibrinogen genotype and the risk of peripheral arteriosclerosis. Lancet 1992; 339: 6935.
  • 5
    Iacoviello L, Di Castelnuovo A, De Knijff P, D’Orazio A, Amore C, Arboretti R, Kluft C, Benedetta Donati M. Polymorphisms in the coagulation factor VII gene and the risk of myocardial infarction. N Engl J Med 1998; 338: 7985.
  • 6
    Reitsma PH. No praise for folly: genomics will never be useful in arterial thrombosis. J Thromb Haemost 2007; 5: 4547.
  • 7
    Ye Z, Liu EH, Higgins JP, Keavney BD, Lowe GD, Collins R, Danesh J. Seven haemostatic gene polymorphisms in coronary disease: meta-analysis of 66,155 cases and 91,307 controls. Lancet 2006; 367: 6518.
  • 8
    Mannucci PM, Asselta R, Duga S, Guella I, Spreafico M, Lotta L, Merlini PA, Peyvandi F, Kathiresan S, Ardissino D. The association of Factor V Leiden with myocardial infarction is replicated in 1880 patients with premature disease. J Thromb Haemost 2010; 8: 211621.
  • 9
    Atherosclerosis, Thrombosis and Vascular Biology Italian Study Group. No evidence of association between prothrombotic gene polymorphisms and the development of acute myocardial infarction at a young age. Circulation 2003; 107: 111722.
  • 10
    Tripodi A, Peyvandi F, Chantarangkul V, Menegatti M, Mannucci PM. Relatively poor performance of clinical laboratories for DNA analyses in the detection of two thrombophilic mutations--a cause for concern. Thromb Haemost 2002; 88: 6901.
  • 11
    Kleinbaum DG, Kupper LL, Morgenstern H (1982). Epidemiological Research: Principles and Quantitative Methods. New York: Van Nostrand Reinhold.
  • 12
    Little J, Higgins JP, Ioannidis JP, Moher D, Gagnon F, Von Elm E, Khoury MJ, Cohen B, Davey-Smith G, Grimshaw J, Scheet P, Gwinn M, Williamson RE, Zou GY, Hutchings K, Johnson CY, Tait V, Wiens M, Golding J, Van Duijn C et al. STrengthening the REporting of Genetic Association studies (STREGA): an extension of the STROBE Statement. Ann Intern Med 2009; 150: 20615.
  • 13
    Rothman N, Stewart WF, Caporaso NE, Hayes RB. Misclassification of genetic susceptibility biomarkers: implications for case-control studies and cross-population comparisons. Cancer Epidemiol Biomarkers Prev 1993; 2: 299303.
  • 14
    Sud S, Douketis J. ACP Journal Club. The devil is in the details...or not? A primer on individual patient data meta-analysis. Ann Intern Med 2009; 151: JC12.