Table 4.
Indirect comparison: results for WOMAC total score change from baseline
| Duloxetine | Ibuprofen | Naproxen | Celecoxib | Etoricoxib | Tramadol | Oxycodone | Hydromorphone | |
|---|---|---|---|---|---|---|---|---|
|
Frequentist analysis |
|
|
|
|
|
|
|
|
| Number of studies |
3 |
2 |
7f |
14f |
5 |
5 |
2 |
2 |
| Change from baseline vs. placebo, mean |
-6.48 |
-8.34 |
-8.27 |
-5.78 |
-11.04 |
-3.99 |
-8.56 |
-2.13 |
| 95% CI |
[-9.09, -3.88] |
[-11.98, -4.71] |
[-10.27, -6.28] |
[-6.86, -4.69] |
[-13.24, -8.84] |
[-6.74, -1.23] |
[-17.23, 0.11] |
[-5.99, 1.72] |
| I2 (%) |
44.35 |
0 |
51.92 |
32.49 |
0 |
58.03 |
71.99 |
63.54 |
| Indirect vs. Duloxetine a |
NA |
-1.86 |
-1.93 |
0.71 |
-4.56 |
2.36 |
-2.07 |
4.35 |
| 95% CIb |
NA |
[-6.33, 2.62] |
[-4.70, 0.84] |
[-2.12, 3.53] |
[-7.97, -1.15] |
[-1.00, 5.73] |
[-11.13, 6.98] |
[-0.31, 9.01] |
|
Bayesian analysis |
|
|
|
|
|
|
|
|
| Number of studies contributing to each compoundc |
3 |
2 |
9 |
16 |
5 |
5 |
2 |
2 |
| Change from baseline vs. placebo, meand |
-6.47 |
-7.85 |
-7.9 |
-6.2 |
-9.53 |
-2.89 |
-7.04 |
-2.19 |
| 95% CI |
[-9.27, -3.7] |
[-11.59, -4.18] |
[-9.54, -6.27] |
[-7.46, -5.03] |
[-11.86, -7.3] |
[-5.41, -0.54] |
[-11.35, -2.95] |
[-5.52, 1.21] |
| Indirect vs. Duloxetinea |
NA |
-1.38 |
-1.43 |
0.27 |
-3.07 |
3.57 |
-0.58 |
4.28 |
| 95% CIb |
NA |
[-6.04, 3.21] |
[-4.65, 1.81] |
[-2.78, 3.28] |
[-6.66, 0.49] |
[-0.17, 7.19] |
[-5.69, 4.32] |
[-0.01, 8.69] |
| Probability Duloxetine is Superior |
NA |
0.28 |
0.19 |
0.57 |
0.04 |
0.97 |
0.41 |
0.97 |
| Number of studies contributing to each compound for adjusted for baseline WOMAC scoree |
3 |
2 |
7 |
14 |
5 |
3 |
1 |
1 |
| Indirect vs. Duloxetine adjusted for baseline WOMAC scoree |
NA |
1.85 |
0.24 |
0.83 |
-0.43 |
4.92 |
-4.67 |
8.19 |
| 95% CIb |
NA |
[-2.13, 5.9] |
[-2.36, 2.87] |
[-1.45, 3.14] |
[-3.4, 2.57] |
[1.51, 8.34] |
[-13.24, 4.07] |
[3.84, 12.56] |
| Probability Duloxetine is Superior | NA | 0.82 | 0.57 | 0.76 | 0.38 | 1 | 0.15 | 1 |
aA positive (negative) result indicates that the compared treatment is worse (better) than duloxetine.
bIf zero does not fall between the upper and lower bounds the null hypothesis (treatments are the same) is rejected.
cThere are fewer studies in the adjusted analyses.
dRandom effects model.
eRandom effects model adjusting for baseline excluding trials with no baseline.
f2 studies without placebo arms were not included in the frequentist analysis.