TABLE 3.
Results of meta‐analyses assessing primary outcome.
| Treatment | Global impression | Depression | Mania | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| k | N | ES | SE | 95% CI | I2 | k | N | ES | SE | 95% CI | I2 | k | N | ES | SE | 95% CI | I2 | |
| Antipsychotics | 6 | 900 | 0.79 | 0.04 | 0.71–0.86 | 0% | 10 | 1131 | 0.75 | 0.09 | 0.56–0.93 | 69% a | 3 | 287 | 1.11 | 0.01 | 0.92–1.30 | 0% d |
| Quetiapine | 4 | 804 | 0.79 | 0.04 | 0.71–0.87 | 0% | 5 | 855 | 0.76 | 0.09 | 0.59–0.92 | 37% a | 1 | 159 | 1.01 | 0.18 | 0.65–1.37 | — |
| Olanzapine/OFC | 1 | 76 | 0.87 | 0.14 | 0.60–1.13 | — b | 3 | 242 | 1.01 | 0.08 | 0.86–1.17 | 0% | 1 | 76 | 1.19 | 0.15 | 0.90–1.49 | — d |
| Aripiprazole | — | — | — | — | — | — | 1 | 14 | 0.06 | 0.27 | −0.47 to 0.58 | — | 1 | 52 | 1.09 | 0.18 | 0.75–1.43 | — |
| Risperidone LAI | 1 | 20 | 0.56 | 0.24 | 0.08–1.03 | — | 1 | 20 | 0.38 | 0.23 | −0.07 to 0.83 | — | — | — | — | — | — | — |
| Studies removed in sensitivity analysis | Tohen (potential effect of including non‐RC patients) | Langosch (RoB/partial remission) | Baldessarini (RoB) | |||||||||||||||
| Mood stabilisers | 5 | 177 | 0.67 | 0.14 | 0.40–0.95 | 59% a | 6 | 192 | 0.83 | 0.13 | 0.57–1.08 | 51% a | — | — | — | — | — | — |
| Lamotrigine | 3 | 75 | 0.58 | 0.21 | 0.17–0.99 | 61% a | 3 | 75 | 0.70 | 0.21 | 0.29–1.11 | 59% a | — | — | — | — | — | — |
| Lithium | 1 | 34 | 0.92 | 0.21 | 0.52–1.33 | — | 2 | 49 | 1.01 | 0.18 | 0.67–1.36 | 0% | — | — | — | — | — | — |
| Divalproex and Na Valproate | 1 | 68 | 0.77 | 0.14 | 0.50–1.00 | — | 1 | 68 | 0.86 | 0.14 | 0.58–1.14 | — b | 1 | 68 | 0.91 | 0.14 | 0.63–1.19 | — c |
| Studies removed in sensitivity analysis | Goldsmith 2003 (partial remission) | Goldsmith (partial remission) & Langosch (RoB/partial remission) | ||||||||||||||||
| Antidepressants | 2 | 65 | 0.90 | 0.19 | 0.52–1.27 | 16% b | 4 | 101 | 1.19 | 0.23 | 0.75–1.64 | 52% | — | — | — | — | — | — |
| Citalopram | 1 | 59 | 0.99 | 0.16 | 0.68–1.30 | — | 1 | 59 | 1.10 | 0.17 | 0.78–1.43 | — | — | — | — | — | — | — |
| Escitalopram | 1 | 6 | 0.48 | 0.43 | −0.36 to 1.33 | — | 1 | 6 | 1.07 | 0.51 | 0.06–2.07 | — | — | — | — | — | — | — |
| Paroxetine | — | — | — | — | — | — | 1 | 24 | 0.90 | 0.24 | 0.42–1.37 | — | — | — | — | — | — | — |
| Venlafaxine | — | — | — | — | — | — | 1 | 12 | 2.48 | 0.58 | 1.33–3.62 | — | — | — | — | — | — | — |
| Hormone treatments | — | — | — | — | — | — | 2 | 27 | 0.60 | 0.24 | 0.14–1.06 | 71% | — | — | — | — | — | — |
| Levothyroxine | — | — | — | — | — | — | 1 | 15 | 1.08 | 0.35 | 0.39–1.76 | — | — | — | — | — | — | — |
| Triiodothyronine | — | — | — | — | — | — | 1 | 12 | 0.20 | 0.32 | −0.42 – 0.83 | — | — | — | — | — | — | — |
| Non‐pharmacological | 2 | 16 | 0.44 | 0.26 | −0.08 to 0.95 | 0% | — | — | — | — | — | — | — | — | — | — | — | — |
| CPT | 1 | 7 | 0.24 | 0.38 | −0.51 to 0.99 | — | — | — | — | — | — | — | — | — | — | — | — | — |
| Bibliotherapy | 1 | 9 | 0.61 | 0.36 | −0.1 to 1.32 | — | — | — | — | — | — | — | — | — | — | — | — | — |
| Controls e | 7 | 347 | 0.67 | 0.06 | 0.55–0.79 | 0% a | 11 | 521 | 0.60 | 0.08 | 0.43–0.76 | 51% a , f | 4 | 126 | 0.33 | 0.09 | 0.15–0.51 | 0% d |
| Placebo | 10 | 496 | 0.60 | 0.09 | 0.42–0.78 | 55% a | ||||||||||||
| Treatment as usual | — | — | — | — | — | — | 1 | 25 | 0.55 | 0.21 | 0.13–0.97 | — | — | — | — | — | — | — |
| Studies removed in sensitivity analysis | Goldsmith, Tohen 03 (above), Bobo (RoB), Cutler (non‐RC) | Goldsmith, Cutler, Muzina 2008 (all mild symptoms). | Baldessarini (above) | |||||||||||||||
Abbreviations: CI, confidence interval; CPT, cognitive psychoeducational therapy; ES, effect size; k, number of studies; LAI, long‐acting injectable; N, number of participants; OFC, olanzapine fluoxetine combination; RC, rapid cycling; RoB, risk of bias.
Removal of study/ies in sensitivity analysis increased ES (and reduced heterogeneity).
The only studies assessed for this treatment were flagged as needing checking in sensitivity analyses. The effect on the category‐level analysis was examined and one of two studies (each) were removed for Valproate (depression) and olanzapine (global impression); in both cases, this was because one of the two studies' removal reduced class heterogeneity and one did not; the latter was kept in. For Valproate, removal of the one study increased the ES (due to patients being in partial remission) and for olanzapine removal reduced the ES (due to a potential effect of non‐RC patients being included in the analysis).
May be over‐estimation of effect (see Supplement 6) but no other studies in category to be compared with.
Removal of study/ies in sensitivity analysis decreased ES (and reduced heterogeneity).
Control arms (more studies), were treated slightly differently in sensitivity analyses; initially, all studies which had previously been flagged were removed; these were then re‐added one by one and were kept in where they did not increase I‐squared. If only the same studies (per outcome) had been removed as in the respective active treatment arms, the heterogeneity was clearly higher but effect sizes were equivalent, just with wider CIs. Specifically, the ES for global outcome was 0.64 (73% heterogeneity), for depression was 0.61 (64% heterogeneity) and the mania studies/results were identical to above.
If excluding two very large outliers (with no a priori reason for exclusion, the results for the control groups would have been k = 9, n = 370, ES = 0.51, 95% CI 0.38–0.64, i2 = 0%. In this case, the control 95% Cis would not have overlapped with lithium, antidepressant category (and within that, citalopram and venlafaxine) in addition to olanzapine.