Clinical Trials

Clinical Trials: A Practical Guide ■❚❙❘Table 1. Cause of death in a trial of magnesium supplementation for acute stroke [3]. Data are numberof patients (%).Cause of death Placebo (n = 1198) Magnesium (n = 1188) P-valueStroke 109 (9) 116 (10) 0.62Coronary heart disease 15 (1) 18 (2) 0.60Cardiac (noncoronary) 2 (0.2) 2 (0.2) 1.00Vascular (noncardiac) 16 (1) 20 (2) 0.51Cancer 0 2 (0.2) 0.25Pneumonia 47 (4) 56 (5) 0.37Other 7 (0.6) 13 (1) 0.16Total 196 (16) 227 (19) 0.086Reproduced with permission from Elsevier (Lancet 2004;363:439–45).The standard tableTable 1 gives an example of a standard outcomes table showing deaths in eachtreatment arm. Glancing at it initially, we are presented with numbers andpercentages by cause of death in a study comparing placebo and the activetreatment of magnesium. It is a trial testing the effect of magnesium supplementson patients who have had an acute stroke within the preceding 12 hours (some ofthis information is summarized in the table heading, which has been expandedfor the purpose of this chapter). The final column gives the P-value. Since there isone P-value per line, these clearly represent comparisons within each row,between the two treatments in each column, ie, between the placebo and themagnesium groups.The authors follow the standard practice of giving both the percentages (useful forcomparing directly between the two groups) and the numbers (useful for givingthe informed reader some idea of the precision of the percentages quoted). Thenumbers of patients recruited are shown at the top, while in each row the numbersof patients experiencing the event are also shown. The reader can thereforecalculate percentages for each outcome in each group for him/herself; thecalculated percentages agree with those shown.We must assume that there are no missing data since the percentages have beencalculated using the same denominators for each column for each cause of death;if there were missing data, then it would be desirable to add the appropriatedenominator in each column beside the absolute number of outcomes(eg, “109/1198 [9%]” for strokes in the placebo group). In this example, there is393