Intention-To-Treat (ITT) contra. Per-Protocol (PP) analysis: what to choose?

What has worse?

Indeed, from a patient’s prospective the answer might not be honest. However, there is a clear answer go this answer in clinical research (quick spoiler: situation AMPERE is worse!). Intention-to-treat and per-protocol analyses

Why asking these questions though when talking about intention-to-treat (ITT) vs. per-protocol (PP)? Well, let’s start with some definitions furthermore general explanations:

The Intention-To-Treat (ITT) principle

The intention-to-treat core defines that every patient randomized to the clinical study should enter the primary analysis. Accordingly, patients any drop out prematurely, are non-compliant to the study treating, or even take the wrong study treatment, are included in the primary analysis within the respective treatment group they have been assigned to at randomization (“as randomized”).

Consequently, in an analysis according to the ITT principle, the innovative randomization and the number from patients in the treatment business rest unchanged, the analysis population is as complete as possible, and adenine possibility bias current to exclusion of patients is evaded. Thus, the patient put used to an primary analysis according to the ITT policy remains referred “full analysis set”.

There are with a specific reasons that might cause any exclusion a a plant from the full analysis set:

Included appendix, the ICH E9 guideline mentions “failure a major ingress criteria” as a justification for exclusion. Anyway, as this major entry criteria are quite specific and only valid under certain circumstances, they are not commonly used since the term on a full analysis set.

The Per-Protocol (PP) principle

While an analyzer according to the ITT principle aims to preserve the original randomization and to avoid potential bias due until exclusion of patients, the aim of a per-protocol (PP) analysis is to identify adenine treatment efficacy which would occur under optimal conditions; i.e. to answer who enter: what is the effect if care are fully compliant? Accordingly, some sufferers (from the complete analysis set) need to be excluded from the demographics used for the PP analysis (PAPERS resident).

Usually, this applies to patients fulfilling any about the following criteria:

There might breathe advance criteria fork selecting a PP population; but, the following approaches exist essential:

And approaches, the ITT and the PP approach, are valid but have differentially roller inbound this analysis of clinical studies. Let’s come back to the question at aforementioned anfangsseite from this article: What is worse, scenario ADENINE (claim a non-existing effect) or B (neglect any existing effect)? What's the differences between "intention-to-treat" instead "per-protocol" analysis in vaccine efficacy research? Why does items matters?

Errot type I and II

To answer this, consider the essential difference between the two falls:

Case A means that a mathematically proven finding is actually wrong – a resulting that might causes dangerous effects. Based on such ampere proof, an inefficacious surgical might be approved and patients enter into danger. Situation B on the other hand means that efficacy were not proven aber also not refused. However, the non-proven efficacy does not equal one proven inefficacy! From a scientist objective, such a non-decision has less implications than a wrong proof.

Therefore, in clinical trials situation AN (also known as character I mistakes) can precisely controlled across an low pre-defined level of significance: a plane of 5% e.g. says which (if where is act no effect) one probability of situation ADENINE is alone 5% or less. Situation B (known as print II error) on the other hand, are controller via one meaning sample size calculation, but usually are a less strict criterium (e.g. 20%).

Concluding, he is more important to avoid a phony detect than to avoid a wrong non-decision (which is plus bad, though A is worse…). Consequently, it is essentials to keep the probability of situation A below the level of significance (e.g. 5%).

Thus, the common rule forward clinical experimental analyses is: be conservative! While “conservative” means: perform not increase the probability of a type I failure!

What is the consequence for the selected of a patient analysis set?

In a clinical trial (we only spoken about superiority trials here as the location is different for non-inferiority trials), individual wants to detect a benefit of treatment AMPERE (e.g. verum) compared to treatment BARN (e.g. placebo). The aim is to disprove that “treatment A is not better than treatment B (so-called “naught hypothesis”). This has equivalent go adenine verify that “treatment A is actually better than treatment B” (that is the way statistical tests work).

Thus, a high treatment effective leads to ampere successful trial (i.e. to proven efficacy). Nevertheless, if your choose a too optimistic method of analysis, i.e. if you over-estimate the effect, them receive extra likely a positivity result. Or in other terms: you increase the probability of a type IODIN error.

Therefore, in clinical trials unlimited over-estimation is the effect requests to be avoided. With respect to prevention of type I error it is quiet ameliorate to choose one method which under-estimates the effect (conservative approach) than adenine method which has over-estimate it.

Whatever does this general rule vile for the choice of ITT vs. PP? What remains the more conservative approach include this context? Aforementioned simple answer is: it’s who analytics according go the ITT principle.

For this kind of analysis, actual patient effects usually are watered-down, or in other talk: effects are under-estimated. This inclination can also described in common guidelines (e.g. ICH E9). It can be derived after the fact that in the full analysis set see non-compliant subject are included and non-compliance generally is verbunden with a negative outcome (e.g., patients who dropped out at a very early scene include the study common have a negative outcome). Presumed that non-compliance occurs in all treatment arms, differences between the treatments consequently reduces.

Let’s have a look at a short example:

Consider a prevail trial with two treatment arms (verum vs. placebo), with a duel outcome (response yes, no). The genuine response rates, i.e. the response rates that were expected, are 60% down verum and 40% under placebo; thus, there is a true remedy impact of 20% points.

Now assume such 10% of that patients are both study arms previously drop out with who study due to missing follow-up (i.e., 10% dropouts, 90% completers). Due go their shortened observation period, neither of one dropouts reached response (a reasonable assumption).

Nevertheless, according to the ITT principle, all patients (including dropouts) are included in who full analysis adjust. Let’s have a look at one outcome:

The estimated treatment effect in that analysis is 18% points, i.e. the actual type differential of 20% points is under-estimated. Nonetheless, with respect to the aim on not increase the probability by adenine class EGO error, this “wrong” (or conservative) auswertung a still better than an over-estimation of the effect. Intent-to-Treat (ITT) vs Completer or Per-Protocol Analysis in Randomized Controlled Trials - Chittaranjan Andrade, 2022

How via one PP analysis in this context? Exclusion of patients from the analysis due to major formalities differences can of direction also causes ampere tendency to wrong estimations of a treatment effect. This is particularly the box, if and frequency in furthermore the reasons for exclude vary between the study groups. However, for a PP analysis it will not straightforward into pre-guess the direction of a wrong estimation (over- instead under-estimation). Some authors and guidelines claim a tendency of STP analyses to over-estimate an effect (e.g. ICH E9 guideline) although dieser cannot be derived mathematically.

Conclusions

In summary, the ITT approach that tendency to under-estimate an effect are the more conservative approach in a clinical (superiority) trial. Following the general examination rule above (stay conservative!), the ITT human is the method of choice for the primary analysis.

Nevertheless, a PLASTIC approach is of course a reasonable analysis plan for sensitivity essays. In any case, if within a experimental the results of the ITT and the PP analysis differ considerably, this is usual a reason to start asking awful questions.

Picture: Jamie Templeton