Intent-to-treat (ITT) vs Observed/on-treatment (OT) analyses
These are two important ways that drug trial results are analysed.
- ITT includes all participants when calculating the response rates.
- OT only calculates the response rates for people still on the randomised treatment.
- 100 people use a trial drug in one arm of a study
- 25 stop treatment before the end of the study for various reasons
- 50 have an undetectable viral load after 48 weeks
- 25 have a detectable viral load after 48 weeks
In an ITT analysis 50% of people got an undetectable viral load using the study drug (50 out of 100 patients).
In an OT analysis, 66% of people got an undetectable viral load using the study drug (50 out of 75 patients).
ITT analyses are more conservative but arguably are most important when looking at overall effectiveness and safety. OT analyses always make a drug look more effective, so you need to check which analysis is being presented.
In-vitro – A study in a test tube.
In-vivo – A study in humans.
Matched sample – ie each group has similar age, gender, ethnicity, etc
Null hypothesis – this sometimes just refers to the hypothesis in a study, but more specifically it refers to the idea that any difference between 2 study groups has only occurred by chance.
Open label – where a patient in a trial knows which treatment they are taking.
Publication bias – refers to the tendency for published results to be different to other trials. For example, trials that show a positive effect are more likely to be reported and published, than trials that find no effect.
Qualitative – where what is being measured either fits one of several categories, or includes descriptive results.
Quantitative – where what is being measured has a numerical value or fits a pre-defined scale or range of responses.
Roll-over study – when patients in one study ‘roll-over’ to a second related study. For example, this can be after a fixed period or after another event (for example, not having a treatment response).
Study population – the group of people studied. What happens to the study population is not guaranteed to happen in every person.
Last updated: 1 January 2023.