Randomized clinical trial (RCT) is the best study design for treatment-related issues, yet these studies may present a number of biases and limitations. The objective of this study is to carry out a qualitative analysis of RCT methodology in the treatment of bipolar depression (BD). A systematic review covering the last 20 years was performed on PubMed selecting double-blind RCTs for BD. The identification items of the articles, their design, methodology, outcome and grant-related issues were all analyzed. Thirty articles were included, all of which had been published in journals with an impact factor >3. While almost half studies (46.7%) used less than 50 patients as a sample, 70% did not describe or did not perform sample size calculation. The Last Observation Carried Forward (LOCF) method was used in 2/3 of the articles and 53.4% of the studies had high sample losses (>20%). Almost half the items were sponsored by the pharmaceutical industry and 33.3% were sponsored by institutions or research foundations. Articles on the pharmacological treatment of BD have several limitations which hinder the extrapolation of the data to clinical practice. Methodological errors and biases are common and statistical simplifications compromise the consistency of the findings.