Author s response to reviews Title: Built environment and physical activity: domain- and activity-specific associations among Brazilian adolescents Authors: Inacio da Silva (inacio_cms@yahoo.com.br) Adriano Hino (akira.hino@pucpr.br) Adalberto Lopes (aadalberto@hotmail.com) Ulf Ekelund (ulf.ekelund@nih.no) Soren Brage (soren.brage@mrc-epid.cam.ac.uk) Helen Gonçalves (hdgs.epi@gmail.com) Ana Menezes (anamene.epi@gmail.com) Rodrigo Reis (reis.rodrigo@wustl.edu) Pedro Hallal (prchallal@gmail.com) Version: 1 Date: 24 May 2017 Author s response to reviews: Response letter Manuscript PUBH-D-17-00556: "Built environment and physical activity: domain- and activityspecific associations among Brazilian adolescents" Dear Editor of the BMC Public Health, Please find enclosed the revised version of the article entitled Built environment and physical activity: domain- and activity-specific associations among Brazilian adolescents. We have responded to each of the reviewer s comments and revised our manuscript accordingly, highlighting all changes. We look forward to hearing from you, Sincerely,
Inácio Crochemore M da Silva On behalf of all authors Reviewer(s)' Comments to Author: Reviewer: 1 Ian Janssen (Reviewer 1): The authors examined the relationship between several built environment factors and physical activity in a large sample of 18-year-olds from a mid-sized city in southern Brazil. As the authors point out in the introduction, more research of this kind has been called for low and middle income countries. With that being said, the selection of the 21 built environment features studied seems to be quite random and it at times seemed like the authors were on a fishing trip. The rationale for choosing these particular built environment features needs to be developed in the Introduction section. More details on some of the built environment measures also needs to be added to the Methods section, particularly the National Demographic Census measures (lines 159-166). For instance, what do you mean by "street garbage"? How was it measured? What is psychometric properties of the measure? What defines low vs. high street garbage? I would also recommend that the authors consider combining some of their built environment measures into a scale. This may help you simplify the findings. This would help reduce the chance of a type 1 error, which I am concerned about given the sheer number of relationships examined in this paper. I am concerned that the authors have at times over interpreted the findings, which may reflect type 1 error. That is, there was as many unexpected findings as there were expected findings, so it is hard to believe the expected findings. Answer: We thank the reviewer for these important comments. We opted in the present study to perform exploratory analyses on built environment characteristics potentially associated to physical activity. We kindly disagree that exploratory analyses in this scenario could be considered as a fishing trip, though we agree with the possibility of type 1 error in our results (as discussed in the limitation of the submitted version of the manuscript). Pelotas is a mid-sized city in the extreme south of Brazil and the population studied is highly-restricted in terms of age (adolescents aged 18 years) To this date there is limited evidence on showing which built and social environment variables are more likely to be associated with patterns of physical activity in populations living in mid-size urban areas in developing countries, which prevents a proper theoretical and or analytical model. Hence, our exploratory approach will provide initial insights for future hypothesis testing. For the same reason, combining built environment measures into a scale based on the international evidence would not be correct due to different environment in which such populations live and would prevent further identification of unique environmental characteristics that could affect physical activity on this population. Finally, any mathematical combination of the items, for example, composite indicators based on Principal Components Analyses, could pool together environment aspects closely related to the socioeconomic position and not aspects closely related to the behaviour evaluated. We have now added a sentence in the introduction section justifying the rationale of such exploratory analyses (page 6, line 111), and
reinforced the possibility of type 1 error in the limitations of the present study (page 18, line 401). Moreover, all results were discussed, including those in the expected and in the unexpected direction. Regarding the detailed assessment procedures required for the environmental assessment, especially the National Demographic Census measures, we added additional information to the methods (page 9, line 185) and added another potential limitation to the discussion section (page 17, line 390). However, using Geographic Information System both strength (of including a large number of variables in the study) and limitation (regarding the specificity or data quality of measurement) must always be considered. It is not clear to me why the authors had to remove the rural dwelling participants, who comprised about 10% of the sample, from the analyses. Certainly, their address could be geocoded and the GIS measures obtained in their neighbourhood. Answer: We thank the reviewer for this comment. Cohort members living in rural areas of Pelotas were not included due to unavailability of environmental characteristics to be assessed using the Geographic Information System. This information was included in the text (page 8, line 176). Moreover, the evaluation of the association between physical activity and built environment in urban and rural areas would require different conceptual framework and assessment of different attributes due to huge differences in the environment and in the physical activity patterns in these two settings. The authors need to justify their use of a 500-meter buffer to define the neighbourhoods. That buffer size would be appropriate for a young child with a limited independent mobility, but would likely be far too small to define the neighbourhood boundaries for an 18-year-old. Answer: The buffer was defined arbitrarily considering that it could mean an approximate distance of a 10-minute walk in the studied context. This information was added to the text (page 9, line 179). The authors considered SES X build environment interactions. They should also look for gender interactions. Answer: We thank the reviewer for this comment. In fact, we tested for gender interactions and it was not identified. This information is now included in the methods and results sections. In the paragraph starting in line 330, the authors provide a number of potential reasons why they did not find many associations. This makes it sound like there are associations, but they just weren't able to detect them for one reason or another. What they fail to recognize is that perhaps there are just not strong or meaningful effects of the built environment constructs measured in their work (i.e., these built environment measures are not important determinants of physical activity in the population and location where the research was conducted). Answer: We thank the reviewer for this important comment. This topic is already included and discussed in the sentence below the mentioned paragraph.
The authors included a very small list of covariates in the analyses and this limitation needs to be recognized. Answer: We thank the reviewer s comment. According to the data availability, the exploratory purposes of the present study and considering a classical definition of confounders factors, we believe that the main covariates were added in the regression models. Final models were adjusted by gender, socioeconomic position, time living in the household and by other environmental variables that remained significantly associated. The English grammar and writing is a bit rough and awkward in places. The paper requires a careful proof by a native English speaker. Also, the authors jump back and forth between the use of American and British spelling, and they need to maintain consistency throughout the manuscript. Answer: We thank the reviewers for this important comment. The revised version was submitted to a careful proof by a native English speaker. Reviewer: 2 Antonio Prista (Reviewer 2): The paper relates to a very interesting topic with a very complex approach including a large number of variables. Problem, aims and methods are well defined and consistent. It is not described if authors previous checked redundancies since a reduction of variables will turn the analysis more clear. I understand that with such a large amount of information is hard to write the results and choose the more relevant information. However, I suggest a revision of the text on those two sections (results and discussion) turning more fluent the text of the results and avoid repetition of the results on discussion text. I consider the paper a very good contribution to the topic but I do not agree that is representative of Brazilian citizens (has described in page 17) neither a low and middle-income countries. Answer: We thank the reviewer for all these important comments. The exploratory purpose of our study is now included in the introduction section, justifying the large number of variables and highlighting the relevance of such descriptive analyses, rather than focusing on just a few information. The results and discussion section were revised accordingly all these comments. Reviewer: 3 Jo Salmon, PhD (Reviewer 3): Review of PUBH-D-17-00556 General comments This is a comprehensive cross-sectional analysis of associations between objectively assessed features of the built environment (BE) and adolescents' self-reported and objective physical
activity. Testing interactions between the BE and active commuting by SES is a novel feature of the paper. The inclusion of physical activity domains is also novel. Because there are a lot of variables and analyses, justifying certain analytic decisions with the data can be challenging. I have made several specific comments about this below. Answer: We thank the all reviewer s comments. Specific comments Background 1. Line 89: " due to the substantial decline in physical activity", please quantify, what is 'substantial'? Answer: We thank the reviewer for this comment. Actually, the sentence was not clear and we have revised the statement to highlight mainly the period of which such decline is usually observed (Page 5, line 87). We opted not to quantify such decline because it varies according to different context and in this paragraph the idea is only to present an overall pattern. Furthermore, in this sentence we have provided two references: (a) a large systematically review which presented decreases in physical activity from childhood to adulthood; and (b) the best local evidence regarding the decline of physical activity objectively measured among comparing children (7 years old), adolescents (18 years old) and young adults (30 years old). 2. Line 92: " operating differentially in the main activity domains ", please provide examples. Answer: We thank the reviewer for this comment. There are many examples of potential physical activity determinants which may play different roles according to different domains of physical activity. In low and middle-income countries, for example, higher socioeconomic status is a correlate of leisure-time physical activity, but it is inversely associated with transport-related physical activity. In the introduction, we opted to present a wider overview about main research questions, especially due the scope of the BMC Public Health. In the discussion, we addressed such differences in terms of physical activity domains focusing on environmental characteristics, taking your comment into account. 3. Lines 99-100: "Research on the association between environmental factors and youth physical activity has grown in recent years." An overview of this growing evidence is needed here. In the following sentences there is a nice definition of environmental influences but nothing about what is known. The sentence on line 110 states that more evidence from LMIC is needed, the reader has to take the authors at their word as this evidence hasn't been described or critically appraised at all. Answer: We thank the reviewer s comment. We added the main environmental characteristics usually associated with physical activity among youth and the gap in the literature (page 6, line
111). However, we opted to present more details of the available evidence in the discussion section, comparing our findings. 4. Line 114: Please add the testing of interactions between the BE and active commuting by SES as an aim. Answer: We agree with this comment and the aim of the study was revised. As we examined sex-interactions in all physical activity domains, we decided to include a general statement about these interactions (page 6, line 120). Methods 5. Lines 117-118: "The city is markedly plane.", I'm not sure what this means - flat? Answer: We thank the reviewer for this note. The sentence was revised and flat is the correct word to be used. 6. Lines 138-140: " analyzed the raw data in 5 second epoch, from which we derived average minutes per day spent in moderate to vigorous physical activity ". More information is needed about management of the accelerometer data. How was non-wear time determined? What was a valid day (how many hours worn as a minimum)? How many days did participants wear the monitors for and how many days were needed in order to calculate average mins/day in MVPA? Also, please add an 's' to the word "epoch". Answer: We agree with this comment. All information regarding accelerometry data collection and data reduction was added in the paragraph (page 7, line 143). Furthermore, we highlighted two references in which other pieces of information were already presented. 7. Line 140: " calculated as time when acceleration was above 100 mg in 10 minutes bouts". This is an unusual requirement for a young (adolescent) population. What is the rationale for this? Is it necessary for adolescents to engage in MVPA in 10+ minute bouts in the physical activity guidelines for youth in Brazil? Answer: We thank the reviewer for this question. We agree that it is not a usual requirement for a young population in accelerometry-based studies. However, it has been often used in wristworn accelerometry based on raw data. Moreover, the bout-criterium might be considered a strategy to evaluate a more structured physical activity, which is in agreement with the research questions of the present study. 8. Lines 143-148: I would like a bit more information about management of the IPAQ data. How was the daily? estimate of mins in each of these domains derived? Answer: We thank the reviewer s comment. The sentence was revised and more information was added (page 8, line 157). As we opted for activity-specific variables, the operational definition
was performed only separating participants between those walking or not, and those practicing moderate to vigorous activities during leisure time or not. 9. Lines 173-175: Is the PARA instrument based on in-person or on-line audits or from GIS data? This is unclear. Answer: We agree with the reviewer s comment. PARA instrument was used based on in-person (in loco) audits. The paragraph was revised accordingly (page 9, line 193). 10. Line 188: "Cycle paths/lanes were measured by a GPS device ". Did participants wear a GPS device or were these data collected some other way? Answer: Participants did not wear a GPS device. Cycle paths/lanes were collected separately and latter included in the ArcGIS software for all analyses. The sentence was revised for proper clarification (page 10, line 213). 11. Line 199: I was surprised that the covariates did not include accelerometer wear time. Surely this varied between participants? Answer: Differently from hip-worn accelerometry, the wrist-worn accelerometry is characterized by a higher compliance. In this follow-up, the members of the Pelotas 1992 Birth Cohort were required to use the accelerometer for a 24h-protocol and there was on average only 5.2% (SD 12.1) of non-wear time. Therefore, we decided not to include this information as a covariate. 12. Lines 212-213: "Associations were tested only when a plausible rationale was found." What does this mean? That only relationships between the BE & MVPA for which there was an a priori hypothesis were examined? This needs to be better explained. Answer: We thank the reviewer s comment. Associations were tested only when there was a plausible rationale in the literature considering environmental attributes and specific physical activity domains. The text was revised accordingly (page 11, line 238). 13. Lines 213-214: With the statistical analyses, was there any clustering at the neighbourhood level that needed to be adjusted for? Also what about the sampling frame (the hospital participants were born in)? Answer: The Pelotas 1993 Birth Cohort included all live births in the city of Pelotas (during the calendar year of 1993) who were born in maternity hospitals (99% of all live births of the city). The unit of analysis was the cohorts members, as we generate a single/specific buffer to evaluate the built environment for each participant. If the sample process or the built environmental assessment were carried out in an aggregated level, it should be considered in the analyses. Results
14. Lines 220-221: "Participants accumulated on average 42.9 (SD 42.7) minutes of MVPA per day as measured by accelerometry". Based on the description in the Methods, this should read "Participants accumulated on average 42.9 (SD 42.7) minutes of MVPA IN 10-MINUTE BOUTS OR MORE per day " Answer: We thank the reviewer s comment. The text was revised accordingly (page 11, line 247). Discussion 15. Line 276: " inverse correlates of commuting physical activity,..". I'm curious to know whether active transport varied by SES. In many developed countries, those living in low SES areas are more likely to actively commute. Perhaps it's the same in Brazil? Answer: In our sample and in other studies conducted in Brazil, active commuting also varied by SES. Low SES is more likely to actively commute. However, the association between built environment and physical activity was not observed in this population group. These results are discussed later in the manuscript and the most likely reason might be attributed for the low variability of built environmental characteristics, as well as the lack of choice (absence of car and other cheaper means of transportation). 16. Line 296: "the built environment might provide more accurate data, and therefore more consistent associations, as compared to subjective techniques". In contrast, studies with adults have also found a mismatch between what people perceive is present in the environment and what is actually there, with perceptions sometimes being most strongly associated with PA. Worth more a bit more discussion? Answer: We thank the reviewer s comment. We agree with this important topic. However, the statement provided is also correct and presented a different discussion. As we have a limited number of words, we opted not to include this discussion in the present study, leaving more space to discuss our findings. 17. Lines 298-300: "Studies among adolescents using objective measures of the environment found that land-use mix and residential density were the most frequent significant correlates of transport-related physical activity." What studies? Need some references here. Answer: The reference had been provided later in the paragraph. We now added a citation in the beginning of this paragraph (page 15, line 328). 18. Line 309: "In addition, a feature possibly related to safety, public street lighting, was associated with leisure-time walking.". Is this positively associated? Please specify. Answer: We thank the reviewer s comment. The information is now provided in the text (positively associated) (page 15, line 337).
19. Lines 314-315: "Built environmental effects on commuting physical activity were modified by SES. Stratified analyses highlighted the lack of environmental influence on commuting ". As this is a cross-sectional study, use terms such as 'effects' and 'influence' are not appropriate. Answer: We thank the reviewer s comment. The paragraph was revised accordingly (page 15, line 342). 20. Lines 317-318: " some unexpected associations among those from THE wealthiest SES group." Please specify the direction of the association. Answer: We thank the reviewer s comment. The paragraph was revised accordingly (page 15, line 337). 21. Lines 318-320: "Commuting physical activity, mainly in low and middle-income settings, might be more related to a need than to an individual choice." As noted earlier this finding should be compared with other studies that have reported on relationships between SES and active commuting. Answer: We thank the reviewer s comment. The paragraph was revised and we highlighted the higher prevalence of commuting physical activity in low and middle-income settings, as well as possible reasons for the weak associations between built environment and physical activity in low SES groups (page 16, line 348). 22. Lines 332-335: "A higher number of public open spaces was found in census tracts with higher average family income, exactly where people use to have a wide range of choice for their leisure-time physical activities." The second half of this sentence is unclear to me, please reword. Answer: We thank the reviewer s comment. The paragraph was reworded accordingly (page 16, line 365). 23. Lines 358-359: "For the current analyses, most environmental information was collected closely in time..". Please reword. Perhaps "within a short time-frame"? Answer: The paragraph was reworded accordingly (page 18, line 397). Conclusion 24. Lines 371-372: "Improvements in the built environment, especially in characteristics like public open spaces, cycle and walking paths "; however, most of the results for public open spaces in the current study were non-significant? Answer: We thank the reviewer s comment. The first part of the conclusion paragraph is based on the large amount of evidence available in the literature and we are clear in our conclusions
later in the paragraph focusing on our findings on beachfront, street lighting, paved streets and cycle paths/lanes. 25. Lines 379-381: "In our study, some environmental attributes like beachfront, street lighting, paved streets and cycle paths/lanes were shown to influence physical activity patterns." Again as a cross-sectional analysis, please use a more appropriate term than 'influence'. Answer: The sentence was reworded accordingly (page 18, line 417). 26. Lines 381-382: "InfrastructurE changes that either improve or deteriorate these environmental factors.". The word deteriorate isn't appropriate here. Answer: We thank the reviewer s comment. The paragraph was revised accordingly (page 19, line 420). 27. Table 2: Please specify the physical activity units here. Average daily minutes of 10+ bouts/day? Answer: We thank the reviewer s comment. The title for Table 2 was revised accordingly. 28. Table 2: I'm confused about the values of the variables without the asterixes in the adjusted column? Are they single environmental variables adjusting for sex, time living at the address and SES? And are the * variables adjusted for these covariates and all other * variables? Answer: All variables were adjusted for sex, time living at that address and socioeconomic status. After this stage, we created a final adjusted model including also other environmental variables that remained associated to the previous adjusted model. 29. Tables 2-5: As already noted above (comment #12), I'm struggling to understand for these analyses how the multivariable models were selected. There are variables which were significantly associated in the crude models that are not included in the multivariable models, and vice versa? Answer: In the adjusted analyses ( ** ) the covariates were sex, time living in that address, and socioeconomic status (SES). However, we opted to perform an additional adjustment for other environmental variables which were associated to the first adjusted model ( ** ). Therefore, the environmental factors that were associated with the specific outcome under investigation in the first adjusted model, were also added in the final model ( * ) [except when environmental variables presented high collinearity (rho 0.6)]. It was described and revised in the methods section and it is now also revised in the tables footnotes. 30. Table 3: As for Table 2, please specify the value of 'walking for leisure', is this none versus some? I also suggest amending the Table 2 title to read "Crude and adjusted associations between built environmental variables and walking FOR LEISURE". Answer: We thank the reviewer s comment. All tables (and titles) were revised accordingly.
Suggested grammatical edits Abstract 31. Change "to" to "with": "proportion of paved streets and buffer's average family income were associated WITH lower MVPA." Remove "al" from "infrastructure": "This suggests that infrastructure interventions" Answer: We thank the reviewer for these important notes. The text was revised accordingly. Background 32. Lines 92-93: "The ecological model adapted for this behavior assumes physical activity IS a consequence ", please change "as" to "is" Answer: We thank the reviewer for this note. The text was revised accordingly. 33. Lines 103-105: "including among other characteristics, streets design and their network, sidewalks, bicycle lanes, green space, public and private..", suggest amending to "including among other characteristics, design and network OF STREETS, sidewalks, bicycle lanes, green space, AND public and private.." Answer: We thank the reviewer for this comment. The text was revised accordingly. Methods 34. Line 185: "We selected these two attributed because..", please amend to "attributes". Answer: We thank the reviewer for this comment. The text was revised accordingly. 35. Lines 186-187: "A variable summing up all attributes to physical activity practice was also generated. ", I suggest amending this sentence to "A variable summing all attributes SUPPORTING physical activity practice was also generated.". Answer: We thank the reviewer for this note. The text was revised accordingly. 36. Lines 193-194: "This information allowed the identification of the number of gyms within each individual". I suggest adding the word "buffer" after "individual". Answer: We thank the reviewer for this note. The word suggested was added in the text.
37. Line 195: "participant's households to any place to physical activity practice", please amend to "participant's households to any place SUPPORTING physical activity practice " Answer: We thank the reviewer for this note. The text was revised accordingly. Results 38. Lines 225-226: "87% of the households showed public street lighting", please amend to read ""87% of the households HAD public street lighting". Answer: We thank the reviewer for this note. The text was revised accordingly. 39. Lines 238-239: "average family income were associated to lower MVPA", please amend to "average family income were associated WITH lower MVPA" Answer: We thank the reviewer for this comment. The text was revised accordingly. Discussion 40. Line 274: "Differently from our expectations,..", suggest amending to "IN CONTRAST TO our expectations ". Answer: We thank the reviewer for this comment. The text was revised accordingly (page 14, line 302). 41. Line 311: "Paved streets proportion, which might be considered ", suggest amending to "The proportion of paved streets within buffers, which might be considered " Answer: We thank the reviewer for this comment. The text was revised accordingly (page 3, line54).