Blood pressure reduction and cardiovascular prevention: meta-regression using ordered categorical (ordinal) event data

Abstract

Blood pressure (BP)-lowering trials studying efficacy mostly assesses binary outcome events, for example stroke/no stroke. Analysis of ordered categorical vascular events (e.g. three levels: fatal stroke/nonfatal stroke/no stroke) provides information on severity and is more powerful statistically than analyses using binary data, as used in previous meta-regression analyses of BP lowering. Summary data on stroke, myocardial infarction, and combined vascular events were obtained from published BP-lowering trials. Ordinal and binary odds ratios were calculated. The relationship between the difference in BP and treatment odds ratio was assessed using meta-regression. Thirty-eight trials involving 180 804 patients were included. A 'U' or 'J'-shaped relationship was found between on-treatment BP difference and three level ordinal stroke, ordinal myocardial infarction, and ordinal combined vascular outcome. Similar relationships were noted for binary stroke, myocardial infarction, and combined vascular outcome. Meta-regression curves for three level ordinal analyses were similar in shape and position to that for binary analyses. Meta-regression using ordinal outcomes in hypertension trials is practical and gives comparable values for the odds ratio as found in analyses based on binary outcomes. However, trials and meta-analyses reporting ordinal outcomes provide additional information in particular that lowering BP reduces both the severity and frequency of fatal and recurrent stroke. Trials should report data so that ordinal analyses may be performed.