Clinical Trials

Clinical Trials: A Practical Guide ■❚❙❘Assessing the size of treatment effect in a two-arm trialfor survival dataIn this section, we introduce some measurements of treatment effect in a two-armtrial and discuss their advantages and disadvantages.Risk difference, risk ratio, and odds ratioRisk is simply measured as the proportion of subjects who have an event ofinterest by a specific time point. The odds are the ratio of patients with an eventcompared to those without the event [5]. The risk difference or ratio and the oddsratio are sometimes used to measure the treatment effect at a specific point infollow-up when an endpoint is time to the occurrence of an event. However, thesemeasurements may be biased for the following reasons:• They are based on the assumption that all patients were followed-upto the end of the study if they had not died. In our example, not allpatients who had not died reached 48 months of follow-up dueto censoring.• Patients with censored events might not be balanced between thetwo groups.• No distinction is made between patients who die at 1 monthand those who die at 48 months.Incidence rate difference and ratioTo take different follow-up times into account, we can calculate an incidence rate,ie, the number of events divided by the number of units of time [5]. By comparingthe incidence rates between treatment groups, we can derive the incidencerate difference and ratios following the procedures described by Kirkwood andSterne [5]. For the pancreatic cancer trial data, the incidence rates are calculatedas 0.9 and 2.9 deaths per 100 person-months for the new treatment group and thestandard treatment group, respectively. The estimates of incidence rate differenceand rate ratio, together with their 95% CI and P-value, are as follows:• incidence rate difference: –2.0, 95% CI (–3.9, –0.1), P = 0.034• incidence rate ratio: 0.30, 95% CI (0.08, 0.94), P = 0.026The above results suggest that the new treatment reduces deaths by two per100 person-months, with a 95% CI of 0.1, 3.9 per 100 person-months, and that theincidence rate for patients in the new treatment group is only about 30% of theincidence rate for those in the standard treatment group.249