We're all about children!
 October 2008
 Green Tea and Obesity: A Google vs. PubMed Throwdown I popped into my Facebook page last week, and saw an advertisement featuring Oprah Winfrey. She was touting the benefits of green tea for weight loss, with a Dr. Nicholas Perricone. Interesting, I thought, since Dr. Maria had just asked me to write something on just this subject! What is the evidence, I wondered, for green tea being an effective weight loss agent? This sounded to me like a perfect opportunity to pit an expert biomedical system against the all that information on the Web. In other words, it seemed like a good question to use for a PubMed - Google throwdown. Where could I get the best answer for this question? How much time would it take me? How reliable would the information I found be? According to a SearchEngineWatch blog, the average Google question consists of four words, up from three words last year. Even so, it seemed a 3-word query would be fine for this question. So going to Google, I typed in green tea obesity. 
Wow. Look at those first two. Look pretty good, don't they? But, we need to find the evidence. Wanting to make this a fair contest, I decided to use some expert search techniques, and attempt to limit to the best studies. 
Then I began to look at what I found. The first hit was an abstract of an article from PubMed, from a Japanese journal, a non-blinded RCT, with intervention patients drinking green tea extract powder for two months. It was difficult to tell if the 66 patients mentioned were all in the intervention group, or half in the intervention and half in the control group. There was not enough additional information to judge the quality of the methodology, and no way to get to the full text of the article. No, not safe to draw conclusions from this. More promising was the next hit, a link to an article in the European Journal of Clinical Nutrition. However, after 3 minutes of searching the sight that was linked, I still had not found the article. I decided to be choosier, and look more closely at the URLs before I clicked on the links. Still on first page of results, I saw a link to Medscape. Medscape I know as the owner of eMedicine, and as a fairly reliable website, even though it is a ".com" commercial site. And, I could see the highlighted term "meta-analysis" in the description. Disappointment awaited me again as I clicked to see that the title of this article was "Green Tea Consumption and Liver Disease." HHMM. Giving Google one more try, I clicked on a link that had RCT and "Effectiveness of green tea on weight reduction" in its description, with a URL that included a PubMed Id number. Great, I thought, Google has pulled up an article indexed in PubMed! Well, sort of. It was an abstract to an unrelated article that had a link to another article by the same authors. Following the link I found the abstract of an article with a promising title: "Effectiveness of green tea on weight reduction in obese Thais: A randomized, controlled trial." But, again, it was difficult to evaluate the methodology of this study than it had been for the first article, and no full text link to find more information. Also, from an evaluation point of view, I was not too thrilled to find advertisements of the sort that appeared at the top of that webpage. 
At that point, I had spent the better part of 15 minutes, and although I had abstracts for two RCT, they were of questionable quality. And I still had questions. Were there other RCTs? Better RCTs? What was a missing? How long did I want to spend looking in Google? Time to go to PubMed. I began with a similar quick keyword search. In seconds, I had a set of 71 articles that could be quickly limited to RCTs just by clicking on a tab. 
Looking at those 13 RCTs, I even found our Thai article, this time with a link to the full text. Next, the expert search. Using the MeSH Database, I composed this search: (("Tea"[Mesh] AND "green tea"[title/abstract]) OR "green tea extract AR25 "[Substance Name]) AND "Obesity"[Mesh] Then, I used the limits tab to limit publication type to RCT, language to English, and humans. 
Because I could use the controlled vocabulary, my results were more relevant. It was much quicker to find the information I needed, and judge the quality of source of the information. Having had this experience, I declare PubMed the hands-down winner of this Green Tea Throwndown! My results are more complete. The time to search was much shorter. I know all of my results are coming from Biomedical journals. I can feel confident that I have found the best evidence available on this question. But, you say, you are a medical librarian. You are bound to be biased in favor of PubMed. In the interest of full disclosure, let me say that I do search PubMed several times a day. I have had National Library of Medicine training in expert search techniques. However, I also use Google several times a day, and I have been searching Google for more years than I have searched PubMed. I have been to professional conference sessions given by Google. I have taught expert Google searching techniques. And I have been tracking Google search algorithms since the beginning - in the days when we called the World Wide Web, the World Wide Wait. Yes, Google is sometimes the best tool for the job. However, for evidence-based, clinical care questions, PubMed is the clear winner. But don't take my word for it! Try it yourself. Take PubMed and Google head to head. And let me know if you don't agree that your PubMed results are more trustworthy, more relevant, and more complete for evidence-based searches than Google's.
|
