Tuesday, May 19, 2009

Author smackdown....Sorry Dr Kari

I just received an email from Annals of Internal Medicine and I see a letter to the Authors from Dr Gulcher and Dr Stefansson. I chuckled to myself because I just finished writing a letter to the editor and had another published in Nature Biotechnology entitled "In Need of a Reality Check"

Why did I laugh? Letters to the author or editor are written to clarify issues or problems with articles in a NON-Peer Reviewed Manner. They are done to make points for the avid readers of the Journals.

It turns out that Doctors Jeff and Kari have found fault with an Article published by Dr. Ridker and Dr. Paynter PhD.....this article showed that the addition of 9p21.3 SNP data to the reynolds risk predictive model, which already includes family history of early heart disease and CRP markers....

This addition of a highly replicated SNP data ADDED ABSOLUTELY NOTHING to the predictive value of this current predictive model.  This is no surprise to me. You essentially covered almost all of the genetic factors that add big risks......blood pressure, cholesterol, family history, smoking proclivity.....so why would this test add anything????

Which brings me back to the issues.
1. Dr Gulcher and Dr Stefannson work for a company selling a 9p21 medical genetics test.
2. They think that the study is skewed because the study population is 50 year old caucasian women. (BTW, this is the hardest population to predict risk for)
3. Gulcher et.al. think that a higher risk population would benefit from their test.

What is funny about letters to the author is the fact that the authors get to defend themselves and often have the last say.....So what did the authors say about these arguments which were posed by the Chief Scientific Officer and CEO of a company which is selling the 9p21 test (Sarcasm included)?

The Author, Dr. Paynter says
1. The argument that reclassification would occur more often in high risk patients has not been proven and is presumptuous.
2. The studies cited for support by Dr Gulcher and Dr Stefansson didn't examine whether the reclassification was more accurate.....i.e. if it is a misclassification, does it count as effective? Not so much.
3. The utility of the Net Reclassification Improvement is valid and that the attack offered up by the deCode team misunderstands what the NRI really is.....

All in all, Dr. Paynter laid the smack down on the Doctors from deCode.......
BTW, this study is great evidence as to why the bar is so very high for reclassifying patients with any risk for heart disease based on anything other than those which increase the risk greater than the current non-genetic risk factors....

The Sherpa Says: I remain convinced, predisease risk prediction based on SNP data is much weaker than current clinical tools.....WHICH HAVE BEEN STUDIED FOR DECADES!!!! Not less than 2 years.....Newly discovered risk markers will require 5 to 7 years to validate and another 5 to be implemented....Plain and simple.

1 comment:

Steve said...

What would interest me, as someone who was adopted and has limited family history, is: how effectively would the SNP data substitute for family history data used in the current disease risk models?

The effectiveness of SNP profiles to substitute for family history would probably be disease-dependent, and so the findings for one disease may not apply to other conditions. But it seems to me that genotype information has the potential to provide more detailed and accurate information than could be provided by family history medical records. Such records likely will have variable quality and depth-of-detail for different patients, or as in my case, almost entirely lacking.