Recent studies have demonstrated that the rate of switching from unipolar depression to bipolar disorder has been underestimated. Frankle et al showed that, over a period of nearly 3 years, out of 160 patients with a major depressive disorder episode, 33 patients receiving antidepressants and 17 patients not receiving antidepressants switched to bipolar disorder. ^{Reference Frankle, Perlis, Deckersbach, Grandin, Gray and Sachs1} The switch rate was 15.2% in patients who did and 17.6% in patients who did not take antidepressants. Holma et al performed a 5-year naturalistic study and found that 29/248 (11.7%) patients with previous unipolar depression switched to bipolar disorder. ^{Reference Holma, Melartin, Holma and Isometsä2} Of these 29 patients, 22 patients switched to bipolar disorder II and 7 patients switched to bipolar disorder I.

Switching from depression to bipolar disorder can involve natural transition or antidepressant-induced transition. Wada et al reported that 7/33 (21.2%) patients developed bipolar disorder during 1 year of antidepressant treatment. ^{Reference Wada, Sasaki, Jitsuiki and Takaishi3} Jin et al carried out a study of the rate of switching from depression to bipolar disorder in patients taking different classes of antidepressants. ^{Reference Jin, Ma, Qiu, Wang, Wang and Fan4} They demonstrated that the overall switch rate was 14.4% over 6 years. The switching rates for the different antidepressant types was as follows: 9.1% for selective serotonin reuptake inhibitors (SSRIs), 22.8% serotonin–noradrenaline reuptake inhibitors (SNRIs), 14.6% for noradrenergic and specific serotonergic antidepressants (NaSSAs), 27.2% for tricyclic antidepressants, and 36% for combination treatment with any two antidepressants. Thus, the rate of switching from unipolar depression to bipolar disorder has been demonstrated to be beyond even our expectation.

We agree with Sugihara & Tajika’s comments regarding the prediction of switch from unipolar depression to bipolar diosrder. It is in fact difficult to identify bipolar disorder when patients present with depression in their first episode. In our pilot study, ^{Reference Li, Zhang, Fan, Yuan, Huang and Chen5} neither plasma BDNF levels nor gene expression level of BDNF alone could differentiate major depressive disorder from bipolar disorder in the first depressive episode; the best model for predicting bipolar disorder in the first depressive episode was the combination of BDNF gene expression and plasma BDNF levels. It indicated that more relative biological biomarkers should be integrated to improve PPV. Furthermore, clinical characteristics are also important predictive factors for bipolar disorder. ^{Reference Holma, Melartin, Holma and Isometsä2,Reference Forty, Smith, Jones, Jones, Caesar and Cooper6} Predicting bipolar disorder becomes more accurate and reliable when we integrate more biological biomarkers and clinical characteristics of patients. Therefore, we are performing another study to construct a model to predict bipolar disorder in the first depressive episode that includes biological biomarkers and clinical characteristics.