Skip to main content

Table 1 Potentially predictive variables for subsequent publication of abstracts presented at an international critical care conference

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

Potential predictor variable Published (n = 62) Non-published (n = 38) p valuea
Type of presentation
 Oral (vs. poster) presentation 14 (23%) 2 (5%) 0.025
 Research location in North Americab 40 (65%) 31 (82%) 0.075
Methodological quality variables
 Randomized 17 (27%) 10 (26%) 0.99
 Blinded 0 (0%) 0 (0%)
 Primary outcome given 13 (21%) 13 (34%) 0.16
 Numbers with denominators 4 (6%) 2 (5%) 0.99
 Number of animals stated 30 (48%) 17 (45%) 0.84
 Sample size calculation 0 (0%) 0 (0%)
Ethical quality variables
 Highest species rodentc 35 (56%) 26 (68%) 0.23
 >19 animals used 15/30 (50%) 5/17 (29%) 0.14
Outcome variables
 Main outcomes positive 57 (92%) 33 (87%) 0.50
 Statistically significant result 35 (56%) 20 (53%) 0.71
Type of animal model
 Sepsis 27 (44%) 13 (34%) 0.41
 Drug used 24 (39%) 15 (39%) 0.99
 Surgery performed 25 (40%) 13 (34%) 0.67
 Animals stated to be killed 34 (55%) 21 (55%) 0.99
Potential predictor variable Data in publication (n = 62) Data in abstract (n = 38) p value
Post-hoc comparisons
 Indicators of publication bias
  Main outcomes positive 62 (100%) 33 (87%) 0.003
  Statistically significant result 58 (94%) 20 (53%) <0.001
 Indicators of methodological quality
  Randomized 24 (39%) 10 (26%) 0.20
  Blinded 4 (6%) 0 (0%) 0.080
  1. a Comparisons made using Fisher’s Exact or Chi square test
  2. b Asia 17 (17%); North America 71 (71%), Europe 15 (15%); Australia/New Zealand 1 (1%), and Africa or South America 0
  3. c Species used were: rodent (61), rabbit (2), farm animal (35), primate (1), other (1: not stated)