Skip to main content
. Author manuscript; available in PMC: 2010 Dec 28.
Published in final edited form as: Ann Intern Med. 2010 Mar 16;152(6):380–390. doi: 10.1059/0003-4819-152-6-201003160-00008

Appendix Table 5.

Potential Predictors of Underlying Risk (Percentage of All-Cause Mortality After 1 Year in Reference Population), Exploring Heterogeneity by Using Categorical and Meta-regression Analyses*

Explanatory Variable Women Men
All Studies Outlying Study Omitted† All Studies Outlying Study Omitted†
R2, % P Value R2, % P Value R2, % P Value R2, % P Value
Categorical meta-analyses
 Region 0.031 0.62 <0.001 0.006
 Control population source 0.012 0.004 0.141 0.149
 Procedure to ascertain death 0.91 0.66 0.64 0.48
Meta-regression analyses
 Latitude 9.7 0.073 2.9 0.34 0.1 0.89 0.3 0.84
 Starting year of study 11.4 0.037 7.8 0.152 2.7 0.53 4.2 0.71
 Duration of study, y 10.9 0.041 2.3 0.39 41.2 <0.001 22.8 0.063
 Included participants with hip fracture, n 2.5 0.44 2.1 0.49 8.0 0.32 8.4 0.33
*

Between-group heterogeneity P value for categorical meta-analyses and model P value for meta-regression analyses. P values are presented without adjustment for multiple testing.

†

The outlying study is the Dubbo Osteoporosis Epidemiology Study, which was done in Australia and had a 15-year observation period (66). The authors acknowledge that there was probably selection bias in their study because participants were healthier than nonparticipants (67).