RSS Feeds
-
In Response:
Submit responseWe appreciate the writers’ interest in our paper and the opportunity to address their questions.
Dr. Gulcher brings up the point of the distribution of risk in the population. While it is indeed correct that the overall risk in our population is low, other variables have improved prediction in this cohort(1). We do not believe that examining the utility of 9p21 genotype in risk classification of a higher risk population would necessarily show different results and we look forward to analyses done in different cohorts. Unfortunately neither of the studies cited in the letter have examined whether the reclassification was more accurate, which we feel is more important than the total number reclassified. The writers also question the utility of the Net Reclassification Improvement (NRI), given the small numbers of events in each cell. However, the NRI does not look at each cell individually – instead it examines whether cases in aggregate are reclassified into higher predicted probability cells more often than into lower and whether controls in aggregate are reclassified in to lower more often than higher. Therefore, we feel the use of the NRI is quite appropriate and not underpowered.
Dr. Kiljanek brings up the concern that the performance of the genetic marker is reduced by the inclusion of C-reactive protein (CRP) and family history of premature myocardial infarction in the prediction score. As shown in Table 3 of our paper, the relative hazard for the genetic effect is not changed with the addition of CRP and family history. When reclassification is assessed using the traditional covariates alone, which do not include either CRP or family history, the improvement shown with the genetic variant is modest at best. Since, as shown in Table 1 of our paper, the 9p21 variant is not associated with CRP, the reduction of the genetic effect from modest to none is largely the result of the inclusion of family history. Our results suggest that, compared to variation at 9p21, a simple question about premature parental MI is equally as good, if not better, at classifying future risk of cardiovascular disease.
Dr. Schwarzenbach raises the question of an interaction between 9p21 or KIF6 genotype and statin therapy response. We agree that this is an important question and while we do not have the ability to examine it directly the Women’s Health Study, we look forward to additional papers on the subject from other groups.
References
1. Cook NR. Comments on 'Evaluating the added predictive ability of a new marker: From area under the ROC curve to reclassification and beyond' by M. J. Pencina et al., Statistics in Medicine (DOI: 10.1002/sim.2929). Stat Med. 2008;27(2):191-5.
-
Paynter study does indeed support use of 9p21 for cardiovascular risk
Submit responseTo the Editor:
Paynter showed that the widely replicated 9p21 genetic risk factor affected risk in WGHS, largely independent of traditional risk factors. They were able to reliably measure this genetic marker years after patient recruitment because DNA is invariant, in contrast to biomarkers like CRP. The power of detecting improvement in cardiovascular event prediction by adding a new risk factor to traditional risk factors depends on the number of events. WGHS recruited women of mean age 52 who had low risk - 2.1% event rate over 10 years. This would be analogous to following a cohort of 42 year old men because events in women peak 10 years later. The authors conclude that the 9p21 marker is not useful because 2.7% of 22,000 patients are reclassified. However, their Table 4 shows that 85% of the cohort (18,500) is biased towards the lowest risk category (1.5% average rate). What the authors fail to point out is that no single risk factor (including CRP, family history, or genetic marker) is sufficient by itself to move such low-risk patients to the next risk category. In contrast, patients in the intermediate and intermediate-high risk categories using either ATPIII or Reynold’s are indeed reclassified by 9p21 at a much higher rate of 12-14%. This is consistent with reclassification rates in those categories of 17-19% reported at AHA 2008 by Ballantyne’s group using the ARIC study with much higher event rates (1350 events out of 10,000) and the 22% published by Humphries' group in NPHS-II (270 events out of 2750) (1, 2).
Patients near or above 5% 10 year risk are the very patients where risk stratification makes the greatest difference in terms of LDL target according to NCEPIII. Furthermore, we would suggest that use of Net Reclassification Improvement is not appropriate here where the limited number of events, further divided into 10 categories, significantly underpowers the assessment – some of the cells include only 7 to 9 events; a misdiagnosis of just one event can lead to 100 percent accuracy in the two cells claimed to be less accurate. In summary, we suggest that the Paynter study is indeed consistent with the conclusion of two other prospective studies that had 4 times more events than their study; that is, addition of 9p21 leads to reclassification of significant numbers of patients, especially in those with 5% or greater 10 year risk, into more accurate risk categories leading to different LDL targets.
References
1. Brautbar, A, Ballantyne, C., Lawson, K., Nambi, V., Chambless, L., Folsom, A., Willerson, J., Boerwinkle, E. Impact of Adding a Single Allele in the 9p21 Locus to Traditional Risk Factors on Risk Classification for Coronary Heart Disease and Implications for Lipid-Modifying Therapy in the White Population of the Atherosclerosis Risk in Communities (ARIC) Study. AHA Meeting 2008, Abstract 5090.
2. Talmud, PJ, Cooper, JA, Palmen, J, Lovering, R, Drenos, F, Hingorani, AD, Humphries, SE, Chromosome 9p21.3 Coronary Heart Disease Locus Genotype and Prospective Risk of CHD in Healthy Middle-Aged Men. Clin Chem. 2008;54:453-5.
Conflict of Interest:
Employment and ownership in deCODE genetics, deCODEMI genetic risk test for MI using 9p21 and its patent
-
Comments on: Cardiovascular Disease Risk Prediction...
Submit responseTo the Editor
I believe a question has to be addressed to the authors of the article. Can't it be possible that rs10757274 risk alleles are already associated with ATP III, elevated C - protein level or family history? This could be the reason why including this genetic marker into prediction didn't improve C-index, couldn't it be ?
Conflict of Interest:
None declared
-
Screening
Submit responseTo the Editor:
The question is, can the 9p21 or even the KIF6 gene be useful in determining who might be more likely to respond to statin therapy? There are probably a lot of people who receive statins who don't need them based on NNT numbers from prior several trials. Can we, using these genetic tests, be more specific as to whom we treat or do not?
Conflict of Interest:
None declared