The association between physical activity and body fat percentage , with adjustment for BMI : a large cross-sectional analysis of UK Biobank

objectives changed to: group.bmj.com on October 19, 2017 Published by http://bmjopen.bmj.com/ Downloaded from

The results of the study indicate that BMI and percent body fat are closely related, and that there is an inverse relationship between physical activity and adiposity, although these are not entirely new findings.Other things that I found odd were that the authors focused their analysis on comparing those with < 5 excess METxhours/week and those with at least 100,000 excess METxhours/week of physical activity.These are wildly divergent groups, and it would also be useful to quantify the cross-sectional relationship between physical activity and adiposity in groups that were not as extreme.I would like to see some statistical analyses run on the less divergent subgroups.
Of course, one of the advantages of using percent body fat rather than BMI as an index of adiposity is that the former takes into account the amount of fat mass and lean body mass (LBM).Given that high levels of physical activity result in oxidation of fat, and could also conceivably build muscle (or decrease age-associated loss of muscle mass), it would be interesting if the authors discussed whether or not their study show differences in the cross sectional relationships between physical activity and adiposity, depending upon which measure of adiposity was employed.

REVIEWER
Pantelis Theodoros Nikolaidis Department of Physical and Cultural Education, Hellenic Army Academy, Athens, Greece REVIEW RETURNED 06-May-2016

GENERAL COMMENTS
General comments This is a very interesting study on an important topic using a very large sample and laboratory assessment methods.The authors have used correctly the scientific methods and present their findings properly using tables and figures of high quality.Given the significance of this study for health practitioners, I would recommend it for publication, but a few issues should be addressed previously.My main concern is that extremely few references (13) have been used for such a well-studied topic.I recommend adding much more references in order to add more information in the Methods 3) Statistics: The criteria for the interpretation of Pearson's r should be presented (e.g. when it is considered small, medium…).
Discussion 4) A paragraph presenting limitations, strengths and practical implications before conclusions is missing.

VERSION 1 -AUTHOR RESPONSE
Reviewer: 1 Comment 1: This research is a nice addition to the literature and will appeal to a broad audience.The "no" response in regards to the methods being indicated completely results form the authors failure to indicate whether participants were given the standard pre-test instructions for body composition assessment via BIA.As these instructions can affect the accuracy of the test, whether they were provided, and whether any measures were taken to determine the degree to which participants followed them would be of interest.
Authors' response: The participants were not given any specific instructions pertaining to the bioimpedance measures prior to attending the assessment centre.We have added this information to the paper (pages 6-7): "Participants were not asked to fast, nor were they given any specific instructions pertaining to the bio-impedance measures prior to attending the assessment centre.Water was available at all times throughout the visit and visits occurred throughout the day (8am-8pm)." We have also added in the following sentence to the discussion (page 16): "Hydration status, exercise and food consumption can have small effects on body fat values measured by bioimpedance; had these factors been standardised between participants, we may have seen slightly stronger associations between body fat percentage and physical activity." Comment 2: The other "no" indicated above is related to the statistical methods utilized.Usage of Cohen's delta would aid tremendously in interpreting the results, particularly since the "n" was so large.Cohen's delta would indicate the size of the difference, rather than simply whether the difference is statistically significant.The conclusion of the authors would be better supported if Cohen's delta was included as part of the analyses.
Authors' response: We have reported in the paper (page 13) the size of the difference in body fat percentage, after adjusting for BMI, in participants who did <5 excess MET-hours per week compared to those who did 100 or more excess MET-hours per week.We have not presented a standardised measure of effect size, such as Cohen's delta, because we felt most readers would find it more useful to have the estimate of the difference on the body fat percentage scale.If the editor feels it is necessary to add another measure of the size of the difference, we can include the Cohen's delta.

Reviewer: 2
Comment 1: The paper has several strengths.Primarily these are the large sample size (N=259,808), and the addition of percent body fat as an additional measure of adiposity.More detail is required on the bioelectrical impedance analyzer (BIA) methods; for instance, were time of day and hydration status controlled?
Authors' response: See response to comment 1 from reviewer 1.
Comment 2: In terms of methodology, if measures of physical activity were obtained in the UK BioBank study with a wearable device, these would provide a more objective measure than the modified IPAQ short form questions.Moreover, repeat measures of physical activity would be preferable given the anecdotal reports that initial and follow-up measures were quite different.

Authors' response:
UK Biobank has collected accelerometer data, however only on a sub-sample of approximately 100,000 participants and currently only the raw data is accessible, thus it is unfortunately not able to be used in the current paper.The same physical activity questionnaire was administered on approximately 20,000 participants in UK Biobank about 5 years after recruitment.The responses to the 2nd questionnaire show regression to the mean and we use this information to help interpret baseline physical activity.
Comment 3: The results of the study indicate that BMI and percent body fat are closely related, and that there is an inverse relationship between physical activity and adiposity, although these are not entirely new findings.Other things that I found odd were that the authors focused their analysis on comparing those with < 5 excess METxhours/week and those with at least 100,000 excess METxhours/week of physical activity.These are wildly divergent groups, and it would also be useful to quantify the crosssectional relationship between physical activity and adiposity in groups that were not as extreme.I would like to see some statistical analyses run on the less divergent subgroups.
Authors' response: It is not possible to give pairwise comparisons between every group, and if we did not use the extreme categories we would have to arbitrarily choose another two categories to compare.We have presented the numbers that correspond to Figure 4 in Supplementary Tables 3 and 4, from which other pairwise comparisons can be derived.Also, although <5 excess MET-hours and 100 or more excess MET-hours seems quite a large difference, the re-measurement of a sub-sample 5 years later indicates that the average physical activity in these groups was approximately 12 excess MET-hours and 82 excess MET-hours, respectively.
Comment 4: Of course, one of the advantages of using percent body fat rather than BMI as an index of adiposity is that the former takes into account the amount of fat mass and lean body mass (LBM).Given that high levels of physical activity result in oxidation of fat, and could also conceivably build muscle (or decrease age-associated loss of muscle mass), it would be interesting if the authors discussed whether or not their study show differences in the cross sectional relationships between physical activity and adiposity, depending upon which measure of adiposity was employed.

Authors' response:
We state in the first sentence of our discussion that "…more physical activity was associated with both a lower BMI and a lower body fat percentage".We also state in the results section (page 13) that "… as shown by the r2 values, age and physical activity explained more of the variation in body fat percentage than the variation in BMI in this study population" Reviewer: 3 General comments: This is a very interesting study on an important topic using a very large sample and laboratory assessment methods.The authors have used correctly the scientific methods and present their findings properly using tables and figures of high quality.Given the significance of this study for health practitioners, I would recommend it for publication, but a few issues should be addressed previously.My main concern is that extremely few references (13) have been used for such a well-studied topic.I recommend adding much more references in order to add more information in the Introduction and Discussion.A recommending additional literature is the following: As stated in response to the previous comment, we have kept the introduction focused on the novel aspect of the paper.
2) All the aims presented in the end of the Introduction have not introduced properly, that is, the relationship between PA with BF and BMI, and the "important lifestyle factors" must be reported in the previous paragraphs.
Authors' response: We have removed reference to the important lifestyle factors in the introductionthis is a sensitivity analysis and does not need to be introduced here.The second paragraph of the introduction we state that "…people who do comparatively more physical activity have a lower BMI than less active people (Ref 4,5).Few large epidemiological studies have directly estimated body fatness…" Methods 3) Statistics: The criteria for the interpretation of Pearson's r should be presented (e.g. when it is considered small, medium…).

Authors' response:
We have added the following to the methods section (page 9): "Pearson's correlation coefficients between BMI and body fat percentage were calculated; values of 0.80 or above are considered very strong, values between 0.60-0.79strong, 0.40-0.59moderate, 0.20-0.39weak, and 0.00-0.19very weak (11)." We have now also described the correlations as strong/very strong (page 12): "The correlation between BMI and body fat percentage was very strong in women (r = 0.85), and strong in men (r = 0.79)."Discussion 4) A paragraph presenting limitations, strengths and practical implications before conclusions is missing.
Authors' response: For the discussion we have followed the order recommended by BMJ open.The 2nd paragraph of the discussion describes the strengths and weaknesses of the study and the 4th paragraph describes the implications of the results (i.e. that our results suggest that other studies that have adjusted for BMI, may not have fully controlled for adiposity).

GENERAL COMMENTS
This is an interesting topic.The topic has been researched considerably previously; however, this study is unusual in its size and due to combining body fat assessment with BIA.Additional reference to previous studies would have been beneficial.Providing pre-test instructions for the BIA assessment and using a more objective measure of physical activity would have greatly strengthened the study.However, objective assessment of physical activity would be difficult in a study of this size.

GENERAL COMMENTS
The authors refused to consider my comments for adding more references and enhancing the discussion.Considering the widely already known conclusions of the abstract that "Conclusions: Physical activity was inversely associated with BMI and body fat percentage.For people with the same BMI, those who were more active had a lower body fat percentage.",I cannot understand why the authors do not discuss their findings with many previous studies.I would expect from a study on such topic to use at least 30-40 references; they cannot conclude the very well known "physical activity was inversely associated with BMI and body fat percentage" without discussing this with many published studies.Thus, I suggest the rejection of the paper.
Nikolaidis, P.T., and Zisimatos, D. (2014): Relationship between body mass index and physical fitness in army cadets.Saudi Journal of Sports Medicine, 14(2):144-150.Nikolaidis, P.T. (2013): Body mass index and body fat percentage are associated with decreased physical fitness in adolescent and adult female volleyball players.Journal of Research in Medical All the aims presented in the end of the Introduction have not introduced properly, that is, the relationship between PA with BF and BMI, and the "important lifestyle factors" must be reported in the previous paragraphs.
Sui, X., Lavie, C.J., Blair, S.N.Body Mass Index, the Most Widely Used but Also Widely Criticized Index Would a Criterion Standard Measure of Total Body Fat Be a Better Predictor of Cardiovascular Disease Mortality?(2016) Mayo Clinic Proceedings, 91 (4), pp.443-455.The Introduction is too short and has few references.The first paragraph needs to be larger.