Who is having the events?

Our final topic here can cause confusion. Consider an outcome that is an event: a stroke, for example. What if the event can happen more than once? A person can have more than one stroke during a period of follow-up. We have to be very careful that we know exactly the nature of the data being reported. For an outcome to be considered as a dichotomous outcome there must be an all-or-nothing distinction between the "yes" and "no" classifications. Thus, if we count that 245 people had at least one stroke and 765 people had no strokes, we have dichotomous data. If, on the other hand, we just know that there were 312 strokes among 1010 people then we may not know who had them. If some people could have had more than one of the strokes we do not have dichotomous data. We cannot treat the data as dichotomous data when entering them into RevMan, unless we can distinguish the number of people having events from the number of events.

The same difficulty occurs when we are analysing adverse events. Suppose we know that there were 120 adverse events among 250 participants in a trial. It may be that 20 people had two events each, accounting for 40 of the 120 events. If the other 80 adverse events occurred as one in each of 80 people, that would mean a total of 100 people had the 120 events. Unless we are told about how often people had different numbers of events, we cannot calculate the number of participants who had any adverse event.

So what can we do? If you think this poses a problem, consult Module 14 or section 'Counts and rates' in the section 8 of the Reviewers' Handbook.