Table 3

Pooled estimates of prostate cancer risk: sensitivity analyses

PooledHomogeneity
GroupingNo. studiesNo. risk estimatesRate ratio95% CIχ2 Woolfp value
No. risk estimates = number of estimators of relative risk. RR, relative risk; PMR, proportional mortality ratio. *The study of Andersen et al (1999), Norway21 was substituted by the study of Kristensen et al (1996)34; the study of Andersen et al (1999), Finland21 was substituted by the study of Pukkala and Notka (1997)40; the study of Andersen et al (1999), Sweden21 was substituted by the study of Wiklund and Dich (1995).48 †Studies included were Settimi et al (1998),43 Thomas et al (1996),46 and Zahm (1997).49 Pooled rate ratios are in bold when the 95% CI does not include 1.
All studies
    All studies excluding PMR1821 1.09 1.00 to 1.19271.805<0.001
Deleting studies with extreme RR values
    Largest value
        Fleming et al (1999) [30]17201.050.96 to 1.14226.675<0.001
    Smallest value
        Andersson et al (1996) [55]1720 1.10 1.01 to 1.20269.724<0.001
Deleting studies with extreme precision values
    Largest value
        Andersen et al (1999) Sweden [21]1820 1.10 1.00 to 1.22258.064<0.001
    Smallest value
        Sperati et al (1999) [45]1720 1.09 1.00 to 1.20271.609<0.001
Substitution by redundant studies*21211.080.98 to 1.20283.062<0.001
Inclusion of studies reporting less than 5 exposed cases†2124 1.09 1.00 to 1.19274.300<0.001