Table 5. Results of sensitivity analyses on selected pair-wise comparisons.
| Comparison | RRR | RRCD | PC1 | PC0 | RRRadj | Percent Bias |
| Etoricoxib versus naproxen | 1.75 | 11.00 | 0.25 | 0.10 | 1.00 | 75.00 |
| Etoricoxib versus ibuprofen | 1.68 | 9.40 | 0.25 | 0.10 | 1.00 | 68.48 |
| Indomethacin versus naproxen | 1.23 | 2.80 | 0.25 | 0.10 | 1.00 | 22.88 |
| Diclofenac versus naproxen | 1.22 | 2.70 | 0.25 | 0.10 | 1.00 | 21.79 |
| Diclofenac versus celecoxib | 1.15 | 2.10 | 0.25 | 0.10 | 1.00 | 14.86 |
| Diclofenac versus ibuprofen | 1.13 | 1.95 | 0.25 | 0.10 | 1.00 | 13.01 |
| Naproxen versus ibuprofen | 0.92 | 0.50 | 0.25 | 0.10 | 1.00 | −7.89 |
RRCD is the association between confounder and disease outcome. PC1is the prevalence of confounder in the exposed. PC0 is the prevalence of confounder in the unexposed. RRRadj is the “true”, or fully adjusted, RRR. Percent bias is the percentage change to the RRR that would be introduced by a hypothetical confounding variable under the assumptions in the table.