Current practice in analysing and reporting binary outcome data—a review of randomised controlled trial reports

Table 3 Reporting of treatment effects, confidence intervals, and p values in the 198 studies that performed a statistical analysis

	Full text (n = 198)	Abstract (n = 198)
Reporting of statistical analysis
Any treatment effect measure¹	109 (55%)	75 (38%)
No treatment effect measure, p value only	75 (38%)	86 (43%)
No statistical analysis result reported	14 (7%)	37 (19%)
Reporting of treatment effects²
Relative treatment effect only (point estimate)	55 (28%)	42 (21%)
Point estimate and CI	55 (28%)	42 (21%)
Absolute treatment effect only (point estimate)	33 (17%)	26 (13%)
Point estimate and CI	30 (15%)	25 (13%)
Both relative and absolute treatment effects reported (point estimate)³	18 (9%)	7 (4%)
Point estimates and CIs	16 (8%)	6 (3%)

Numbers refer to any analyses for the primary binary outcome in the report and are not limited to the only principal analysis
CI confidence interval
¹Including papers that reported an estimate of an absolute or relative effect measure (point estimate and/or confidence interval)
²Three studies reported confidence intervals, but no point estimate (full text only). These studies were counted in ‘reporting of statistical analysis’, but not in ‘reporting of treatment effects—relative/absolute treatment effects only (point estimate)’ and ‘both relative and absolute treatment effects reported (point estimate)’
³Where both absolute and relative treatment effect estimates were presented, different statistical methods were used to obtain these estimates; no papers described that transformations to obtain an absolute effect from a relative one, or vice versa, were used

ISSN: 1741-7015