Linda and Ramses the Great
His chief wife was named Nefertari.
Of her he wrote, "She for whom the sun duth rise"
I know how he felt.
Well, after an abnormally warm and pleasant summer the Bellingham we all know and love has reasserted itself today. I awakened to hard rain, followed by a wind that blew all
my plastic chairs into a heap on the north side of the deck, banged things around
so loudly that my cats hid under the couch, and threatened to take the door off
my greenhouse. Thank God global warming
hasn’t made our climate too comfortable; otherwise masses of people would move
here and I would have to move to Montana.
With plenty of excuse to stay inside, I set myself to
figuring out what ROCA means. You
may remember ROCA, which stands for Risk of Ovarian Cancer Algorithm. I mentioned it once or twice before, as – in
a sense – a rival to the research being done by the Fred Hutch group I
try to help. In common with our trial,
the ROCA trial involves following the changes of the concentration of the antigen
CA 125 in the blood of individual women, and basing decisions on whether or not
to send them for TVS (trans-vaginal ultrasound) on how the concentration
changes. That’s how our group does it too,
only, instead of using ROCA, we use our own home-cooked statistic, the “parametric
empirical Bayesian longitudinal algorithm.”
I have also written of this puzzling object before, largely because its
name is funny. I think I know what PEB means – the B stands for Bayes, a
statistician of a bygone century, who developed a method to re-compute the
probability of something happening as additional evidence pile up. Thus, if a woman’s CA 125 score suddenly
increases, our PEB will tell us what that means, and whether or not to send her
to ultrasound.
Well, ROCA must work something like that. The group working with it says that it has
high “sensitivity” – it does a good job of catching the condition in women who
have OVCA, and gives few false positives.
Another way of saying more or less the same thing is that it has a high
“positive prediction value”, which means that of those who test positive, most
have the disease. It remains to be seen
whether our method works better.
Anyway, I spend most of a rainy morning dredging through the
Internet, trying to find out just how the devil you calculate ROCA. I finally ran it down to a simple formula: if
a woman’s CA 125 score exceeded her "nadir" by a certain amount she landed in
the high risk category, if it exceeded nadir by a lesser amount she was in the
elevated risk category, and if it didn’t exceed her nadir at all she was in the normal
risk group. The one thing I couldn’t
figure out was – what the hell is a “nadir”?
Several hours later I gave up and went to practice the piano.
Anyway, I still put my money on cancer sniffing dogs (see
blog entry for 9/6/13)>
I blogged about ROCA yesterday. I thought I understood it, although I came a cropper over the word “nadir”. I know what nadir generally means, but it doesn’t seem to be appropriate in this context. As it seems to be used in cancer studies, nadir refers to the lowest point in a person’s blood count after undergoing a chemo treatment. I spent most of today researching ROCA, and now I understand it much better – although I couldn’t do the fancy statistics entailed, of course. The key aspect seems to be locating a thing called a “change point” For you who are mathematically inclined; I take “change point” to be equivalent to an abrupt change in the second derivative of a function – that is, a significant break in slope. ROCA seems to be a way to determine whether a woman’s serial CA 125 counts adhere more closely to a change point curve than to one that is steady.
ReplyDeleteSo, big deal. The reason I am writing this is to point out that the basics of this method was published in 1995! Why, then, in hell hasn’t it been implemented, after 18 years! Is the FDA or the NIH dragging their collective, bureaucratic feet? Are the researchers themselves ridiculously frightened of the possibility that they might make a mistake? How much does it cost to do a CA 125 measurement, anyway? Not much, I suspect. I think much of the cause lies in the fragmented nature of our health care system, as well as in deficiencies in the ways advances in medical research are disseminated to front line doctors. I never thought I’d say this, but maybe we need a single-payer system, a la the Brits and the Canucks. If Linda had been covered routinely by ROCA-based monitoring she might still be alive.