The Search Lab
Find the published research and the clinical trials for a condition, and learn to correct your own queries.
Nothing you type is stored or sent to us. Searches go directly from your browser to Europe PMC and ClinicalTrials.gov.
About this course
This course is built for anyone who has to find out what is published on a condition without a medical librarian at hand: advocacy group staff and volunteers, research and clinical teams new to searching, students, people living with a condition, and family caregivers.
You do not need a specialist background, a university login, or a subscription. The two collections used here are free and public: Europe PMC, which holds the same abstracts as PubMed plus a large body of free full text, and ClinicalTrials.gov, the United States registry of clinical studies.
Learn
Nine short lessons on how a search is built and why the same words give one person nine results and another nine thousand.
Practice
Two live benches. Type a real search, run it, and get specific feedback on what to change next.
Take with you
A report of every search you ran and every study you saved, ready to print or email before a meeting, a class, or an appointment.
How to use the two benches
The Search Lab and the Trial Lab both work the same way. You describe your condition in the form, the tool assembles a proper search string, and you run it. Then a feedback panel tells you whether your result count is usable and offers one-click repairs. Every repair explains itself before you accept it, so you finish the session knowing why the search changed.
What to have ready
- The name of the condition, spelled the way a clinician writes it.
- Any other names for it: an older name, an abbreviation, a gene name, a brand name for a drug.
- One question you actually want answered. "Does this treatment help with the falls?" beats "tell me about the condition."
Three places to look, and what each one holds
People often describe all of this as "searching PubMed." Three separate collections are involved, and knowing which is which saves a great deal of frustration.
| Collection | What is in it | Use it when |
|---|---|---|
| PubMed pubmed.ncbi.nlm.nih.gov | Around 38 million citations and abstracts from the biomedical literature, run by the US National Library of Medicine. Mostly abstracts, not full articles. | You want the reference standard, or a clinician asked you to send them a PMID. |
| Europe PMC europepmc.org | Everything in PubMed, plus preprints, patents, and several million free full-text articles you can read on the spot. | You want to actually read the paper, not only its summary. This is what the Search Lab queries live. |
| ClinicalTrials.gov clinicaltrials.gov | Registered clinical studies worldwide, including ones still recruiting. Not articles: study plans, sites, contacts, and eligibility rules. | You want to know what is being tested now and whether anyone near you can join. |
Why this course runs on Europe PMC
Europe PMC indexes the same records as PubMed and allows a tool like this one to query it directly from your browser. Your searches stay private, results appear in a few seconds, and free full text is flagged so you know what you can read without paying. Every search you build here also converts into PubMed's own syntax with one click, so nothing you learn is stuck inside this page.
The paywall question
Some articles cost money to read. Before you pay anything, try these in order: the free full text badge in your results, PubMed Central, the author's own copy on a university website, an email to the corresponding author asking for a copy (this works more often than people expect), and your local public or hospital library, which can often request an article for you at no charge.
What none of these hold
Registries of people with your condition, patient community discussions, and the practical knowledge families build over years do not appear in any of the three. They are real evidence of a different kind. Use the literature to inform the conversation rather than to replace what your community already knows.
Collect your words before you search anything
Most disappointing searches are word problems, not skill problems. The literature may describe your condition in four different ways, and a search that uses one of them finds one quarter of the evidence.
The six kinds of words to gather
| Kind | Example for one condition |
|---|---|
| The clinical name | Lennox-Gastaut syndrome |
| The abbreviation | LGS |
| An older or historical name | petit mal variant |
| The gene or biological marker, if there is one | SCN1A, CDKL5, SOD1 |
| Brand and generic drug names | Epidiolex, cannabidiol |
| The symptom in plain words | drop attacks, falls, tonic seizures |
Where to find the other names
- The Introduction section of any recent review article on your condition. Authors list alternative names in the first paragraph almost as a rule.
- The condition's page on a rare disease reference such as GARD, Orphanet, or MedlinePlus.
- Your own records, where you can see how the diagnosis is coded and worded on a discharge summary or a visit note.
- Your advocacy group. Ask which terms the researchers in that field use now, and which are out of favor.
Build your word list now
Type the terms you have collected. This list feeds straight into the Search Lab, and it appears in the report you can print at the end.
Add abbreviations, older names, gene names, and both spellings, one at a time.
Written in your own words. This is a note for you, not part of the search.
AND, OR, and quotation marks
Using punctuation to improve your search results.
OR makes the pile bigger
Use it between words that mean the same thing. LGS OR "Lennox-Gastaut" finds records using either name.
AND makes the pile smaller
Use it between different ideas. epilepsy AND cannabidiol finds records about both, not either.
Quotation marks hold words together
"quality of life" searches the phrase. Without them you get every record containing the word quality anywhere.
The pattern that works nearly every time
Group your synonyms with OR inside parentheses, then join the groups with AND. One idea per set of parentheses.
Read it as: any of these names for the condition, together with any of these ways of asking my question.
Three habits that fix most bad results
| Symptom | Likely cause | Repair |
|---|---|---|
| Tens of thousands of results | Only one broad idea, no second concept | AND a second group: the treatment, the symptom, or the outcome you care about |
| Zero results | A typo, a phrase in quotation marks that nobody writes that way, or too many ANDs | Remove the quotation marks, then remove one AND group at a time |
| Results about the wrong disease | An abbreviation that means several things | Pair the abbreviation with the full name using OR, and add a second concept with AND |
Try it in your head first
Look at the three searches below and decide which one returns the most records. The answer is at the Search Lab, where you can run all three in under a minute.
Filters that cut the pile
Filters trim a search without changing its words. Used well they take a search from unusable to readable in one click. Used carelessly they hide the very paper you needed.
| Filter | What it does | When to be careful |
|---|---|---|
| Date range | Keeps only recent work. Five years is a good first cut for treatment questions. | Foundational papers and rare disease case reports are often older. Widen the range if a search goes quiet. |
| Review articles | Keeps papers that summarize many studies. The fastest way into an unfamiliar topic. | Reviews reflect their author's reading and can lag two or three years behind new results. |
| Systematic reviews | Keeps the subset built by a documented, repeatable method. The strongest single place to start. | They are uncommon in rare conditions, so expect few or none. |
| Clinical trial reports | Keeps papers reporting a trial result. | These describe finished trials. Use the Trial Lab for studies still enrolling. |
| Free full text | Keeps only what you can read today without paying. | This filters by access, not by quality. Turn it off when you want the complete picture. |
| Title and abstract only | Requires your words in the summary rather than anywhere in the article. | The single most effective filter for cutting a huge result count, though it is occasionally too strict. |
How many results is the right number?
A search you can actually use returns something between roughly twenty and two hundred records. Below twenty you have probably cut something you wanted. Above a few hundred you cannot read them, and you will read the first ten instead, which is a decision made by a sorting algorithm rather than by you.
The shaded band is the readable range. The Search Lab shows you where each of your searches falls on this scale and offers the specific repair that moves it into the band.
Order of operations when you have too many
- Add a second concept with AND. This is the largest single cut and it keeps the search honest.
- Restrict to title and abstract.
- Apply a date range.
- Only then restrict by article type.
Do them one at a time and watch the count after each step. That habit alone will make you better at this than most people who search occasionally.
Reading a result list without a science degree
You are not going to read two hundred papers. You are going to triage two hundred titles down to five, and read those properly.
Work through the list in three passes
Pass one: titles
Ten seconds each. Does this involve my condition and my question? Most titles answer that on their own.
Pass two: abstracts
Read only the last two sentences first. That is where authors state what they found. Then read the methods sentence.
Pass three: full text
For the handful that survive. Read the discussion section before the results. It explains what the numbers were supposed to show.
What the parts of a record tell you
| Part | What to take from it |
|---|---|
| Year | How current the thinking is. In a fast-moving area, anything older than five years needs checking against newer work. |
| Journal | A rough signal, not a verdict. Specialty journals in your condition are often more useful than famous general ones. |
| Authors | Names that keep appearing are the people working in your condition, so note them. You can search by author later, and you can write to them. |
| Study type | Case report, cohort study, randomized trial, review. This tells you how much confidence the design supports. |
| Number of participants | Usually in the abstract. Six people and six hundred people support very different claims. |
| Free full text badge | Whether you can read it now. |
A trick for rare conditions
When a search returns almost nothing, find the one paper closest to your question and look at two things: the reference list at the end, and the "cited by" count on the article page. Older relevant work appears in the first, newer relevant work appears in the second. Two or three rounds of this often finds more than another hour of keyword searching.
Words you will meet, in plain terms
| Term | Plain meaning |
|---|---|
| Retrospective | The researchers looked backward at records that already existed. |
| Prospective | They decided what to measure first, then followed people forward. |
| Randomized | Chance decided who got which treatment, which is how you separate the treatment from everything else. |
| Placebo controlled | Some participants got an inactive version so the comparison holds. |
| Open label | Everyone knew who was getting what. Easier to run, easier to be fooled by. |
| Primary endpoint | The one result the study was designed to measure. Everything else is a secondary finding and has less weight. |
| Statistically significant | The difference is unlikely to be chance alone. It does not mean the difference is large or that it changes daily life. |
Is this study any good?
Everything you find has passed through some kind of filter, and the filters differ enormously. These are the checks that take a minute and change how you read the rest.
Peer reviewed, preprint, or neither
A peer-reviewed article has been read and challenged by other researchers before publication. A preprint has been posted publicly before that happens, which makes it fast and unvetted. Europe PMC labels preprints clearly. Preprints are legitimate and often hold the newest findings in rare disease, and they can also be wrong in ways review would have caught. Read them with that in mind, and check later whether a peer-reviewed version appeared.
Four questions to ask of any single study
- How many people were in it? Look for the number in the abstract. A case report describes one person and is a signal, not evidence of effect.
- Compared with what? A study with no comparison group can tell you what happened, not what would have happened otherwise.
- Who paid for it, and who wrote it? Funding and conflict of interest statements are near the end. Industry funding does not disqualify a study, and it is best read as context.
- Does the conclusion match the result? Compare the last sentence of the abstract against the actual numbers. Language sometimes runs ahead of findings.
Reading a trial record
A ClinicalTrials.gov record is a study plan written for regulators, so it reads strangely at first. Six fields hold most of what you need.
| Field | How to read it |
|---|---|
| Recruitment status | Recruiting means they are enrolling now. Not yet recruiting means soon, and contact details are usually already there. Active, not recruiting means the study is running but closed to new participants. Completed means enrollment and follow-up are finished, and results may be posted on the record itself. |
| Study type | Interventional means something is being given or done and tested. Observational means researchers are watching and measuring without assigning treatment. Natural history studies and registries are observational, and they are often the most accessible way to contribute. |
| Phase | Phase 1 tests safety in a small group. Phase 2 looks for early signs of effect. Phase 3 compares against current care in a large group. Phase 4 follows an approved treatment in ordinary use. |
| Eligibility criteria | Two lists, inclusion and exclusion, and you should read both slowly. Age ranges, prior treatments, and required test results are the usual reasons someone does not qualify. |
| Locations and contacts | Sites, cities, and a named contact with a phone number or email. Contact details are there because they want to hear from people. |
| Sponsor | Who is running it: a company, a university, a government institute, or an advocacy group. |
The NCT number
Every registered study has an identifier beginning with NCT, and you should write it down. It is the fastest way to pull up the same record again, to ask a clinician about a specific study, and to find published results later by searching that number in the literature.
What to ask before joining anything
- What is asked of me: how many visits, how far, over how long a period?
- Are travel and time reimbursed?
- What happens to my current treatment during the study?
- What is the chance I receive placebo, and is there an option to receive the treatment afterward?
- Who do I call if something goes wrong at two in the morning?
- Will I be told the results, individually or as a group?
Alerts, so the search keeps working after you close the page
The best search you will ever build is the one that emails you when something new appears. Setting this up takes about five minutes, once.
A saved search in PubMed
- Build the search here and use Open in PubMed.
- Create a free NCBI account if you do not have one.
- On the results page, choose Create alert under the search box.
- Choose how often you want email and whether you want the abstracts included.
An alert in Europe PMC
- Use Open in Europe PMC from either lab.
- Create a free account, then save the search from the results page and turn on email notification.
A trial alert
On ClinicalTrials.gov, run your search, then save it to your account to receive notifications when new studies match. This is the one most families find valuable, because a new recruiting study in a rare condition can appear at any time and close within months.
How many alerts to run
Be careful not to set too many alerts. Consider one broad alert on the condition name alone, one narrow alert on the condition plus your specific question, and one trial alert. Alert fatigue is real and it sometimes results in ignoring all of your alerts.
Other ways to keep current
- Your advocacy group's research newsletter. Someone there is already reading this literature and summarizing it.
- The reference and citation trails described in lesson five, revisited every few months.
- Conference abstracts, which appear months before papers. Search the condition name together with the year and the word abstracts.
Search Lab: the published research
Build a search, run it against Europe PMC, and read the feedback. Every repair offered here explains itself and applies with one click.
Step one: the condition
These are joined with OR, so a record using any one of them is found.
Step two: what you want to know
Choose one to start. Each one you add is joined with AND, which makes the search narrower.
Step three: filters
The search string
You can edit this directly. Editing switches the tool to your own wording until you choose Rebuild from the form.
Run a search to see how many records match and what to change next.
Results
Trial Lab: ClinicalTrials.gov
The same method applied to studies rather than papers. Search what is registered, see which studies are recruiting now, and save the ones to ask about.
Step one: the condition
Registry records use the clinical name. If nothing comes back, try a broader parent term, for example epilepsy instead of one syndrome.
A drug name, a gene, a device, or a symptom. Leave this empty on your first run.
Step two: narrow it
A country, state, or city. Leave it empty first, then add it and watch what the count does.
Search to see how many registered studies match, and what to change if the answer is none or hundreds.
Studies
Five challenges
Each one checks itself against what you have actually done in the two labs. Work on your own condition rather than an example, so that finishing leaves you with something you can use.
One. Two names, one search
In the Search Lab, add at least two other names for your condition and run the search. This is the habit that finds the evidence other people miss.
Two. Into the readable band
Get one literature search to return between 20 and 200 results. Use the repairs offered in the feedback panel rather than guessing.
Three. Watch a repair work
Apply at least one suggested repair and look at the before and after counts. Knowing the size of the change each filter makes is most of the skill.
Four. Save three papers
Choose three records to keep. Prefer free full text so you can read them, and prefer recent reviews if you are new to the topic.
Five. Find one study to ask about
In the Trial Lab, save at least one registered study. It does not have to be one you would join. It has to be one you can raise at your next appointment or meeting.
My report
Everything you have done in this session, in one page you can print, save as a PDF, or email to yourself before an appointment, a meeting, or a class.
A name or a note to yourself. It appears at the top of the printed page and goes nowhere else.
One question per line.
Your words
Search log
Every search you ran, with the count it returned. This is the record a librarian or a researcher would keep, and it is the reason you can repeat your work later. Remove any line you do not want on the printed page.
Saved papers
Saved studies
Clear the report
This empties the log, the saved papers, the saved studies, and the questions box, and leaves your condition and your word list in place so the two labs still work. Each button asks you to press it twice.
Quick reference
The syntax, the links, and the fixes on one screen. Print this section on its own if you like.
Search syntax in both dialects
| What you want | Europe PMC | PubMed |
|---|---|---|
| Either of two words | (LGS OR "Lennox-Gastaut") | (LGS OR "Lennox-Gastaut") |
| Both ideas present | epilepsy AND cannabidiol | epilepsy AND cannabidiol |
| An exact phrase | "quality of life" | "quality of life" |
| In the title or abstract only | (TITLE:"x" OR ABSTRACT:"x") | x[tiab] |
| Published since 2021 | PUB_YEAR:[2021 TO 2026] | ("2021"[dp] : "3000"[dp]) |
| Reviews only | PUB_TYPE:"Review" | review[pt] |
| Free full text | OPEN_ACCESS:Y | "free full text"[sb] |
| By author | AUTH:"Smith J" | Smith J[au] |
| By journal | JOURNAL:"Epilepsia" | Epilepsia[ta] |
| A word and its endings | caregiv* | caregiv* |
When the count is wrong, in order
| Count | Do this first | Then this |
|---|---|---|
| 0 | Check the spelling and remove quotation marks | Remove one AND group, then widen the years |
| 1 to 19 | Add another name for the condition with OR | Remove the date and type filters |
| 20 to 200 | Read it | Save what you want and set an alert |
| 201 to 1000 | Restrict to title and abstract | Add a date range, then reviews only |
| Over 1000 | Add a second concept with AND | Then title and abstract, then dates |
Direct links
| Site | Address | What to use it for |
|---|---|---|
| PubMed | pubmed.ncbi.nlm.nih.gov | The reference index, saved searches, and email alerts |
| Europe PMC | europepmc.org | Free full text, preprints, citation trails |
| PubMed Central | ncbi.nlm.nih.gov/pmc | Free full-text archive |
| ClinicalTrials.gov | clinicaltrials.gov | Studies recruiting now, eligibility, contacts |
| WHO ICTRP | trialsearch.who.int | Trials registered outside the United States |
| MedlinePlus | medlineplus.gov | Plain-language background written for the public |
| GARD | rarediseases.info.nih.gov | Rare disease overviews and alternative names |
| Orphanet | orpha.net | Rare disease nomenclature, genes, and expert centers |
| OpenAlex | openalex.org | Citation trails and author profiles |
| Unpaywall | unpaywall.org | Finding a legal free copy of a paywalled article |
Ten habits that make searching easier
- Write the question down before you type anything.
- Search one idea at a time and add the second only after you see the first count.
- Keep a log, because the count you got last Tuesday is information.
- Use OR generously for names and AND sparingly for ideas.
- Read the last two sentences of an abstract first.
- Follow the references backward and the citations forward.
- Note the author names that keep appearing, then read what else they wrote.
- Set three alerts and no more.
- Ask your advocacy group what they already have. Someone may have done this search last month.
- Bring the paper, not the summary, to the appointment or the meeting.
Help, privacy, and a word about safety
If a search does not run
- Check that you are online, then choose Run this search again. Both services occasionally pause for maintenance.
- Some workplace, school, and hospital networks block outside connections. Try the same search from a home network or a phone.
- Very long search strings can fail. Remove one group and try again.
- If the tool still cannot reach a service, use Open in Europe PMC, Open in PubMed, or Open on ClinicalTrials.gov. Your search string travels with the link and everything you have learned still applies.
Privacy
This page keeps everything in the browser tab you are looking at. Nothing is saved to your device, no account is created, and no analytics of your searches are collected here. When you run a search, your search words travel to Europe PMC or ClinicalTrials.gov in the same way they would if you typed them on those sites yourself. Closing or refreshing the tab clears your work, so download the report before you go.
Safety
Research findings describe groups of people, and you are one person with a particular history. A result that holds for a study population may not hold for you, and stopping or changing a treatment on the strength of something you read can cause real harm. Take what you find to a clinician who knows your case.
If a website offers an experimental treatment for a fee, promises a cure, or asks you to travel abroad for an unregistered procedure, slow down. Check whether the study appears on ClinicalTrials.gov, ask your clinician, and ask your advocacy group whether they have heard of it.
Who made this
Built by Danielle Boyce, MPH, DPA, of Boyce Data Science. More free tools for advocates and research teams are at boycedatascience.com. If this helped, the printable companion workbook, The Search Field Guide, is available in the ThePreparedPage shop on Etsy and puts the same method on paper for use away from a screen.