The Arthritis Research Institute of America
300 S. Duncan Ave, Suite 188
Clearwater, FL 33755
(727) 461-4054
October 27, 2008
Joseph Dunckley
Assistant Editor
BMC-series journals
Re: MS 1838846535217525
Dear Mr. Dunckley,
Thank you for your comments on our manuscript The association of BMI and knee pain
among persons with radiographic knee osteoarthritis: A cross-sectional study. We did
implement your suggestion to use a copy editing service. We also have provided a point
by point response to all reviewers’ comments.
The ‘Statistics’ paragraph has been re-written and several references added (17, 22 – 27)
to address the concerns of the second reviewer regarding the Uniform Requirements.
You will also notice minor changes to the Discussion and Conclusion sections. The copy
editor caught a small error in that I had stated in the conclusion that we were the first to
look at this relationship. Actually, we are the first to look at in this manner (risk ratios
for each BMI category) but not the first to report results related to this question. I had
referenced three papers in the Background that had looked at this question in some
manner. I now mention this in the first part of the Discussion section and have removed
reference to it in the Conclusions.
Specific replies to reviewers’ comments are attached. Please let me know if anything
further is required for this revised submission.
Sincerely,
Matthew W. Rogers, MS
MS: 1838846535217525
The association of BMI and knee pain among persons with radiographic knee
osteoarthritis: A cross-sectional study Matthew W Rogers and Frances Vaughn Wilder
Reviewer: Max Reijman
Reviewer's report:
Major Compulsory revisions:
The study population consisted of subjects with radiographic signs of knee OA,
determined by the Kellgren & Lawrence criteria. (³ grade 2). As stated by the authors is
knee pain related with radiographic severity. So the relation between knee pain status
and BMI should also be adjusted for the severity of ROA of the knee.
Absolutely. This factor was included in the adjusted model. The revised statistics
now reflect the consideration of RKOA disease severity.
Minor essential revisions:
The authors reported that there is a positive linear relationship between the 4 categories
of BMI and pain. Did the authors test for linearity?
Another reviewer of this manuscript felt that linearity could be confusing in
reporting our analyses. The text has been re-worded as follows:
OLD: Unadjusted and adjusted odds ratios demonstrated a positive, linear
association between BMI group and pain for each successive BMI category.
REVISED: Unadjusted and adjusted odds ratios demonstrated a consistent, positive
risk for pain for each successive BMI category.
OLD: For the four BMI groups, a positive linear relationship with pain is
demonstrated in the unadjusted and adjusted analyses
REVISED: For successively higher BMI groups, a positive increase in the risk of
pain is demonstrated in the unadjusted and adjusted analyses
Footnote Table2: The adjusted odds ratio models should be: pain = bmi + age + gender
(+ severity of ROA)
The footnote now reads: BMI = Pain + age + gender + RKOA grade.
Figure 2 presented similar results as Table 2, so one can be deleted.
Only the old Table 2 (now Table 3) has been kept in the current revision.
Reviewer: Annette W-Dahl
Reviewer's report:
1. The main reason for major compulsory revision is the lack of description of the
statistical methods used, with reference to the Uniform Requirements. It’s not possible to
follow how the association of pain and BMI has been analyzed and thereby limited
possibility to interpret the findings.
Numerous changes have been made to our description of the statistical methods
used. We reviewed the Uniform Requirements section on statistical methods shown
below.
“IV.A.6.c. Statistics Describe statistical methods with enough detail to enable a
knowledgeable reader with access to the original data to verify the reported results.
When possible, quantify findings and present them with appropriate indicators of
measurement error or uncertainty (such as confidence intervals). Avoid relying
solely on statistical hypothesis testing, such as the use of P values, which fails to
convey important information about effect size. References for the design of the
study and statistical methods should be to standard works when possible (with
pages stated). Define statistical terms, abbreviations, and most symbols. Specify the
computer software used.”
The revised text below shows that we:
Described in more details our general statistical approach;
We noted our reporting of confidence internals for the odds ratios;
We stated clearly our cross-sectional study design and provided a reference.
Using our sample of subjects with RKOA, our cross-sectional study aimed to
quantify the association between BMI and knee pain among subjects with
radiographic evidence of knee OA;
We added clarity to the statistical tests and concepts noted and referenced
them in the text. This includes odds ratios, logistic regression, adjustment for
confounding adjustment, and the kappa statistic; and
We specified the computer software used, “SAS/STAT® software version 9.2
was used for all analyses.”
“Frequencies, percentages, and means were calculated to provide descriptive data
about our study sample. To estimate the strength of the association between BMI
and knee pain, we computed odds ratios (OR) [Kleinbaum DG, Kupper, LL, and
Morgenstern, H. Epidemiologic Research Principles and Quantitative Methods. John
Wiley & Sons, New York 1982, pp243-250.] using logistic regression [DawsonSaunders B, Trapp, RG. Basic and Clinical Biostatistics, 2nd ed. Norwalk, CN:
Appleton & Lange; 1994; 222-223; SAS Institute Inc., Logistic Regression Examples
Using the SAS System, Version 6, First Edition, Cary, NC; SAS Institute Inc., 1995.] .
Knee pain status (yes/no) was the outcome factor. With BMI as the exposure factor,
subjects with normal BMI (18.5 to 24.9 kg/m2) served as the referent group for the
analyses (e.g. Normal vs. Obese III). To address potential confounding when
assessing the association between BMI and knee pain, we ran an adjusted statistical
model (Pain = BMI + age + gender + RKOA grade). Logistic regression was used as
we had a binary outcome and several explanatory factors (e.g. gender) in the
statistical model [Walker, GA. Common Statistical Methods for Clinical Research with
SAS® Examples Cary, NC: SAS Institute Inc. 1997, pp 225-226.]. Confidence
internals were reported for the odds ratios. Every tenth subjects’ assembled
radiographs were independently interpreted by a non-affiliated radiologist blinded
to the results of the first reading. The inter-observer variability of x-ray
interpretations was calculated using the kappa coefficient [20] measuring the
amount of agreement that is above random chance. SAS/STAT® software version
9.2 was used for all analyses [SAS/STAT software, Version 9.2, SAS System for
Windows. Copyright © 2008. SAS Institute Inc. Cary, NC, USA.].”
a. The potential confounders, except age and gender, have only been described but not
analyzed.
We included the following information in the Results section, “Evaluation for
potential confounding showed no significant differences, by Pain Group, for
occupation (p=0.85), education (p=0.73), and self-reported disease status for
diabetes (p=0.63), heart disease (p=0.53), nor stroke (p=0.37). Although the mean
age was similar for each group, a significant group difference was evident for
gender (p < 0.01). Based on preliminary analyses, as well as consideration of known
factors related to OA, we built our adjusted model. BMI, age, gender, and disease
severity (as measured by OA grade) were included in the final adjusted model.
It might also be considered if there are other potential confounders that might be
interesting and necessary. For example grade of radiographic knee OA might be
interesting to include in the adjusted analyses.
RKOA disease severity, as measured by grade, was a solid addition to the study.
This factor was included in the adjusted model. The revised statistics now reflect
the consideration of RKOA disease severity. The final adjusted model is: BMI =
Pain + age + gender + RKOA grade.
Disease status is not presented in table 1 or elsewhere. Further it would be informative to
know how many of the patients in each BMI category who had pain and no pain
respectively.
Good point. This information has been added to the manuscript. Shown below, a
new table was created to display these data. (Note previous Table 2 is now Table 3).
Table 2
BMI Category and OA Grade counts, by pain status
among subjects (N = 576) with radiographic knee osteoarthritis a
Pain
No Pain
n
(N=329)
(N=247)
Normal BMI b
155
66
89
Pre-Obese
208
113
95
Obese I
119
75
44
Obese II
61
47
14
Obese III
33
28
5
RKOA Grade 2
331
155
176
RKOA Grade 3
150
100
50
RKOA Grade 4
95
74
21
a
Kellgren and Lawrence scale for radiographic osteoarthritis (Grades 2+).
b
BMI = body mass index (kg / m2) WHO Classification: Normal 18.5-24.9; Pre-Obese
25.0-29.9; Obese I 30.0-34.9; Obese II 35.0-39.9; and Obese III 40.0 +.
b. My guess is that ¨Pain/No pain¨ has been used as dependent variable in the regression
analysis and BMI category as independent variable as well as age and gender. However
table 2 says; BMI=pain+age+gender it would then rather be Pain= BMI+age+gender.
Our error. The footnote (for what is now Table 3) now reads: BMI = Pain + age +
gender + RKOA grade.
c. In the result part the authors write ¨a positive linear relationship with pain …..¨ .The
analyses are paired comparison and not linear regression which make the statement that
there is a linear relation not possible.
The text has been re-worded as follows:
OLD: Unadjusted and adjusted odds ratios demonstrated a positive, linear
association between BMI group and pain for each successive BMI category.
REVISED: Unadjusted and adjusted odds ratios demonstrated a consistent, positive
risk for pain for each successive BMI category.
OLD: For the four BMI groups, a positive linear relationship with pain is
demonstrated in the unadjusted and adjusted analyses
REVISED: For successively higher BMI groups, a positive increase in the risk of
pain is demonstrated in the unadjusted and adjusted analyses
2. There is a lack of definition/description that makes the purpose of the study unclear.
a. The use of the word ¨quantification¨ in the background of the abstract is confusing.
Quantification has two distinct meanings, one in mathematics and empirical science and
one in logic. Do the authors mean to quantify the association between pain and BMI
among RKAO individuals as stated as the aim of the study by reporting the OR (is OR 1.6
an expression of quantifying the association between pain and BMI?). However
transforming the continuous data of BMI to categorical data and analyze pain with
categorical data become then confusing. The choice of words should be considered and
well defined (major compulsory revision).
Our goal was communicate to the reader as clearly as possible. The purpose of the
study has been re-worded in the abstract and main text. In the main text it now
reads, “Our current study will help clarify these findings by reporting odds ratios to
further quantify the relationship between pain and BMI among persons with RKOA.
Using our sample of subjects with RKOA, our cross-sectional study [Hennekens CH,
Buring JE. Epidemiology in Medicine. Boston, MA and Toronto: Little, Brown and
Co. 1987; 108-112.] aimed to quantify the association between BMI and knee pain
among subjects with radiographic evidence of knee OA. “
b. The use of the term ¨ symptomatic¨ includes more than pain. Why not use the terms
¨pain¨ and ¨no pain¨ as it is pain that is of interest (minor essential revision).
Good point, thank you. The terms have been changed to Pain and No Pain
throughout the manuscript and tables.
3. The purpose of the study is not stated in the abstract (major compulsory revision).
Purpose has been more clearly stated in the abstract as follows: The purpose of the
study is to explore the relationship of BMI and knee pain among persons with RKOA.
Old: As researchers seek a clear profile of which factors contribute to knee pain,
quantification of the role of BMI is vital.
4. The background includes written errors and/or wrong use of words (minor essential
revision).
a. The cause of RKOA may be the cause of knee OA.
For this revision, the manuscript was sent to a copy editing service. (Comment 4a
appears to be a quoted typographical error, but we were not able to locate this
phrase in our manuscript.)
b. High-risk occupations of what?
“High-risk occupations” has been changed to “certain occupations” (as described in
the listed references; certain occupations have been linked to the development of
OA)
c. What is RKOA knee pain? Knee pain in individuals with RKOA?
Wording has been changed to include RKOA related knee pain
d. Considerations of the last sentence, page 3 ¨Subsequently ….¨ most patients seek
medical care due to pain and disabilities not due to RKOA which makes the clinicians
possibility to delay and prevent the onset of pain and impairment in these individuals
unrealistic. Further, it’s only few of the patients with knee OA who need knee
replacement, not ¨often¨ as written but rather in worse cases.
Thank you, this is a good point. More realistically clinicians (e.g. physicians,
physical therapists, fitness professionals) might counsel their overweight and obese
patients on weight loss (as one example of a modifiable risk factor, and relevant to
our study) to prevent or reduce knee pain. “Often” was changed to “may”. This
text was modified as follows: As there is no cure for OA on the horizon, it is important
to identify what factors influence the risk of symptomatic knee OA. Subsequently,
clinicians could guide their patients to take steps (e.g., weight loss) to prevent or delay
the onset of pain and impairments which may lead to work disability, daily living
disability, and joint replacement surgery.
e. In the background, the first part of page 4 includes partly author interpretations that
are irrelevant to referred studies. The authors’ logical connection of BMI as a cause of
knee OA becomes illogical when considers that even normal-weighted individuals suffer
from knee OA and the authors may distinguish between cause and risk factor.
Normal weight subjects report knee pain at lower rates vs. overweight and obese
persons. This paragraph has been modified as follows: It is known that persons with
higher body mass indexes (BMI; kg/m2) are more likely to report idiopathic knee pain
and accompanying disability in comparison to persons with normal BMI [11-13]. This
is suggestive of elevated BMI as a potential risk factor for knee pain among persons
with RKOA. Despite this logical connection, a paucity of research has been conducted
in this area.
Further the present study has scanty any possibility to clarify the relationship between
pain and BMI among patients with RKOA neither by reporting the OR for each BMI
category in a cross-sectional study nor with the very narrow perspective used (major
compulsory revision).
Per your recommendations as noted in 1a, and those of another reviewer, these
relationships have now been reported.
The result, discussion and conclusion sections are difficult to interpret as the methods of
statistics are scanty described as well as the purpose of the study is unclear. The authors
have an excellent cohort and together with a clear defined research question, with a
wider perspective of a systemic disease and
appropriate, well described statistical methods the paper has the potential to be
interesting and may contribute to increasing knowledge.
These issues have been addressed and modified in response to reviewers’ comments.
******