Re: Misclassification Rate = FIFTY PERCENT, NOT 550%

[Guest guest]

>

> This is the same data presented earlier, calculated for the

> misclassification rate of a positive urine EtG representing proof of

> alcohol ingestion.

>

> From the Department of Clinical Neuroscience, Section of Alcohol and

> Drug Dependence Research, The Karolinska Institutet, Stockholm, Sweden

> published initially in a dissertation written earlier this year by Helen

> Dahl <http://diss.kib.ki.se/2006/91-7140-644-1/thesis.pdf> . The

> results of her thesis have subsequently been published in several

> articles contained in peer reviewed journals listed by PubMed and

> Medline. I did not seek permission to reproduce her data as it is

> publicly available on the internet.

>

> On page 24 of Ms. Dahl's dissertation she reports, " During the placebo

> period two of the subjects showed positive results for urinary EtG,

> albeit at low levels. "

>

> Using Ms. Dahl's data, the following statistical table is generated. At

> the time of this writing Ms. Dahl's data represents the only published

> peer reviewed measurements available for calculation.

>

>

>

>

>

> ACTUAL

>

> OUTCOME

>

>

>

>

>

>

>

> EtOH ingested

>

>

> No EtOH ingested

>

> TOTALS

>

> PREDICTED

>

> EtOH ingested

>

> 9

>

> 2

>

> 11

>

> OUTCOME

>

> No EtOH ingested

>

> 0

>

> 7

>

> 7

>

>

>

> TOTALS

>

> 9

>

> 9

>

> 18

>

>

>

> Misclassification Rate = 50%

>

>

> (a + c) / (a + b + c + d)

>

> Misclassification rates are calculated to assess the risk of harm during

> situations where overtreatment may occur. If the intervention is not

> harmful to the person, then it's not a significant issue. If the

> intervention carries a risk of harm with it, such as bleeding from

> arterial access during angiography, or dye reactions, then it becomes

> very significant. With respect to urine EtG testing, there is clear

> risk of harm to an unaffected individual if the same intervention is

> performed upon them as for an affected individual.

>

> None of these kinds of calculations have been performed to date in the

> literature surrounding urine EtG use. Dang it! If I could secure some

> protected time I could get a worthy publication by subjecting the

> available urine EtG data to metaanalysis and rigorous statistical

> examination and peer review. Anyone got access to grant money? (I'm

> not entirely kidding. None of it would go into my pocket, but it would

> allow for protected time from active ward/teaching rotations to author a

> paper with some world famous experts in the field of applied clinical

> statistics.

>

> I remain firmly convinced the use of urine EtG testing will become a

> case study in future lectures about the misapplication of clinical

> statistics. It's shameful that a test like this can get as far as it

> has without an enormous outcry from the American College of Pathology

> and the American Society of Addiction Medicine. I believe the boards

> will begin to spontaneously back off simply to not appear stupid, which

> they are not. I also think the clock speed for all of this is slow.

>