Likda K's Thread cont/hepatitisresarcher/Willing/Ina/Others

[jmjm530]

Willing: Perhaps this seems like quibbling over many RNAs fragments can hide on the head of a pin
---------------------------------------------------------
Probably more HCV RNA on the head of a "needle" as opposed to "pin" per the Talmudic expression :)
-------------------------------
Anyway, Thought I'd open this up to continue the interesting dialogue with hepatitisresearcher and others. Hopefully, Linda K's original question got answered along the way :)

[Hepatitis Researcher]

I forgot to answer your question regarding the fibroscan: Non fasting will shift the results towards higher median stiffness.
While I discussed that with Echonsens at the AASLD 06 I am sure it will not cause any protocol change now that they have progressed so nicely. I certainly do not want them to be mad at me, since I need them to service my machine and give me software updates.

[jmjm530]

Given that I assume most labs don't separate their steps by city blocks, I can understand better why there may be so many false positives. Actually, my worry was more about false negatives and that's why I ordered redundant tests post treatment. What a mess if it turned out that my Heptimax was positive and my HCV RNA qualitative was negative or vice versa. Fortunately, it didn't happen or maybe it did and the lab computer picked it up :)

As to Shiffman, assuming the half log variance in either direction, if someone tested pre-tx at 1 million IU/ml, then in actuality (if my math is correct) they could be anywhere from 450,000 IU/ml to 5,000,000 IU/ml. If so, this seems to destroy the theory of pre-tx viral load as being a predictor of SVR in the sense that the cutoff of "low" VL is usually around 600,000 IU/ml. So is this person low or not? Maybe it all balances out statistically with large populations but an individual has to wonder where they really stand.

-- Jim

[jmjm530]

Thanks for getting back and what you say makes sense. It is also a bit reassuring since I took my second scan non-fasting and either took my first scan fasting or non-fasting can't remember. In either event I dropped close to a stage or possibly more given one of the fast/not fast scenarios.

That said, my understanding is that part of these trials is to come up with correlations in the American "model" between the computer generated Fibroscan number, and fibrosis stage as measured by needle biospsy.

Assuming that these subjects are a mishmash of fasting and non-fasting participants I assume this would throw the final fibrosis scores off for any one individual, and yet, these scores are being compared to relatively recent liver biopsies with apparently very good correlations. I assume this is because the differences between fasting and non-fasting readings are relatively small? For example, how much of a difference have you seen between let's say someone who reads stage 3 in a fasting state and then re-tests same day in a non-fasting state?

Thanks again for all your information and insights.

-- Jim

[willing]

not quite: 1 million IU is 6 log units so +- 0.5 log units puts the range at 5.5 to 6.5 in logIU or 316,000 to 3,162,000 in IU, but I agree that it seems a very wide range ((that's actually one of the examples in Shiffman's review)). As a counterexample to Shiffman, look at abstract 350 from the recent AASLD. T Berg reported that low-VL as an SVR predictor was distinctly superior when a cutoff of 400,000 was used rather than the older 800,000. However, per Shiffman if your "true" VL is 800,000 readings down to 251,000 are within the margin of error.. (I tend to believe Berg here)

[Hepatitis Researcher]

The Shiffman 06 paper was not accessible in full text through UCLA, the o1 paper did not contain the .5 log accuracy info that you are referring to and the link to the clinical care options did not work and the search for Shiffman on the internet proves that it is a very common name in the US.
Thus I cannot yet comment on the .5 log Shiffman accuracy in discussion. I have to check NGIs database to see where NGIs superquant PCR accuracy stands. It should be much better than .5 logs. bDNA for values over a Million are BTW very reliable. I do know that Amplicor was doing badly in the past and could explain at length why that is - a very technical issue.

[jmjm530]

I guess the lower your acutal number, the lower the range, so maybe statistically it all works out with large numbers of patients, but appears to leave some doubt on a single patients viral load result. Fortuantly, as viral load decreases toward zero, this becomes less of a problem or we'd really have a mess in dx those non-detectible. For example, a half log variance at 10 IU/ml would be -- I'm approximating so don't kill me here :) -- anywhere between 5 IU/ml and 50 IU/ml. Certainly a much smaller absolute spread.