| Study (reference), Year |
Assessed Quality? | Included Trials | Methods | Years of Followup | Relative Risk (95% Confidence Interval) |
Number Needed to Screen |
|---|---|---|---|---|---|---|
|
Larsson et al,50 1997; Nystrom et al,32 1993 |
No | 5 Swedish trials | Weighted relative risks | 12.8 | 0.77 (0.59-1.01) | |
|
Cox,51 1997; Elwood,52 1993 |
No | All 8 trials | Fixed effects | 10 | 0.93 (0.77-1.11) | |
| Glasziou and Irwig,53,54 1997 |
Yes. All studies were "good." Rated Malmö and CNBSS highest and Two-County trial and Gothenburg lowest |
All 8 trials | Variance-weighted | 13.13 | 0.85 (0.71-1.01) | |
|
Hendrick et al,55 1997; Smart et al,56 1995 |
No | All 8 trialsa | Fixed effects | 12.7 | 0.82 (0.71-0.95) | 1,540 |
| Kerlikowske,57,58 1995,1997 | No | All 8 trials | Fixed effects | Approximately 12 | .84 (0.71-0.99) | 2,500 |
| Berry,30 1998 | No | All 8 trials | Random effectsb | 12-15 | .82 (0.49-1.17) | |
| Olsen and Gotzsche,8 2001 | Yes. Excluded 6 trials rated "flawed" or "poor" | Canadian, Malmö | Fixed Effects | 13 | 1.03 (0.77-1.38) | |
| Current study, 2002 | Yes. Rated Edinburgh "poor" and others fair or better | 7 trials, excluding Edinburgh | Random effects | Approximately 14 | 0.85 (0.73-0.99) | 1,792 |
aIncluded an additional 17 000 subjects from the Malmö II trial.
b Hierarchical Bayes model; estimates are for the "next trial" analysis.
Note: For multiple publications, data from the most recent update are recorded in the table.
CNBSS indicates Canadian National Breast Screening Study; NR, not reported.