Downloaded from http://bmjopen.bmj.com/ on June 16, 2017 - Published by group.bmj.com
PEER REVIEW HISTORY
BMJ Open publishes all reviews undertaken for accepted manuscripts. Reviewers are asked to
complete a checklist review form (http://bmjopen.bmj.com/site/about/resources/checklist.pdf) and
are provided with free text boxes to elaborate on their assessment. These free text comments are
reproduced below.
ARTICLE DETAILS
TITLE (PROVISIONAL)
AUTHORS
The interaction of socioeconomic position and type 2 diabetes
mellitus family history: a cross-sectional analysis of the Lifelines
Cohort and Biobank Study
van Zon, Sander; Snieder, Harold; Bultmann, Ute; Reijneveld,
Sijmen
VERSION 1 - REVIEW
REVIEWER
REVIEW RETURNED
GENERAL COMMENTS
Vera Tsenkova
University of Wisconsin-Madison, USA
31-Jan-2017
This is a well-written paper that examines the interplay between SEP
and family history of diabetes on risk for diabetes. Strengths include
a large dataset, biomarker-based diagnosis of diabetes, and a novel
research question.
On p. 4, Authors state: ―This interaction between low SEP and
T2DM family history may have large consequences for prevention
and clinical care.‖ Similarly, the authors repeatedly state that SEP is
modifiable but do not provide any specific ways in which information
from their study can be used to tailor intervention work. What is it
about low SEP (as indexed here by education), particularly in the
context of a genetic predisposition to diabetes, that is especially
harmful to glucoregulation?
The authors use a family history of diabetes classification that
includes siblings, which introduces bias inherent in different family
structures (e.g., number of siblings). I would suggest the authors
investigate whether their findings remain consistent if they create a
―parental history of diabetes‖ category. Given the large number of
cases, it will be beneficial to compare ―maternal history only‖ with
―paternal history only‖ and ―both parents history.‖
In order to stratify by gender, a significant effect modification by
gender is usually documented first. Was a significant interaction
documented in this paper? Relatedly, the justification for gender
differences is underdeveloped—why did the authors expect gender
differences in these associations and underlying mechanisms?
Minor edits.
In the manuscript, the formatting of commas, periods, and citations
is unusual: the in-text citations are placed after the period throughout
the manuscript. See example below.
Type 2 diabetes mellitus (T2DM) is a common chronic health
condition with an estimated prevalence of 9.8% in males and 9.2%
in females.[1,2]
Downloaded from http://bmjopen.bmj.com/ on June 16, 2017 - Published by group.bmj.com
REVIEWER
REVIEW RETURNED
GENERAL COMMENTS
Jo-Anne Manski-Nankervis
Department of General Practice
University of Melbourne
Australia
06-Feb-2017
Thank you for the opportunity to review this paper which raises an
interesting question – is it socioeconomic status or family history, or
an interaction between the two – which results in increased risk of
developing type 2 diabetes? I have some comments which the
authors may wish to consider
1. Measures and procedures, page 7/31, line 5 and Page 19/31, last
paragraph: Can the authors comment on whether they think other
groups may have been included in the study because of the use of
an A10B medication? For example, metformin may be used in
women with polycystic ovary syndrome or in people with prediabetes and GLP1 agonists in people with obesity, but not diabetes.
Similarly for hypertension (page 8/31) – some patients will be taking
antihypertensive medications for reasons other than hypertension.
eg. propranolol for migraine prophylaxis and ACE inhibitors for
microalbuminuria. Is there a reason that this wasn‘t cross-checked
with patient self report of hypertension as you have done for
diabetes classification?
2. Page 9/31 line 3: The authors state the analyses were adjusted
for ―age and age2.‖ For the general reader it would be good to
explain this – is it related to risk of diabetes with age being nonlinear?
3. Page 9/31, paragraph 2: I would prefer that a statistician comment
on the methodology described here. For the general reader it would
be helpful if you could comment how moderate SEP was accounted
for in the regression analysis (The model was specified as i = 1
when low SEP was present and 0 when high SEP was present –
what about moderate SEP? Was i=0.5?)
4. Missing data – The authors state that participants with missing
SEP data were excluded. Was there any missing data for the clinical
measures, and if so, how was this dealt with? Page 30/31 of
STROBE 14b which addresses this has not been completed.
5. Page 18/31, paragraph 1: The authors write ―Behavioural and
clinical risk factors, especially weight status, partly explained these
gender differences, as well as the associations underlying the
interaction in females.‖ By interaction you mean the interaction
between family history and SES? I think that this should be stated
for clarity.
6. Page 18/31, paragraph 2: I agree with the authors that further
research leading to the gender difference found is required. I am not
clear what the authors are referring when they say ―The gender
difference in this interaction effect may be due to females with a high
SEP and a family history of T2DM being better capable to adapt
their behaviour based on their familial predisposition.‖ What is the
evidence that women are more able to change their behaviour than
men – and what are the authors referring to by the term ―familial
predisposition‖? The comment that physiologic response to stress
may differ in men compared to women is clearer.
7. Page 19/31, paragraph 2: Have the authors considered whether
SES status may result in epigenetic changes (eg affecting gene
expression) rather than, or in addition to, a genetic risk score
combining multiple loci. (See Loi et al (2013) Public Health Ethics
6(2):142-153)
8. Are there any limitations to the authors using only education level
Downloaded from http://bmjopen.bmj.com/ on June 16, 2017 - Published by group.bmj.com
as a reflection of socioeconomic position? Were there any other
variables in Lifelines that could have been utilised?
VERSION 1 – AUTHOR RESPONSE
Reviewer: 1
Reviewer Name: Vera Tsenkova
Institution and Country: University of Wisconsin-Madison, USA Competing Interests: None declared.
COMMENT 1. This is a well-written paper that examines the interplay between SEP and family history
of diabetes on risk for diabetes. Strengths include a large dataset, biomarker-based diagnosis of
diabetes, and a novel research question.
[RESPONSE]
We thank the reviewer for the kind words regarding our manuscript.
COMMENT 2. On p. 4, Authors state: ―This interaction between low SEP and T2DM family history
may have large consequences for prevention and clinical care.‖ Similarly, the authors repeatedly state
that SEP is modifiable but do not provide any specific ways in which information from their study can
be used to tailor intervention work. What is it about low SEP (as indexed here by education),
particularly in the context of a genetic predisposition to diabetes, that is especially harmful to
glucoregulation?
[RESPONSE]
We thank the reviewer for this question. Our assumption/hypothesis is that the higher prevalence and
clustering of adverse health behaviours (e.g. low physical activity) and clinical risk factors (e.g.
obesity) for T2DM among people with a low SEP may aggravate the predisposition related to their
family history. Our findings show that this is indeed the case for females but not that much for males.
Hence, we may need to focus interventions at women with a low SEP and a family history of T2DM.
We agree with the reviewer that we could have been more specific regarding possible interventions.
We have therefore added some additional information to the Discussion section. The revised text is:
Discussion, page 24, lines 359-368
―Future studies need to investigate whether interventions targeted specifically at low SEP females
with T2DM family history contribute to a reduction of T2DM related morbidity and mortality. Prevention
and intervention strategies should include both individual (e.g. health behaviour) and contextual
factors (e.g. food availability). Prevention and intervention efforts should further take into account that
people with a low SEP often have low health literacy.46 Low health literacy may hamper the
translation of knowledge about risk factors into healthy behaviour. Future studies should further
examine contributing pathways for the interaction effect in females, as this may offer insight into which
behavioural or clinical risk factors are most important to target at in this specific group of women.‖
COMMENT 3. The authors use a family history of diabetes classification that includes siblings, which
introduces bias inherent in different family structures (e.g., number of siblings). I would suggest the
authors investigate whether their findings remain consistent if they create a ―parental history of
diabetes‖ category. Given the large number of cases, it will be beneficial to compare ―maternal history
only‖ with ―paternal history only‖ and ―both parents history.‖
[RESPONSE]
We thank the reviewer for raising this point. The classification for family history of type 2 diabetes
including siblings is in line with the many other studies that we have cited (Hilding 2006 and van ‗t Riet
2010, for example). We have performed additional analysis using ―parental history of type 2 diabetes‖
Downloaded from http://bmjopen.bmj.com/ on June 16, 2017 - Published by group.bmj.com
only (i.e. no family history through siblings or children). The results of these additional analyses were
basically the same, except for some small differences in the magnitude of the associations and
interactions. In addition, we have performed separate analyses for ―maternal‖ and ―paternal‖ type 2
diabetes histories. Again, the main findings were the same but the exact magnitudes of associations
and interactions differed. We have added information on these sensitivity analyses to the Methods
section, and have added the results to the Results section, including one additional table. We have
also performed additional analysis using ―both parents have type 2 diabetes histories‖. However, we
do not present the results of these analyses because the estimates of the associations and
interactions are rather inaccurate, because of a low T2DM prevalence using this categorization,
leading to wide confidence intervals. The added text is:
Methods, page 10, lines 199-201
―In sensitivity analyses, we examined whether the results were consistent if we based family history
on parental T2DM only (i.e. siblings and children not included in family history). We also repeated the
analyses for maternal and paternal T2DM separately.‖
Results, page 19, lines 267-273
―Sensitivity analyses showed that the main findings remain consistent when using different
categorizations for family history (Table 5). The first part of table 5 shows the results of repeating the
main analyses with family history based only on parental T2DM (i.e. siblings and children not included
in family history). Except for a small difference in the magnitude of associations and interactions,
results were virtually identical to the main analyses. The second part of Table 5 shows that findings
are also essentially similar when repeating the analyses for maternal and paternal T2DM separately.‖
COMMENT 4. In order to stratify by gender, a significant effect modification by gender is usually
documented first. Was a significant interaction documented in this paper? Relatedly, the justification
for gender differences is underdeveloped—why did the authors expect gender differences in these
associations and underlying mechanisms?
[RESPONSE]
We have now documented the p-value (p=0.001) for the interaction term. We further added to a
justification for the reason why gender differences in the examined associations and interaction may
be expected to the Introduction section. The revised text is:
Introduction, page 5, lines 79-85
―The possible aggravation of familial predisposition for T2DM by low SEP may differ between males
and females through fundamental biological gender differences in genes and hormones in the
development of T2DM.21 For example, some aspects of the control of metabolic homeostasis are
regulated differently in males and females.21 Gender differences in the aggravation of familial
predisposition for T2DM by low SEP may be important for the development of prevention and
intervention programs and therefore need to be taken into account.‖
Methods, page 8-9, lines 162-164
―Potential modification of gender was assessed by adding an interaction term to the model. As
significant modification was found (p = 0.001), all analyses were stratified by gender.‖
Minor edits.
COMMENT 5. In the manuscript, the formatting of commas, periods, and citations is unusual: the intext citations are placed after the period throughout the manuscript. See example below.
Type 2 diabetes mellitus (T2DM) is a common chronic health condition with an estimated prevalence
of 9.8% in males and 9.2% in females.[1,2]
Downloaded from http://bmjopen.bmj.com/ on June 16, 2017 - Published by group.bmj.com
[RESPONSE]
We thank the reviewer for noting. We have now changed the citations to the style of BMJ Open.
Reviewer: 2
Reviewer Name: Jo-Anne Manski-Nankervis
Institution and Country: Department of General Practice, University of Melbourne, Australia
Competing Interests: None declared
Thank you for the opportunity to review this paper which raises an interesting question – is it
socioeconomic status or family history, or an interaction between the two – which results in increased
risk of developing type 2 diabetes? I have some comments which the authors may wish to consider
COMMENT 1. Measures and procedures, page 7/31, line 5 and Page 19/31, last paragraph: Can the
authors comment on whether they think other groups may have been included in the study because of
the use of an A10B medication? For example, metformin may be used in women with polycystic ovary
syndrome or in people with pre-diabetes and GLP1 agonists in people with obesity, but not diabetes.
Similarly for hypertension (page 8/31) – some patients will be taking antihypertensive medications for
reasons other than hypertension. eg. propranolol for migraine prophylaxis and ACE inhibitors for
microalbuminuria. Is there a reason that this wasn‘t cross-checked with patient self report of
hypertension as you have done for diabetes classification?
[RESPONSE]
We thank the reviewer for this question. In Lifelines, chronic health conditions have been classified by
expert groups in each field. This includes T2DM and hypertension. For consistency, we have followed
these classifications for both T2DM and hypertension. We have added two references extensively
describing T2DM diagnosis (Amini 2017) and hypertension diagnosis (Amini 2017, Meems 2015). We
further added the possibility of misclassification as a limitation of the study to the Discussion section.
The revised text is:
Methods, page 6-7, lines 108-116
―Participants were categorized as having T2DM if they had a measured fasting plasma glucose (FPG)
≥7.0 mmol/L,23 and/or a measured glycated haemoglobin (HbA1c) ≥6.5% (48 mmol/mol),23 and/or
self-reported T2DM (i.e. ―Do you have diabetes mellitus?‖, ―If you have diabetes, what type of
diabetes do you have? With answer categories: type 1, type 2, other, I do not know) in combination
with self-reported medication use (i.e. only tablets, only insulin, tablets and insulin, only a diet), and/or
recorded T2DM medication use (i.e. anatomical therapeutic chemical (ATC) codes A10A and
A10B24). The classification of T2DM diagnosis was made by a group of experts and was also used in
previous studies that used Lifelines data.25,26‖
Methods, page 8, lines 152-155
―Participants with a measured BP >140/90 mm Hg30 and participants with recorded antihypertensive
medication (i.e. ATC codes C02, C03, C07, C08, C0924) were categorized as hypertensive.
Hypertension diagnosis was made by a group of experts and was also used in previous studies that
used Lifelines data.25,26‖
Discussion, page 23-24, lines 349-354
―Third, we categorized participants using medication with ATC codes A10A and A10B as having
T2DM. We might have diagnosed participants as having T2DM while they are actually being treated
for pre-diabetes, with metformin (A10B) for example. However, pharmaceutical treatment of prediabetes is uncommon in the Netherlands. General practitioners are advised to use life-style related
interventions in the treatment of pre-diabetes‖
Downloaded from http://bmjopen.bmj.com/ on June 16, 2017 - Published by group.bmj.com
COMMENT 2. Page 9/31 line 3: The authors state the analyses were adjusted for ―age and age2.‖ For
the general reader it would be good to explain this – is it related to risk of diabetes with age being
non-linear?
[RESPONSE]
Indeed, we adjusted the analyses for age and age2 because the logit of the risk for T2DM may have a
non-linear association with age. We have now added this to the method section. The revised text is:
Methods, page 9, lines 166-168
―These and following analyses were adjusted for age and age2. The age2 term was added because
the logit of the risk for T2DM may have a non-linear association with age.‖
COMMENT 3. Page 9/31, paragraph 2: I would prefer that a statistician comment on the methodology
described here. For the general reader it would be helpful if you could comment how moderate SEP
was accounted for in the regression analysis (The model was specified as i = 1 when low SEP was
present and 0 when high SEP was present – what about moderate SEP? Was i=0.5?)
[RESPONSE]
We thank the reviewer for pointing out the lack of clarity regarding the description of the statistical
analyses. Actually, we performed all analyses twice. In the first round, low SEP was compared with
high SEP. In the second round, medium SEP was compared with high SEP. For the calculation of the
RERI it is not possible to have more than two categories in one analysis. We have adjusted the text of
the Methods section to clarify this issue. The revised text is:
Methods, page 9, lines 173-176
―The model was specified as i = 1 when low SEP was present and 0 when high SEP was present, and
j = 1 when T2DM family history was present and 0 when T2DM family history was not present.
Participants with a medium SEP were thus not included in these analyses (i.e. for them, separate
analyses were performed).
Methods, page 9, lines 182-184
―All analyses were repeated with medium SEP as risk factor (i.e. comparing medium with high SEP,
with medium SEP coded as 1 and high SEP as 0).‖
COMMENT 4. Missing data – The authors state that participants with missing SEP data were
excluded. Was there any missing data for the clinical measures, and if so, how was this dealt with?
Page 30/31 of STROBE 14b which addresses this has not been completed.
[RESPONSE]
We thank the reviewer for noting this un-clarity. In the fully adjusted model (model 3), we had 12.8%
missing data. This was mainly due to missing values on behavioural risk factors that were obtained
from questionnaires (i.e. 4.0% on alcohol use, 8.0% on smoking, 6.2% on physical activity). We had
only 21 missing values (0.0%) on the clinical measures; these were obtained at the Lifelines research
centres, which highly reduced the number of missings. We have added information about these
missing data in the method and discussion sections. The revised text is:
Methods, page 10, lines 189-198
―Fifth, to examine whether behavioural and clinical risk factors explain the associations and
interactions of SEP and T2DM family history with T2DM, we adjusted the basic model including age
and age2 stepwise for behavioural and clinical risk factors, including smoking status, alcohol
consumption, and physical activity (model 1); and further for general weight status and abdominal
obesity (model 2), and finally also for hypertension (model 3). Participants with missing data on
Downloaded from http://bmjopen.bmj.com/ on June 16, 2017 - Published by group.bmj.com
behavioural or clinical risk factors were omitted from the analyses. In the final model we had 12.8%
missing data, mainly due to missing data on alcohol use (4.0%), smoking (8.0%), and physical activity
(6.2%), with for some participants having missing data on several behaviours. Data on clinical risk
factors was mostly complete (i.e. only 21 (0.0%) missing values).‖
Discussion, page 24, lines 354-356
―Fourth, we had some missing data on behavioural risk factors. This may have resulted in an underadjustment of the associations and interactions of SEP and T2DM family history with T2DM.‖
COMMENT 5. Page 18/31, paragraph 1: The authors write ―Behavioural and clinical risk factors,
especially weight status, partly explained these gender differences, as well as the associations
underlying the interaction in females.‖ By interaction you mean the interaction between family history
and SES? I think that this should be stated for clarity.
[RESPONSE]
Indeed, we meant the interaction between family history and SEP. We have revised the text to clarify
this issue. The revised text is:
Discussion, page 21, lines 284-286
―Behavioural and clinical risk factors, especially weight status, partly explained these gender
differences, as well as the associations underlying the interaction between low SEP and T2DM family
history in females.‖
COMMENT 6. Page 18/31, paragraph 2: I agree with the authors that further research leading to the
gender difference found is required. I am not clear what the authors are referring when they say ―The
gender difference in this interaction effect may be due to females with a high SEP and a family history
of T2DM being better capable to adapt their behaviour based on their familial predisposition.‖ What is
the evidence that women are more able to change their behaviour than men – and what are the
authors referring to by the term ―familial predisposition‖? The comment that physiologic response to
stress may differ in men compared to women is clearer.
[RESPONSE]
We agree with the reviewer that our explanation for the gender difference in interaction effect was
somewhat unclear. We have therefore revised the text. The revised text is:
Discussion, page 21, lines 293-301
―A possible explanation for the gender difference in this interaction effect may be that men and
women have different coping styles in response to the knowledge of having a family history of T2DM.
In general, women are more likely to engage in problem-focused coping styles.36 Females with a
medium or high SEP and a family history of T2DM may therefore make better use of the knowledge of
having a T2DM family history than their male counterparts in adapting their preventive behaviour.
However, the pathways leading to this gender-difference definitely deserve further study.‖
COMMENT 7. Page 19/31, paragraph 2: Have the authors considered whether SES status may result
in epigenetic changes (eg affecting gene expression) rather than, or in addition to, a genetic risk score
combining multiple loci. (See Loi et al (2013) Public Health Ethics 6(2):142-153)
[RESPONSE]
We thank the reviewer for this interesting remark. We did not consider this option but we agree with
the reviewer that it is an interesting topic for future research. We have therefore added information on
this topic to the Discussion section. The revised text is:
Downloaded from http://bmjopen.bmj.com/ on June 16, 2017 - Published by group.bmj.com
Discussion, pages 23, lines 330-336
―Although T2DM family history and a GRS for T2DM both measure T2DM predisposition, future
studies should examine if, and to what degree, the use of a GRS would lead to different results
compared to this study. In addition, future studies should focus on epigenetic changes related to low
SEP as these may play a role in T2DM development.43 For example, living in a deprived area has
been related to global DNA methylation which in turn has been associated with inflammation markers
and cardiovascular disease.44 Epigenetic changes related to low SEP might also play a role in T2DM
development.‖
COMMENT 8. Are there any limitations to the authors using only education level as a reflection of
socioeconomic position? Were there any other variables in Lifelines that could have been utilised?
[RESPONSE]
We thank the reviewer for this question. It is well-known indeed that the magnitude of associations
between different indicators of socioeconomic position (e.g. income, occupational status) and health
outcomes may differ. However, educational level is least sensitive for reverse causation which is of
relevance given the cross-sectional nature of our study. In general, the highest educational level is
achieved long before type 2 diabetes develops. Moreover, education has been shown to best capture
the effects of socioeconomic position in studies on other chronic diseases in the Netherlands, e.g. in
chronic kidney disease (Vart 2013). We have added some lines to the Discussion section on the
possible difference in results by indicator of socioeconomic position. The revised text is:
Discussion, page 22, lines 312-323
―A systematic review on the relation between SEP and T2DM incidence showed that people with low
SEP have an increased risk for T2DM, irrespective of the SEP indicator used.8 Although our findings
might have differed somewhat for other indicators of SEP, it is likely that the main conclusions would
be similar because all indicators for SEP measure a certain underlying social stratification.38
Moreover, education has been shown to best capture the effects of SEP in studies on other chronic
diseases in the Netherlands, e.g. in chronic kidney disease.39 T2DM may also affect SEP through a
decrease in work participation, absenteeism, and early retirement,40 but such a reverse causation is
unlikely to explain our findings. Our measure of SEP, educational level, has usually been established
long before T2DM develops and is therefore the SEP indicator least sensitive for reverse causation.
This is important given the cross-sectional nature of our study.
VERSION 2 – REVIEW
REVIEWER
REVIEW RETURNED
Vera Tsenkova
University of Wisconsin-Madison, USA
06-Mar-2017
GENERAL COMMENTS
The authors did an outstanding job of answering all concerns raised
by Editor and reviewers.
REVIEWER
Jo-Anne Manski-Nankervis
Department of General Practice, University of Melbourne, Australia
12-Mar-2017
REVIEW RETURNED
GENERAL COMMENTS
Thank you for the opportunity to review this paper again after the
revisions made by the authors. I feel that my comments have all
been addressed and recommend that the paper be accepted for
publication.
Downloaded from http://bmjopen.bmj.com/ on June 16, 2017 - Published by group.bmj.com
The interaction of socioeconomic position
and type 2 diabetes mellitus family history: a
cross-sectional analysis of the Lifelines
Cohort and Biobank Study
Sander K.R. van Zon, Harold Snieder, Ute Bültmann and Sijmen A.
Reijneveld
BMJ Open 2017 7:
doi: 10.1136/bmjopen-2016-015275
Updated information and services can be found at:
http://bmjopen.bmj.com/content/7/4/e015275
These include:
References
This article cites 43 articles, 14 of which you can access for free at:
http://bmjopen.bmj.com/content/7/4/e015275#BIBL
Open Access
This is an Open Access article distributed in accordance with the Creative
Commons Attribution Non Commercial (CC BY-NC 4.0) license, which
permits others to distribute, remix, adapt, build upon this work
non-commercially, and license their derivative works on different terms,
provided the original work is properly cited and the use is
non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/
Email alerting
service
Receive free email alerts when new articles cite this article. Sign up in the
box at the top right corner of the online article.
Notes
To request permissions go to:
http://group.bmj.com/group/rights-licensing/permissions
To order reprints go to:
http://journals.bmj.com/cgi/reprintform
To subscribe to BMJ go to:
http://group.bmj.com/subscribe/