|
There seems to be some confusion concerning the ideas of double-blind studies, particularly that EVERY scientific study NEEDS to have a double blind study in order to be valid.
This is wrong on many levels, and I will try to explain in simple terms later on.
For the record, I'm not a scientist. I'm not a Doctor. I'm not a statistician.
However, before I became a nurse, I worked for a leading center of cancer research, and I was involved in 2 studies that dealt with cancer.
Both studies (one dealt with Ovarian, the other with Breast cancer) were aimed at trying to find a better way to diagnose ovarian and breast cancer earlier than our current detection allows. Especially for ovarian cancer, which is known as a silent killer because it often manifests no symptoms or very vague symptoms (bloating, gas, abdominal upset, etc) that can be attributed to other, mostly benign, causes.
For breast cancer, there is a genetic marker--CA-125, but this is known to have false positives and false negatives and can be a guiding point to warrant further diagnostics (in the case of a high CA-125), but it isn't foolproof.
The research study I worked on had multiple components:
1) enrolling women who had known 1st and 2nd degree relatives with breast and or ovarian cancer (since the two are very closely related), but had no breast or ovarian cancer themselves
2) enrolling women who had known 1st and 2nd degree relatives with breast and or ovarian cancer, but did have breast and or ovarian cancer themselves
3) enrolling women whose 1st or 2nd degree relatives had no breast and or ovarian cancer, and the participants had no breast or ovarian cancer
4) enrolling women whose 1st or 2nd degree relatives had no breast and or ovarian cancer, yet the participants DID have breast and or ovarian cancer.
Participants in the study agreed to long-term monitoring through blood draws over the course of 2 years, each blood draw 6 months apart.
Participants also all filled out a standardized multi-page questionnaire which went through their family history (going beyond first and second degree relatives), their ethnic background (Azenkazi Jews have a higher rate of certain cancers relative to the general population), the number of pregnancies they have had (as well as the number of term births, miscarriages, still births, and abortions), when they started to menstruate, the average length of menstruation, and when (if) they started menopause, and the symptoms (if any) of menopause. Also we were interested in medications they have taken, especially those with hormonal and steroidal components. What kind of birth control did they use, and how long did they use it? Have they ever had any other kind of cancer? What treatment did they get? Any chemo or radiation? What about adjuvant therapy? HRT?
All of the women enrolled in the study were given a randomly-generated number that was attached to their questionnaire. The questionaire did not have a spot for their name or any other identifying information aside from the randomly generated number.
Their blood samples were also labeled with just this number (and the date and time of the blood draw) but no other identifying information.
One of my roles in the study was to keep a database of the name of the study participant, and the randomly generated number that was assigned to them. This database was not able to be accessed by any of the researchers.
Then, in a separate database that WAS accessable to one set of the researchers, I entered the information from the questionnaires.
Then, in a THIRD database that was accessable to another set (but not the first set) of researchers, there was information about the blood samples obtained from the patients, again, using the same randomly generated number that was attached to the pt's name and questionnaire.
All three of these databases were separate, and only accessable to those working on THAT part of the study: the questionnaires, or the blood, or the names and randomly generated numbers.
THen, another component of the study was for women who had filled out the questionnaires, and who had given blood, but were also undergoing surgery to remove tissue---either through a mastectomy in the event of breast cancer, or a hysterectomy (removal of only the uterus and cervix) or a total abdomominal hysterectomy, salpingo-oophorectomy (removal of the uterus, cervix, ovaries, and fallopian tubes).
My job was to take the tissue removed from the operation and dissect parts of the tissue, cut it up, and freeze it in a variety of medium for study. Each tissue sample was marked with the same randomly generated number that was assigned to the participant, the blood, and the questionnaire.
Each surgery would land us about 10 different samples of ovary and fallopian tube or breast tissue. They were labeled, put in boxes, and put in liquid nitrogen for study.
The location, box, and date of collection was recorded (by moi) in a FOURTH database that was only accessable to the scientists working on the tissue.
----
So this is a double-blind, and this is why double-blind is needed in this kind of study. The scientists working on the blood had no idea what the risk factors were for the women whose blood we were taking. The scientists working on the tissue had no idea what the bloodwork showed, and the statiticians had no idea what correlation the questionnaire information had on the results of blood or tissue samples.
Each researcher was working independently of any pre-conceived notion of the woman who was part of the study. The only information the researchers had prior was:
1) this was a woman 2) this woman was above the age of 18
---
Then, say a researcher found a genetic marker in several samples of blood that may or may not have pointed to a certain cancer. They would enter that into the database. Later, the separate researcher would find something in the tissue that may or may not have pointed to a certain kind of cancer. They'd enter THAT into the database. The people working with the study will have rated the patient's information based on risk factor from KNOWN genetic or environmental risk factors.
ANOTHER set of people would go through the information entered by the blood people, the tissue people, and the questionnaire people and would parse that out to see what that information meant.
FINALLY, at the end of the study (we're not there yet, and I was working on the study in 2004-2005), someone will go through and put ALL the information together---the name, the number, the answers to the questions and what the researchers parsed from that, the blood and what the researchers parsed from that, and the tissue and what those researchers parsed from that. These FINAL people put it all together, find the statistical relevant information, find out what the outliers are and why they are there, and the study is published.
THIS is a double blind study. No one knows anything EXCEPT for the raw data they are processing and finding.
We were also involved with about 5 other Cancer Research Centers around the world that were doing the EXACT SAME STUDY with the EXACT SAME METHODOLOGY so that it would be a combined, large-sample comprehensive study on risk factors (hopefully updated from what we currently know) and HOPEFULLY genetic markers that can be more easily, and correctly, identified so that women can get early treatment.
---
Okay----so that's when a double-blind is necessary. It keeps the researchers from being biased (in this case) from extrapolating information that they may not have extrapolated had they known something about the research participant. If a researcher knew that the woman's blood he was testing was positive for breast cancer, he may look at that blood a little differently, or reach pre-conceived conclusions based on his knowledge of the pt's cancer status. However, not knowing how many blood samples came from people with or without cancer, they are able to look at ALL blood with the same non-bias.
The same with medicine---a double blind in the case of testing medicine gets rid of the bias of the researcher who knows that Medicine A is a placebo, but Medicine B has an anti-inflammatory process. That knowledge could bias the resarcher as far as followup questions, or how the information at the end of the study is extrapolated.
Same with the patient taking Medicine A---if they know it's a placebo, they will define their answers of questions based on the knowledge that they know the medicine is useless (therapeutically). If they know that they are receiving a medication with anti-inflammatory process, they will define their answers to questions on effecacy based on this knowledge.
---
But a study doesn't HAVE to be double-blind in order to be valid.
If I am a researcher and want to know how people view the quality of mail service in their area, I don't need to double-blind the study. I can ask the questions, the participants can know what they're being asked, and answer appropriately. I don't necessarily need to TELL Them that I am doing a study on mail satisfaction, and I can add "innocent" questions in the questionnaire about their garbage service, or the quality of the roads, just so they don't find the survey as a way to "rant" about mail service but rather to give honest answers about their feelings.
If I wanted to know what percentage of people in a given area code vote Republican, I can ask that question right out without hiding the reason for the study, or being hidden from knowing what the study was about.
--- Just because it's double-blind doesn't mean it's invalid. There *ARE* types of studies that can BE invalid because they weren't double blind, but double-blind is not a REQUIREMENT for every single type of study that is conducted.
:falling on deaf ears:
|