Author's response to reviews Title: Associations between depression and different measures of obesity (BMI, WC, WHtR, WHR) Authors: Jörg Wiltink (joerg.wiltink@unimedizin-mainz.de) Matthias Michal (matthias.michal@unimedizin-mainz.de) Philipp S Wild (philipp.wild@unimedizin-mainz.de) Isabella Zwiener (isabella.zwiener@unimedizin-mainz.de) Maria Blettner (blettner@imbei.uni-mainz.de) Thomas Münzel (tmuenzel@uni-mainz.de) Andreas Schulz (andreas.schulz@unimedizin-mainz.de) Yvonne Kirschner (yvonne.kirschner@unimedizin-mainz.de) Manfred E Beutel (manfred.beutel@unimedizin-mainz.de) Version: 2 Date: 6 May 2013 Author's response to reviews: see over
Klinik und Poliklinik für Psychosomatische Medizin und Psychotherapie Editor of BMC Psychiatry Direktor: Univ.-Prof. Dr. med. Dipl.-Psych. Manfred E. Beutel PD Dr. med. Dipl.-Psych Jörg Wiltink Geb. 207, 1. OG, Zi. 1.206 Langenbeckstr. 1 55131 Mainz Telefon: +49 (0) 6131 17-7289; Telefax: +49 (0) 6131 17-477289 E-Mail: joerg.wiltink@unimedizin-mainz.de http://www.klinik.uni-mainz.de/psychosomatik Mainz, 27.04.2013 Manuscript submission: Wiltink et al. Associations between depression and different measures of obesity (BMI, WC, WHtR, WHR) Dear Editor, we appreciate the opportunity to resubmit our article Associations between subtypes of depression and different measures of obesity (BMI, WC, WHtR, WHR) for review and publication to BMC Psychiatry. The reviewers comments were very helpful for us. We considered them carefully and changed our manuscript accordingly. Our changes are typed in colour with the former text at the margin of the page. On behalf of the co-authors. With best regards, Jörg Wiltink UNIVERSITÄTSMEDIZIN der Johannes Gutenberg-Universität Mainz. Körperschaft des öffentlichen Rechts Vorstand: Univ.-Prof. Dr. med. Norbert Pfeiffer (Vorsitzender und Medizinischer Vorstand), Univ.-Prof. Dr. med. Ulrich Förstermann (Wissenschaftlicher Vorstand), Evelyn Möhlenkamp (Pflegevorstand), Götz Scholz (Kaufmännischer Vorstand), Vorsitzende des Aufsichtsrates: Doris Ahnen Langenbeckstr. 1. 55131 Mainz. Telefon +49 (0) 6131 17-0. www.unimedizin-mainz.de. Bankverbindung: Sparkasse Mainz BLZ 550 501 20 Konto-Nr. 75
Seite 2/6 Editors Comments: 1. Copyediting: After reading through your manuscript, we feel that the quality of written English needs to be improved before the manuscript can be considered further. We advise you to seek the assistance of a fluent English speaking colleague, or to have a professional editing service correct your language. Please ensure that particular attention is paid to the abstract. We carefully revised the manuscript with the help of a fluent English speaking colleague. Reviewer: Giovanni Viscogliosi Major Compulsory Revisions None Minor essential revisions The proportion of elderly subjects as well as the age at onset of depression have not been taken into account. In my opinion it could represent a main limitation since cardio-vascular risk factors are mostly associated with late-onset depression. It would be interesting to assess whether the age at onset would have affected the associations found. Unfortunately we did not assess age at onset of depressive symptoms. This would be an interesting issue for future work. The use of cardio-metabolic medications such as anti-diabetes, anti-platelets, statins... should be taken into account in the multilogistic regression analyses. In order not to overcorrect our models, we decided to use only psychiatric medication for correction and leave cardio-metabolic medications out from our analyses. The authors considered 140/90 mmhg as the cut off of hypertension. This value may underestimate the proportion of hypertensives, especially because many young subjects were included. In my opinion, a cut off value of 130/85 (NCEP-ATP III, 2001 criteria) would be more appropriate. We decided to use the cut-off of 140/90 mmhg for caseness of at least mild hypertension following the WHO criteria which are identical with The Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure. We agree that a lower cut-off would be more appropriate especially when high normal/ prehypertension (cut-off 130/85 mmhg) is in focus e.g. to determine MetS. Our systolic cut-off has recently been advocated by the American Diabetes Association for patients with diabetes (Diabetes Care 2013; 36: Supplement) Why did the authors put "never smokers" and "former smokers" in the same category? Current smoking is one of the classical risk factors for CVD. In order to adjust for this risk factor we decided to control for the binary variable current smoking rather than differentiating smoking status. For a more in depth analysis regarding smoking in the study sample see Michal et al. (2013) Association of mental distress with smoking status in the community: Results from the Gutenberg Health Study. Journal of Affective Disorders 146:355-360.. The authors should reflect on the evidence that diabetes leads to sarcopenia, especially in the elderly. This makes BMI less specific in this kind of subjects. This is is a good issue for further research. We added this point to the discussion section (further research) d) focus on the mechanisms relating depression and abdominal obesity (e.g. genetics, pro-inflammatory cytokines) and analysis of moderators (characteristic of specific disorders, e.g. sarcopenia in the elderly with diabetes) potentially influencing the relation between measures of obesity and depressive symptoms in a longitudinal approach
Seite 3/6 Even if the authors included age and diabetes in the multilogistic models, it is impossible to understand the individual impacts of age and diabetes on the associations. We totally agree that the individual impact of single factors can not be the focus of our research in a representative sample. A longitudinal approach with a larger sample would have enough power to answer questions regarding smaller subgroups (elderly with diabetes and e.g. sarcopenia). Discretary revisions None Reviewer: Radboud Marijnissen This paper examines the association between BMI and different measures for abdominal obesity to somatic vs cognitive depression and different mental conditions in a cross-sectional population based study. MAJOR COMPULSORY REVISIONS 1* The question is not really original as it was examined already before. However previous research concerned only participants aged 50 to 70 years, so the study of Wiltink is original in the fact that also younger people are included. They should pay attention about this topic. We added this point (inclusion of younger participants) to the discussion section as a strength of our study. The main question is why they don t use longitudinal data. We agree with the reviewer that prospective data would be desirable in this field. Longitudinal data from our study are not yet available. 2*The aim of the study in the abstract is too vague: what is meant by different mental conditions? More important: the aim of the study in the background section (page 4) is not well defined: Wiltink et al compare BMI and different measures of abdominal obesity in what context? According the abstract they compare the associations between BMI and different measures of visceral obesity with somatic vs cognitive depressive symptoms? In the results and the conclusion there is nothing about the different mental conditions. The authors must have a clear aim of the study We have focused the relation between BMI and different measures of abdominal obesity and somatic vs. cognitive depressive symptoms. We omitted other mental conditions from the text. 3* In the hypothesis: Page 5 Following the vascular depression hypothesis. Etc.. Why do they refer to the vascular depression hypothesis? I don t understand why they expect a closer relation according the vascular depression hypothesis. Again mental conditions is too vague. All of a sudden (page 5) also anxiety and Type D are mentioned as included, but in the introduction there was nothing told about any relationship between obesity and personality or anxiety from literature. There are no references about the vascular depression hypothesis of Alexopoulos. It is better to leave the vascular depression hypothesis. We agree to the reviewer and focused the theoretical background on depression and avoided vague formulations like mental conditions. As suggested by the reviewer we decided to leave out the theoretical background regarding the vascular depression hypothesis and changed the text accordingly. 4** The conclusion in the abstract should be based on the main findings of the paper. The important finding of somatic vs cognitive depressive symptoms is missing here. We changed the abstract accordingly.
Seite 4/6 5* The data are chaotic and proper interpretation is very difficult. They should have used the variables as continuous variables. Its unclear what the mean depression scores are for the whole sample, probably very low, as it is a population based study. In discussion there is no attention for that. So it is not possible to interpret the data well. The conclusions are not all supported by the data. So conclude only that what is supported by the data. We rearranged table 1 regarding the following aspects and think it is much clearer now: - Results are shown for the whole sample (not separately for male and female participants) - Quartiles (Q1 and Q3) for the cognitive-affective and somatic-affective symptoms were added, which were used as continuous measures. - We inserted the median (Q1 and Q3) for PHQ (total score). 6* The methods are also chaotic and not well described. The methods should be reduced to only the relevant matters (see below). We rearranged the methods and we deleted the descriptions of panic, social anxiety, GAD and type-d that are no longer part of the analysis. MAJOR COMPULSORY REVISIONS Specific comments 1* Throughout the paper its is disturbing what is the focus of the study: different clusters of depression, other mental conditions etcetera. We clarified the focus throughout the paper. 2* The authors mention: subtypes of depression or somatic depression. It is better to use somatic affective depressive symptoms and cognitive affective depressive symptoms as its not quite clear if there is a subtype of somatic depression. We agree with the reviewer and changed wording throughout the manuscript. BACKGROUND 3* The background should focus more on depressive symptoms and obesity. Suddenly (sentence 19) they mention that visceral adipose tissue seems to play. Depression.. please add references. There is no reference about the vascular depression hypothesis.. In the background the authors should introduce this hypothesis, also about the other potential mechanisms linking obesity and depression (page 3/ 4) The authors should explain what they mean with anthropometric measures, this is not clear for the readers. Throughout the paper, some Germans words are included like und in the sentence 20 : A dutch study with 1284 participants aged between 50 und 70 years. We modified the background section focusing only on depressive symptoms. Relevant references were added. As suggested above we left the term vascular depression hypothesis from the section. Anthropometric measures were defined and wording (German) was corrected throughout the paper. METHODS Sample 4* Persons with physical or mental inability were excluded: it is important to know what kind of physical or mental inability? We specified disabilities in the methods section. Persons with insufficient knowledge of German language, or those who reported that they were not able to visit the study center on their own (due to their physical and/or mental condition) were excluded. 5*The methods should be reduced to make is less chaotic. The reader only wants to know: what is the sample, what are the primary outcome measures (and how were they measured?) and what are the potential confounders. After that describe the statistical methods.
Seite 5/6 Outcome measures were sorted (Primary outcome, confounders) as suggested. The statistical section was placed as suggested. 6*Primary outcome measures: *depressive symptoms: Focus on PHQ-9. PHQ measures depressive symptoms and a score of 10 or above is seen as indicative for diagnosis of major depression. The factoranalyses in previous studies was done in a sample with coronary heart disease. It would strengten the paper to conduct a factoranalysis of the PHQ-9 in this large population bases sample. For more clear paper its better to quit personality and other mental conditions. Its totally unclear why physical activity, the likert scale is introduced and why (page 7) the quartiles of physical activity is introduced. The problem is that the covariates are not described well. We added a description to this point to the method section (PHQ): Despite being aware, that dimensions of depression (cognitive-affective and somatic-affective) in the community might differ from those in cardiovascular settings, we used the same dimensions for comparison purposes and due to their high face validity and comparability. The description of physical activity For analysis the activity score was presented in quartiles with Q1 denominating the lowest quartile of physical activity and Q4 the highest. is from an earlier version of the manuscript. We omitted this sentence and apologize for this mistake. * Obesity: page 7 well described. 7* The potential confounders are missing and should be described well on basis of literature (age, sex, lifestyle (smoking/alcohol/physical acitivity, psychotropic drugs know to affect body weight), somatic comorbidity). We reorganized this section as described above. 8* Why do they exclude participants with underweight? (page 8) I suppose it is because it is known that there is a U shape relationship between obesity and depressive symptoms and because the authors want to examine the linear relationship between depressive symptoms and obesity? Please explain with adequate references! We added a relevant reference for the reason of the exclusion of underweight participants in the method section (subjects). 9* Statistical analysis It is should be clear why they use quartiles and not continous variables. Its better to mention here previous factor analyses and preferably their own factoranalysis because this is another sample than cardiac. As stated above quartiles (Q4 vs Q1+Q2) were used for comparisons to detect meaningful differences. We added the following description to the method section: While we were aware, that dimensions of depression (cognitive-affective and somatic-affective) in the community might differ from those in cardiovascular settings, we used the same dimensions for comparision purposes and due to their high face validity. 10*RESULTS The results section should start with the sample characteristics. They exclude only 30 people. Were the no missing items or was there no violating the rules for reliable measurement? What is the final study sample? 5000-30=4970? Table 1 is about 4970 participants? We now start the results section with sample characteristics and changed the sample characteristics regarding the total N of 4970 (5000 without underweight participants). 11* Table 1 should be in the results section. There are too much results in the table. Why do the authors give the results for female and male? The sample characteristics should be about the whole sample and only show those results which are necessary to show. What is the meanscore on PHQ-9 for whole
Seite 6/6 sample. You need this to interpret the results. We followed the suggestions and simplified Table 1. See 5*. The mean score of PHQ-9 is now given in Table 1. 12* Figure 1: not necessary for the main aim of the study (if this is in the abstract)\ We omitted figure 1 as suggested. 13* Then about the final results: the depressive scores and anthropopmetrics should have been used as continuous varibables. I dont understand why only age in model 1. It is not clear in text or table-legend why they adjust for all those covariates like depersonalization, type D etcetera. The legend is not clear.. what is psychotropic medication N06AA? Etcetera. Explain FH etcetera. Depression scores were continuous variables. We decided to use quartiles (Q4 vs Q1+Q2) for anthropometrics to detect meaningful differences. As most papers do, we also calculated a model without further corrections (except for age) to identify crude associations between anthropometrics and depression. We rearranged both tables, and we hope that covariates have become much clearer now. In order not to over correct our models mental conditions like depersonalization were removed from the analyses. DISCUSSION AND CONCLUSION 14* This section should start with the main conclusions according to the aim of the study. On page 11 it is explained that the analyses were fully adjusteds for age cardiovascular risk factors.. etcetera.. physical activity is missing. The authors start with repeating the results section and a kind of explaining table 2. We rearranged this section as suggested by the reviewer. 15* On page 11 they conclude that the finding of association between somatic depression and WC in men is consistent with report demonstrating an increased risk of depression in initially non depressed men with high visceral fat. This is not demonstrated in the papers used as references. We are grateful, that the reviewer found this error in citations. Our statement of course is only true for the study of Vogelzangs et al. 2009. 16* The connection with the vascular depression hypothesis is not clear at all, even the connection with inflammation. In the discussion the authors should compare with the literature and its not clear why they make the connection. We followed the reviewers suggestion (above) and deleted the vascular depression hypothesis. We hope our discussion including inflammation is clearer now. 17*The main concern indeed is the high risk of finding relations by chance, the main strength the large sample size. An important limitation is also measuring visceral fat by WC, CT would have been better. There might be also an underlying latent variable that may explain both variables like childhood abuse or genetics. We added these important concerns to the discussion section.