Publication bias in animal research presented at the 2008 Society of Critical Care Medicine Conference

Table 2 Changes in reporting from abstract to subsequent publication of animal research presented at an international critical care conference

Variable	Prevalence in abstracts n = 100	Prevalence in publications n = 62	Change from abstract (A) to publication (P)
Randomized	27 (27%)	24 (39%)	12/62 (19%): non-R in A; R in P
Method of randomization	0 (0%)	0 (0%)	0 (0%)
Allocation concealment	0 (0%)	0 (0%)	0 (0%)
Blinding (possible)	0 (0%)	4 (6%)	4/62 (6%): no mention in A; blinding of some outcomes in P
Sample size calculation	0 (0%)	0 (0%)	0 (0%)
Primary outcome stated	26 (26%)	3 (5%)	10/62 (16%): 9 stated in A, not stated in P; 1 stated primary outcome was different between A and P
Numbers with denominators	6 (6%)	21 (34%)	17/62 (27%): no denominators in A; denominators in P
Main outcomes positive	90 (90%)	62 (100%)	5/62 (8%): negative in A; positive (or excluded) in P
Number of animals stated^a	47 (47%)	35 (56%)	13/62 (21%): in the P the number was smaller in 3 (5%) and larger in 10 (16%)
Statistically significant result of main outcomes^b	55 (55%)	58 (94%)	23/62 (37%): not significant (or not stated) in A; significant in P

A abstract, P publication, R randomized
^a In abstract: median 18 [IQR 11–24] (range 1–60), total 957 animals used. In publication: median 20 [IQR 14–35] (range 5–125), total 993 animals used. When smaller in publication: by 3, 4, and 6 animals. When larger in publication: by median 14 [IQR 5–25] range 4–54, total 213 animals. Reasons for change in number were due to: new control group (1), different numbers in both control and intervention group (9), different numbers in the only group in the study (1), new reason animals required (2), or not clear (1)
^b In the 23 that changed in statistical significance from A to P: the animal numbers did not change in 5 [these numbers did change in 4 (larger number in 2, and smaller number in 2), and change could not be determined in the rest because numbers were not stated in P in 9, and were not stated in the A in 5]; the main outcomes changed in 2; and denominators changed in 3 [change could not be determined in 19 others because denominators were not reported; thus we could be sure that denominators did not change in only 1]