BMJ - Decision on Manuscript ID BMJ

Transcription

1 BMJ - Decision on Manuscript ID BMJ

2 Body: 19-Feb-2018 Dear Mr. Lee Manuscript ID BMJ entitled "Predicted lean body mass, fat mass, and all-cause and cause-specific mortality in men: results from a prospective US cohort study" which you submitted to BMJ, Thank you for sending us your paper, manuscript.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. Yours sincerely, Jose Merino jmerino@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: John Fletcher (chair), Jonathan Deeks (statistical consultant), Elizabeth Loder, Sophie Cook, Daoxin Yin, Tiago Villanueva, José Merino Decision: Put points Detailed comments from the meeting:

3 First, please revise your paper to respond to all of the comments by the reviewers. You will notice that the reviews included below are those that you previously saw when we rejected the paper before your appeal. Even though you already sent us a rebuttal, please provide it again and also explain any changes that you made or will make as a result of these comments. Please also respond to these additional comments by the committee: * Obesity and the obesity paradox are topics of great interest to many. This paper tries to explain the paradox using prediction equations in a well characterized cohort study with good quality followup information and a large number of outcome events. We think that the paper could be improved if in the introduction and discussion it highlights the the obesity paradox and it includes more discussion on the clinical and public health implications of using the idea of a "lean compartment" and "fat compartment" rather than BMI as providing a better measure of adiposity. * Include actual risk values in the abstract. *We need additional details about the validation of the predictive equations. The validation of the predictive equations is based around correlation coefficients and does not take into account discrepancies in measurement. Do the equations predict with equal error across range of LBM and FM? Correlation doesn t tell us this. Given that the impact of the findings is largest at the extremes of LBM and FM, if the errors in the measurements are greater at these points that could be a problem. The validation paper does not provide this information, and it should be included in the current manuscript. * The data come from the Health Professionals follow-up study It is a large dataset of men. The original sample was so there was a 26% loss to follow-up. Why? * We were glad to see some analysis of continuous measures rather than categories. * Lag time analysis is poorly described. Please provide details. 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 the external peer reviewers** Reviewer: 1 Recommendation: Comments: This study evaluated the associations of predicted lean body mass, fat mass and all-cause and cause-specific mortality in men from the Health Professionals Follow-up Study. The paper is well-written and the data analyses are comprehensive. The limitations of the study had been addressed in the discussion. The findings provide important insight on the obesity paradox. I only have a few minor comments as below: 1. Abstract: please give the age range of the cohort at baseline. 2. Page 8, please give the number of participants with predicted LBM and FM, as well excluded due to caner or CVD, or BMI <12.5 or >60 kg/m2 at baseline.

4 3. Page 9, line 24, add 2 years after in the early follow-up period. 4. Supplementary table 5, please give the number of participants included in the analysis for each model. 5. Supplementary table 6, it would be helpful to the readers if BMI data for each decile of fat mass, lean body mass and BMI are given in the table. It also seems this table is not mentioned in the results section. Additional Questions: Please enter your name: Kun Zhu Job Title: Adjunct Associate Professor Institution: University of Western Australia 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=' lists/declaration-competing-interests'target='_new'> (please see BMJ policy) </a>please declare them here: Reviewer: 2 Recommendation: Comments: Title of paper: Predicted lean body mass, fat mass, and all-cause and cause-specific mortality in men: results from a prospective US cohort study. The submitted paper examines the association between predicted lean body mass and predicted lean fat mass and all-cause and cause-specific mortality in a large cohort (n=38,021) US men, with the ultimate aim to explain the J or U-shaped association between BMI and mortality. The topic is of interest to general readers interested in obesity or undernutrition. A strength of the study is the large sample size. A major weakness of the study is that predicted measures of body composition are used instead of direct measures of body composition. Although the authors argue in their methods section and discussion that the applied regression equations to assess body composition show a high predictive ability without systematic bias, these data are not convincing.

5 To argue the research gap of the present study, the authors state that the relationship between body composition and mortality is still unknown, as direct measure of body composition is difficult in epidemiological studies. In their introduction, reference is made to studies that use less accurate surrogate measures, such as arm circumference. However, there are studies (not included in the introduction) that do use direct measurements of body composition (for example in Gerontology 2012;58:32 40 ). The authors state in the discussion that there are a limited number of studies that examined mortality in relation to directly measured body composition, but no references are provided in the text, nor are these results discussed in the light of the current study results. This should be corrected. The research question is clear, but the BMI-mortality association should also be included in the research question/aim (abstract, introduction) as this result is also used for drawing conclusions (explanation of BMI-relationship). The use of predicted body composition instead of direct measures of body composition is a major weakness of the current study. The study uses predicted lean body mass (LBM) and predicted fat mass (FM) based on previously derived regression equations in the NHANES study (n=7,531). Based on DXA as dependent variable, a linear model was developed with age, race, height, self-reported weight and (self-measured) waist circumference as independent variables. The standard error of estimate was 2.6 kg for both predicted LBM and FM. The main concern of this study is that this measurement error is quite large at an individual level. It is also not clear if these measurement errors are similar across different levels of LBM and FM. It is not clear how these results should be interpreted. The authors are running a model with (differently weighted) age, race, height, self-reported weight and (self-measured) waist circumference as independent variables and mortality as dependent variable, also adjusting for height and race. It is not convincing that these results can be compared to a model with true measures of body composition and mortality. The methods, including the study population, are adequately described. With regard to the statistical methods: page 9, line 5/6: with excluding those with a low LBM, you may also exclude many with a low FM. It is therefore not clear how to interpret the results of this analysis step. The results are adequately presented with some room for improvement. I would also present the BMI-mortality association with quintiles, to be able to better compare the results. Likewise, I would also include BMI in Figure 1. The shape of the associations (and subtle changes therein) is difficult to read from Table 2. Is the U shape really stronger for LBM in the mutually adjusted model? Perhaps the HR s can be displayed in a Figure. The conclusion is too strong given the fact that predicted LBM and FM were used. The authors state (page 19, line 31-42) that only a limited number of studies uses direct measures of body composition and that these studies have major limitations like a small sample size. No references are provided and these earlier study results are not discussed. Given the major limitation of the current study (no direct measurement of body composition), it is especially important to carefully compare the results to all these previous studies. The abstract and key messages reflect accurately what paper says. See earlier comments on the aim (adding BMI). Minor: abbreviations LBM and FM are already introduced in objective (so in results use abbreviations); what does monotonic association mean? (=linear?).

6 Additional Questions: Please enter your name: Hanneke Wijnhoven Job Title: Assistant professor Institution: Vrije Universiteit Amsterdam, Netherlands, Department of Health sciences 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=' lists/declaration-competing-interests'target='_new'> (please see BMJ policy) </a>please declare them here: Reviewer: 3 Recommendation: Comments: There have been many previous studies, most comprehensively the global mega-analysis published in the Lancet in 2016, that have shown a U- or J-shaped relationship between BMI and mortality even after careful consideration of residual confounding by smoking and reverse causality due to weight loss. This study corroborates these previous observations and explores the relationship between two components of body composition predicted fat mass and lean body mass and mortality. Direct measures of body composition in large cohort studies are rare. There are some smaller studies that have looked at other surrogate measures of body composition. For example, a Danish cohort has published on body composition measured by bioimpedance and all-cause mortality in ~ 55,000 middle-aged men and women followed for an average of 5 years with a total of 1800 deaths accruing over 5 years of follow up [Bigaard et al. Obesity 2004;12:1042]. Another cohort study measured body composition via skinfolds on ~20,000 middle-aged men with a total of 428 deaths accruing during an average follow-up of 8 years [Lee et al. AJCN 1999; 69:373]. The present study uses predictive equations (developed in NHANES participants, and validated in an independent sample) to estimate body composition in ~ 40,000 men with a total of ~12,000 deaths accruing over a maximum of 25 years of follow-up. Thus, the long follow-up of the present study allowed for a large number of deaths to accrue and therefore this study provides much more

7 information than previous studies. It also allows examination of the association by different lengths of follow up time which allows us to look at reverse causality. A general journal is a good fit for this work the interpretation of the U-shaped association between BMI and all-cause mortality that has been shown in previous studies has been debated. It is very important that this association is examined in more detail because of the potential public health messages and implications of these observed relationships. This study explores whether differential relationships between components of body composition (predicted fat mass and predicted lean body mass) and mortality help to explain the higher risk of mortality in those with a low BMI. The results suggest that a low lean body mass may be underlying the higher risk of mortality in those in the lower range of BMI. The work is relevant for clinicians, patients and policy makers. The overall design of the study is adequate. The study relies on predictive equations to estimate fat mass and lean body mass, and a direct measurement would be preferable, but less feasible on a large cohort. The equations were developed in the NHANES participants and validated on an independent sample. The participants are described well. A couple of details are missing: Pg 6, line 36, what was the sample size of the independent validation group? What were some of the obesity-related biomarkers that were used in this study? Pg 7, lines At what timepoint was information on race collected? I have no ethical concerns about the study. The methods are generally described well, although some details are missing from the text of the manuscript: Pg 8, line 17. What is the underlying time variable in the Cox regression model? Attained age? The figure legend for Figure 1 states the the knots for the non-linear model are placed at the 5th, 50th, and 95th percentiles, but I couldn t find this information in the text of the methods section. Similarly, in the methods text the authors state that they adjust height for LBM and FM. The footnotes of the tables give more detail: Height was adjusted by including height as a continuous variable for fat mass and by regressing out variation due to height for lean body mass.. I m still not completely clear how height was adjusted for in the lean body mass model, I think by adding the residuals in the model? This could be explained more clearly in the footnotes and the methods text. Pg 9, line 24. The authors describe a sensitivity analyses where they used right censoring for age but the cut-off age is not stated in the text, although is given as age 85 in the footnotes of the relevant table (Supplementary Table 5). Pg 8, statistical analysis section. How did the authors deal with the repeated measurements of predicted LBM and FM? Was a cumulative average used? The results are well presented. The cubic spline figures are useful to visualize the relationships. The authors could also consider presenting the main results in Table 2 in figure format I think U-shape/J-shaped relationships are easier to visualize in a figure rather than a table. A flow diagram or timeline would be useful to see how many participants were excluded at each stage and how many had complete lean body mass and fat mass

8 measurements. Or otherwise numbers excluded need to be stated in the text of the manuscript (i.e. how many participants were excluded due to prior diagnosis with cancer or CVD? Or because of extremely low or high BMI? What was the average duration of follow up? Table 1. I found it very striking that men with a BMI < 18.5 were, on average, 6cm or more taller than every other category of BMI, and I think this should be mentioned in the results text. They also have a higher waist circumference than those in the next highest BMI category, although after that the relationship between waist and BMI is monotonic. Pg 10, lines I don t think this last sentence best describes the respective associations. Fat mass went down slightly in the 2nd category of BMI, and the associations with physical activity and AHEI do not appear to be linear, peaking in the 3rd category of BMI. What is the correlation between fat mass and lean body mass in the participants in the current study? Pg 12, lines In a mutually adjusted model including both predicted FM and LBM, the association between predicted FM and allcause mortality became slightly stronger. I don t agree with this statement the point estimate for quintile 5, goes from 1.33 in model 2, to 1.35 in model 3 (the mutually adjusted model). The point estimates and CIs are very similar. Pg 12, line 19. Please include 95% CI around the point estimate of 35%. Table 3. For model 3, where the bottom 2.5% of participants with low lean body mass are excluded there are no men left with a BMI < 18.5, therefore the J-/U-shape disappears, although I accept that the point estimates were raised a bit in the 2nd two categories and in model 3 they aren t. However, I think the interpretation of this analysis is limited. Pg 16. Cause-specific mortality. I think it is worth mentioning that the observed association with BMI is U-shaped for CVD mortality but inverse with respiratory mortality and positive with cancer mortality. The discussion is very well-written and the authors have clearly communicated the message of the paper. In the reference list, reference no.1 doesn t appear to be a full reference. I was expecting reference no. 30 to be emphasized more, and mentioned in the introduction as well as the discussion. The mega-analysis of over 200 prospective studies is the most comprehensive study on BMI and all-cause mortality. The What this paper adds box is a clear and accurate reflection of the key messages of the paper. Abstract, Pg 2, line 12. Please put the age range of the men at recruitment. Abstract, Pg 2. Please include the total number of deaths during follow up. Abstract, Pg 2, lines It would be helpful to put the 95% CI around the point estimates in the abstract.

9 Abstract, Pg 2, line 36. The abstract mentions the obesity paradox in the conclusions section, but does not explain what the paradox is. Given that the BMJ is a general journal, it would be helpful to briefly explain it. There is no background section for BMJ abstracts, but perhaps in the conclusion section the authors could mention that they found a U-shaped association between BMI and mortality which is in agreement with other previous studies. Additional Questions: Please enter your name: Kathryn Bradbury Job Title: Nutritional Epidemiologist Institution: University of Oxford 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=' lists/declaration-competing-interests'target='_new'> (please see BMJ policy) </a>please declare them here: