OBJECTIVES: Even in late-phase trials, it is common practice to exclude patients with certain pre-existing conditions based primarily on clinical knowledge. Such exclusions can result in low recruitment and associated consequences. Comorbidities are common in chronic diseases and for a specific outcome, data on the risk attributable to each can help inform decisions on inclusion/exclusion. We present a graphical methodology for obtaining such empirical evidence METHODS: A retrospective cohort of 23,881 patients aged 50+ in the UK-GPRD at incident COPD diagnosis between 1990-1998 provided our setting. Each death patient was matched to as many survivors from the same practice as possible, of same age, sex and COPD duration. Some 18 binary comorbidities measured at the time of death were analysed in relation to mortality. Using conditional logistic regression model, we estimated hazard ratio (HR) for each comorbidity, adjusted for key baseline characteristics as well as its prevalence at the time of COPD diagnosis (PRC) and in 1998 (PR98). We plotted the HRs against the PRCs and PR98s as graphs A and B respectively RESULTS: Some 2,938 dead patients were matched to 5792 survivors. The most contributors to mortality risk were- CHF (HR: 3.3 p<0.0001; PRC=15.6%, PR98=18.2%), lung cancer (HR: 20.4 p<0.0001; PRC=0.7%, PR98=1.2%), other cancers (HR: 12.3 p<0.0001; PRC=0.7%, PR98=1.8%), and CVD (HR: 4.1 p<0.0001; PRC=3.2%, PR98=4.7%). Newly diagnosed moderate/severe liver disease (<1 year) was rare but with a high risk (HR: 16.7 p=0.014; PRC=0.3%, PR98=0.4%), suggesting such patients could be excluded in a trial. In contrast, diabetes-without-complication was common but with little effect on risk (HR: 1.2 p>0.37; PRC=5.1%, PR98=7.2%), suggesting such patients could be included in a trial with recruitment concerns CONCLUSIONS: The information provided by the tool can assist trial planning on sample size estimations as well as improve our understanding of how comorbidities influence outcome