Web Trials

Thanks to Rey Arbolay, at the Shangri-La Diet forums, the eternal question “will this help?” is being answered in a new way. The specific question is “will the Shangri-La Diet help me lose weight?” The new way of answering it is that people are posting their results with the diet in the Post Your Tracking Data Here section of the forums. What they post is standardized and numerical enough that ordinary statistical methods can be used to learn from them. I’ll call this sort of thing a web trial.

It’s a lot better than nothing or a series of individual cases studied separately. I learned a lot from my most recent analysis — for example, that people lose at a rate of about 1 pound per week after Week 5. I couldn’t have done a good job of predicting where any of the fitted lines on the scatterplots would be or the size of the male/female difference. Nor could I have done a good job predicting the variability — the scatter around the lines.

It’s a lot worse than perfection. It would be much better if a comparison treatment (in the case of SLD, a different way of losing weight) was being tested in the same way. Then results from the two treatments could be compared and you would be closer to answering the practical question “what should I do?” (That modern clinical trials — very difficult and expensive — still use placebo control groups although placebos are not serious treatment options is a sign of . . . something not good.)

I can imagine a future in which people with a health problem (acne, insomnia, etc.) go to a website and enroll in a web trial. They told about several plausible treatments: A, B, C, etc., all readily available. They are given a choice of (a) choosing among them or (b) being randomly assigned. They post their results in a standardized format for a few weeks or months. Then someone with data analysis skills analyzes the data and posts the results. As for the participants, if the problem hasn’t been solved they could enroll again. This would be a way that anyone with a problem could help everyone with that problem, including themselves. The people who set up the trials and analyze the results would be like the book industry or Wikipedia insiders — people with special skills who help everyone learn from everyone.