Sunday, April 24, 2011

Thoughts on Goetz's Decision Tree

Overall, this book by Thomas Goetz was extremely interesting. It provides a lot of information about measuring oneself, and sets forth the premise that it is becoming increasingly possible and advantageous to collect personal information and such measurements offer many health benefits. However, to take advantage of this it is imperative to know a little bit of biostatistics and epidemiology, which Goetz describes in straightforward language. He makes a number of useful points. First of all, he introduces several classic clinical studies to the reader, such as the Whitehall Study in England and the Massachusetts-based Framingham Heart Study, now in its third generation, which is known for developing many of the well-known associations we take for granted such as those involving blood pressure and stroke or the distinction between good and bad cholesterol.

Goetz also talks about how disease can be traditionally thought of as a manifestation of particular symptoms, but that it is increasingly being thought of in terms of specific biomarkers and even in terms of specific risk factors, which do not indicate that someone actually has a disease but rather his or her probability of getting it. He illustrates this with diabetes, which is now understood in terms of a single risk factor - namely, blood sugar level -- and talks about how in the future we might predict a person's diabetes risk not from this but from a constellation of subtle biomarkers. Goetz introduce a company, TETHYS Bioscience in California, which is developing such a test.

Then Goetz goes on to talk about some new diseases such as metabolic syndrome, which are essentially just constellations of risk factors for other diseases such as heart disease. He presents a particularly good case where we did not appreciate the risk factors of disease. This is the case of FDR who had a measurably elevated blood pressure level that went up during his time in office and which is clearly associated with the stroke he died from, but which at the time was not appreciated as a risk factor.

The book also talks a bit about new technologies for quantifying oneself and identifying correlates, such as personal genomics and imaging. In particular he talks about a company called Proteus, which is developing a site to integrate all of this information.

The book of course extols the benefits of early screening, particularly for cancer, but also talks about some of the statistical paradoxes with doing this -- that is someone might appear to live longer after the cancer is detected by early screening, but it is imperative to somehow back out from this calculation the additional amounts of time the cancer remained undetected before it was discovered and then calculate how long one would have lived after that.

Overall the book extols the benefit of measuring one's health and even points out a theory called the Hawthorne effect that suggests that one might see improvement just by the mere act of measuring, essentially another version of the well-known placebo effect.

The book rails to some degree against the paternalism of modern medicine and the way doctors work against this patient-centric approach, seeking to control information, and in the process do a worse job for the patients and drive up the costs of health care. He seems to argue for a future more dominated by small entrepreneurial companies and websites like rather than the paternalism of the AMA.

Overall I found this book a very good read and would highly recommend it.

Above is my review, which is also on Amazon, viz:

Some more random thoughts on the book:

* The book describes the interesting "decision tree" facing Eddy Curry
-- whether to take a DNA indicating whether he a risk allele for
hypertrophic cardiomyopathy (HCM).

* The book talks about the prevalence of FPs in screening tests - and
the paradox of PKU where treatment of many FPs is actually worse net
than no test.

* The book has nice quote from Haseltine on the problem of dark data - data
on drugs that never make it to market that might still be useful.

* An analysis of health searches by Microsoft shows that people
searching for medical problems tend to gravitate towards the worst and
most unlikely outcomes.

* Throughout, the book has some useful language on genetic tests: "only
testing to treat" and "toxic knowledge".

* The book describes imaging (CT scans) as a providing a surrogate for
proximity. It describes the remarkable way that Hounsfield at EMI
capitalized on revenue from the Beetles in the '60s to do the basic
research to invent the CT scanner. (Unfortunately, despite this
remarkable technology, the book implies that CT scans now represent a
market failure and are significantly adding to the cost of medicine in
the US.)

Above are thoughts on:
The Decision Tree: Taking Control of Your Health in the New Era of Personalized Medicine
by Thomas Goetz

My tag is [[* DecTree0mg *]]