Author's response to reviews Title: Elevated depressive symptoms in metabolic syndrome in a general population of Japanese men: a cross-sectional study Authors: Atsuko Sekita (atsekita@med.kyushu-u.ac.jp) Hisatomi Arima (harima@envmed.med.kyushu-u.ac.jp) Toshiharu Ninomiya (nino@intmed2.med.kyushu-u.ac.jp) Tomoyuki Ohara (ohara77@envmed.med.kyushu-u.ac.jp) Yasufumi Doi (doi@intmed2.med.kyushu-u.ac.jp) Yoichiro Hirakawa (you1@envmed.med.kyushu-u.ac.jp) Masayo Fukuhara (fukuharm@envmed.med.kyushu-u.ac.jp) Jun Hata (jhata@georgeinstitute.org.au) Koji Yonemoto (yonemoto@envmed.med.kyushu-u.ac.jp) Yukiko Ga (yuki-ko@med.kyushu-u.ac.jp) Takanari Kitazono (kitazono@intmed2.med.kyushu-u.ac.jp) Shigenobu Kanba (skanba@npsych.med.kyushu-u.ac.jp) Yutaka Kiyohara (kiyohara@envmed.med.kyushu-u.ac.jp) Version: 2 Date: 14 June 2013 Author's response to reviews: see over
Thank you for your favorable comments and useful suggestions. We have attempted to address your suggestions as follows: Editor (Dr Annemieke van Straten) The association between metabolic syndrome and depressive symptoms has been described before. However, less so in Asian populations. Therefore I do think this study is relevant providing that the authors discuss similarities and discrepancies with other populations. The paper needs clarifications on a number of points (as described by the reviewers) which the authors should be able to provide. Thank you for your useful suggestion. We have discussed similarities and discrepancies with other populations in the 2 nd and 3 rd paragraphs of the Discussion (pages 11-12). We have also attempted to address the reviewers comments as described below. Reviewer 1 (Dr Kai Kahl) Comment 1. Please add the following references (introduction): Lamers et al. 2010, J Clin Psychiatry, Identifying depressive subtypes in a large cohort study: results from the Netherlands Study of Depression and Anxiety (NESDA) Kahl et al. 2012, Eur Arch Psychiatry Clin Neurosci., Prevalence of the metabolic syndrome in unipolar major depression. Thank you for your useful suggestion. Because we revised the Introduction according to a comment from Reviewer 2, the recommended references no longer fit into the logical progression of the passage. We therefore added the references to the Discussion section instead (references No. 23 and 27) (page 11, line 4). Comment 2. Current results from the NESDA study point to an important role of the autonomous nervous system mediating the relationship between MetS and depression. You may add this in the discussion (Licht et al. 2010, J Clin Endocrinol Metab. Increased sympathetic and decreased parasympathetic 1
activity rather than changes in hypothalamic-pituitary-adrenal axis activity is associated with metabolic abnormalities) We added the reference as suggested (reference No. 36) and discussed the autonomous nervous system as a possible mechanism underlying the link between MetS and depression in the 4 th paragraph of the Discussion section as follows: Chronic stress has also been shown to increase the risk of metabolic disorders through elevated sympathetic activity (page 12, line 18-19). Reviewer 2 (Dr Nicole Vogelzangs) Comment 1. Background: Recently, a systematic review on metabolic syndrome and depression was published (Pan et al., Diabetes Care 2012). This should be addressed in the Background. According to your suggestion, we revised the Background section and referred to this systematic review as follows: Recently, a systematic review of observational studies demonstrated a link between metabolic syndrome (MetS) and depressive symptoms. (reference No. 3; page 5, line 15-16). Comment 2. Methods, Study population and design: Which participants were eligible for the Hisayama Study? All residents of Hisayama of 40 years and older? So, the town of Hisayama holds 4328 residents of 40 years or older? Or were there any additional selection criteria? All 4,330 residents aged 40 years or older in the town of Hisayama were invited to participate in the examination. Among them, 3,376 participants provided consent to participate (participation rate 78.0%). We have clarified these facts in the Study Population and Design section of the Methods (page 6, line 4-7). Comment 3. Methods: Structuring of the Methods section would be more clear if covariates were separated from metabolic syndrome measures. The term Risk factor measurement is confusing as the paragraph describes both covariates as metabolic syndrome measurements. Measurements of e.g. blood pressure, waist circumference and lipids could be described under the subheading 2
Metabolic syndrome. Thank you for your useful suggestion. We now separately describe Metabolic syndrome and Other covariates in the Methods. Measurements of blood pressure, waist circumference, serum lipid concentrations and fasting blood glucose levels are now under the subheading of Metabolic syndrome in the Methods (page 6, line 10 - page 7, line 20). Comment 4. Methods: A description of how marital status was defined is missing. According to your suggestion, we have clarified that marital status was classified as either having a spouse on a family register or not (e.g., single, divorced) in the Other Covariates section of the Methods (page 7, line 12-13). Comment 5. Methods: Also, a more precise description of how other covariates were assessed should be included to help the reader understand how well these factors were measured. We have tried to provide more detailed information on other covariates in the Other Covariates section of the Methods (page 7, line 13-20). Comment 6. Methods: How was medication use (antihypertensive, antidiabetic, antidepressant) ascertained? The current use of medications (antihypertensive, antidiabetic, and antidepressant) was collected using a self-administered questionnaire and confirmed using the consumer drug information by trained staff. These facts are described in the Metabolic syndrome section (page 6, line 11-13) and the Elevated depressive symptoms section of the Methods (page 8, line 1-2). Comment 7. Methods, Measurement of depressive symptoms: Was the 20-item version of the CES-D used? Please specify. Also, has the CES-D been validated in Japan? We used the Japanese 20-item version of the CES-D, the reliability of which has 3
been validated. We have added this information to the Elevated depressive symptoms section of the Methods (page 7, line 23-25). Comment 8. Methods, Statistical analyses: Please specify the predictor and the outcome variables. Also, elaborate more on the different analyses that have been conducted (with different predictors) and add that analyses were performed sex-specific. According to your suggestion, we have specified the key risk factor and the outcome of the paper. We have provided detailed information on different analyses and specified the covariates included in the multivariable-adjusted models. We have also specified that all the analyses were performed separately for men and women. All these changes were made in the Statistical Analyses section of the Methods (page 8, line 5-18). Comment 9. Methods, Statistical analyses: Please specify how age-adjusted prevalences were calculated. The direct method is too ambiguous. The prevalence of elevated depressive symptoms in each subgroup defined by MetS or its components was standardized for age distribution of the total study subjects by the direct method using 10-year age groupings. This information is described in the Statistical Analyses section of the Methods (page 8, line 7-9). Comment 10. Methods, Statistical analyses/results: It would be good to formally test and report on sex-interactions. The differences in the association of MetS with elevated depressive symptoms between men and women were evaluated by adding interaction term(s) to the logistic regression models as described in the Statistical Analyses section of the Methods (page 8, line 15-18). The results (p values for homogeneity) are described in the 3 rd and 4 th paragraphs of the Results section (page 10, line 4-5 and line 16). Comment 11. Results, Table 1: Please include antidiabetic medication. We have added information on antidiabetic medication in the 1 st paragraph of the 4
Results section (page 9, line 7) and Table 1 (page 21). Comment 12. Results, Tables 1-4: Please include in a footnote the statistical tests that have been used. The statistical tests used in the tables are now described in the footnote of Tables 1-4 (pages 21-24). Comment 13. Results: Were the significant findings of Table 2, still significant after adjustment for all covariates? These findings could be reported within the text. This association of low HDL cholesterol (multivariate-adjusted OR 2.44 [95% CI 1.04-5.69]) and elevated fasting plasma glucose (1.93 [95% CI 1.07-3.49]) with elevated depressive symptoms remained significant even after adjustment for age, marital status, history of cardiovascular disease, smoking habits, alcohol intake, and regular exercise. We reported these findings in the 2 nd paragraph of the Results section (page 9, line 15-19). Comment 14. Results/Discussion: After adjustment for multiple covariates, the OR s hardly change. Considering the fact that covariates do not seem to be measured very precisely, it is possible that there is residual confounding. This should be addressed in the limitations section. : As you pointed out, the OR of MetS and the number of its components did not change substantially after adjustment for other covariates. This finding may have been partly due to residual confounding factors, because the accuracy of some covariates may have been somewhat limited due to self-reporting. We have added a sentence describing the limited accuracy of self-reported covariates as a study limitation in the 6 th paragraph of the Discussion section (page 13, line10). Comment 15. Discussion, 2nd paragraph: rather positive associations have been reported in studies evaluating mid- to late-life depressive symptoms. This sentence is followed by the discussion of several studies showing positive results. However, it is not clearly indicated whether these individual studies have 5
been conducted within older populations. Using phrases such as in older men and women in stead of in men and women could help to clarify. Also, some negative findings in older populations have been found and these studies should be discussed as well. Thank you very much for your useful suggestion. As you pointed out, a further detailed review of the literature revealed inconsistent results for middle-aged and elderly individuals as well as for young adults. We now recognize that there is limited use in comparing between younger and older groups. We sincerely apologize for the rather inadequate discussion of this matter in the previous version of the manuscript. We have revised the 2 nd paragraph of the Discussion section so that it focuses exclusively on the association between MetS and elevated depressive symptoms in general (page 11, line 2-19). Comment 16. Discussion: Authors should check the reference list of the recently published meta-analysis on metabolic syndrome and depression (Pan et al., Diabetes Care 2012) for more studies to be included into their Discussion. For instance, more studies within older samples and more studies within Asian/Japanese samples. Thank you for your useful suggestion. We have carefully reviewed all the references in the meta-analysis you mention, and we have updated the 2 nd and 3 rd paragraphs of the Discussion section accordingly (page 11, line2 page 12, line14). Comment 17. Discussion: The systematic review could also be used to give a more complete picture of previously found gender differences. According to your suggestion, we have carefully reviewed all the references in the systematic review and have added a passage regarding the gender differences in Western and Asian populations in the 3 rd paragraph of the Discussion section (page 11, line 20 page 12, line 14). Comment 18. Discussion: An important aim for the present study as stated in the Background was to examine the association between metabolic syndrome and depressive symptom in an Asian population (as opposed to Western 6
populations). It would be good then to include a paragraph in the Discussion considering this issue. Are findings similar? Are there any differences, for instance with regard to sex-differences? Thank you for your useful suggestion. We now discuss the gender differences separately for Western and Asian populations in the 3 rd paragraph of the Discussion section (page 11, line 20 page 12, line 14). Comment 19. Discussion, Conclusions, last sentence: Please replace and seems able to provide with and might be able to provide. We have changed and seems able to provide to and might be able to provide in last sentence of the Conclusion (page 13, line 17-18). Comment 20. Abstract, Methods: MetS was defined according to the joint interim statement. This is unclear. Please be more specific. We added the following sentence to the Methods section of the Abstract: MetS was diagnosed when a subject had three or more of the following components: 1) central obesity (waist circumference 90 cm for men, 80 cm in for women); 2) elevated blood pressure ( 130/85 mmhg or current use of antihypertensive medication); 3) hypertriglyceridemia ( 1.7 mmol/l); 4) low HDL cholesterol (<1.0 mmol/l for men, <1.3 mmol/l for women); and 5) elevated fasting plasma glucose ( 5.55 mmol/l or current use of antidiabetic medication). (page 4, line 8-14) Comment 21. Throughout paper: Multivariate should be replaced by multivariable. Multivariate means multiple outcomes, but that is not what is being meant here. We sincerely apologize for the misuse of the word multivariate. All instances of multivariate have now been changed to multivariable. Comment 22. Discussion, 1st paragraph: Being the first and summarizing paragraph, this paragraph should include a sentence stating there were no associations in women. 7
According to your suggestion, we have added a sentence stating that there were no associations in women in the first paragraph of the Discussion section (page 10, line 24 page 11, line 1). Comment 23. Typo s etc.: * Abstract, Results, line 3: delete the word it. * Keywords: depressive symptoms (add an s). * Keywords: population-based (add an d). * Methods, Risk factor measurement, line 6: exertion should be exercise. * Methods: Measurement of depressive symptoms, line 3: current use of antidepressant medication (add of). * Methods, Statistical analyses, line 3: a logistic regression model in stead of the logistic regression model. * Discussion, last line page 9: In regard to should be With regard to. Thank you very much for finding these typos. We have corrected them as you indicated. Comment 24. Methods: Why was BMI described and included in Table 1? It is not included in any of the analyses. As you suggested, we have removed BMI from the text and Table 1. Comment 25. Methods /Results: Why was total cholesterol included as a covariate? Theoretically, this would be overadjustment in my opinion. We have excluded total cholesterol from the multivariable-adjusted statistical models as suggested. Comment 26. Methods, Statistical analyses/results: Authors could consider to additionally include associations between continuous metabolic syndrome components and elevated depressive symptoms and/or to include analyses with continuous depressive symptoms as the outcome (linear regression). Analyses based on continuous measures in general have more power to detect associations and can show evidence for a dose-response relationship. 8
According to your suggestion, we have evaluated the effects of the number of MetS components as a continuous variable on elevated depressive symptoms. Increase in 1 component was associated with a 35% (95% CI 8-67%) increase in elevated depressive symptoms among men but not among women (OR 0.9, 95% CI 0.76-1.06). These results are now described in the 4 th paragraph of the Results section (Page 10, line 11-16). In preparation for conducting a linear regression analysis, we checked the distribution of the CES-D score. However, the distribution of the score was highly skewed even after log-transformation. Furthermore, the CES-D score of treated depression was in the normal range (mean 15) and is likely to have been underestimated. Therefore, we are afraid that a linear regression analysis would be too difficult to conduct. 9