Intent vs. Per Protocol: Understanding the Difference

This article is a summary of the YouTube video ‘Intent-To-Treat & Per-Protocol’ by Epidemiology Stuff

Written by: Recapz Bot

Written by: Recapz Bot

AI Summaries of YouTube Videos to Save you Time

How does it work?
Randomized control trials assign participants to different treatment groups; non-compliance, or contamination, can occur, and two approaches to handling it are Intent to Treat (ITT) that ignores non-compliance, and Per Protocol (PP) that excludes data from participants who switched groups, with ITT being considered the gold standard for assessing treatment effectiveness.

Key Insights

  • Key Insights from the video transcript:
  • Randomized control trials (RCTs) involve assigning participants to different treatment groups.
  • Non-compliance can occur when participants fail to adhere to their assigned treatment.
  • Non-compliance is referred to as contamination, where individuals from one study group affect the other.
  • Two approaches to handling non-compliance are discussed: Intent to Treat and Per Protocol.
  • Intent to Treat (ITT) is considered the gold standard and involves ignoring non-compliance.
  • Under ITT, participants who switch treatment groups are still analyzed in their original assigned group.
  • Per Protocol (PP) involves excluding data from participants who switched treatment groups.
  • PP makes participants essentially choose their own treatment, resembling an observational study.
  • There are potential biases introduced in PP due to differences between those who stayed in their original group and those who switched.
  • PP is also known as an efficacy trial or as treated approach, focusing on strict adherence to the treatment.
  • ITT is more realistic for assessing treatment effectiveness in real-world scenarios where non-compliance occurs.
  • ITT accounts for the fact that people may stop treatment, providing a more comprehensive evaluation of effectiveness.
  • ITT preserves the intent of randomization by analyzing participants based on their original treatment assignment.

Seedless Grapes: Are They GMOs?

Annexation of Puerto Rico: ‘Little Giants’ Trick Play Explained

Android Hacking Made Easy: AndroRAT Tutorial

Andrew Huberman’s Muscle Growth and Strength Workout Plan

AMG Lyrics – Peso Pluma

Alex Lora: Rising Passion


This is part of my series on randomized control trials, and this one’s gonna talk about how to handle non-compliance.

So when we assign a treatment to participants, it’s possible that they might not stick to that treatment. They might stop taking it, or they might switch to the other treatment arm. The example that I’m gonna be using in this video is an exercise program, so the people assigned to the exercise group would be maybe going through intense bouts of exercise like on a daily basis, while the control group would not be doing that. Someone in the exercise group might stop exercising, and in that sense, they both stop taking the intervention and switch to the control arm. This is sometimes called contamination in the sense that people from one arm of the study are contaminating the other arm. So it’s basically a question of how do we handle this?

So just kind of following that example before, we have the exercise group and the non-exercise group. This is what they were initially assigned. And five years later, we see that four people from the exercise group now aren’t exercising, and one person from the non-exercise group is. So how do we handle that basically?

Intent to treat is the gold standard. And with intent to treat, you basically ignore that non-compliance occurred. So these four people that are now in the non-exercise group, we just pretend that they were always in the exercise group and that didn’t happen. And same with that one person who switched to the exercise group. This seems counterintuitive that that’s the gold standard, but basically it preserves the intent of randomization. And that’ll become clear when I talk about the other option, which is per protocol.

So with per protocol, we basically get rid of the, well, like we don’t count the data for people who switched arms. Basically, we make it a criteria for you to stay in the study that you stayed in your treatment arm. So these four people who switched treatment arms, their data is just deleted basically. And now participants are essentially choosing their treatment. And that’s how an observational study works. In this setting, we’re trying to do a randomized control trial where treatment is assigned. And we have basically made it so participants can choose their own treatment because everyone who, like as soon as people are switching arms and we’re analyzing accordingly, those who stayed in their treatment group essentially chose their treatment group because switching treatment groups was an option. And there are now two distinct populations. There’s those who stayed in their treatment group, they chose it. And those who did not stay in their treatment group, they did not choose it. And now all of a sudden we have these two populations and the people who stayed in their treatment group and those who didn’t might systematically differ in ways that are introducing bias.

Because when people self-assign treatment, it can be for a variety of reasons. People who exercise and those who don’t are probably different in many ways besides the fact that they exercise. So one group might make healthier decisions in all kinds of ways. Like the exercise group, they might have a better diet and they might just kind of live a healthier lifestyle. And since we’re basically permitting an element of choice, those who remain in the exercise group might have those traits compared to those who switched from the exercise group to the non-exercise group. And in that way, we’ve introduced bias.

So this per protocol way of doing things is also called an efficacy trial or as treated. And yeah, so it’s not quite as realistic as the other option, but depending on certain settings, like if you’re trying to test the efficacy of some drug or some vaccine, then that can be valuable in the sense that you really wanna know how well this works in a very controlled setting. But it’s not as realistic because in reality people do switch and we need to assess its effectiveness in the real world.

So yeah, basically yeah. The intent to treat option is more realistic of like the effect a medication will have in the real world because when you really do prescribe a medication, some people will stop taking it. And that’s basically factored into our effectiveness that we end up determining how effective the medication is. Where with an efficacy trial, we might overestimate its effectiveness because now we have this like group that is strict adherence to the medication which has really isolated the effectiveness of the medication because now we definitely know how effective it is. But that’s not necessarily an indication of how effective it will be in the real world because in the real world, some people stop taking it and that is part of its effectiveness that to account for how effective it’s gonna be in practice, accounting for the fact that people switch and intent to treat does that in addition to preserving the intent of randomization. So basically intent to treat is the better option.

This article is a summary of the YouTube video ‘Intent-To-Treat & Per-Protocol’ by Epidemiology Stuff