Labour Q1 pain – poorly analysed and reported: a systematic review

Table 3 Statistical methods and presentation of results in 133 labour pain studies with repeated measures design

		Number	Percent (%)
Most advanced statistical method to analyse labour pain	End-point analysis^a	97	72.9
	rANOVA	26	19.5
	Mixed effect models	9	6.8
	Not clear	1	0.8
Clear how between-groups comparisons are conducted?	No	34	25.6
Clear how between-groups comparisons are conducted?	Yes	99	74.4
Are comparison conducted within groups (between time points)?	No	87	65.4
	Yes	46	34.6
Clear how within group comparisons are conducted?	No	19	41.3
Clear how within group comparisons are conducted?	Yes	27	58.7
Are numbers of valid observations used in analysis for each group and time-point stated?	No - only number of individuals at one time-point, e.g. baseline is stated	112	84.2
	Yes - number of individuals/valid measurements at each time-point is stated	21	15.8
Is normality assumption tested/mentioned?	No	120	90.2
Is normality assumption tested/mentioned?	Yes	13	9.8
Clear agreement between statistical methods and results presented?	No	77	57.9
	Yes	56	42.1

^aSeveral end-point analyses could be used in each of these 97 studies: t-tests were used in 59 (44%) studies, ANOVA in 41 studies (31%), Mann-Whitney in 20 studies (15%), Wilcoxon in 6 studies (5%), and ANCOVA in 1 study (1%)
Illustrative binomial exact 95% confidence intervals for percentages when sample size is 133: 1% (0 to 4%); 5% (2 to 10%); 10%(5 to 16%); 25%(18 to 33%); 50%(40 to 58%); 90(84 to 95%)

ISSN: 1471-2393