Intention-to-Treat vs. Per-Protocol

Methodology Published Jul 31, 2026

Intention-to-Treat vs. Per-Protocol

A way to read trial results by assignment or adherence.

Also known as

ITT analysis · intention to treat analysis · modified intention-to-treat · mITT · per protocol analysis · PP analysis · as treated analysis · full analysis set · protocol adherent population

It changes how much trust you should place in a study result, especially when people miss doses or drop out.

4 min read · 832 words · 3 sources

In brief

In brief

Intention-to-treat compares participants by original random assignment, while per-protocol compares only participants who closely followed the study plan; the difference matters because exclusions can make results look cleaner than real life.

  • Intention-to-treat keeps randomization intact by analyzing participants in assigned groups, even after missed doses or dropouts 1.
  • Per-protocol narrows analysis to participants who followed the protocol closely, which can raise apparent efficacy.
  • Per-protocol can select a more adherent group, so stronger results do not prove a better treatment.

Deep dive

How it works

ITT estimates the effect of being assigned to an intervention strategy. Per-protocol estimates the effect under sufficient adherence, but only if the comparison properly handles why people did or did not adhere. Simple per-protocol analyses can be biased because adherence is often related to symptoms, side effects, motivation, or early response.

When you'll see this

The term in the wild

Scenario

You are reading a randomized trial of a named probiotic strain, Lactobacillus rhamnosus GG, and the abstract says the ITT result was not statistically significant but the per-protocol result was.

What to notice

Notice that the larger result came after narrowing the group to people who followed the capsule schedule and completed follow-up. That does not prove the probiotic failed or worked. It tells you adherence changed the analysis question.

Why it matters

This keeps you from treating a cleaner subgroup result as the main real-world answer.

Scenario

A paper’s CONSORT flow diagram starts with 240 randomized participants, but only 178 appear in the per-protocol analysis.

What to notice

The missing 62 people matter. If dropouts were uneven between groups, or if people left because of side effects or no benefit, the per-protocol result may look better than the experience of all people who began the trial.

Why it matters

The flow diagram lets you judge whether the result survived real participation, not just perfect participation.

Scenario

A sports nutrition study of creatine says “modified intention-to-treat” included only participants with at least one post-baseline strength test.

What to notice

Modified ITT is not always the same as full ITT. It often excludes some randomized people, commonly those with no follow-up data. The paper should define exactly who was included.

Why it matters

The phrase sounds rigorous, but the definition determines whether the analysis still protects randomization well.

The full picture

The result can change before the supplement changes

Imagine a probiotic trial reports two numbers. In the main table, the benefit is modest. In a second table, the benefit looks much larger after researchers include only people who took at least 80 percent of capsules and finished every stool diary. The probiotic did not change between tables. The study question changed.

That is the core difference between intention-to-treat and per-protocol analysis. Intention-to-treat, often shortened to ITT, keeps people in the group they were originally assigned to, even if they missed doses, stopped early, crossed over, or did not follow the plan perfectly. The International Council for Harmonisation describes this as following and analyzing people regardless of how well they complied, because that preserves the benefit of random assignment.

Per-protocol, often shortened to PP, narrows the analysis to people who followed the study plan closely enough. It asks a different question: what was the effect among participants who actually took the intervention as planned and avoided major rule breaks? Cochrane frames the difference as the effect of assignment to an intervention versus the effect of adhering to it.

Why ITT is usually the first result to trust

Random assignment is powerful because, at the start, it spreads known and unknown differences across groups by chance. ITT protects that starting balance. If people are removed after randomization, the groups may no longer be comparable. The people who finish a demanding supplement protocol may be healthier, more motivated, less symptomatic, or less bothered by side effects than the people who leave.

That does not make per-protocol useless. It can answer a practical adherence question. For example, if a magnesium trial requires nightly dosing, a per-protocol analysis may help estimate what happened among people who actually took it nightly. But it is more vulnerable to a hidden sorting problem: the analysis may partly compare people who could stick with the plan against people who could not.

CONSORT, the major reporting guideline for randomized trials, pushes authors to show participant flow, exclusions, and numbers analyzed so readers can see how many people disappeared between randomization and the final result. That flow diagram is not paperwork. It is where the trial shows whether the cleaner result came from the intervention or from narrowing the group.

The decision to make today

When you read a randomized supplement trial, treat the ITT result as the headline estimate unless the paper clearly explains why another analysis was primary. Then look at the per-protocol result as a follow-up clue about adherence. If ITT shows little effect but per-protocol looks strong, do not jump to “it works if used correctly.” A safer reading is: “the effect may depend on adherence, but the cleaner analysis may also have selected a different kind of person.”

Modern trial guidance also uses the word “estimand,” which means the exact treatment question the study is trying to answer. ITT and per-protocol are not rival scorekeepers. They are answers to different questions, and the honest paper tells you which question came first.

Myths vs reality

What people get wrong

Myth

ITT means everyone took the treatment exactly as assigned.

Reality

ITT means people are counted in their assigned groups even when real life interfered. Missed doses, stopping early, or switching treatments can still remain in the analysis.

Why people believe this

The name sounds as if it describes what participants intended to do. In trial methods, it describes how researchers keep the randomized groups intact.


Myth

Per-protocol is always more accurate because it removes messy participants.

Reality

Per-protocol can be useful, but removing people after randomization can change who is being compared. The result may partly reflect who was able to follow the plan.

Why people believe this

Supplement marketing often highlights the cleanest responder-style result, while the CONSORT flow diagram and exclusions receive less attention than the abstract conclusion.


Myth

If ITT and per-protocol disagree, the study is useless.

Reality

Disagreement can be informative. It often says the effect may depend on adherence, missing data, tolerability, or how easy the intervention is to use.

Why people believe this

Readers are taught to look for one final answer, but trial analyses can answer different practical questions.

Why this keeps coming up

This comes up in supplement and nutrition trials because real participants often do not follow the plan perfectly, and that can change the result you see.

How to use this knowledge

A common failure mode is reading only the abstract. Before changing a supplement routine based on a trial, open the full paper and find the analysis population table or flow diagram. If many randomized people were left out of the result being promoted, lower your confidence in that result.

What to do with this

  • Start with the intention-to-treat result when you want the most realistic estimate.
  • Read the per-protocol result as a clue about what happens with close follow-through.
  • Compare how many people were randomized with how many were analyzed.
  • Check why people were left out before treating a cleaner result as the main answer.

Frequently asked

Common questions

Which result should I use when deciding whether a supplement is worth trying?

Start with the ITT result because it best reflects what happened after people were assigned to use the supplement. Use the per-protocol result as a clue about whether consistent use might matter.

Why do some trials report both ITT and per-protocol?

They answer different questions. ITT asks about assignment in real conditions, while per-protocol asks about outcomes among people who followed the plan closely.

What does modified intention-to-treat mean?

It means the authors changed the full ITT set in some way, often by excluding randomized people with no follow-up data. The term is not enough by itself. The paper must define who was removed.

Can per-protocol analysis ever be the main analysis?

Yes, especially when the scientific question is about the effect of actually adhering to the intervention. But the paper should say this in advance and explain how it handled bias from excluding people.

What is the fastest way to spot a problem?

Compare the number randomized with the number analyzed. If the promoted result uses far fewer people, read the exclusion reasons before trusting the size of the effect.

Want personalized recommendations?

Show me what works for me