Has there ever been a double-blind study of homeopathy in which the null hypothesis was rejected?
I'm not looking for a blanket validation. Just a single study where the null hypothesis (that there is no difference between homeopathy and placebo) was rejected.
Yes, loads. Here is one:
I don't know enough about medical trials to tell if this is common or not, but it seems weird to me. They first put the patients on open-label homeopathic treatment. Those who respond well to it then goes on to a double-blind cross-over study.
Sounds like the biggest selection bias this side of the Rio Grande to me. And this is the problem with most of these studies, they have a selection bias, or they have a small sample size, etc. And most damning of all: Nobody seems able to reliably repeat them. Although few even try.
And here is a meta-study that lists studies, some of them also positive:
Homeopathy for Childhood and Adolescence Ailments: Systematic Review of Randomized Clinical Trials
The way to find these studies is probably through meta-studies, of which there are many (which do not show better than placebo effects).
These studies, as well as cold fusion, show the importance of the word "repeatable" in "Controlled, repeatable experiments".
I recently stumbled across a 2005 paper in Chest (Frass, M., Dielacher, C., Linkesch, M., Endler, C., Muchitsch, I., Schuster, E., & Kaye, A. (2005). Influence of Potassium Dichromate on Tracheal Secretions in Critically Ill Patients. In Chest (Vol. 127, Issue 3, pp. 936–941). Elsevier BV. https://doi.org/10.1378/chest.127.3.936 ) that as far as I can see has not been previously discussed here on Skeptics, so I figured it might be worth resurrecting this question. The methodology is quite a bit better than the ADD paper (but not perfect as I'll discuss below.) In short, a C30 dilution of potassium dichromate significantly reduced secretions, intubation time, and time in the ICU in a sample of 25 COPD patients vs. 25 on a well-designed placebo.
Abstract:
Background: Stringy, tenacious tracheal secretions may prevent extubation in patients weaned from the respirator. This prospective, randomized, double-blind, placebo-controlled study with parallel assignment was performed to assess the influence of sublingually administered potassium dichromate C30 on the amount of tenacious, stringy tracheal secretions in critically ill patients with a history of tobacco use and COPD.
Methods: In this study, 50 patients breathing spontaneously with continuous positive airway pressure were receiving either potassium dichromate C30 globules (group 1) [Deutsche Homöopathie-Union, Pharmaceutical Company; Karlsruhe, Germany] or placebo (group 2). Five globules were administered twice daily at intervals of 12 h. The amount of tracheal secretions on day 2 after the start of the study as well as the time for successful extubation and length of stay in the ICU were recorded.
Results: The amount of tracheal secretions was reduced significantly in group 1 (p < 0.0001). Extubation could be performed significantly earlier in group 1 (p < 0.0001). Similarly, length of stay was significantly shorter in group 1 (4.20 ± 1.61 days vs 7.68 ± 3.60 days, p < 0.0001 [mean ± SD]).
Conclusion: These data suggest that potentized (diluted and vigorously shaken) potassium dichromate may help to decrease the amount of stringy tracheal secretions in COPD patients.
Chest's publication of a homeopathic study was understandably criticized, most notably in this letter to the editor by David Colquhoun in which he also neatly rips apart homeopathic theory.
It surprises me that CHEST would publish an article (March 2005)1 on the effect of a therapeutic agent when in fact the patients received none of the agent mentioned in the title of the article. ... It is one thing to tolerate homeopathy as a harmless 19th century eccentricity for its placebo effect in minor self-limiting conditions like colds. It is quite another to have it recommended for seriously ill patients. That is downright dangerous.
However this response did not address the actual results of the study. To answer that, we need to look at the distribution of the patients after randomization. Although the study's assignment methods seem sound this table shows that by several measures the placebo group may have been less healthy than the treatment group before intervention.
Of course, none of these differences are statistically significant, but taken together they are certainly very suggestive that the randomization was not well-distributed. Additionally, this table show greater variation in the placebo group among the significant outcomes, suggesting a few outliers could have skewed the results. In a paper making extraordinary claims such as this one, such inconsistencies demand that the results be re-examined and reproduced. Since I can find no more supporting studies on this topic since 2005, it seems unlikely that the treatment meets the reproducibility standard.
