Comment on Inappropriate subgroup analysis

[Guest guest]

A few comments on my prior post are required.

Correlations between the characteristics of patients and better outcomes

are certainly a VALID basis for what to study further, particularly when

there are plausible explanations for the association, such as normal genetic

characteristics specific to interactions between the agent and the tumor.

Point taken . and this is also a valid way towards achieving personalize

medicine - getting the right drug to the right patient.

However, I think it's inappropriate for a company, in this case BIOVEST, to

advance this type of analysis as evidence of efficacy without testing it

prospectively . to write:

" Biovest conducted this subset isotype analysis and discovered a

**fundamental relationship** between vaccine isotype and the efficacy of

BiovaxID " .

I asked for confirmation about my concern from an expert advisor, who

responded:

" If they [biovest] ever re-run such a trial, complete accrual, and

prospectively confirm this finding, then it is one thing. Otherwise, is just

a retrospective analysis of a largely incomplete set of patients.

.. For years investigators told us that V/V 138 FcRIII predicted response to

vaccine (twice in JCO). Then, when they tried to verify it prospectively in

the Genitope trial and... oops, untrue. "

In the report on subset analysis the word " LUDICROUS, " hurt some feelings,

understandably .

.. But, it was the correlation to date of birth that the author's described

as ludicrous (an obvious non-plausible explanation for a better outcome),

NOT that it's ludicrous to find subset analysis persuasive!

..Indeed, the second article makes the point that even physicians (very smart

people) are sometimes influenced by such analysis:

Sleight writes:

" Plausible explanations can usually be found for effects that are, in

reality, simply due to the play of chance. When clinicians believe such

subgroup analyses, there is a real danger of harm to the individual

patient. " <http://www.ncbi.nlm.nih.gov/pmc/articles/PMC59592/>

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC59592/

A core part of PAL's mission is to learn how to read reports critically,

which includes looking at the limitations of the methods used to reach the

author's or sponsoring company's conclusions.

In this case, a correlation with date of birth helps to prove the point

about subtype analysis . but it does not speak to the plausibility of

vaccines, which is obviously high because vaccines have been tested in large

scale studies three times.

Finally, without standards for evidence the practice of medicine would be

based on theory and sales pitch. In such a world we'd have thousands of

choices, but no basis for making a helpful one, and no foundation on which

to make real scientific progress.

Karl

PAL www.lymphomation.org

Evidence based information, independent of health industry funding.