We thank Kuppili, Singhai and Nebhinani for their recent comments on our article ‘Mood stabilisers and risk of stroke with bipolar disorder’.^{Reference Chen, Tsai, Pan, Chang, Su and Chen1} Their comments drew attention to several confounding factors that could have influenced our study findings.

The association between anxiety disorders and risk of stroke has recently received increased attention because of its high prevalence in bipolar disorder.^{Reference Pavlova, Perlis, Mantere, Sellgren, Isometsä and Mitchell2} In addition, this association might be observed between seizure disorders and risk of stroke because evidence suggests a link between bipolar disorder and epilepsy.^{Reference Wotton and Goldacre3} With these considerations, we examined the association between these two types of comorbidities and risk of stroke in our patients with bipolar disorder; however, results were not significant. The crude risk ratios of anxiety disorder and seizure disorders for the risk of stroke were 1.21 (95% CI 0.74–1.96, P = 0.446) and 2.18 (95% CI 0.35–13.49, P = 0.403), respectively, based on the case–crossover study.^{Reference Chen, Tsai, Pan, Chang, Su and Chen1} These findings suggested that the association between acute exposure to mood stabilisers and risk of stroke in patients with bipolar disorder may not be confounded by anxiety and seizure comorbidities.

We agree with the comment that the role of psychosocial factors in stroke should be addressed in patients with bipolar disorder. Information on these variables was unavailable in the National Health Insurance Research Database (NHIRD) of Taiwan; this is one of the limitations of this study. However, the design of a case–crossover study has the advantage that study participants serve as their own controls and therefore this may minimise the effects of such unmeasured variables.

We would also like to address the limitation issues indicated by Kuppili and colleagues. In Taiwan, the prescription of oral contraceptive pills is not covered by national health insurance. Therefore, the effect of oral contraceptives on the association between valproic acid use and risk of stroke cannot be excluded. In addition, medication adherence is not available in the NHIRD. Therefore, these points should be considered as limitations of our study.

Finally, we defined the use of carbamazepine, valproic acid, lithium or lamotrigine as the use of any mood stabiliser. Guidelines have suggested that combination therapy is an acceptable strategies for treating bipolar disorder.^{Reference Yatham, Kennedy, Parikh, Schaffer, Beaulieu and Alda4} Similar to the results of our prior study,^{Reference Yang, Liao, Liu, Chen, Chen and Kuo5} we believe that combination therapy for bipolar disorder may have contributed to the gap between the number of patients receiving any mood stabiliser and the sum of patients as per the numbers given separately for carbamazepine, valproic acid, lithium and lamotrigine in our study.