Bowden 1994.

Methods	Study design: randomised, double‐blind, parallel‐group study
Participants	Diagnosis: manic disorder Method of diagnosis: met Research Diagnostic Criteria based on the structured interview and rating scale of SADS Age: for lithium, median = 39.1 (SD = 11.2) years; for divalproex, median = 40.4 (SD = 12.8) years; for placebo, median = 39.0 (SD = 10.0) years; range = not specified Sex: lithium 28% women; 72% men, divalproex 48% women; 52% men, placebo 43% women; 57% men. Location: Audie L. Murphy Memorial Veterans Hospital, San Antonio, Texas and San Antonio State Hospital (44 participants); Harris County Psychiatric Center, Houston, Texas (49 participants); Larue D. Carter Memorial Hospital, Indianapolis (19 participants); Emory University Hospital, Atlanta, Georgia (6 participants); University Hospitals of Cleveland (Ohio) (22 participants); Veterans Affairs Medical Center, Dallas, Texas (13 participants); Jackson Memorial Hospital, Miami, Florida (7 participants); and Illinois State Psychiatric Center, Chicago (16 participants). Co‐morbidities: not described Adjunctive therapy: none Adjunctive medication: the protocol allowed the use of adjunctive chloral hydrate or lorazepam as needed for control of agitation, irritability, restlessness, insomnia, and hostile behaviours.
Interventions	Participants were randomly assigned to either: Experimental arm ‐ lithium N = 35 Duration: 3 weeks Treatment protocol: lithium carbonate was administered at an initial dose of 900 mg/d, dispensed in divided doses three times daily. On day 3, the total daily dosages of divalproex or lithium were increased to 1000 mg and 1200 mg respectively, and trough serum concentrations of both drugs were determined. On day 5, an unblended physician at each centre reviewed the serum concentration and adjusted the dosage of active medication. Thereafter, trough concentrations were measured on days 8, 10, 12, 15, and 18. Medication adjustments were made on days 7, 10, 12, 14, and 17. Drug dosage was raised on each adjustment day unless precluded by an adverse event or a serum concentration of valproate or lithium exceeding 1041 μ/L (150 μg/mL) or 1.5 mmol/L, respectively. Therapist/face‐to‐face contact: not described Comparator arm ‐ divalproex sodium N = 68 Duration: 3 weeks Treatment protocol: divalproex sodium was administered at an initial dose of 750 mg/day, dispensed in divided doses three times daily. On day 3, the total daily dosages of divalproex or lithium were increased to 1000 mg and 1200 mg respectively, and trough serum concentrations of both drugs were determined. On day 5, an unblended physician at each centre reviewed the serum concentration and adjusted the dosage of active medication. Thereafter, trough concentrations were measured on days 8, 10, 12, 15, and 18. Medication adjustments were made on days 7, 10, 12, 14, and 17. Drug dosage was raised on each adjustment day unless precluded by an adverse event or a serum concentration of valproate or lithium exceeding 1041 μ/L (150 μg/mL) or 1.5 mmol/L, respectively. Therapist/face‐to‐face contact: not described Comparator arm ‐ placebo N = 73 Duration: 3 weeks Treatment protocol: placebo, dispensed in divided doses three times daily. Comparable adjustments were made in the dosage of placebo according to blinded protocol‐specified dosing schedules. Therapist/face‐to‐face contact: not described
Outcomes	Timepoints for assessment: 5, 10, 15, and 21 days. Primary outcome: MRS Secondary outcome: Withdrawal At least 50% improvement, in the manic syndrome subscale score Mean change in MRS(subgroups ‐ previous lithium response) Adverse events/ Side effects
Notes	Date of study: not clear Funding source: this study was funded in part by a grant from Abbott Laboratories, North Chicago, 111. Declarations of interest among the primary researchers: not described
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"A separate randomization schedule for each center was generated prior to the study start. Randomized in blocks of 5"
Allocation concealment (selection bias)	Low risk	"The post‐study treatment was individually determined for each patient without breaking the study blind".
Blinding (performance bias and detection bias) All outcomes	Low risk	"On day 5, an unblinded physician at each center reviewed the serum concentration and adjusted the dosage of active medication. Whenever possible, the same blinded investigator rated the patient throughout the study".
Incomplete outcome data (attrition bias) All outcomes	High risk	Sixty‐four percent of participants in the placebo group and 61% in the lithium‐treated group failed to complete all 21 days of treatment compared with only 48% in the divalproex‐treated group
Selective reporting (reporting bias)	Unclear risk	Incomplete data provided
Other bias	Unclear risk	None identified