Table 3.
Results for the osteoarthritis of the knee data.
| Treatment | Parameter | New procedure | Previous Bayesian analysis | ||
|---|---|---|---|---|---|
| Estimate | P (best) | Estimate | P (best) | ||
| A: standard care | — | 0.00 | 0.00 | ||
| B: placebo | δ AB | 0.04 (0.20) | 0.00 | 0.04 (0.23) | 0.00 |
| C: no medication | δ AC | 0.60 (0.32) | 0.00 | 0.59 (0.35) | 0.00 |
| D: acupuncture | δ AD | −0.78 (0.16) | 0.08 | −0.78 (0.19) | 0.07 |
| E: balneotherapy | δ AE | −0.46 (0.25) | 0.00 | −0.49 (0.30) | 0.01 |
| F: braces | δ AF | −0.15 (0.46) | 0.02 | −0.15 (0.50) | 0.02 |
| G: aerobic exercise | δ AG | −0.59 (0.22) | 0.03 | −0.57 (0.25) | 0.03 |
| H: muscle exercise | δ AH | −0.37 (0.11) | 0.00 | −0.36 (0.15) | 0.00 |
| I: heat treatment | δ AI | −0.03 (0.30) | 0.00 | −0.02 (0.33) | 0.00 |
| J: insoles | δ AJ | −0.01 (0.35) | 0.00 | 0.00 (0.41) | 0.00 |
| K: tai chi | δ AK | −0.28 (0.29) | 0.01 | −0.28 (0.34) | 0.01 |
| L: weight loss | δ AL | −0.35 (0.26) | 0.01 | −0.35 (0.29) | 0.01 |
| M: sham acupuncture | δ AM | −0.25 (0.23) | 0.00 | −0.28 (0.28) | 0.00 |
| N: ice/cooling | δ AN | −0.25 (0.37) | 0.01 | −0.25 (0.39) | 0.01 |
| O: interferential | δ AO | −1.11 (0.48) | 0.52 | −1.11 (0.51) | 0.49 |
| P: laser | δ AP | −0.25 (0.36) | 0.00 | −0.24 (0.42) | 0.01 |
| Q: manual | δ AQ | −0.29 (0.30) | 0.00 | −0.29 (0.32) | 0.01 |
| R: NMES | δ AR | 0.46 (0.55) | 0.00 | 0.45 (0.58) | 0.00 |
| S: PES | δ AS | −0.70 (0.32) | 0.08 | −0.73 (0.36) | 0.09 |
| T: PEMF | δ AT | 0.01 (0.32) | 0.00 | 0.01 (0.38) | 0.00 |
| U: static magnets | δ AU | −0.78 (0.57) | 0.24 | −0.78 (0.61) | 0.23 |
| V: TENS | δ AV | −0.63 (0.22) | 0.01 | −0.61 (0.24) | 0.01 |
| Heterogeneity | τ β | 0.42 | 0.42 (0.06) | ||
| Inconsistency | τ ω | 0 | 0.14 (0.10) | ||
NMES, neuromuscular electrical stimulation; PES, pulsed electrical stimulation; PEMF, pulsed electro‐ magnetic fields; TENS, transcutaneous electrical nerve stimulation.Two sets of results are shown, those using the new ‘DerSimonian and Laird’ method and those from the previous analysis using WinBUGS are also shown for comparison. For the new procedure, estimates are followed by standard errors in parentheses. For the previous WinBUGS analysis, the estimates are posterior means, which are followed by posterior standard deviations in parentheses. P (best) is the probability that each treatment is best; these probabilities are included so that the results using the new method can be compared with those obtained previously and are not adequate for the full probabilistic ranking of the treatments.