Somebody sends you a protocol. It runs sixty pages and the first thing on page one is a table.

The call is in nine days. You have been asked what you think.

Here is what nobody says out loud: you cannot say anything useful about a study until you know what kind of study it is. Not the topic. The shape.

A comment that is exactly right about an early safety study is beside the point in a real-world effectiveness trial. Same words. Wrong room.

So this is a field guide to shapes.

I am not a methodologist. I am a person with pain who has sat at these tables long enough to learn to ask what I am looking at. That question — what kind of study is this? — has never once made me look uninformed. It slows the room down and makes someone explain, which is usually what the room needed anyway.

Start with the biggest fork: is anyone being assigned anything?

“Clinical study” is the umbrella term. Two families sit under it.

• Interventional study — a clinical trial. Investigators assign participants to receive something, according to a written protocol. That something might be a drug, a device, a procedure, a change in behavior, or a different way of delivering care.
• Observational study. Investigators watch and measure. They do not assign. Cohort studies, case-control studies, cross-sectional surveys, and patient registries all live here.

Both are real research. Both are registered on ClinicalTrials.gov. But the questions worth asking are not the same.

In an observational study, “why is the comparison group getting nothing?” is not a question. In a trial, it is often the most important question in the room.

What the trial is for

Every registered study declares a primary purpose. The common ones:

• Treatment. Does this thing help the condition?
• Prevention. Does this thing keep the condition from starting or coming back?
• Diagnostic. Does this test identify the condition accurately?
• Screening. Does looking for this, in people who feel fine, do more good than harm?
• Supportive care. Does this improve comfort, function, or quality of life alongside whatever else is happening?
• Health services research. Does this way of organizing or delivering care work better than that way?
• Basic science. What is actually going on in the body?

Pain research is spread across all seven. That matters more than it sounds. Supportive care and health services research are where a great many of the outcomes we care about actually live, and they are the categories most often treated as the soft ones.

Phases — and when there are none

Phases come from FDA definitions, and they apply to drugs and biologic products. Five labels, in order:

• Early Phase 1 (once called Phase 0). Very small, very exploratory, tiny doses. Nobody is looking for benefit here.
• Phase 1. Safety and dose. FDA describes a typical size of twenty to one hundred people, often healthy volunteers rather than people with the condition.
• Phase 2. Does it appear to do anything, and what else does it do? More safety data, and the material used to design the next stage.
• Phase 3. The pivotal one. FDA describes three hundred to three thousand participants, and says roughly a quarter to a third of drugs that reach this stage move on from it. This is where most safety data comes from, because rare harms need numbers to show up.
• Phase 4. After approval. Long-term safety and how the thing behaves out in ordinary use.

Now the line that matters most to us.

A great many pain studies have no phase at all. ClinicalTrials.gov labels them “Not Applicable.” Device trials, behavioral trials, exercise programs, psychological therapies, and care-delivery models are all in that bucket. Nothing is missing from them. The phase ladder was built for drug development and was never meant to describe a physical therapy protocol.

If someone tells you a non-drug study is “only Phase 2,” ask what they mean by that. Sometimes it is shorthand. Sometimes it is a tell.

How people are assigned

• Randomized. Chance decides which group you land in. The point is not fairness in the moral sense. The point is that the groups start out alike in ways nobody thought to measure, so a difference at the end is more likely to be the treatment and less likely to be the people.
• Non-randomized. Assignment follows some other rule — by site, by preference, by order of arrival.
• Single group. Everyone gets the same thing. There is no comparison group, so there is a limit to what the result can tell you.

How the groups are arranged

• Parallel. Two or more groups run at the same time, each getting one thing. The default shape.
• Crossover. Each person gets both treatments, one after the other, usually in a randomized order. You serve as your own comparison. Common in pain research, and it raises a question worth asking out loud: what happens during the washout period between treatments, and who is comfortable coming off something that is working?
• Factorial. Two questions at once. Half the participants get the drug or not, and independently, half get the therapy or not. Efficient, and harder to explain on a consent form.
• Cluster randomized. Whole clinics or communities are randomized rather than individuals. Used when you cannot sensibly randomize person by person inside one waiting room.
• Stepped wedge. Everyone eventually receives the intervention; what is randomized is when. This design deserves to be better known in our community, because it answers the objection that somebody has to go without.

What it is being compared against

This is the choice I would look at first if I only had time to look at one.

• Placebo. An inactive comparison.
• Active comparator. A real treatment already in use.
• Usual care. Whatever the clinic would have done anyway.
• No intervention. The group gets nothing added.
• Historical or external control. Comparison against people from a previous study or a database rather than from this one.

A treatment that beats a sugar pill has cleared a low bar. A treatment that beats what you are already taking has told you something you can use on a Tuesday morning in a clinic.

One more thing about placebo, since this is a pain publication. The placebo response in pain trials is large and it is real. That is a fact about how pain works, not an accusation against people who report relief. When a trial fails because the placebo group improved too, the honest reading is that the design struggled — not that anyone lied.

Who knows what

Masking, which you will also hear called blinding, describes who is kept from knowing the assignment. ClinicalTrials.gov counts up to four groups: the participant, the care provider, the investigator, and the person assessing the outcome. A study is listed as single, double, triple, or quadruple masked depending on how many of those four are kept from knowing. Open label means nobody is.

Behavioral and procedural trials often cannot mask the participant. You know whether you did the exercise. That is a limitation to name plainly in the paper, not a flaw to bury in the discussion section.

Explanatory or pragmatic — the fork that matters most to us

Two very different questions hide behind the same word “trial.”

• Explanatory. Can this work under good conditions? Narrow eligibility, and monitoring no ordinary clinic could sustain. The answer is clean and it is about a population that may not include you.
• Pragmatic. Does this work in the clinic I actually go to, for the people who are actually in it? Broad eligibility, ordinary settings, and data pulled from records that were being kept anyway.

PCORI describes its comparative clinical effectiveness research as sitting near the pragmatic end — large studies comparing two or more real alternatives that people and clinicians are already choosing between, in real-world settings, with eligibility loosened on purpose so the findings travel.

Here is the test I use. Read the exclusion criteria and count how many people in my Tuesday support group would be screened out. If the answer is most of them, that study is not going to answer our question, however well it is run.

And a correction to the easy version of this: pragmatic is not automatically better. Loosening the controls buys reach and costs precision. Both designs are honest. They are answering different questions, and the trouble starts when a result from one gets quoted as though it came from the other.

Newer shapes you will start to see

• Adaptive. Rules written in advance let the trial change while it runs — dropping an arm that is not working, or shifting more participants toward one that is. The changes are prespecified, not improvised.
• Platform, umbrella, and basket trials. One piece of infrastructure tests several treatments, or one treatment across several conditions, against a shared control group. Slower to set up and much faster after that.
• N-of-1. One person, cycling between treatments in a randomized order, measured each time. This is the design that comes closest to the question a person with pain actually asks, which is not “does this work” but “does this work for me.”

Where a person with lived experience actually changes something

Every fork above is a decision. Each one gets made months before the first participant is recruited, usually in a room with no PWLE in it.

These are the six places where our presence changes the study rather than decorating it:

• Eligibility criteria. Every exclusion is a group of people who will not be represented in the answer.
• The comparator. What the new thing is measured against is a values question at least as much as a statistical one.
• Outcome measures, and how often they are collected. Pain intensity at twelve weeks is not the same question as whether someone got their life back.
• Participation burden. Travel, and unpaid time away from work.
• Washout periods in crossover designs. Someone is being asked to come off a treatment that is working, in service of the design.
• What happens when the study ends. Whether people who improved keep access, and whether anyone tells them what was found.

If you are a researcher reading this, that list is the ask. Bring those six to your partners before the protocol is written. Afterward, what we give you is a comment.

Learning the shapes will not make you a methodologist. It will make you harder to hand a finished protocol and call that engagement.

Patient engagement is an invitation to advise. Patient empowerment is holding a real vote.

Knowing what kind of trial it is tells you which one you were offered.

— Tom