Talk:Acupuncture/dblblind

The following material has been moved, for further editorial discussion, from Osteoarthritis of the knee under the article's "Examples of controlled studies" subsection. It may be original research and/or a content fork that belongs in an article on the general subject of blinding. - Jim Butler^(talk) 07:40, 14 October 2006 (UTC)[reply]

But unblinded blindstudies are problematic: "Success of blinding is a fundamental issue in many clinical trials. The validity of a trial may be questioned if this important assumption is violated." (Bang, H. et al., Assessment of blinding in clinical trials, Controlled Clinical Trials 25 (2004) 143-56). "When unblinded, participants may introduce bias through use of other effective interventions, differential reporting of symptoms, psychological or biological effects of receiving a placebo (although recent studies show conflicting evidence), or dropping out..") (Devereaux PJ, Bhandari M, Montori VM, Manns BJ, Ghall WA, Guyatt GH, Double blind, you have been voted off the island!. McMaster University, Hamilton, Ontario, Canada. Evidence-Based Mental Health. 5(2):36-7, 2002 May). This applies also to the three other GERAC studies about headache, migraine and low back pain because study details were freely available for patients before the study even started [1] and [2] and [3] .

- Readers should note that almost all scientific double-blind studies require research subjects be informed of (i.e., to know and understand) the experimental protocol explicitly in writing when they signed the informed consent form. This does not mean that the study was "unblind" before the study is proceeded. Almost all Institutional Review Boards (IRB) (the governing body overseeing the approval of any use of experimental subjects in a research institute) require explicit informed consent detailing the experimental protocol of what will be done on them before the study is allowed to proceed, particularly in double-blind studies because the subjects not only have the rights to know what may be done on them, but also the fact that they will not know which procedure is done on them because of the randomized blind nature of the study. This is NOT unblinding.
- Unblinding is "revealing exactly which procedure is done on the subject," which is not done in any of the mentioned studies above or below. It should not be confused with what "blinding" is in a double-blind study.
- "Blinding" in double-blind means "not knowing which treatment/procedure is done on them." The assignment of which procedure is used is randomized, and not be revealed until after the study is concluded.
- Single-blind means only the subject doesn't know which procedure is done on them, but the experimenter knows; nonetheless, the subjects are informed of which procedure may be done on them, except that they won't know until after the study is completed.
- Double-blind means neither the subject nor the experimenter knows which procedure is done. In case where the experimenter has to know which procedure is done on the subject (such as in acupuncture or surgery), then the evaluator/assessor of the outcome of the treatment (i.e., the clinician that evaluates the outcome of the treatment) is blinded, i.e., doesn't know which procedure the experimenter had performed on the subject. In either case, informing the subject DOES NOT imply unblinding.
- Only when the treatment is revealed, then it is "unblinded." An example is the recent hormone replacement therapy (HRT) study on the effects of estrogen on menopausal women, the double-blind studies were halted before the studies were completed because analysis showed that the risk outweighted the nature of blindness, and they revealed what treatments were done on the subjects — that is unblinding — but all the subjects are informed of the experimental protocol before they receive either estrogen or placebo, before this unblinding is done.
- Triple blind means neither the subject nor the experimenter nor the analyst knows which procedure is done on who. In most cases, studies are done double-blind rather than triple-blind, as in the hormone replacement therapy study where the analysts (the statisticians) know which treatment is done on who, which is why they aborted the study when they considered the risk is deemed too great to be unethical to continue the study. It is only then that the study is unblinded.
- Thus, revealing the experimental protocol to the subjects (or patients) by informed consent, or publishing the protocol to the public on the web would not in any way invalid the results or conclusion. In fact, revealing the protocol to the subjects (patients) before the study is critical and essential to be scientifically valid because of the statistic assumptions that all subjects will have the same knowledge about what the experiment is about. Otherwise, the study would be skewed when some subjects have more knowledge or expectations than others, which means they are not on equal grounds. The only thing they don't know is which treatment is done on them, which they are required to know of this uncertainty to be statistically valid. Thus, disclosing the experimental protocol IS part of the protocol if any double blind study is done correctly and scientifically. It is not "unblinding," unless the acupuncturist disclosed to the patient whether he/she inserted a real needle or a sham needle at the acupoint or some other random non-acupoint location on the skin, which is unlikely if the acupuncturist followed the protocol instructions.
- Most importantly, contrary to most misconception, double-blinding DOES NOT eliminate biases because human is inherently biased, this is the fact-of-life. What it does is merely subtract out the biases, if it exists, using the statistical analysis. That is why double-blinding requires randomized trials. The randomization is essentially what makes the analysis possible by filtering out the biases using statistical methods.
- Simply put, if everyone is biased (i.e., even if both subjects and experimenters are totally biased), they will bias the sham control (placebo) the same way they bias the real treatment. Since neither one knows what is done on who (because of the randomization), the bias will be canceled out in the analysis when you subtract the two out.
- By the same token, because the experimental protocol requires selection of random subjects and random experimenters from a pool, statistics show that some will be believers and some will be non-believers, so on average, the believer-effect will counterbalance the non-believer-effect, negating both biases! So in the final analysis, the statistics always prevail, i.e., any biases will all be averaged out no matter how extreme the subject's or experimenter's beliefs are.
- In fact, even if the experimenters (acupuncturists) fudge the data, the fudge factor will all be eliminated by the statistics because the experimenters will fudge the placebo trials the same way they fudge the real trials; because they don't know which is which, it will all be subtracted out in the statistics. (Just like in political polling, the extreme views will always be washed out in the statistics, if the sample is a scientific sample, i.e., drawn from a randomized pool.)
- That is the power of statistics. Using a good experimental design with randomized trials, all biases will be washed out. It’s in the statistical analysis that eliminates the biases, not by eliminating the biases of the subjects or experimenters — because it is impossible to do that or change anyone’s belief system; never does, and never will.