Breast Cancer - Unknown vs Known Family History
September 3, 2019 2:00 PM   Subscribe

While filling out the mammogram questionnaire, I noticed there was no way to indicate "Unknown" for any type of family history. Not just whether or not someone was diagnosed, but also when they were diagnosed. As a data person this concerns me - Unknown is not the same as No. Is this a thing (for Drs and researchers)?

The familial risk factors have changed over the years, and what I thought I needed to know 20 years ago is different than what medical professionals want to know now. For me personally, there are reasons I may never know some details, and it got me wondering about various scenarios of what is known.

First, I think there's the scenario of you know your maternal grandmother had breast cancer but you don't know what age she was diagnosed or if she died of it. This is one of my scenarios. It was important to know that my maternal grandmother had breast cancer so that is all I knew growing up. Not saying "Yes" to "diagnosed under 50" implies no she wasn't, but that's not the same thing as not knowing.

Second is that the people who know these things aren't around anymore. People die. Or are estranged. You may not ever find out when your maternal grandmother got cancer, or you might not even know she had cancer because your parents died young or whatnot. Esp as researches hone in on what is really a factor, the more detailed the questions will get, and it seems like the more blanks people will have in their family medical histories.

A third scenario is you don't know your blood family medical history for whatever reason. They died too young and that knowledge wasn't passed down. You were adopted. Your biological father isn't who you think it is. I would think this is important to know, so why was this not on the questionnaire?

Another piece of this, which is a bit different, is what if you don't have any aunts and/or sisters? Again, there was no way to indicate this, and yet it seems important to say they don't exist vs no because no implies they exist and didn't get breast cancer.

All of these things seem like they would be important distinctions for researchers, or at least for patients wanting a clearer picture of their risk. But are these things in fact important? Has it been looked at and dismissed? or were they not considered at all? Does anyone happen to know? Or know who I'd ask?
posted by evening to Health & Fitness (9 answers total) 3 users marked this as a favorite
Best answer: As someone who used to work in health outcomes measurement, yes, these things are important, yes, they are looked at by researchers, and getting people to change the surveys they use in clinical practice is very difficult. There’s a whole lot that happens in medicine because it’s always been done that way.
posted by coppermoss at 2:52 PM on September 3, 2019 [7 favorites]

Best answer: I can't speak for breast cancer research, but I do epidemiological and health outcomes research in Parkinson disease, and yes this is a Thing We Think About. (Not so much on the genetic side, that's not my area, but environmental risk/protective factors definitely.) There is a lot of "Well, doc, I had a summer job 40 years ago spraying for mosquitoes; I know it wasn't DDT but I don't know what was in it, and no we didn't wear masks back then." From a research perspective, I would code that as "possible" under Pesticides.

When I do write out family histories clinically, I usually word things as "No known family history of Huntington disease or any other choreiform disorder" or "Maternal grandmother diagnosed with Alzheimer at 80, symptom onset late 70s" or "Mr. X does not have any daughters" or "Mr. X has one daughter who does/does not have a son with intellectual disability." (We get down into the nitty-gritty in neurology, although I admit I have not yet managed to ask some 70+ year old guy if his middle-aged daughter is still having periods.)

I'm surprised that whatever questionnaire you're completing doesn't have an option for N/A, at least. Our questionnaires (both clinical and research) have an option for "adopted" and for "unknown" under family history. People who respond adopted/unknown are excluded when we run our analyses, precisely because we can't say one way or the other. In your fourth example, someone with no sisters and whose mother does not have breast cancer would simply not be included in an analysis of first-degree relatives. Clinically, I take their risk to be about the same as the baseline population, unless I have compelling evidence to the contrary.

I hope that makes sense. But yeah, it's something most doctors care about and we try to account for the scenarios you present.
posted by basalganglia at 3:04 PM on September 3, 2019 [4 favorites]

Best answer: As someone who has worked on the writing of questionnaires for diagnostics used in clinical practice (I think that's what you are filling out), I can add that this part of test development seems to be something that doesn't always get the amount of attention, thoughtfulness, and care that it should.

It may make very much make sense to have an "Unknown" or "N/A" option, but if the person/team responsible didn't have the experience, the time, or just the concern enough to put it in, it might have gotten overlooked.

And then once diagnostics are developed and in use, it's very difficult to change inputs....
posted by Tandem Affinity at 3:57 PM on September 3, 2019 [2 favorites]

If you don't fill out that question, what happens? Do you actually talk to a doctor? If this is to guide your care - as opposed to for research - there is often an opportunity later to add nuance.
posted by Lady Li at 9:00 PM on September 3, 2019 [1 favorite]

If you don't fill out that question, what happens? Do you actually talk to a doctor? If this is to guide your care - as opposed to for research - there is often an opportunity later to add nuance.

This is the exact answer I got the other day, when I asked the doctor almost the same question because I couldn't remember exactly when my sister was diagnosed with cancer, and it took her a while to reply my text message.
I have the BRCA2 gene mutation, and my data are very well mapped for research. The doctors even found an entire "new" branch of our family in a different part of the country*. The questionnaire I had filled out was for my care, not for research purposes.
posted by mumimor at 11:32 PM on September 3, 2019 [1 favorite]

Information can also evolve over time. We always thought my grandma had a/some strokes. Then two of her four adult children were diagnosed with forms of Parkinson’s and people started saying wait a minute, what if the ‘series of strokes’ was actually ‘progressing Parkinson’s’ and we didn’t know? No way to find that out at this point...
posted by ficbot at 4:51 AM on September 4, 2019 [1 favorite]

If the question is whether to mark "yes" or "no" given only these choices, I'd mark "yes" or leave these blank. A "yes" or blank should sort to "further discussion is needed" whereas a "no" is likely to sort to a "no further discussion is needed unless this is a positive/indeterminate result." But all of this is changing as recommendations are updated and labs are able to have their internal systems updated to comply. And highly nuanced details may not be needed---it often is enough to know the decade of dx. Info that might be valuable today may not even have been recognized as important during that relative's care (like even 10 yrs) ago and so wasn't even thought to be documented.

From a clinical care perspective, detailed family history (famhx) needs will vary. For mammogram that is absolutely negative or absolutely positive, the radiologist may never particularly review your hx. And if the mammogram is indeterminate, your clinician will be talking to you further anyway. And for those who have cancer, treatment options will be driven by other factors than famhx.

The molecular genetics of a FULL cancer gene analysis are somewhat similar; clearly positive or clearly negative results may benefit by knowing famhx but a "variant of unknown significance" may need famhx for final classification. A targeted screen for only select variants (like 23andMe) is more iffy because famhx does play a much larger role in determining if that screen was sufficient.
posted by beaning at 8:26 AM on September 4, 2019

Response by poster: My questionnaire was for both patient care and research - there was a box at the end to opt out of my info being used for research (or at least shared with the local cancer center).

I've had a hard time trying to find research on known vs unknown history, etc. I could only find this study where about 40 out of 1500 or so patients that didn't know their family history or were adopted, and the history included the age that family was diagnosed with breast cancer. The idea that out of the half of patients who had a family history of breast cancer, that all of them knew the age of diagnosis just seems suspicious to me.

This is not about what I should put down or not put down for my questionnaire, but rather about the research and how they are making these distinctions. You have the patient filling out the form, the person inputting the form into the computer, and then the people analyzing the data. Especially between the first and second there can be a real difference depending on how things are worded and instructions (or lack thereof). And then not talking about that in the paper makes it hard to tell how different scenarios are categorized and then how that might affect the end results. In my own non-research data work these are all things I'd make sure my audience knew about as definitions can be really important in making more accurate assumptions from the data.
posted by evening at 8:47 AM on September 4, 2019

Best answer: Generalizing a bit from my experience working in health care research: Generally, anyone giving a questionnaire that's designed to be used for research should have been trained on how to answer questions/clarifications by the patient. (Sometimes, that training might literally be "you're not allowed to give them any particular clarification, just tell them to do whatever seems best" but it's training, anyway.) Definitely feel free to ask all the questions you want, and they should be able to help clarify what they're looking for. But some people will ask for clarifications and some won't, and some will just guess, and the data will be messy, and that's - just the nature of health data research. Patients who came in with all the details of their family history would have been great, but that basically never happens, so sometimes all you can do is say "take your best guess, do you think your mom was between 30-50"?

In particular, in a clinical-research setting, it might be useful to understand that often the questionnaires are not what the researcher would ask, given a free hand and the ability to design their own forms. The forms are potentially the ones the hospital already uses, and the researchers can't get the hospital to agree to the additional work of asking new questions so they've accepted subpar forms rather than none at all. Or the forms are being used at twenty different research sites and your local doctors have absolutely no control over the wording even though they file complaints with the central site periodically to let them ask useful questions. Or they're trying to do a follow-up study on research from thirty years ago and they're carefully using the same wording for comparability even if it's not what they'd ask for a brand new study now. Or they've put in a review board request to change the form wording and ask more questions but it's stuck in review hell and has been bouncing back and forth between reviewers for three months. Or they have no intention of actually looking at this particular form and don't care about the question that much, but someone on the team has a vague idea that having it in addition to the other data might make for an interesting graduate student project someday so they threw it in there and it'll be two more years before anyone notices that the data is nigh-useless.

Whatever the reason is, they may or may not be given enough of a word limit to explain it in the published article. Ideally they would, but in practice there's a tight space limit and a detailed methods section is often sacrificed to make space for a longer discussion/results section.

What you describe is messy as hell but absolutely standard for health care research.

That said, many researchers will absolutely talk to you more in-depth about a paper if you contact them (the corresponding author on a paper, the PI of the research you're participating in) and say, hey, I'm really curious about XYZ, as a partient and research participant. Can you point me toward any studies? Or: I'm trying to understand your paper and it would be really helpful to know what question was actually asked. You may not get a quick response but you may well get one that can help you understand more than the authors were able to cram into the actual paper.)

....and now that I'm reading this back and your original question, I think I've veered away from it a bit. I'm still gonna leave the middle bit in case it's useful to understand some of the context for why these questions may not be written the way your data-brain thinks (and is right!) that they should be. But I guess the most applicable advice is at the end there. People who work in health care research are used to hearing directly from people with the conditions they research, and many of them will take some time to email with you or offer you some pointers in fruitful research directions, and they may be able to tell you a bit more about their own studies than their papers described if you have specific questions.
posted by Stacey at 9:18 AM on September 4, 2019 [2 favorites]

« Older A week in Provence & Cote d'Azur: Travel...   |   Good Introductory Book on UK Parliamentary Law Newer »
This thread is closed to new comments.