In our recent paper on adolescent cannabis use and risk of psychosis we found out that heavy cannabis use was associated with increased risk of any psychotic disorder or condition during 15 years of follow-up. Colleagues Savya Saluja and Hitesh Khurana made some important comments on this paper and we will try our best to clarify the concerns they raise.

The major argument in their letter was that affective psychoses should not be included with other psychoses, and they raise concern that this would not provide accurate results in general. In our study the objective was to study the associations between cannabis use and any psychotic condition and not to focus only on non-affective psychoses. We disagree that F30.2 (mania with psychotic features), F31.2 (bipolar disorder with psychotic features) and F32.3 (major depressive episode with psychotic features) should be considered solely as mood disorders; indeed ICD-10 also categorises them as severe psychotic states.¹ The risk factors for psychotic disorders are likely to be pluripotent and further increase the risk across all psychotic disorders.^{Reference Suetani, Siskind, Scott and McGrath2} Furthermore, the broad definition of psychosis has also been used in other studies as a primary or secondary outcome.^{Reference Bechtold, Hipwell, Lewis, Loeber and Pardini3–Reference Di Forti, Sallis, Allegri, Trotta, Ferraro and Stilo5} It should also be noted that many of the previous studies on adolescent cannabis use and psychosis outcomes have focused on psychotic symptoms instead of clinical diagnoses, which is the major strength in our article.

We reported in the paper also that when the psychosis diagnoses were analysed separately, both schizophrenia spectrum disorder (HR = 11.18, 95% CI 3.16–39.62) and psychotic depression (HR = 9.74, 95% CI 3.83–24.73) are associated with cannabis use, whereas similar associations were not found for schizophrenia, bipolar disorder with psychotic episodes and other psychosis. We agree that the number of individuals with psychosis was small in the subgroup analyses, as we noted in the limitations of the study. Whether substance use contributes to an increased risk across different psychosis categories differentially is a matter requiring replication, and we think that in future it would be worthwhile studying affective psychoses under their own category if sample size permits.

There were also some minor comments. In the Method section they pointed out correctly that ‘dots’ in ICD-10 psychosis diagnosis codes are missing, which is unfortunately true, and should be there, for example F30.2 instead of F302. They also point out some discrepancies between the figures and the main text and raise concern that the figures are not self-explanatory. The key point in these figures is that participants with both prodromal symptoms (cut-off ≥3 items in PROD-screen) and cannabis use are more likely to develop psychosis during the 15-year follow-up than participants with just prodromal symptoms or cannabis use alone, suggesting that cannabis use might be more harmful for those with preceding psychotic experiences. We hope that this answers the questions.