Thursday, November 6, 2014

Warning: snarkiness ahead.

It's probably not a good month to be my stagiaire. 
On the surface: calm, subdued, not at all excitable.
But just scratch a bit, and PAF.
Case in point - you would expect a third year graduate student studying the involvement of the BRCA2 gene in hereditary breast cancer would have some idea of what hereditary breast cancer looks like. That's easy. It happens early, happens to close relatives, and is often accompanied by ovarian cancer. You can hardly navigate the BRCA literature without being beaten over the head with this, and you might be incited to include prostate and pancreatic cancer as part of the spectrum.

So the student shows up with a bunch of samples - 40 familial, 200 sporadic, and 150 healthy controls to do a case-control study to see if BRCA2 is involved in familial breast cancer in Tunisian women. As if the thousands of papers on the subject from all over the world - including Tunisia - did not already show that null mutations in BRCA2 are associated with 10-fold breast and ovarian cancer risks, and that neutral polymorphisms are just that - neutral.

Well, if she's here to reinvent the wheel, at least let's make sure it's a round wheel.

How were the familial cases chosen?
They're familial.
Yes, but what criteria did they meet to be defined as familial breast cancer cases? (Age at diagnosis matters, as does the degree of relatedness between affected relatives.)
Um, we asked the case if she had a relative with breast cancer, and we put it into a computer program that told us more than 50%.
Ah. What computer program was this ? (there are several commonly used)
So much for the familial cases. What about the sporadic cases? 
They're not familial.
Yes, but did you assign a cutoff for age at diagnosis? Did any of them have relatives with ovarian cancer?
Not only doesn't she know, but she doesn't think it important to know. Has she read anything at all? I don't even get into the control samples. In a case-control study you really have to vet the controls and be sure they represent the absence of what you're looking for in the cases - a control with two sisters sick with breast cancer is not a control here. Plus, if your controls don't have breast cancer but they're only 30 years old, well, that's not any good either when your cases are all 50+.
OK. Here's one of the papers that reviews the BRCA genes and hereditary cancer risk. It has a handy table in the back that we use to quickly score families for the likelihood of carrying a BRCA mutation. How about you calculate the score for each of your cases, and then we'll take a look together and reassign familial versus sporadic cases accordingly.
Uh, I can't. We didn't ask at what age the relatives had cancer.
Great. In the Eisinger scale, only female breast cancer is weighted by age at diagnosis, so if she has the age of the index case, assigning a fixed weight to each relative just according to type of cancer shouldn't get us too far off track.
She's only here for 2 months, which is just time to test about half the samples she brought. So we need to decide which half. The project is poorly designed to begin with, and just cutting the groups in half randomly will exacerbate the effect of mistakes in assigning samples to one or another group. The study can be improved by selecting the real familial case group by applying the Eisinger scale to all her cases, then identifying a sporadic group selected for the absence of an enlarged definition of family history plus age at diagnosis around the average for Tunisia, and then taking only the oldest of the women in the control group (excluding personal or family history of ovarian cancer, though I doubt she has that information).

It bugs me that a 3rd year grad student knows essentially nothing about her subject. What's she been doing for 2 years? Who I'd really like to throttle is her boss, who came up with this nonsense project, who has not insisted that the student read and understand the literature on the subject, and who cranks out lame little papers that give the appearance of saying something when they really don't because the methodology just isn't there.

No comments: