For starters

The simplest trial design to analyse is called a parallel group design, where participants are randomized to receive one of two interventions, and outcome measurements are made at one time point at the end of the treatment period.

In this module we will look at some other types of trial design, which depart from the simple trial design above. These departures lead, among other problems, to mismatches between numbers of randomized units and the number of included participants (such as twenty participants but forty eyes) or outcome observations. These can lead to confusion, and errors. We call these errors 'unit of analysis errors', and the issues discussed in this module are generally known as 'unit of analysis issues'.

To determine which issues are relevant to your review, consider whether you have, or anticipate, any trials of the following sort:

• Crossover trials, in which patients are randomized to a sequence of interventions. In crossover trials, all patients might receive both (or more, if there are more than two) interventions.
• Cluster randomized trials, in which groups of participants are randomized.
• 'Body-part randomization' designs, in which different parts of a person are randomized to be treated differently. For example, a split-mouth design in dentistry might randomize the left and right halves of the mouth to different topical applications. Or, a physiotherapy trial might randomize two limbs of the same individual to different types of exercise.

Four other related topics can occur with standard parallel group designs:

• 'Body part analyses' of standard trial designs, in which a person is randomized to an intervention, but outcomes are measured for several different body parts. For example, in dentistry, a person might be randomized to use a particular toothpaste, but outcomes are collected on every tooth (i.e., cavity: yes/no for each tooth). Or eczema severity may be measured separately on both arms, even though they both received the same treatment.
• Trials with three or more treatment groups. An experimental intervention may be compared with both a standard intervention and with a placebo. Alternatively, two or more experimental interventions might be tested against a standard intervention
• Trials with the same outcome assessed at several time points
• Trials where an event outcome may occur more than once in each participant. A common example is the number of adverse events - participants may have more than one adverse event. Another example, in sub-fertility studies, a woman might have more than one pregnancy during a period of treatment. Or, in cardiovascular studies, a patient might experience more than one stroke
>