*** PLEASE NOTE: This is a two-step process. After clicking on the link, you will be directed to a webpage to confirm. ***

Transcription

1 23-Feb-2018 Dear Dr. Gafoor Manuscript ID BMJ entitled "Antidepressant utilisation and incidence of weight gain during 11 years follow-up. Population-based cohort study" Thank you for sending us your paper. We sent it for external peer review and discussed it at our manuscript committee meeting. We recognise its potential importance and relevance to general medical readers, but I am afraid that we have not yet been able to reach a final decision on it because several important aspects of the work still need clarifying. We hope very much that you will be willing and able to revise your paper as explained below in the report from the manuscript meeting, so that we will be in a better position to understand your study and decide whether the BMJ is the right journal for it. We are looking forward to reading the revised version and, we hope, reaching a decision. Please remember that the author list and order were finalised upon initial submission, and reviewers and editors judged the paper in light of this information, particularly regarding any competing interests. If authors are later added to a paper this process is subverted. In that case, we reserve the right to rescind any previous decision or return the paper to the review process. Please also remember that we reserve the right to require formation of an authorship group when there are a large number of authors. Best regards Kristina Fišter kfister@bmj.com *** PLEASE NOTE: This is a two-step process. After clicking on the link, you will be directed to a webpage to confirm. *** **Report from The BMJ s manuscript committee meeting** These comments are an attempt to summarise the discussions at the manuscript meeting. They are not an exact transcript. Members of the committee were: José Merino (Chair), Jamie Kirkham (Statistics advisor), John Fletcher, Elizabeth Loder, Tiago Villanueva, Wim Weber, Sophie Cook, Daoxin Yin, Amy Price, Kristina Fišter. Decision: Put points Detailed comments from the meeting: First, please revise your paper to respond to all of the comments by the reviewers. Their reports are available at the end of this letter, below. Please also respond to these additional comments by the committee: 1) What was the purpose of the random sample of 30,000 from each BMI/Gender boundary? Why not use the full cohort? What was the size of the full cohort.

2 2) Why was a 5% weight gain (BMI) chosen as being the important year on year difference? Could you also consider analysing other relevant outcomes as suggested by both our as well as the Annals reviewers? 3) The imputation methods seems unusual. If missing BMI values were imputed using LOCF for up to 3 years then the BMI is the same - this will clearly impact on the outcome of interest (5% BMI yearly weight gain). This is therefore potentially biased. The primary analysis therefore needs to consider complete cases perhaps? 4) There is some suggestion that rarely used ADs are excluded - surely this is not an issue for the main analysis as the exposure is based on AD prescription not type of AD. 5) Table 1. Where are we told about the missingness as described in the text. The baselines are tabulated by male/female - this would be more informative if it was tabulated by AD/Non-AD as in Table 2, with M/F included in the left column. 6) Pg 7 (line 50 onwards) - this seems a mixture of methods and results. The sensitivity aspectbest/worst case scenario needs to be described and reported better. 7) Figure 1 doesn't seem all that informative. 8) There are attempts at computing NNH - how these are computed needs to be in the methods. 9) Figure 5 - title correctly labelled? RR or IRR? 10) How should these findings be communicated to patients? Have you considered including an at a glance infographic? 11) It is difficult to disentangle the effect of anti-depressants from behaviours and changes in lifestyle that may come with depressive symptoms (i.e. being less physically active, eating more etc). Have you controlled for diet and physical activity? How does this impact on the interpretation of the results? 12) What are the implications of these findings? What does the paper add to what is already in the literature and guidelines? 13) The inclusion criteria - you say those included more than three measures of BMI, but the flowchart refers to all BMI values missing, more clarity is needed here. 14) What proportion of the general population fulfils these inclusion criteria? Could this be a select sample of people for whom doctors are worried about weight gain? How may this influence the interpretation of this paper? 15) We thought you should also adjust for depression in the analysis. 16) Could you also present population figures to show general population weight gain over 11 years. 17) Please move the declaration on patient involvement to the last section in the methods. We also thought it would be good to include a thank you to the patients for the use of the data in the acknowledgments. In your response please provide, point by point, your replies to the comments made by the reviewers and the editors, explaining how you have dealt with them in the paper. Comments from Reviewers Reviewer: 1

3 Recommendation: Comments: The authors present important data on the long-term weight effects of anti-depressant use (or rather prescription). The manuscript is generally well written and I only have a few remarks. With their valuable data at hand, I wonder if the manuscript could even provide more evidence when additional analyses are run. General remarks: The authors cautiously drew their sample based on obesity/weight categories, but do not differentiate their results regarding these. Becoming overweight/obese (e.g. shifting from one weight category into another), too, is a major public health concern. I therefore would like to see BMI specific analyses as patients with lower BMIs may be more prone to subtle (5%) weight changes than patients with (very) high BMI values already. Results 1. Page 10, ll. 12: How was the time frame for exclusion for women chosen? Based on what assumption? 2. Some info in the results section seems to be redundant to information given in the methods section. Please consider making an effort to clearly differentiate between the two. Lastly, the discussion may benefit from some information on the correlation of depressive symptoms and anti-depressant use. If I am not mistaken, the point prevalence of depression ought to be around 10% at any point in time, however, this includes all levels of severity. Guidelines of treatment do not seem to be fully adhered to, when such a higher level of prescriptions is found. Maybe the authors could comment on this issue as well. Additional Questions: Please enter your name: Claudia Luck-Sikorski Job Title: Professor for Psychological Health and Psychotherapie Institution: SRH University of Applied Health Sciences, Gera Reimbursement for attending a symposium?: No A fee for speaking?: No A fee for organising education?: No Funds for research?: No Funds for a member of staff?: No Fees for consulting?: No Have you in the past five years been employed by an organisation that may in any way gain or lose financially from the publication of this paper?: No Do you hold any stocks or shares in an organisation that may in any way gain or lose financially from the publication of this paper?: No If you have any competing interests <A HREF=' mpeting-interests'target='_new'> (please see BMJ policy) </a>please declare them here: Reviewer: 2

4 Recommendation: Comments: The present manuscript aims to investigate the long-term population impact of antidepressant prescribing on weight gain in a very large cohort study (Analyzed subjects ) using primary care electronic health records. Authors found that in participants who were prescribed antidepressants, 10.4% of person years for men and 12.1% for women, were associated with weight gain (adjusted IRR 1.25). Furthermore, they reported that the risk of weight gain remained elevated during a maximum of 11 years' follow-up. Differences on weight gain effects among different antidepressant drugs were reported as well. Thus, Authors concluded that the widespread utilization of antidepressant may be contributing to long-term increases in risk of weight gain at population-level. The issue is certainly of great interest since the long term risk of weight gain for antidepressant drugs is still under debate, as well as the specific risk for each antidepressant. Thus, the present manuscript adds interesting insight about the long term risk of weight gain, supporting the relevance of weight gain monitoring and management during antidepressant treatments, as already suggested by current guidelines. Nonetheless, there are a number of methodological issues that mitigate the potential impact of the present paper, raising some doubts about the associations reported. In particular, as partially stated by Authors in the "Strengths and Limitations" section, several recruitment biases could happen in primary care, leading to a misinterpretation of the results. However, apart from the possible bias identified by Authors, in my opinion some other relevant ones have not been taken into account sufficiently. For example, antidepressant drugs are currently prescribed for a number of disorders (e.g., anxiety disorders, personality disorders, eating disorders, negative symptoms in schizophrenia, depressive symptoms due to medical conditions such as following a heart attack or a stroke, etc.), thus a causal relationship between antidepressant prescription and depression cannot be ensured. Further, some differences in weight gain risk among the different antidepressants are already known and reported in the most used guidelines (e.g., in the Maudsely Prescribing Guideline). Thus, the prescribing doctor may avoid antidepressant with higher risk of weight gain in overweight/obese subjects or in patients affected by other co-morbidities such as diabetes or CHD, leading to a potential relevant recruitment bias. As a matter of fact, the observed gradual switch in antidepressant prescriptions from TCA to SSRI/SNRI indicated that primary care physicians carefully selected the antidepressant drugs prescribed accordingly to current guidelines. Other potential biases were underlined below. More in detail, the manuscript should be revised according to the following suggestions. Introduction A brief introduction about the relationship between depression and overweight/obesity should be added (e.g., (Jung et al. 2017; Quek et al. 2017; Rajan and Menon 2017; Schachter et al. 2017)). Similarly, a brief introduction about the possible mechanisms linking antidepressant treatment with weight gain should be added (e.g., (Himmerich et al. 2015)). A little more information from the meta-analysis by Serretti & Mandelli should be added (i.e., Authors should report the antidepressant identified as at lowest and highest risks of weight gain, at least). In my opinion, a retrospective cohort study as the present one is not suitable to test a causal hypothesis (i.e., "that antidepressant prescribing is associated with increased risk of weight gain over time"). Consistently, Authors recognized in the discussion that a causal link between antidepressant prescribing and weight gain cannot be demonstrated. Thus, I suggest to delete the last sentence from the introduction. Methods Participants and cohort selection Authors stated that "The sample was drawn from the population of patients registered with CPRD between November 1, 2004 and October inclusive [...]" but in the title and abstract the reported a 11 years follow-up. Please, clarify this discrepancy. Authors stated that they "selected a random sample of up to a maximum of participants [...] from each category of BMI and gender [...]". However, they did not list in the BMI categories considered the underweight one (i.e., BMI ), which is however reported in table 1 (although the subjects included in this BMI category are very few compared to the other categories). Furthermore, in the BMI category from 25-0 to 29.9, participants are listed in the male subsample (similarly to the BMI category ). However, authors stated that the maximum of participants for category was after random selection. Please, clarify these discrepancies.

5 Moreover, taking into account that the extreme BMI categories included very few subjects compared to the other ones, the comparability among the different categories cannot be granted, at least for the BMI category Thus, Authors should perform an exploratory analysis excluding these categories to confirm current findings. Main measures Since current antidepressant prescriptions are usually for 6 months or more, I do not understand why Authors assumed "a 90 days in duration from the date entered in the CPRD database". Please, clarify and better justify this arbitrary choice. Authors stated that "The mean BMI value for each participant year was used". If I correctly understand, this means that if a participant had more than one BMI value recorded in one year, the mean was used. However, more than one BMI record by year likely suggests an attention by the General Practitioner and/or by the subject himself to weight gain. Thus, appropriate measures to control weight gain in these subjects cannot be excluded (e.g., diet advices to avoid weight gain under antidepressant treatments, concomitant medical problems underlying current weight gain such as hypothyroidism, etc.). Thus, I suggest to perform an exploratory analysis excluding subjects with more than one BMI record by year to control for this possible bias. The imputation with the method of last observation carried forward for up to three years is another font of possible bias. Thus, in my opinion, an exploratory analysis without this imputation should be performed to confirm current findings. Given the known risk of weight gain after smoking cessation (e.g., (Tian et al. 2015)), smoking cessation should be considered separately as possible confounding factor. Furthermore, taking into account the poor effectiveness of interventions for smoking cessation (e.g. (Cabezas et al. 2011)), Authors should better specify how they considered subjects under smoking cessation interventions about the smoking status. Please, clarify. However, every attempt to decrease or to stop smoking is likely related to an increased risk of weight gain. For this reason, in my opinion, subjects trying to stop smoking should be considered separately from current smokers in the analysis. Statistical analysis "The extent of missingness in the variables BMI and smoking status" is not easily understandable from Table 1. Please, provide the missing values in table 1 in a clearer way. I did not understand the "extreme scenarios" described in the text. If I correctly understand, in the first scenario Authors set missing values in the following way: all BMI as "Normal weight", deprivation scores to "most deprived" and smoking status to "Current smoker", while in the second one: all BMI as "most overweight", missing deprivation scores to "Least deprived" and smoking status as "non smoker". However, these seem to not be the extreme scenario to me. Indeed, the worst scenario should be "Most overweight", "Most deprived" and "Current smoker", while the best one should be "normal weight", "Least deprived", and "non smoker". Please, better clarify and justify your choice. Results Authors should state whether at the baseline, patients who had been prescribed an antidepressant differ about weight from those who had not been treated with these drugs. This could help to verify the association between depression and overweight/obesity in this sample before the antidepressant prescription. Authors did not report here or elsewhere if any measure of compliance was considered or even recorded. Taking into account that poor compliance through antidepressant treatment has been reported to be as higher as 40% (e.g., (Demyttenaere and Haddad 2000)), this potential bias should be discussed. Taking into account the great impact of antipsychotic and antiepileptic drugs on weight gain, subjects with these co-prescriptions should be excluded from the analysis, at least in an exploratory one. The finding that "IRRS for weight gain tended to be slightly lower for participants with diabetes mellitus, coronary heart disease or stroke compared with those without comorbidity" suggests a prescription bias by General Practitioners, as previously hypothesized. At least, some interventions to control weight gain are likely applied to these subjects. This is another bias that should be better discussed. Also the findings about TCA antidepressants and paroxetine (which are usually associated with higher weight gain risk) suggested a prescription bias. Furthermore, amitriptyline is used, at very low dosages, also in the migraine prevention, partially explaining the high frequency found by authors for an old antidepressant, which is not yet included among the first line antidepressants. However, low dosages of amitriptyline (i.e mg/day) likely do not induce significant weight gain, representing another potential bias. Discussion

6 Discussion should be revised taking into account the comments above. In particular, several potential biases described above should be considered and discussed. The limitations underlined should be stated and addressed as well. References Cabezas C, Advani M, Puente D, Rodriguez-Blanco T, Martin C (2011) Effectiveness of a stepped primary care smoking cessation intervention: cluster randomized clinical trial (ISTAPS study) Addiction 106: doi: /j x Demyttenaere K, Haddad P (2000) Compliance with antidepressant therapy and antidepressant discontinuation symptoms Acta psychiatrica Scandinavica Supplementum 403:50-56 Himmerich H, Minkwitz J, Kirkby KC (2015) Weight Gain and Metabolic Changes During Treatment with Antipsychotics and Antidepressants Endocrine, metabolic & immune disorders drug targets 15: Jung SJ et al. (2017) Association between body size, weight change and depression: systematic review and meta-analysis The British journal of psychiatry : the journal of mental science 211:14-21 doi: /bjp.bp Quek YH, Tam WWS, Zhang MWB, Ho RCM (2017) Exploring the association between childhood and adolescent obesity and depression: a meta-analysis Obesity reviews : an official journal of the International Association for the Study of Obesity 18: doi: /obr Rajan TM, Menon V (2017) Psychiatric disorders and obesity: A review of association studies Journal of postgraduate medicine 63: doi: /jpgm.jpgm_712_16 Schachter J et al. (2017) Effects of obesity on depression: A role for inflammation and the gut microbiota Brain, behavior, and immunity doi: /j.bbi Tian J, Venn A, Otahal P, Gall S (2015) The association between quitting smoking and weight gain: a systematic review and meta-analysis of prospective cohort studies Obesity reviews : an official journal of the International Association for the Study of Obesity 16: doi: /obr Additional Questions: Please enter your name: Alessandro Serretti Job Title: professor of psychiatry Institution: bologna university Reimbursement for attending a symposium?: No A fee for speaking?: No A fee for organising education?: No Funds for research?: No Funds for a member of staff?: No Fees for consulting?: No Have you in the past five years been employed by an organisation that may in any way gain or lose financially from the publication of this paper?: No Do you hold any stocks or shares in an organisation that may in any way gain or lose financially from the publication of this paper?: No If you have any competing interests <A HREF=' mpeting-interests'target='_new'> (please see BMJ policy) </a>please declare them here: none

7 Reviewer: 3 Recommendation: Comments: The submitted manuscript reports findings from a cohort study on the association between specific antidepressant scripts and risk of weight gain (5% or more) per calendar year in the UK based on general practice medical records. The authors found that there was an increased risk (25%) of weight gain associated with antidepressants and concluded that their widespread use may contributing to long term population level weight gain. This is paper has potential for contributing new information on this issue. I present several suggestions for improvement. General: Respectfully, the paper is not well-written. The use of inappropriate references is ubiquitous. For example, references in the first paragraph: (2) reports trends in BP to support trends in BMI; (3) is a secondary source on overweight and obesity % in the US; (5) is a narrative review to support mortality risk; and (6) is a paper based on the same cohort to support a global statement on Concerted efforts..limited success. The clinical implications of the study findings are difficult to interpret. Investigations of dose-response associations between antidepressants and risk of weight gain may help in this regard Abstract: The results for number needed to harm seem inconsistent with the IRR in figure 4. Methods: The method for computing number needed to treat/harm (values in the abstract) has not been described. The scenario to determine risk of bias from missing values is a result. Roles of funding should not be in the text. Results: Risk of bias from missing and excluded data has not been properly assessed (i.e. IRR in each group). Please consider examining dose-response effect of the number of scripts and weight gain. Please clarify if the results in figure 3 & 4 were adjusted for covariates. Discussion: A discussion on the results in the context of previous studies is lacking. A recently published paper was missed: SSRI antidepressant use potentiates weight gain in the context of unhealthy lifestyles: results from a 4-year Australian follow-up study Z Shi, E Atlantis, AW Taylor, TK Gill, K Price, S Appleton, ML Wong,... BMJ open 7 (8), e The results in figure 3 suggest the increased risk of weight gain decreasing after 2-3 of antidepressant use. Perhaps the substantially lower risk of weight gain from specific antidepressants (paroxetine, TCAs) including clinical implications are worth discussing. I hope that my comments are received in a constructive spirit, as intended. Additional Questions:

8 Please enter your name: Evan Atlantis Job Title: Senior Research Fellow Institution: Western Sydney University Reimbursement for attending a symposium?: No A fee for speaking?: Yes A fee for organising education?: No Funds for research?: No Funds for a member of staff?: No Fees for consulting?: No Have you in the past five years been employed by an organisation that may in any way gain or lose financially from the publication of this paper?: No Do you hold any stocks or shares in an organisation that may in any way gain or lose financially from the publication of this paper?: No If you have any competing interests <A HREF=' mpeting-interests'target='_new'> (please see BMJ policy) </a>please declare them here: I received honoraria from Lilly Diabetes Australia in Reviewer: 4 Recommendation: Comments: Suggestion every table and figure should be interpreted without having to look at the article, e.g. Legend for Figure 1: Flowchart of Subject CPRD Clinical Practice Research Datalink (abbreviation should be fully explained) Table 1: Baseline characteristics of participants with antidepressant prescribing. Figures are frequencies (row percent) except where indicated. Suggestion Table 1: Baseline characteristics of participants with antidepressant prescribing (AD). Figures are frequencies (row percent) except where indicated. Legend for Figure 2: Forest plot showing incidence rate ratios (95% confidence intervals) for the association of antidepressant prescribing with 5% body weight gain for the whole cohort and for relevant subgroups Suggestion Legend for Figure 2: Forest plot showing incidence rate ratios (IRR) (95% confidence intervals) for the association of antidepressant prescribing with 5% body weight gain for the whole cohort and for relevant subgroups Legend for Figure 3: IRR of Relative risk of weight gain by duration of antidepressant therapy Suggestion

9 Legend for Figure 3: Incidence rate ratios of Relative risk of weight gain by duration of antidepressant therapy Etc. Additional Questions: Please enter your name: Zlatko Ulovec Job Title: Assistant Professor at the Chair of Social Medicine and Epidemiology Institution: School of Dental Medicine University of Zagreb Reimbursement for attending a symposium?: No A fee for speaking?: No A fee for organising education?: No Funds for research?: No Funds for a member of staff?: No Fees for consulting?: No Have you in the past five years been employed by an organisation that may in any way gain or lose financially from the publication of this paper?: No Do you hold any stocks or shares in an organisation that may in any way gain or lose financially from the publication of this paper?: No If you have any competing interests <A HREF=' mpeting-interests'target='_new'> (please see BMJ policy) </a>please declare them here: Reviewer: 5 Recommendation: Comments: Good study...shows broad spectrum of consideration needed to be given to this issue at GP and clinical level evident by conclusion/ breadth of research. Very topical and patient and public awareness needed. Additional Questions: Please enter your name: Kai Branch Job Title: Patient Institution: - Reimbursement for attending a symposium?: No A fee for speaking?: No A fee for organising education?: No Funds for research?: No

10 Funds for a member of staff?: No Fees for consulting?: No Have you in the past five years been employed by an organisation that may in any way gain or lose financially from the publication of this paper?: No Do you hold any stocks or shares in an organisation that may in any way gain or lose financially from the publication of this paper?: No If you have any competing interests <A HREF=' mpeting-interests'target='_new'> (please see BMJ policy) </a>please declare them here: No **Information for submitting a revision**