Childhood Obesity and Maternal Personality Traits: A New Point of View on Obesity Behavioural Aspects
, Luigi Vetri 1,3, Pierluigi Marzuillo 4,*, Valentina Lanzara 1, Ilaria Bitetti 1, Margherita Siciliano 1, Emanuele Miraglia del Giudice 4, Maria Esposito 1, Nicola Santoro 5,6 and Marco Carotenuto 1Round 1
Reviewer 1 Report
I congratulate the authors for the extensive work developed to produce this paper. I insert my comment to improve the work:
1 Line 38-40, Use a more incisive language (e.g. delete 'may').
2 Line 87, Please detail the meaning of Type A.
3 Line 90-91, Please briefly explain the MMPI-2 scores used.
4 Tables There is an error in Table numeration because 2 tables 1 are present. In the table's legends insert the meaning of NS =not significant.
5 Line 175 please, insert the square brackets for reference 10.
Author Response
I congratulate the authors for the extensive work developed to produce this paper. I insert my comment to improve the work:
A: We thanks Reviewer 1 for the appreciation of our manuscript and study design.
1 Line 38-40, Use a more incisive language (e.g. delete 'may').
A: We deleted the term.
2 Line 87, Please detail the meaning of Type A.
A: We specified the Type A meaning.
3 Line 90-91, Please briefly explain the MMPI-2 scores used.
A: We detailed the MMPI-2 scores used.
4 Tables There is an error in Table numeration because 2 tables 1 are present. In the table's legends insert the meaning of NS =not significant.
A: We corrected the errors.
5 Line 175 please, insert the square brackets for reference 10.
A: We modified the text as requested.
Reviewer 2 Report
The present study investigates personality characteristics of obese children’s mothers by comparing them with the personalities of lean children’s mothers. The authors find higher scores of obese children’s mothers on many dimensions of the MMPI-2. This study focuses on an important topic. However, I have some concerns with this paper in its present form.
There is a clearer description of the sample needed. What is a “third-level Endocrinologic Center” (l. 60) and why were mothers enrolled there although they and their children were not overweight?
l. 62 reads as though obese mothers were excluded from the sample of 258 mothers. However, they had been excluded beforehand and a sample of 258 remained. This sentence should be reformulated to make this clear.
ll. 103-107 should be moved to the description of the sample. What does, for example, “+ 0.9“ mean? Is it the standard deviation? If so, please state (and it should then be +/- 0.9). What is BMI-SDS?
Maybe the most important concern is that it is unclear to which extent the two groups of mothers were really comparable. How strictly were they matched? Only few covariates seem to have been considered. How was matching on age, BMI, and education exactly performed? Was a statistical approach such as propensity score matching used? If so, please report its results.
I also wondered why no comparisons with the MMPI standard values were made. A comparison between obese children’s mothers and females in the norm sample might also allow making meaningful statements. Although comparisons with norm samples have their own drawbacks, they might be fruitful in this study, especially if the comparability of both recruited groups is unclear.
The Method section might be divided in subsections, one describing the sample, one describing the instrument, and one describing the analyses.
The MMPI should be illustrated in more detail for readers who might not be familiar with it. For example, there is not one Clinical Basis Scale, but ten (i.e., the ten scales the authors describe). Denoting them as one scale might be misleading (see also l. 108). How were the Content Scales derived and what do they mean? What are the L scale, the F scale, etc.? Readers who are not familiar with this instrument cannot know that. A short explanation would be helpful.
How was mothers’ education measured?
The statistical analysis can also be improved. As the mothers were matched, the question is why the authors applied t-tests for independent samples. Normally, matched data are dependent samples and should be treated as such. But that might also depend on how the authors assured that the mothers really matched (see above).
Further, as a lot of statistical tests were performed, type I error accumulation is a problem here. Bonferroni correction should be applied to adjust the nominal alpha level. This is all the more important as the authors approach seems to be quite exploratory. Reading Tables 1 and 2 (by the way, Table 2 is denoted as Table 1), I noticed that the authors stated to report Bonferroni-corrected p values. However, the p values are the p values, Bonferroni correction just cannot change them. Rather, the nominal alpha level is corrected, and the authors do not report the adjusted alpha level. This leads to questions. For example, in Table 1 a p = .0193 seems to be denoted as significant. However, if we apply Bonferroni correction just to the ten statistical tests related to the Clinical Basis Scales, then the adjusted alpha is .005 and a p = .0193 should not be denoted as significant. The same is true for Table 2, where even 15 test results are reported. Taken all statistical tests together, alpha = .05 should be divided by 25, resulting in an adjusted alpha = .002.
In addition, effect sizes (e.g., Cohen’s d) should be reported. Without having effect sizes reported, the practical significance of the differences is hard to judge.
Tables 1 and 2 can be improved. Why not include the full scale names in the tables instead in the title and just delete the abbreviations? Please also clearly indicate which value is the mean and which is the standard deviation (I guess it is the standard deviation and not the standard error or confidence interval. However, readers should not be forced to guess). The t-values and degrees of freedom should be reported, too.
The authors conclude that their findings suggest that mothers’ personality could play a role for their children’s obesity. I do not say that this reasoning is implausible, but the design of this study heavily limits such conclusions and, although the authors discuss some limitations, I missed a critical discussion of this point, which is in my opinion the most important one. The design is quasi-experimental in nature, and quasi-experiments always lack internal validity. They do especially so with decreasing number of control variables, and this study has a considerable paucity of variables that might have been used for matching purposes. The authors controlled for mothers’ weight, age, and education (although it is unclear how education was operationalized). However, there might be a myriad of additional variables that might affect children’s weight (e.g., parenting style, children’s physical and mental health, television consumption, peer relationships, etc.), and these variables might also be related to mother’s personality (mothers’ education might tap some of them, but certainly not all). The interplay between those variables is probably very complex, and just picking one of them while controlling for only a few of them is certainly problematic. The authors can probably not change that, but this point should at least be extensively discussed.
l. 131-132: How are “self- esteem, body image, and social mobility impairment in obese children” related to the present study? Why is this aspect mentioned here?
Minor comments:
- 44, 51, 135: I would not denote studies published between 2007 and 2010 as “recent”.
- 104-107: Please be consistent in number of decimals reported throughout the manuscript.
Author Response
The present study investigates personality characteristics of obese children’s mothers by comparing them with the personalities of lean children’s mothers. The authors find higher scores of obese children’s mothers on many dimensions of the MMPI-2. This study focuses on an important topic. However, I have some concerns with this paper in its present form.
A: We thanks Reviewer 2 for the appreciation of our manuscript and study design. We thank Reviewer 2 for the valued suggestions.
There is a clearer description of the sample needed. What is a “third-level Endocrinologic Center” (l. 60) and why were mothers enrolled there although they and their children were not overweight?
A: We thank Reviewer 2 for the invaluable suggestion. “Third-level Endocrinologic Center” corresponds to a university endocrinology center. There are many clinical reasons for consulting a third level endocrinology center dedicated to the developmental age. In fact, we specified the recruitment modality by pointing out that the mothers of non-obese children were recruited in the Pediatric Unit in which the Endocrinologic Center for Childhood Obesity is specific section of the clinic service. Therefore, mothers of non-obese children were recruited among the outpatients and in-patients for pediatric problems different from obesity or overweight (i.e. gastrointestinal, growth, problems etc.). Finally, in developmental age for systemic pathologies such as obesity or overweight (but not only) it is mandatory to take overall care of the family and the patient in order to improve the outcome as much as possible. For this reason, it was considered important to carry out the personality survey of the mothers of the children recruited. We specified and detailed the recruitment modality in the Materials and Methods section.
- 62 reads as though obese mothers were excluded from the sample of 258 mothers. However, they had been excluded beforehand and a sample of 258 remained. This sentence should be reformulated to make this clear.
A: We selected only non-obese women in order to exclude other putative confounding factors. We specified the inclusion and exclusion criteria in the Participants section.
- 103-107 should be moved to the description of the sample. What does, for example, “+ 0.9“ mean? Is it the standard deviation? If so, please state (and it should then be +/- 0.9). What is BMI-SDS?
A: We are sorry for the mistakes that we corrected. We specified in the Participants section the terms BMI-SDS indicated the standard deviation BMI score (BMI-SDS).
Maybe the most important concern is that it is unclear to which extent the two groups of mothers were really comparable. How strictly were they matched? Only few covariates seem to have been considered. How was matching on age, BMI, and education exactly performed? Was a statistical approach such as propensity score matching used? If so, please report its results.
A: We are sorry we were not clear. The groups were not matched but two independent groups comparable for demographics. In this revised version of the manuscript, we better specified statistical analysis and results.
I also wondered why no comparisons with the MMPI standard values were made. A comparison between obese children’s mothers and females in the norm sample might also allow making meaningful statements. Although comparisons with norm samples have their own drawbacks, they might be fruitful in this study, especially if the comparability of both recruited groups is unclear.
A: We thank the reviewer for the request. The time devoted to recruitment made it possible to obtain an absolutely comparable sample between the two groups. It is suggested that a comparison be made between mothers and females of obese children in the normal sample. This procedure would not be feasible because the MMPI-A test (for example) cannot be administered before the age of 14. Above all, we do not find a reason to carry out this type of assessment to children, especially considering that self-administered personality tests do not exist in pre-pubertal age. Finally, just as the Reviewer 2 correctly points out, we deliberately did not use the reference values of the MMPI-2 test, preferring to carry out a case-control study.
The Method section might be divided in subsections, one describing the sample, one describing the instrument, and one describing the analyses.
A: We thank Reviewer 2 for the suggestion. We divided the Method section in the following subsections: Participants; Methods; Ethical statement.
The MMPI should be illustrated in more detail for readers who might not be familiar with it. For example, there is not one Clinical Basis Scale, but ten (i.e., the ten scales the authors describe). Denoting them as one scale might be misleading (see also l. 108). How were the Content Scales derived and what do they mean? What are the L scale, the F scale, etc.? Readers who are not familiar with this instrument cannot know that. A short explanation would be helpful.
A: We thank the Reviewer for the valued suggestions. We added the suggested changes.
How was mothers’ education measured?
A: We measured the maternal education level according with the scholarly years.
The statistical analysis can also be improved. As the mothers were matched, the question is why the authors applied t-tests for independent samples. Normally, matched data are dependent samples and should be treated as such. But that might also depend on how the authors assured that the mothers really matched (see above).
A: Following the reviewer suggestion, the statistical analysis has been greatly implemented and clarified through the tables. The groups were not paired but two independent groups. We are sorry for this blunder.
Further, as a lot of statistical tests were performed, type I error accumulation is a problem here. Bonferroni correction should be applied to adjust the nominal alpha level. This is all the more important as the authors approach seems to be quite exploratory. Reading Tables 1 and 2 (by the way, Table 2 is denoted as Table 1), I noticed that the authors stated to report Bonferroni-corrected p values. However, the p values are the p values, Bonferroni correction just cannot change them. Rather, the nominal alpha level is corrected, and the authors do not report the adjusted alpha level. This leads to questions. For example, in Table 1 a p = .0193 seems to be denoted as significant. However, if we apply Bonferroni correction just to the ten statistical tests related to the Clinical Basis Scales, then the adjusted alpha is .005 and a p = .0193 should not be denoted as significant. The same is true for Table 2, where even 15 test results are reported. Taken all statistical tests together, alpha = .05 should be divided by 15, resulting in an adjusted alpha = .002.
A: We thank the reviewer for raising this point. Following the reviewer’s suggestion, to reduce the chances of obtaining false-positive results (Type I errors) when multiple pairwise tests were performed on a single set of data, Bonferroni corrections (Bonferroni type adjustment) were applied by dividing the p value by the number of comparisons being made.
In addition, effect sizes (e.g., Cohen’s d) should be reported. Without having effect sizes reported, the practical significance of the differences is hard to judge.
A: We agree with the reviewer. Without having the effect sizes reported the practical significance of the differences is hard to judge. So, following the reviewer suggestion, we reported the effect sizes.
Tables 1 and 2 can be improved. Why not include the full-scale names in the tables instead in the title and just delete the abbreviations? Please also clearly indicate which value is the mean and which is the standard deviation (I guess it is the standard deviation and not the standard error or confidence interval. However, readers should not be forced to guess). The t-values and degrees of freedom should be reported, too.
A: We modified the Tables 1 and 2 as requested. We thank Reviewer for the valued suggestion.
The authors conclude that their findings suggest that mothers’ personality could play a role for their children’s obesity. I do not say that this reasoning is implausible, but the design of this study heavily limits such conclusions and, although the authors discuss some limitations, I missed a critical discussion of this point, which is in my opinion the most important one. The design is quasi-experimental in nature, and quasi-experiments always lack internal validity. They do especially so with decreasing number of control variables, and this study has a considerable paucity of variables that might have been used for matching purposes. The authors controlled for mothers’ weight, age, and education (although it is unclear how education was operationalized). However, there might be a myriad of additional variables that might affect children’s weight (e.g., parenting style, children’s physical and mental health, television consumption, peer relationships, etc.), and these variables might also be related to mother’s personality (mothers’ education might tap some of them, but certainly not all). The interplay between those variables is probably very complex, and just picking one of them while controlling for only a few of them is certainly problematic. The authors can probably not change that, but this point should at least be extensively discussed.
A: We thank Reviewer for the suggestion. We modified the conclusion accordingly to the real conceptual dimension of the present study.
- 131-132: How are “self- esteem, body image, and social mobility impairment in obese children” related to the present study? Why is this aspect mentioned here?
A: We added a sentence to clarify the concept.
Minor comments:
- 44, 51, 135: I would not denote studies published between 2007 and 2010 as “recent”.
A: We corrected the text as suggested.
- 104-107: Please be consistent in number of decimals reported throughout the manuscript.
A: We corrected the text as suggested.
Reviewer 3 Report
Interesting approach to evaluating the serious issue of what contributes to development of obesity in childhood and focusing on maternal personality trait. However, The MMPI-2 does not evaluate habits, does not evaluate the home environment. Implying that "peculiar" personality profile on the MMPI-2, that is not intended to be used as you are using it, "could play a key role in the development of obesity in children" is not scientifically supported by your study. Since when a personality profile causes behavior? Recommend a science based psychologist to advice you next time.
Author Response
Interesting approach to evaluating the serious issue of what contributes to development of obesity in childhood and focusing on maternal personality trait. However, The MMPI-2 does not evaluate habits, does not evaluate the home environment. Implying that "peculiar" personality profile on the MMPI-2, that is not intended to be used as you are using it, "could play a key role in the development of obesity in children" is not scientifically supported by your study. Since when a personality profile causes behavior? Recommend a science based psychologist to advice you next time.
A: We thank Reviewer 3 for his valued comments.
We rewrite all the sentences that could be reason of misunderstanding or of weak association.
We thank again for the suggestion for next similar personality study.
Round 2
Reviewer 3 Report
This manuscript has too many methodological problems and potential sources of error.
Author Response
Thank you for your comment. We extensively revised the text and the statistical analysis.
Round 3
Reviewer 3 Report
This is a much improved manuscript that documents and explains methods, procedures and findings much better. I still have concerns about the use of the MMPI-2 since it evaluates pathology and then sees obesity as a pathology induced problem as opposed to and social/environmental/cultural and behavior based problem. Not all obesity is directly related to psychopathology. These issues are addressed but it may need further clarification. Also many paragraphs in the manuscript seem to long and may benefit from further editing and breaking them into shorter paragraphs.
Author Response
We thank Reviewer 3 for the precious suggestions.
We deleted the ambiguous sentences and shortened paragaphs.
Moreover, we clarified the real weight of our findings, especially at lines 63-65, 199-202, 208-213, 271-273, 275-277 of the new version of the manuscript.
