Keto made easy

How to Search PubMed for Diet Studies

Quick answer: Go to pubmed.ncbi.nlm.nih.gov, search a few specific words (the diet, the outcome, and "trial" or an author's name), open the most relevant result, and read the abstract. Check four things: who was studied, how many, for how long, and what the result was in actual units. That's how the KetoDick evidence pages were checked, and it takes about ten minutes once you've done it twice.

On this page

Why PubMed instead of a forum

A forum gives you opinions about studies, often secondhand, often from someone who read a headline. PubMed gives you the study itself, or at least its abstract: the short summary of what researchers did and found. Searching it is free.

You won't understand every word in an abstract, and you don't need to. You need to find a handful of facts and compare them with the claim you were told.

Step by step: finding and reading a study

  1. Go to PubMed. The address is pubmed.ncbi.nlm.nih.gov. No account needed.
  2. Search specific words. Not "is keto good." Try the diet plus the outcome plus a study word: ketogenic diet weight loss randomized trial, or low carbohydrate diet LDL cholesterol. If you know an author or a study nickname, use it: Gardner DIETFITS.
  3. Narrow it if you get thousands of hits. The filters on the left let you limit by article type (for example clinical trials or meta-analyses) and by year. Start with trials and reviews in humans.
  4. Open the result whose title matches your question. The title tells you a lot: "randomized," "meta-analysis," "cohort" and "in mice" are all big clues to what kind of evidence you're looking at.
  5. Read the abstract and pull out four things: - Who: adults, people with diabetes, athletes, mice? - How many: 17 people or 100,000? - How long: a week on a research ward, or two years at home? - What happened, in units: kilograms, mmHg, grams per day. Not "significantly improved."
  6. Compare with the claim. Does the study say what you were told? Often the answer is "sort of": right direction, smaller effect, different people.
  7. Write down the PMID or DOI. The PMID is PubMed's ID number; the DOI is the paper's permanent link. Either lets anyone, including your doctor, find exactly what you read.

A tip for step 5: an observational study (researchers watched what people ate) can show that two things go together, not that one caused the other. A trial (researchers assigned the diet) is better at cause. See observational vs randomised studies.

A worked example: DIETFITS

Say a friend tells you a big study proved low-fat and low-carb are the same.

Search Gardner low-fat low-carbohydrate 12 month and you'll find the 2018 JAMA paper known as DIETFITS. Pull the four facts:

So your friend is roughly right about the headline. What the four facts don't tell you, and the evidence page adds, is that the low-carb arm was a moderate low-carb diet, not a ketogenic one. That's the kind of detail you only get by reading past the headline. More in DIETFITS explained.

Google Scholar, and the "add pubmed" trick

Google Scholar (scholar.google.com) searches across journals, and it shows how many later papers cite each one, which is a rough sign of how much attention a paper got. Highly cited doesn't mean correct, but it helps you find the papers everyone else is arguing about.

KetoDick's version is even simpler: go to Google Scholar, or when you google something, add "pubmed" to the end (video 1JbK1_ClDmY). That pushes the actual studies up past the blogs.

What the research says

Here's how the evidence pages on this site were put together and checked, so you can see the method at full size:

Another check the pages make visible is what an abstract can and can't give you. The two-year DIRECT trial's abstract reports mean losses of 2.9 kg (low-fat), 4.4 kg (Mediterranean) and 4.7 kg (low-carb) in 322 adults (Shai et al., 2008). That's an average. It can't tell you what you'd lose, and no abstract can. See reading studies.

KetoDick's take

He's said since 2017 that he reads PubMed and NIH publications for most of his information and cross-checks blog ideas against medical research, and that if you ask, he'll find you the actual studies so you can judge for yourself whether there were enough people or the work was too old (video xyEZL0HuejI). By 2018 he was putting sources in the comments so you and your doctor could check them (video UygKGoP_BtA).

His reason for PubMed specifically is that a study has to report exactly what it told people to do, because it needs to be repeatable (video 1JbK1_ClDmY). So if a magazine says keto bans vegetables, you can check whether any study's protocol actually said that.

Key takeaways