PEB - Pacific Edge Ltd

[psychic]

Accept that's some scary stuff for anyone who was left wondering, (thankfully not me, was a bad UTI) but if the test isn't funded by the local DHB's its hard to see it getting much traction here.

Well, no question that the NZ DHB's have little to throw around on new tests. Whether they will run with it later we will see. They don't use the other established biomarkers and given the market size here, it is not hard to see why PEB place little emphasis on it.

With growing awareness, I reckon there will be a few here interested in paying for it privately - am sure I would.
Wonder where Sth X etc are up to with approval process?

[Snow Leopard]

Yes, I understand that, but this gives you the prevalence of the disease only within that 485 presenting with Haematuria. Is this group representative? Are there other factors that come into play when calculating the PPV? Just askin....

It is this group of 485 on which the trial was conducted and it is this group from which the results are derived, and thus should be the basis for PPV as it is for Sensitivity etc.
If PPV is calculated on a different group (from everything else) and they are not providing any information that it is then frankly such a thing would be very unethical.

So although I believe they are wrong again I believe it is just another mistake.

Part of the theory for calculating a sensitivity and specificity is that this is independent of the particular prevalence of the condition in the group and these values can then be taken and the numbers crunched for a different population. (The Spanish, we are told, have a higher prevalence than many other nations for instance).

Now for the rub:

If you wander over to this Online Diagnostic Test Evaluation Calculator you can bung some numbers in:

This is what you get for the cytology bit from the test numbers (I could be one or two out but it is not important):

Attachment 7119


and this is what you get when you make the test ten times larger with the exact same ratios:

Attachment 7120

The only difference is the confidence in the accuracy of the results as a result of the larger sample size (basic statistics). So the 95% confidence level for sensitivity goes from it is probably between 43.3% to 68.3% to a narrow 52.2% to 59.9%.

So in summary. The particular prevalence matters for the quoted PPV and it would be really good to have some results from other trials preferably based on a larger sample size.

I will leave you to play with the above calculator and the CxBladder results.

Best Wishes
Paper Tiger

[psychic]

Nice Link!! Better than my back of the envelope, but gobsmacked I got it right..! lol

But wait. it then says

If the sample sizes in the positive (Disease present) and the negative (Disease absent) groups do not reflect the real prevalence of the disease, then the Positive and Negative predicted values cannot be estimated and you should ignore those values.
Alternatively, when the disease prevalence is known then the positive and negative predictive values can be calculated using the following formula's based on Bayes' theorem:


Well, I'm not taxing myself further with this Bayes Theorem thing - sounds ghastly. But does this not suggest that the PPV should abe calculated using a known wider prevalence. Presumably if the same prevalence is used, then it allows comparison of tests..?

Ducks for cover.....

[Minerbarejet]

It is this group of 485 on which the trial was conducted and it is this group from which the results are derived, and thus should be the basis for PPV as it is for Sensitivity etc.
If PPV is calculated on a different group (from everything else) and they are not providing any information that it is then frankly such a thing would be very unethical.

So although I believe they are wrong again I believe it is just another mistake.

Part of the theory for calculating a sensitivity and specificity is that this is independent of the particular prevalence of the condition in the group and these valuescan then be taken and the numbers crunched for a different population. (The Spanish, we are told, have a higher prevalence than many other nations for instance).

Now for the rub:

If you wander over to this Online Diagnostic Test Evaluation Calculator you can bung some numbers in:

This is what you get for the cytology bit from the test numbers (I could be one or two out but it is not important):

Attachment 7119


and this is what you get when you make the test ten times larger with the exact same ratios:

Attachment 7120

The only difference is the confidence in the accuracy of the results as a result of the larger sample size (basic statistics). So the 95% confidence level for sensitivity goes from it is probably between 43.3% to 68.3% to a narrow 52.2% to 59.9%.

So in summary. The particular prevalence should not matter and it would be really good to have some results from other trials preferably based on a larger sample size.

I will leave you to play with the above calculator and the CxBladder results.

Best Wishes
Paper Tiger

Thanks for that PT.
We await the release of further trial statistics conducted of which I believe there are a few in the offing.
One thing of note though in your offering was the line that unless the prevalence is known then PPV can largely be ignored.
This raises the question , was the expected prevalence of the group known before the trial or was the PPV established as a result of the trial.
Hmmm. Further work required
Cheers
Miner

[Snow Leopard]

Nice Link!! Better than my back of the envelope, but gobsmacked I got it right..! lol

But wait. it then says

If the sample sizes in the positive (Disease present) and the negative (Disease absent) groups do not reflect the real prevalence of the disease, then the Positive and Negative predicted values cannot be estimated and you should ignore those values.
Alternatively, when the disease prevalence is known then the positive and negative predictive values can be calculated using the following formula's based on Bayes' theorem:


Well, I'm not taxing myself further with this Bayes Theorem thing - sounds ghastly. But does this not suggest that the PPV should abe calculated using a known wider prevalence. Presumably if the same prevalence is used, then it allows comparison of tests..?

Ducks for cover.....

Thanks for that PT.
We await the release of further trial statistics conducted of which I believe there are a few in the offing.
One thing of note though in your offering was the line that unless the prevalence is known then PPV can largely be ignored.
This raises the question , was the expected prevalence of the group known before the trial or was the PPV established as a result of the trial.
Hmmm. Further work required
Cheers
Miner

What that means is:
The prevalence it calculates is the prevalence of the condition in the numbers you input and
if the numbers you input do not have the same prevalence as the condition in the real world then the calculated prevalence is not the prevalence in the real world.

So the prevalence in the test was what it was and the PPV for the test is what it was
BUT in the real world the prevalence may be different and thus the PPV may be different which is why the really important numbers are the sensitivity and the specificity (which are affected more by sample size).
BUT
PEB decided to throw a PPV out there.

OK - that was definitely the last post from me for a while on this subject.

Next up, on another day, will be why 100% sensitivity is only half the story and we need to talk more about specificity.

Best Wishes
Paper Tiger

[Absolute144]

No - those pages have a footer which contains within the HTML "Last updated: 16 July 2014"
You are just making an assumption.

Best Wishes
Paper Tiger (Last Updated: 1-Jan-2000)

Thanks for the scrutiny

[Minerbarejet]

Thanks for your input PT.
Seems that prevalence has little relevance except in the context of the test involved
Which makes it unusable to relate to another set of parameters from another test.
Perhaps it is about now that the algorithms of triage come into play with demographics, size, shape, smoker, age, etc all affecting the prevalence of any trial conducted.
Interesting times
Miner

[Mista_Trix]

No commentary re: the price sensitive trading halt we just went through?

[blobbles]

European patent granted. Covering most countries in Europe... I believe this was expected though, shouldn't ruffle too many feathers?

[Schrodinger]

I think they are announcing they have tens of thousand tests oops that's been done before.