Skip to main content

Table 2 Bivariable associations between trial characteristics and reporting of any exploration of subgroup analysis and reporting of subgroup analysis using appropriate methods

From: No improvement in the reporting of clinical trial subgroup effects in high-impact general medical journals

Characteristic Number Reports subgroup analysis, n (%) N = 270 Uses appropriate methods, n (%) N = 185
Journal of publication   p = 0.01 p = 0.22
Annals 37 21 (57) 18 (86)
British Medical Journal 48 21 (44) 12 (57)
Journal of the American Medical Association 76 44 (58) 33 (75)
Lancet 123 76 (62) 52 (68)
New England Journal of Medicine 153 108 (71) 70 (65)
Year of publication   p = 0.27 p = 0.046
 2007 119 74 (62) 57 (77)
 2010 144 79 (55) 54 (68)
 2013–2014 174 117 (67) 74 (63)
Biostatistician as coauthor   p = 0.07 p = 0.38
 No biostatistician as coauthor 183 104 (57) 68 (65)
 Biostatistician as coauthor 254 166 (65) 117 (70)
Medical condition under study   p = 0.003 p < 0.001
 Cardiovascular 101 73 (72) 63 (86)
 Infectious disease 82 52 (63) 25 (48)
 Cancer 59 45 (76) 24 (53)
 Psychiatry/neurology 40 20 (50) 11 (55)
 Autoimmune, including diabetes 37 22 (59) 15 (68)
 Pulmonary/critical care 29 14 (48) 11 (79)
 Obstetrics/gynecological 21 10 (48) 8 (80)
 Other chronic disease 41 24 (59) 19 (79)
 Other, uncategorized 27 10 (37) 9 (90)
First author’s region   p = 0.14 p = 0.19
 North America 185 123 (66) 84 (68)
 Europe 188 113 (60) 82 (73)
 Other 64 34 (53) 19 (56)
Funding   p < 0.001 p = 0.005
 Industry funding 188 141 (75) 86 (61)
 No industry funding 249 129 (52) 99 (77)
Significance of the primary outcomea   p = 0.013 p = 0.003
 Not significant 153 106 (69) 85 (80)
 Significant 277 158 (57) 100 (63)
Study design   p = 0.002 p = 0.53)
 Parallel 425 268 (63) 18 (69)
 Crossover 12 2 (17) 1 (50)
Sample size   p < 0.001 p = 0.01
 Quintile 1 (median = 69) 88 29 (33) 13 (45)
 Quintile 2 (median = 234) 87 44 (51) 30 (68)
 Quintile 3 (median = 507) 88 58 (66) 39 (67)
 Quintile 4 (median = 1080) 87 65 (75) 48 (74)
 Quintile 5 (median = 5455) 87 74 (85) 55 (74)
  1. For articles that report on the appropriate use of methods for subgroup analysis, the denominator used is the number reporting any subgroup analysis
  2. Chi-square tests were used for categorical variables. In the case of small cells, we used Fishers exact test. A test for trend was used for the year and sample size
  3. an = 7 trials were excluded for not reporting a statistical test for the primary outcome