Social support, social network and salt-reduction behaviours in children: a substudy of the School-EduSalt trial

Objectives Healthy behaviour changes, such as reducing salt intake, are important to prevent lifestyle-related diseases. Social environment is a major challenge to achieve such behaviours, but the explicit mechanisms remain largely unknown. We investigated whether social networks of children were associated with their behaviours to reduce salt intake. Design An ancillary study of a school-based cluster randomised controlled trial to reduce salt intake in children and their families (School-EduSalt), in which salt intake of children was significantly reduced by 25%. Setting 14 primary schools in urban Changzhi, northern China. Participants 603 children aged 10–12 years in the intervention arm. Primary and secondary outcome measures We developed a score assessing salt-reduction behaviours (SRB score) of children based on self-administered questionnaires. The SRB score was validated by the changes in salt intake measured by 24-hour urine collection in a random sample of 135 children. A 1-unit increase in SRB score was associated with a 0.31 g/day greater reduction in salt intake during the trial (95% CI 0.06 to 0.57, p=0.016). Results Children from families with more family members not supporting salt reduction had significantly lower SRB scores (p<0.0001). Children from a class with a smaller size and from a class with more friendship connections, as well as children having more friends within the class all showed higher SRB scores (all p<0.05). Children whose school teachers attended the intervention programme more frequently also had higher SRB scores (p=0.043). Conclusion Social networks were associated with the behaviours to reduce salt intake in children. Future salt-reduction programmes may benefit from strategies that actively engage families and teachers, and strategies that enhance interconnectivity among peers. Trial registration number NCT01821144; post-results.

is a change score). Were the Qs included for SRB score part of the original planned trial and data collection? • Line 6, data analysis section: additional adjustment of BMI too across quartile SRB score • Line 53: Is patient ? a typo, perhaps participant • Table 1& 3: why is it not possible to separate out mother from grandparents i.e. the 'others'; fathers was collected separately and information can then be provided about the importance or not of support from this family member. Is there a reason why mothers was not specifically examined? Can this be stated in the method. Particularly as this is also a discussion point at line 41, page 12

GENERAL COMMENTS
The reviewer completed the checklist but made no further comments.

GENERAL COMMENTS
The methods are correctly used. I recommend to give the references of the techniques used and not only the software employed.

Yipu Shi
Public Health Agency of Canada Canada REVIEW RETURNED 25-Mar-2019

GENERAL COMMENTS
Comments to the statistical procedures: 1. Well designed and methodological vigour school based intervention study.
2. Mixed linear models were used to appropriately account for the clustering effect of children in the same class. It's a methodological strength however, it seems that the association of SRB with each aspect of the children's social networks was assessed individually, then the joint effect of these networks on SRB was assessed in a new model by choosing one most influencing variable from each network (page 8 line 46), can't this be done in one mixed model so the independent effect of all variables of the network are assessed simultaneously?
3. It's not clear how the family support (partial vs. full) was defined in the joint mixed linear model, looks like it's defined differently in the individual family support model (Table 3) 4. Are the children having their 24-hour urine collected before and after the trial used for the validation of the SRB score (n=135) the same as those used for the sensitivity analysis (n=135)? Suggest the reduction in 24-hour urine used in the mixed model in Supplementary table S1 be treated as positive so that the direction and strengths of the association are comparable with Table 3.
5. The authors used advanced statistical model and social network analysis techniques in addition to the traditional method to carefully examine the role of peer influence in salt reduction, which is another technical strength and contribution to the methodologies in this field.

VERSION 1 -AUTHOR RESPONSE
Reviewer #1 Tatsuo Shimosawa, International University of Health and Welfare, Japan The authors analyzed School-EduSalt trial to find social environment impact on salt reduction in children and address important issue for population approach to reduce salt intake.
1.The study is well-organized and data are properly analyzed except for one minor concern I have. In Method section, the authors stated that both children and parents urine were analyzed for sodium excretion but I could not find data of urinary sodium excretion changes in parents. Does it correlate with children's urinary sodium excretion?
Response: We thank the reviewer for raising this relevant point. In the main trial, 24-hour urine collection was made for children and two adult family members of each child at both baseline and the end of the trial. In the current study, we did not provide data on the change in sodium intake of parents because this study aimed to examine the social network factors associated with sodium consumption of children.
To answer the reviewer's question, we have performed the following analyses. The change in salt intake in children from baseline to the end of the trial (post-pre) was -0.7g/d (95% confidence interval: -1.1 to -0.3). The change in salt intake in adults was -2.2 gram/d (95% confidence interval: -2.9 to -1.6). The change in sodium intake of children was closely related to the average change in sodium intake of their family members, and the correlation coefficient was 0.70 (P<0.001). Among the family members recruited, 73% were parents and 27% were grandparents. Sodium intake of children and that of their parents were also significantly correlated, but with smaller magnitude (correlation coefficient r=0.40, P<0.001).

Reviewer #2 Carley Grimes, Deakin University, Australia
This is an interesting paper which makes use of existing data collected within a randomised trial targeting dietary salt reduction among Chinese families. The data presented are novel, indicating that select social support structures are important factors to help children change behaviours to lower dietary salt intake. This information can be used in the development of future salt reduction education and behavioural programs. The paper is well-written and information clearly presented. I have some minor comments to address below.
1.Can it please be clarified in the methods when the questions for the SRB score were administered to children. Was this done just once, at the end of the study? i.e. there was no pre score and therefore change in this score? (Unlike salt intake for which there is a change score). Were the Qs included for SRB score part of the original planned trial and data collection?
Response: We thank the reviewer for this important comment. The post-trial survey, in which information on SRB score was collected, was conducted nine months after the end of the trial. This information is provided in the method section (Line 6-11 on Page 6). We do not have data on SRB score at baseline because this study was proposed after the main trial was completed, as an ancillary study of the main trial, with the aim to better understand why the intervention worked for some children but failed for others so that future interventions can be better informed. This study was not part of the original planned trial. Clearly, this is not a perfect design, but given the time it took to develop the protocol, to obtain ethical approval, and to coordinate field work, we had strived to ensure a rigorous and timely implementation. During the stage of data analyses, we carefully assessed the validity of the SRB score using the change in salt intake measured by the gold-standard method, i.e. two consecutive 24-hour urinary sodium excretions. We also performed the sensitivity analysis using change in sodium intake as the outcome, which showed consistent findings, suggesting the robustness of the findings. Nevertheless, this is a limitation of our study, and we have acknowledged this in the discussion section (Line 26-28 on Page 13 and line 1-3 on Page 14).
2.Line 6, data analysis section: additional adjustment of BMI too across quartile SRB score Response: We thank the reviewer for this relevant comment. Our primary analysis was performed among all 603 students within the 14 classes. We didn't adjust for BMI (body mass index) because weight and height were measured only for a subsample of 135 children who collected 24-hour urine.
To account for the potential confounding effect of body weight, we performed the following analyses. First, the SRB score was validated by the change in salt intake from baseline to the end of trial measured by 24h urine, with the adjustment of age sex and BMI. The SRB score was significantly associated with the reduction in salt intake (β=0.31, 95%CI: 0.06-0.57, P=0.016), suggesting that a higher SRB score was associated with a larger reduction in salt intake independent of baseline BMI. Second, in the sensitivity analysis using reduction in salt intake, measured by 24h urinary sodium, as the outcome among random sample of children, the association estimates were obtained after adjusting for age, gender and baseline BMI. These association showed consistent findings with the primary analysis, suggesting the robustness of the findings. We have clarified the relevant descriptions on the validation analyses (Line 27 on Page 6 in the method section; Line 13 on Page 10 in the result section) and the sensitivity analyses (Line 6-7 on Page 9 and supplementary table S1).

3.Line 53: Is patient ? a typo, perhaps participant
Response: We thank the reviewer for this comment. It is per the journal style that we should include a patient and public involvement statements. We agree with the reviewer that "participant involvement" is more appropriate for the current study, and we would like to change it accordingly if it follows the style of BMJ open.
4. Table 1& 3: why is it not possible to separate out mother from grandparents i.e. the 'others'; fathers was collected separately and information can then be provided about the importance or not of support from this family member. Is there a reason why mothers was not specifically examined? Can this be stated in the method. Particularly as this is also a discussion point at line 41, page 12.
Response: We are grateful to the reviewer for the very helpful comment. It is possible to separate the role of mother from grandparents and we have revised it accordingly. Now we present the data for mother, father, grandmother and grandfather separately throughout the manuscript, including table 1,  table 3, main text (Line 18-23 on Page 10) and Supplementary table S1. Previously we grouped the findings together based on the analyses for each family member, as we found that the association remains consistently significant with the largest magnitude for the support from the father, while the association for the support from mother, or grandparents attenuated substantially in the sensitivity analysis. We agree with the reviewer that we should separate these roles to provide enough information for the readers to understand the relationship in a clearer way. Given that the discrepancy in the association estimates could also be due to differential statistical power, we have toned down the related discussion and interpreted the findings with extra caution (Line 12-13 on Page 12).