Table 5

 Independent predictors of the assessment of publication bias yielded by a stepwise logistic regression analysis

VariableAll reviews (n = 225)Reviews with ⩾8 studies (n = 115)
OR (95% CI)pOR (95% CI)p
*⩾8 (ref<8) for the all sample, ⩾16 (ref 8–15) for the sample with ⩾8 studies.
Type of journal of publication (ref cardiology)
    Medicine in general0.96 (0.26 to 3.53)0.9551.00 (0.21 to 4.69)0.997
    Epidemiology/public health3.59 (0.47 to 27.73)0.2205.41 (0.59 to 49.56)0.135
    Other specialties0.14 (0.02 to 0.96)0.0450.12 (0.01 to 1.13)0.064
    Cochrane Library0.06 (0.01 to 0.69)0.024
Statistician among the authors (ref no)3.59 (0.89 to 14.45)0.0723.84 (0.73 to 20.23)0.112
Number of primary studies*5.40 (1.36 to 21.44)0.0171.23 (0.33 to 4.60)0.761
Use of several databases (ref 2):
    35.51 (1.07 to 28.41)0.04111.71 (1.16 to 118.1)0.037
    ⩾48.58 (1.73 to 42.62)0.0097.89 (1.00 to 62.13)0.050
Heterogeneity ascertainment (ref no)5.01 (0.46 to 54.36)0.1905.14 (0.38 to 68.72)0.216
Type of studies (ref experimental)3.60 (1.04 to 12.49)0.0443.61 (0.79 to 16.43)0.096
Year of publication (ref ⩽1998):
    20007.14 (1.50 to 33.96)0.01412.47 (1.83 to 84.79)0.010
    20025.73 (1.16 to 28.36)0.0326.18 (0.87 to 43.65)0.068
Statistical model (random or fixed effects) mentioned (ref no)4.17 (0.84 to 20.65)0.0802.91 (0.45 to 18.93)0.264