Author's response to reviews Title:Continued cannabis use at one year follow up is associated with elevated mood and lower global functioning in bipolar I disorder Authors: Levi R Kvitland (l.r.kvitland@medisin.uio.no) Ingrid Melle (ingrid.melle@medisin.uio.no) Sofie R Aminoff (s.r.aminoff@medisin.uio.no) Christine Demmo (christine.demmo@medisn.uio.no) Trine V Lagerberg (t.v.lagerberg@medisin.uio.no) Ole A Andreassen (o.a.andreassen@medisin.uio.no) Petter A Ringen (p.a.ringen@medisin.uio.no) Version:5Date:17 December 2014 Author's response to reviews: see over
Dear Editor Enclosed you will find a revised version of the manuscript Continued cannabis use at one year follow up is associated with elevated mood and lower global functioning in bipolar I disorder: Kvitland L R et al. The reviewers comments were most helpful in improving the paper, and we hope that this version meets the questions raised by the reviewers. As stated earlier, we are not able to fully respond to one of the comments of reviewer 2, regarding the temporal sequencing of cannabis use and mood symptoms over the follow-up period. To have reliable data on this would require a more comprehensive follow-along with several measurement points or the use of ecological measurement. We have added this as a limitation, and hope our results will be of interest - and perhaps the argument and basis for more in-depth studies in the future. Beneath you will find our specific comments, in italics, to the reviewers reports including details of the changes made in the paper. The paper is not substantially lengthened. We hope these improvements are satisfactory, and that you find the paper of interest for the readers of BMC Psychiatry. On behalf of the authors, Sincerely Levi Kvitland
Reviewer 1. Continued cannabis use at one year follow up is associated with elevated mood and lower global functioning in bipolar I disorder The authors study a topic with great clinical relevance: cannabis use in bipolar I disorder. However, there are important limitations that the authors need to take into account in order to publish this study. Please see the comments below. 1) Major concern: My primary concern is the low number of subjects. Only 6 patients used cannabis at both time points (baseline and one year follow up) and there was a large difference in sample size between groups. I highly recommend to draw with more caution the results and especially the discussion section because there is a substantial chance that the results are induced by II error type. Answer 1: The reviewers concerns about the low number of subjects are valid. We have made changes to the Discussion section and to the conclusion to clearer state the limitations of a low number of participants 2) It would be helpful to add, in table 2, the number of patients in each group. Answer 2: We agree with the reviewer. N is now presented for each group in table 2. 3) It is not clear the reason to include gender in the correlation analyses (table 3), since gender is a dichotomous variable and correlation is only useful with continuous variables. Please, clarify this point in the methods and results section. Answer 3: We agree that gender should not have been part of the correlation analyses and have thus removed the variable from these analyses. 4) Table 3 is unnecessary and can be summarized within the text. Answer 4: We agree with the reviewer. We have removed table 3 and instead added a summarization in the Results section. 5) The results showed a positive correlation between YMRS and GAF. However, the authors did not discuss this result in the discussion section. This result was obtained due to the small sample size? Or is there another reasonable explanation? Answer 5: We agree with the reviewer that the correlation between YMRS and GAF-F at baseline should be discussed. We apologize for rather confusingly writing that it was positive: Since the direction of the two scales is opposite (high YMRS score indicates poor symptomatology while high GAF indicates good functioning) a statistically negative correlation indicates a relationship where good symptom scores
are associated with good functioning. We have now expressed this in the statistically correct way and have furthermore added a description in the results section as well as in the discussion. 6) The authors did not include the YMRS score at the baseline. Maybe both groups had these differences at the baseline and it is not associated with cannabis in this sample. I highly recommend taking this into account to better describe the sample. Answer 6: We agree with the reviewer that both YMRS (and GAF-F) at baseline could interfere with the results. We had explored this in our analyses, but since there were no significant differences we did not included it in the results. In the results section we have now included the lack of baseline differences on YMRS and GAF-F between the groups, and why we think this makes our assumptions stronger. 7) The authors include in the same group, patients that did not use cannabis, patients that used cannabis at the baseline and patients that started use cannabis at the follow-up period. It would be important to exclude patients that used cannabis at the baseline and those that started to use at the follow-up to analyze a more homogeneous sample. It is possible that the results are interfered by the presence of patients that used cannabis during the period of one year, in the without continued used cannabis group. Another option is to use these information (cannabis use at baseline or started use cannabis at the follow-up period) as a covariate. Answer 7: The reviewers concern about the possible interference of those with cannabis at baseline (but not at follow up) and those with cannabis at follow up (but not at baseline) is valid. As use of cannabis in the follow up period (for those without continued use) could possibly induce a mood elevating effect, one could hypothesize such an addition could strengthen our results, and this have now been taken into the text both in the methods and in the results section by adding follow-up analyses contrasting the continued users with strictly non-users Reviewer 2. The authors did not control by other potential confusing factors like alcohol or cocaine and drug treatment adherence. Answer 1: The reviewers concern about potential confusing factors like alcohol, cocaine or drug treatment adherence is valid. There were no associations with the outcome measures in terms of continued alcohol, cocaine or amphetamine use or drug treatment adherence; this is now clearly stated in the manuscript. In despite of the lack of an association, we also performed a follow-up analysis where we added alcohol in the second step in the regression analysis which actually slightly strengthen our results, but because of the initial lack of association this is not added to the tables.
They state that the inclusion of outpatients and inpatients makes the study stronger, but they did not adjust for these factors. Answer 2. We agree that this aspect should have been more clearly stated in the text. Hospitalization in bipolar disorder is only necessary for some, during episodes of mania or depression with marked functional impairment. Most patients are briefly hospitalized during the height of an episode and then treated as outpatients for the remainder of the time. Level of functioning and hospitalization status will also be highly correlated. The in-patient population at different time points (here baseline and follow-up) will thus be different. However, recruiting only from inpatient settings as many studies are doing, would limit the study s recruitment basis and we see a broad recruitment as one of the study s strengths. We however have no hypothesis that our study outcomes would be influenced by patients treatment status since this to some extent at a given moment is random. This is now more clearly accounted for in the Methods section. The interpretation of results is well explained and unbiased. The references provided are in line with the discussion. I think the authors should explain in greater depth why the results do not support the self-medication hypothesis. Do they have data to test the self-medication hypothesis (like cannabis use after t0 and the onset or continuous symptoms)? Answer 3. We agree that we do not have enough empirical data to argue against the selfmedication hypothesis and have toned this down in the manuscript. The method is well described. The main variables, like global functioning and psychopathology, are of interest. The authors did not explain one of the more important variables, the use of cannabis, well enough. How was it measured? What does continued cannabis use mean)? Is it the same for one subject who smokes once in 12 months as another who is a daily cannabis smoker? Answer 4. We agree that we should have explained this variable better. A clear definition and a better description is now added to the Methods section. The strengths are very good use of statistical methods, well designed with low dropout rate in the follow-up. Another strong point, as commented by the authors themselves, is the inclusion of outpatients and inpatients. The weak point is how they measure cannabis use. Was it self-report? This is an important limitation. Did the authors use urine screening? If not, memory bias should be taken into account as a limitation of the study. Answer 5. We agree that self-reported cannabis use should be stated as a limitation. We used urine screening at baseline and follow-up. Our data indicates that patients reports at these points of time corresponds well with the urine results. We do however not have follow-along data and this is now clearly pointed out in the limitation section.
One strong point about the results is that the difference between cannabis use and no cannabis use is clinically relevant by 17 points on a scale of 100. I think this is important in the clinical field. The authors should include effect size in Table 2 Answer 6. We agree that estimates of effect size is of interest, especially in the clinic it is important. We have now added Cochen s d to show the (considerable) effect sizes in table 2. MAJOR COMPULSORY REVISIONS: The following data are required to improve the paper: # Data about which start first during the follow up, cannabis use or mania symptoms? Answer 7. We agree with the reviewers concern about the lacking temporal sequencing of cannabis and mania in the follow up period. As all patients were identified in relation to their first treated mania and the continuous group was characterized by using cannabis at both time-points, from one point of view both phenomena had started out before the actual follow-up. It is however very probable that both fluctuate in intensity over time. Unfortunately, as we only have cross-sectional data from the two time points, we do not possess data that could document any sequencing. # Did the mania symptoms precede the onset of cannabis use? Answer 8. We agree that the question about what come first, mania or cannabis, is a valid concern, as stated above. YMRS is a scale showing mania symptoms last 48 hours while the continuous cannabis users had been using cannabis over a longer period of time. We could thus argue that it is reasonable to assume that the level of mania measured at follow up could be directly related to the continued cannabis use. However, the lack of temporal sequencing data make it hard to rule out that the patients who had been using cannabis continuously could have done this as a result of manic symptoms. We have now added an account of this to the limitations section. # Data on how many subjects did not recover during the 12 months of follow up? Answer 9 We agree that rate of recovery is of interest. Recovery is a concept that is difficult to define, but for severe mental illness a GAF-score over 61 has been used for cut-off (Stone MH. Exploratory psychotherapy in schizophrenia-spectrum patients. A reevaluation in the light of long-term follow-up of schizophrenic and borderline patients. Bull Menninger Clin 1986 May;50(3):287-306). We have now added information about recovery defined as GAF over 61 for both groups in the Methods and Results section. We have, however, data on symptomatic recoverey/remission at follow-up (as defined as no affective or psychotic symptoms last 6 months). An account of this is now added to the Methods and Results sections.
# Data about the pattern of cannabis use and the use of other legal (alcohol) or illegal drugs (cocaine)? Answer 10. We agree this should have been more clearly stated. As stated in answer 1 and 4 we have now added information on how continued use was defined, and we have commented on continued alcohol and continued stimulant use in the Methods section.