There are three distinct outcome types which are commonly reported in clinical trials:
- Continuous outcomes, such as blood pressure, weight or 6 minute walking distance.
- Binary outcomes recording "events" such as death, response to treatment or relief of symptoms.
- Survival, or time-to-event, outcomes such as time to death, time to relief of symptoms or time to discharge from hospital.
Most outcomes are of one of these types. Depending on how they are summarised, they might lead to more than one type of data. For example, a trial of cholesterol lowering drugs vs diet might report the difference in the mean changes in cholesterol at a given timepoint, or they might report how many participants achieved a cholesterol of
Another type of outcome might be seen in trials which seek to establish prognostic or diagnostic information. Prognostic research would generally use a regression model to assess the impact of a number of potential prognostic factors being considered on the outcome(s) of interest, with the results sometimes being converted into a prognostic index to allow a simple assessment to be made of the overall outlook for an individual. Regression models and diagnostic tests are outside the scope of this unit.
Occasionally you will come across a trial with an outcome that does not fit simply into one of these boxes, for example trials in conditions which involve repeated events over the period of follow-up, such as epilepsy, rheumatoid arthritis or asthma. There are some fairly sophisticated methods for analysing repeated events which are beyond the scope of this tutorial. In practice, many trials in conditions of this kind will use a simpler form of analysis, based on the more familiar methods which will be discussed in the rest of these pages. Where an unfamiliar or non-standard method of analysis has been used, it is wise to seek the advice of a statistician to check that it was appropriate.