Objectives: To model the effect of measurement of only one parameter for the outcome of spontaneous remitting illness.

Methods: Trials of intervention directed at symptoms for spontaneously remitting illnesses usually use time-to-recover (eg. time for a set percentage of people to recover; or percentage of people recovered at a set time). However, the amount of illness can be represented by severity as well as time (eg. severity x time). We demonstrate graphically the differences between two methods.

Results: Estimating differences between two-dimensional constructs by single dimensions is likely to yield under-estimates if severity contributes an important amount to the illness. Examples are drawn from interventions for acute respiratory infections.

Conclusions: It may be, therefore, that past trials have under-estimated the effects of many interventions for symptoms. In future trials we should collect daily severity data to enable such analyses.