Joint Bone Spine 79 (2012) 291–297

Original article

Body mass index and susceptibility to knee osteoarthritis: A systematic review

and meta-analysis
Liying Jiang a , Wenjing Tian a , Yingchen Wang b , Jiesheng Rong c , Chundan Bao a ,
Yupeng Liu a , Yashuang Zhao a,∗ , Chaoxu Wang d,1
a
Department of Epidemiology, Public Health College, Harbin Medical University, Harbin, Heilongjiang Province, PR China
b
Department of Hygienic Microbiology, Public Health College, Harbin Medical University, Harbin, Heilongjiang Province, PR China
c
Department of Orthopedics Surgery, The Second Affiliated Hospital of Harbin Medical University, Harbin, Heilongjiang Province, PR China
d
Department of Nutrition and Food Higiene, Public Health College, Harbin Medical University, Harbin, Heilongjiang Province, PR China

a r t i c l e i n f o a b s t r a c t

Article history: Objective: Excess bodyweight, expressed as increased body mass index, is associated with osteoarthritis
Accepted 16 May 2011 risk, especially in weight bearing joints. However, the strength of the association was inconsistent. The
Available online 30 July 2011 study was conducted to quantitatively assess the association between body mass index and the risk
of knee osteoarthritis and investigate the difference of the strength stratified by sex, study type and
Keywords: osteoarthritis definition.
Obesity Methods: We used published guidelines of the Meta-analysis of Observational Studies in Epidemiology
Osteoarthritis
Group (MOOSE) to perform the meta-analysis. The search strategy employed included computerized
Meta-analysis
bibliographic searches of MEDLINE, PubMed, EMBASE, The Cochran Library and references of published
manuscripts. Study-specific incremental estimates were standardized to determine the risk of knee
osteoarthritis associated with a 5 kg/m2 increase in BMI.
Results: Twenty-one studies were included in the study. The results showed that body mass index was
significantly positive associated with osteoarthritis risk in knee site. A 5-unit increase in body mass index
was associated with an 35% increased risk of knee osteoarthritis (RR: 1.35; 95%CI: 1.21, 1.51). Magnitude
of the association was significantly stronger in women than that in men with significant difference (men,
RR: 1.22; 95%CI: 1.19, 1.25; women, RR: 1.38; 95%CI: 1.23, 1.54; p = 0.04). The summary effect size was
1.25(95%CI: 1.18, 1.32) in case-control studies and 1.37 (95%CI: 1.19, 1.56) in cohort studies (p = 0.28).
Body mass index was positively associated with knee osteoarthritis defined by radiography and/or clinical
symptom (RR: 1.25, 95%CI: 1.17, 1.35) and clinical surgery (RR: 1.54, 95%CI: 1.29, 1.83). The latter tended
to be stronger than the former (p < 0.01).
Conclusion: Increased body mass index contribute to a substantially increased risk of knee OA. The
magnitude of the association varies by sex and OA definition.
© 2011 Société française de rhumatologie. Published by Elsevier Masson SAS. All rights reserved.

1. Introduction Obesity has become a global problem leading to excess morbid-

ity and mortality. According to the latest World Health Organization
Osteoarthritis (OA) is the most common joint disease and is (WHO) report, more than 1.6 billion adults (aged 15 years old)
one of the most prevalent symptomatic health problems for older are overweight [3]. Obesity has drawn interest in recent studies
individuals. The etiology of OA is multifactorial, including inflam- because of its modifiable status and its association with OA. There is
matory factors, metabolic factors, and mechanical factors [1]. OA considerable evidence indicating that obesity plausibly represents
have incurred substantial disease burden and influenced the quality one of the most important risk factors for particular peripheral joint
of daily life for the elderly. Knee joint is the most clinically affected sites, predominantly the knee site and the hip site [4–11].
site, and knee OA is the main indication for a large number of knee Standard therapeutic modalities can alleviate symptoms and
replacement surgeries performed annually [2]. improve function but cannot alter the disease process [10].
Therefore, much attention has been invested in improving the
recognition of epidemiology and in elucidating possible factors pre-
disposing to OA development. Mechanical stress resulting from a
∗ Corresponding author. Tel.: +86-0451-87502823; fax: +86-0451-87502881.
high body mass index (BMI) is known to be a risk factor for the
E-mail addresses: zhao [email protected], [email protected]
(Y. Zhao), [email protected] (C. Wang). development of knee OA, and better understanding of the positive
1
Tel: +86-0451-87502870; Fax: +86-0451-87502881. effect of obesity on OA development is likely to be valuable in the

1297-319X/$ – see front matter © 2011 Société française de rhumatologie. Published by Elsevier Masson SAS. All rights reserved.
doi:10.1016/j.jbspin.2011.05.015
292 L. Jiang et al. / Joint Bone Spine 79 (2012) 291–297

campaign against osteoarthritis. To our knowledge, many epidemi- reference or expressed as a slope per incremental BMI, covariates
ologic studies have been performed to investigate the association controlled in the study.
between obesity with knee OA, and obesity is an unequivocal risk For measuring quality of observational studies no universal scale
factor for the onset and progression of knee OA [9,12,13]. However, was available, therefore, we followed the guidelines of the Meta-
the link between obesity and the risk of knee OA lack quantitatively analysis of Observational Studies in Epidemiology (MOOSE) group
assessment. and assessed the quality of the included studies based on the fol-
It is probable that once sufficient information is provided lowing criteria [14]:
through systematic studies, we will be able to obtain a better under-
standing of the cause of OA as a whole and perhaps develop new • clear description of the population and setting;
interventions to lower the prevalence of obesity and knee OA in • sufficient sample size;
the future. Therefore, we conducted a meta-analysis and meta- • appropriate measurement of outcomes;
regression to quantitatively clarify the association between BMI, • response rate of the original study was higher than 80%;
as a measure of overall obesity, and the risk of developing knee OA • completeness of follow-up or questionnaires;
based on available studies to date. • appropriate consideration and adjustment for potential con-
founders;
• knee OA was assessed identical in studied population;
2. Methods
• exposure was measured identical for cases and controls;
• information on completers vs. withdrawals;
2.1. Search strategy and inclusion criteria
• exact statistical and analytical methods.
The search strategy included computerized bibliographic
searches of PubMed (1966–2010), EMBASE (1974–2010) and refer- Throughout this assessment, when information about a specific
ences of published manuscripts. The Cochrane Library of Systematic item was consistent with the criteria, it was scored as “1”; other-
Reviews also was queried. Other websites searched included wise, it was scored as “0”. Studies that fulfilled more than eight
National Institute for Clinical Excellence, National Electronic criteria were considered high quality (i.e., scored ≥ 8).
Library for Health- musculoskeletal specialist library, Arthritis
Research Campaign, Arthritis Care, Arthritis and Musculoskeletal 2.3. Statistical analysis
Alliance, and Arthritis Foundation National Office. The search terms
included obesity, overweight, BMI, adiposity, arthritis, osteoarthri- Meta-analysis was performed for each individual study using
tis and OA. Studies in humans of the association between BMI and Stata 8.0 (Stata Corp, College Station, TX). Briefly, the analysis
knee OA were included if they met the following criteria: software produced forrest plots as a schematic description of the
meta-analysis results. Summary random effect estimates were
reported using pooled odds ratios (OR), and 95% confidence inter-
• the study was English-language articles;
vals (CI) that were calculated around each summary effect estimate.
• the study was of a cohort, or a (nested) case-control study
The random effect model assumed that included study was a ran-
reported the association with corresponding confidence intervals
dom sample of a hypothetical population of study.
(95%CI) for at least three categories of exposure (BMI), or must
Study-specific OR and corresponding 95%CI was chosen rather
provide sufficient data to estimate them. Height and weight or
than weight to explore the association because of the wide variation
BMI must be provided in those studies;
in average weight and height across races. The cut-off for expo-
• the study was on primary or idiopathic knee OA, while not sec-
sure categories varied between studies included in the study. In
ondary OA
order to keep all studies on a coincident scale, category-specific risk
• the study was on the relationship between obesity and knee joint
was transformed into risk estimates associated with per 5 kg/m2
replacements.
increments in BMI for each study instead of per 1 kg/m2 incre-
ment, which would be impractical and non-significant. Therefore,
No specific limits of ethnicity, the type of knee OA (femoro- for practical reasons per 5 kg/m2 increments in BMI has been used
tibial and patello-femoral OA), and the site of knee OA (unilateral as a surrogate for the change of bodyweight. These estimates were
or bilateral knee OA) were set. calculated on the assumption of a linear trend between the natural
The titles and abstracts were scanned to exclude studies that logarithm of OR and increased BMI for different exposure cate-
were clearly irrelevant. The full texts of the remaining articles were gory [15]. Scaled OR was estimated using the method described
read to determine whether they contained information about our by Greenland [16]. As most studies have reported effect size and
interest. In addition, the electronic searches were supplemented their corresponding 95%CI, we used these data as summary statis-
by scanning reference lists from retrieved articles to identify addi- tics for each study. The value assigned to each BMI category was the
tional studies. For studies with same population resources or mid-point for closed categories. For open-ended categories (e.g.,
over-lapping datasets, the most complete one was included. The BMI > 30), we estimated the median (assuming a normal distri-
excluded studies included reviews, conference abstracts, letters to bution for BMI) [17]. Unless special statement, the adjusted risk
editors and studies on the progression of OA. estimate from each study was used.
Heterogeneity testing using Q statistics was performed to eval-
2.2. Data extraction and quality assessment uate variance between studies [18]. If the between-study variance
was large enough to make the test of heterogeneity significant
Data were extracted using a standardized data extraction form (P < 0.05), random effects models were considered most appropri-
and the information of studies was assessed by two investigators ate. As these tests of heterogeneity were relatively insensitive, a
independently. Discrepancies were resolved by discussion and more conservative P-value of less than 0.10 was used. Finally, the
repeated examination of the included articles The following infor- potential publication bias was examined with funnel plots, which
mation was used for each study: The first author’s name, year of plot the log of effect sizes against the inverse standard error of the
publication, country where the study was performed, participant effect size. The Egger’s regression test was used to test funnel plot
population, study design, OA definition/assessment, risk estimates asymmetry [19]. A P-value < 0.05 was considered significant for
with corresponding 95%CI either with one BMI category as the publication bias. We also conducted sensitivity analysis omitting
 L. Jiang et al. / Joint Bone Spine 79 (2012) 291–297 293

each study to assess whether result was influenced excessively by

a single study [20].

3. Results

3.1. Studies included in the meta-analysis

The primary literature search had identified 414 potentially

relevant studies. After review of these titles and abstracts, we
retrieved full articles for further assessment. A total of 44 stud-
ies have been identified. Twenty-three studies were excluded for
unavailable or incomplete data [7,21–42]. Finally, 21 unique studies
were available for this meta-analysis, including 15 cohort studies
[9,11–13,43–53] and six case-control studies [54–59]. The number
of included studies whose definition of OA was based on clinical
surgery was six studies [13,46,50,55,57,59], on radiography seven
studies [11,12,45,48,51,53,58], on radiography and clinical symp-
tom six studies [9,43,44,52,54,56], and on self-reported two studies
[47,49]. Fig. 1 shows the search and selection process. Table 1 sum-
Fig. 2. The forest plot of the association between BMI and knee OA using random-
marizes the characteristics of included studies. effects model.

3.2. Meta-analysis of increased BMI and the risk of knee OA

The analysis involved 872 717 subjects from 18 studies. A sum- for Q test; I2 < 0.001% for heterogeneity) in men with sig-
mary of meta-analysis for the association between obesity and nificant difference (P = 0.04) (Table 2). Examined case-control
susceptibility to knee OA was shown in Fig. 2. A 5-unit increase studies and cohort studies separately, we found that the sum-
in BMI was significantly related to an increased risk of knee OA mary association was 1.25(95%CI: 1.18, 1.32; P = 0.15 for Q test;
(RR: 1.35; 95%CI: 1.21, 1.51; P < 0.001 for Q test; I2 = 99.20% for het- I2 = 44.40% for heterogeneity) in case-control studies and 1.37
erogeneity). This suggested that every 5-unit increase in BMI was (95%CI: 1.19, 1.56; P < 0.001 for Q test; I2 = 99.40% for heterogeneity)
associated with a 35% increased risk of knee OA. in cohort studies with no statistical significant difference (P = 0.28)
We also carried subgroup analysis to examine whether the (Table 2). To assess the association in more details, we further
association was different because of stratification by sex, study conducted the analysis according to the definition of OA (clin-
design and definition of knee OA. For different sex, the associa- ical surgery versus clinical symptom or radiography or clinical
tion was 1.38(95% CI: 1.23–1.54; P < 0.001 for Q test; I2 = 97.00% symptom + radiography). The end-point was precisely categorized
for heterogeneity) in women and 1.22 (95%CI: 1.19, 1.25; P = 0.71 into non-clinical surgery (radiography and/or clinical symptom)
and clinical surgery. The association was 1.25(95%CI: 1.17, 1.35;
P < 0.001 for Q test; I2 = 93.20% for heterogeneity) defined by radiog-
raphy and/or clinical symptom and 1.54(95%CI: 1.29, 1.83; P < 0.001
Articles identified after citation (n=414) for Q test; I2 = 98.60% for heterogeneity) defined by clinical surgery,
respectively. The latter tended to give larger effect than the former
(P < 0.01) (Table 2).
Excluded by screening of titles and/or abstracts (n=210)
3.3. Heterogeneity and publication bias
Potentially relevant articles identified and screened for retrieval (n=204)
Significant heterogeneity between studies was noted in cohort
Excluded due to no related factors we studies, while moderate heterogeneity in case-control studies. It
studied (n=160) was difficult to correlate the funnel plot, which is usually used to
detect publication bias, because the number of studies included in
Potentially appropriate articles to be included in this the analysis was relatively small. No publication bias was detected
meta-analysis (n=44) by Egger’s regression (P = 0.95), but Begger methods had tested
bias opposing to Egger’s regression (P = 0.03). Trim and fill anal-
ysis showed no study might have been missing. Influence analysis
articles excluded after review of full text produced similar summary risk estimates and did not affect the
(n=23): strength of evidence for increased BMI and the risk of knee OA.
1. the effect on OA the progression of
(n=2) .
2. study on prognostic factors (n=1). 4. Discussion
3. special population (n=2).
4. twin study on genetic factors(n=1). The meta-analysis shows that the risk of knee OA increase with
5. cross-sectional study (n=1) BMI and a dose-response relationship exist. A 5-unit increase in
6. No category, OR and 95%CI, or BMI is significantly associated with an increased risk of developing
number of cases and controls to knee OA (RR: 1.35; 95%CI: 1.21, 1.51). That is, every 5-unit increase
calculate (n=16). in BMI is associated with a 35% increased risk of knee OA. Our find-
ings support the notion that there is a positive association between
21 articles included in this meta-analysis increased BMI and the risk of knee OA. Meanwhile, our results pro-
vide a quantitative risk estimates and a continuous scale that is a
Fig. 1. Flow chart of article selection. reliably steady result. BMI is analyzed on the continuous scale, and
 294
Table 1
Characteristics of studies on obesity and knee osteoarthritis risk.

Author, year Country Study participants (% of Study design Quality of No. cases Mean age of Definition OA site Adjustments
(ref.) women) study patients (SD)

Hart et al., England 1003 women aged 49–60 (100) Prospective 9 95 55.30 (5.60) Radiographya Knee Social classic, smoking, ERT,
1999 [11] cohort study hysterectomy, knee pain, physical
activity

Coggon et al., England 1050 men and women Case control 8 525 NA Radiography and Knee Age, gender
2001 [56] aged ≥ 45 (60.95) study clinical diagnosisb
Grotle et al., Norway 1675 men and women aged Prospective 9 114 NA Self-reported (based on Knee Age, gender, work, leisure time
2008 [49] 24–66 (56.30) cohort study clinical diagnosis) activities

Cooper et al., England 354 men and women aged ≥ 55 Prospective 9 45 NA Radiographyc Knee Age, gender, knee pain
2000[45] (72.03) cohort study
Reijman et al., Netherlands 1372 men and women Prospective 9 76 66.30 (6.70) Radiographyc Knee Age, gender
2007 [48] aged ≥ 55 (5.5) cohort study
Sandmark Sweden 1173 men and women aged Case control 8 625 NA Clinical surgeryd TKR Smoking, hormones, physical loads,
et al., 1999 [55] 55–70 (49.79) study housework, sports activities

Jarvholm et al., Sweden 320192 men aged 15–67 (0) Prospective 9 502 NA Clinical surgeryd TKR Age, smoking
2005 [46] cohort study

L. Jiang et al. / Joint Bone Spine 79 (2012) 291–297

Oliveria et al., Unites states 268 women aged 20–89 (100) Case control 8 134 60.8 Radiography and Knee Estrogen use, smoking, health care
1999 [54] study diagnosise
Niu et al., 2009 United states 2623 men and women aged Prospective 9 163 NA Radiographyc Knee Age, gender, race, bone mineral
[51] 54–70 (59.40) cohort study density, knee injury

Liu et al., 2007 England 490532 women aged 50–69 Prospective 9 974 NA Self-reported (based on TKR Age, region of recruitment, deprivation
[47] (100) cohort study clinical surgery) index

Holmberg Sweden 1650 men and women aged Case control 8 825 NA Radiographyc Knee Heredity, knee injury, smoking, sports
et al., 2005 [58] 51–70 (57.20) study activity

Wang et al., Australia 41528 men and women aged Prospective 9 541 60.30 (6.80) Clinically surgeryd TKR Age, gender, country of birth, education
2009 [13] 25–75 (58.95) cohort study
Felson et al., United states 217men and 381women Prospective 9 93 NA Radiography and Knee Age, gender, smoking, physical activity,
1997 [9] (63.71) cohort study diagnosisc knee injury, weight change

Hochberg et al., United states 298 men and 139 women Prospective 9 NA NA Radiographyc Knee Age, gender, smoking
2004 [12] aged ≥ 20 (31.81) cohort study
Franklin et al., Sweden 2576 men and women aged Case control 8 431 73.70 Clinical surgeryd TKR Age, occupation
2009 [59] 63–92 (57.10) study
Lohmander Sweden 11026 men and 16934 women Prospective 9 471 NA Clinical surgeryd TKA Age, gender, smoking, physical activity
et al., 2009 [50] aged 45–73 (60.57) cohort study
Manninen Finland 6647 men and women aged Prospective 9 126 NA Radiography and Knee
et al., 1996 [43] 40–64 (54.10) cohort study diagnosisc
Manninen Finland 805 men and women aged Case control 8 281 68.87 Clinical surgeryd TKR Age, gender
et al., 2002 [57] 55–75 (75.78) study
Shiozaki et al., Japan 1191 women aged 40–65 (100) Prospective 9 NA NA Radiography and Knee Physical exercise, knee injury
1999 [44] cohort study diagnosisc
Toivanen AT Finland 8000 men and women Prospective 9 94 NA Radiography and Knee Age, gender
et al., 2010 [52] aged ≥ 30 (54.54) cohort study diagnosisc
Nishimura A Japan 261 men and women aged ≥ 65 Prospective 9 57 70.80 (20.40) Radiographyc Knee Age, gender, BMI, knee pain,
et al., 2010 [53] (63.22) cohort study osteoporosis, Heberden’s nodes

TKR: total knee replacement; TKA: knee arthroplasty due to osteoarthritis; NA: no available
a
Radiography: Kellgren and Lawrence index of grade ≥ 1.
b
Listed for total anthropathy.
c
Radiography: Kellgren and Lawrence index of grade ≥ 2.
d
The first clinical surgery for primary osteoarthritis.
e
Radiography-based definition suggested by the American College of Rheumatology.
 L. Jiang et al. / Joint Bone Spine 79 (2012) 291–297 295

Table 2 way of BMI report. Begg methods considered that this test had
Summary effect sizes of knee osteoarthritis per 5-unit increase in BMI by sex, study
larger test performance when the number of involved studies were
type and OA definition.
more than 75, while moderate ability when 25 studies. Egger et al.
Subgroup No. of studies Odds ratios (95%CI) Heterogeneity used linear regression method to detect the symmetry of funnel
p for Q test I2 (%) plot, and its statistics performance was higher than Begg methods.
So we used trimming and filling analysis to further evaluate publi-
Female 10 1.38 (1.23–1.54) < 0.001 97.00
Male 7 1.22 (1.19–1.25) 0.71 < 0.001 cation bias. Comprehension of trimming and filling result did not lie
Case control study 4 1.25 (1.18–1.32) 0.15 44.40 in how many studies should be included, and the result remained
Cohort study 14 1.37 (1.19–1.56) < 0.001 99.40 similar after being adjusted by trimming and filling analysis. Influ-
Subtotal 18 1.35 (1.21–1.51) < 0.001 99.20 ence analysis suggested that no single studies affected significantly
Surgery 4 1.54 (1.29–1.83) < 0.001 98.60
No surgery 14 1.25 (1.17–1.35) < 0.001 93.20
the summary estimate.
Subtotal 18 1.35 (1.21–1.51) < 0.001 99.20 There are some limitations considered in this study. First, smok-
ing seems to be a major potential confounder [29]. However, we
could not determine the effect of smoking because few studies were
stratified by smoking status. Second, radiographic OA cases based
effect size is converted to be per five units to reflect both over- on traditional criteria using Kellgren-Lawrence scales are generally
weight BMI of 27 against normal BMI of 22 and obese BMI of 32 defined. Alternative classification scales for OA have been used [11].
against overweight BMI of 27. It is also possible that any point of Studies using both the traditional Kellgren-Lawrence and newer
categories for BMI among normal distribution of population can modified individual scales might account for the variation. Third,
interpret the increased risk of knee OA. And, the magnitude of the specific occupational groups whose jobs requiring repetitious tasks
association was slightly stronger in women than men with signif- show high predisposition to developing OA in repetitively used
icant difference in our meta-analysis. A meta-analysis evaluated joints [25,39]. BMI, defined by anthropometric measures or self-
data on the relationship between obesity and the risk of knee OA reported, may contribute to different results. All these could be
[60]. However, the report provided no obvious evidence as a contin- additional bias and need further research. Finally, some impor-
uous scale for risk estimates of knee OA. Compared to Blagojevic’s tant confounders have not been measured with sufficient precision,
published meta-analysis on observational studies, we used strict or have not been measured at all in original studies. Since BMI
diagnostic criteria for knee OA in order to determine true asso- seemed to predict slightly stronger for bilateral OA than unilat-
ciation between the exposure and the outcome, and provided an eral OA [43]. The positive association of BMI with bilateral knee
obvious evidence of dose-response relationship as a continuous OA was even more obvious. Unfortunately, few studies analyzed
scale. such a differentiation [43]. We could not rule out the possibility
Meta-analysis of observational studies are prone to bias, and of those potential confounders and brought about heterogene-
confounding factors that are inherent in original studies brought ity. In addition, anthropometric measures other than BMI, such
about different extent of bias. Case control studies are more prone as waist-to-hip ratio and waist circumference, also serve as better
to recall bias such as recalling BMI. In case-control studies patients measures of adiposity. However, few studies provided such infor-
with symptomatic knee OA might exaggerate the risk of developing mation to permit comprehensive analysis of associations across
OA associated with prior weight because they are more likely to studies.
overestimate earlier body weight. We thus restricted our analysis WHO endorses the use of BMI (kilogram per square
to cohort studies and case control studies separately. However, the meter) to define obesity as BMI ≥ 30 kg/m2 and overweight as
magnitude of the association remained similar after being stratified BMI ≥ 25 kg/m2 . Anthropometric measurement or self-reported of
by different study type. This indicated that there was few recall BMI at baseline could not represent that of changes during follow-
bias. up. If BMI is generally obtained from the baseline year, the analyses
OA is generally diagnosed by radiography and clinical symp- do not allow for that a large period of time have elapsed between
tom. Radiographic OA is defined by X-ray based on traditional the ascertainment of exposure and the development of outcome.
criteria of Kellgren-Lawrence scales (graded 0–4, where 0 = none; However, fewer articles reported the association between BMI ear-
1 = possible osteophytes only; 2 = definite osteophytes and possible lier in life and the risk of OA in the life course [58]. Moreover, the
joint space narrowing; 3 = moderate osteophytes and/or definite prevalence of OA increases with age, which is further aggravated by
joint space narrowing; and 4 = large osteophytes, severe joint space obesity. Risk estimates assessed by BMI at age 18 years were signif-
narrowing, and/or bony sclerosis). Clinical OA (Symptomatic OA) is icantly greater than that by “recent” BMI. Crucial questions remain
diagnosed on the basis of disease histories, symptoms and clini- unresolved about the cumulative effect of excess bodyweight over
cal findings according to standard criteria evaluated by specially several decades, the effect of key weight-change periods in the life
trained physicians. The validity of study depends on the accu- course of individuals, and interactions with other potential risk
racy of case definition, which may be one of the reasons that factors.
risk estimates are different across studies. In fact, OA is diffi- Although with some limitations, our meta-analysis provides an
cult to assess because it combines both symptoms and pathologic obvious evidence of dose-response relationship as a continuous
changes, the latter is usually recognized by radiography. Radiologi- scale for the association between increased BMI with the risk of
cal grounds probably mean more sensitivity and include mild cases, developing knee OA. We have transformed category-specific risk
whereas joint replacement surgery indicates severe end-stage OA. estimates into a quantitative risk estimate associated with every
Some papers involved in the study have reported the association 5 kg/m2 increase in BMI. Most importantly, we have modeled a
based on different OA definition. More and larger studies that 5 kg/m2 increase in BMI, and the result may be considerably prac-
report BMI-OA risk associations separately for OA definition are tical and theoretically important. Obesity is particularly important
anticipated. for developing OA in weight-bearing joints, and modifying obesity
We found significant heterogeneity in our meta-analysis, can prevent osteoarthritis-related pain and disability. The find-
whereas it did not reduce substantially in cohort studies and case ing further supports the significance of obesity in the etiology of
control studies after subgroup was conducted. This may be because osteoarthritis. The prevention of obesity is important, as it has
of the variations in OA definition, different population, diverse con- adverse effects on health. Also, change of body weight is useful and
founders, the irregularities of category-specific information and the feasible.
296 L. Jiang et al. / Joint Bone Spine 79 (2012) 291–297

Disclosure of interests [26] Cicuttini FM, Spector T, Baker J. Risk factors for osteoarthritis in the tibiofemoral
and patellofemoral joints of the knee. J Rheumatol 1997;24:1164–7.
[27] Sahlstrom A, Montgomery F. Risk analysis of occupational factors influencing
The authors declare that they have no conflicts of interest con- the development of arthrosis of the knee. Eur J Epidemiol 1997;13:675–9.
cerning this article. [28] Cheng Y, Macera CA, Davis DR, et al. Physical activity and self-reported,
physician-diagnosed osteoarthritis: is physical activity a risk factor? J Clin
Epidemiol 2000;53:315–22.
Acknowledgements [29] Lau EC, Cooper C, Lam D, et al. Factors associated with osteoarthritis of the
hip and knee in Hong Kong Chinese: obesity, joint injury, and occupational
The study was supported by Heilongjiang Province office of edu- activities. Am J Epidemiol 2000;152:855–62.
[30] Sturmer T, Gunther KP, Brenner H. Obesity, overweight and patterns of
cation foundation (No. 11531166). The study received generous osteoarthritis: the Ulm Osteoarthritis Study. J Clin Epidemiol 2000;53:307–13.
assistance from Yan Liu in Public Health College in Harbin Medical [31] Sutton AJ, Muir KR, Mockett S, et al. A case-control study to investigate the
University. relation between low and moderate levels of physical activity and osteoarthritis
of the knee using data collected as part of the Allied Dunbar National Fitness
Survey. Ann Rheum Dis 2001;60:756–64.
References [32] Lachance L, Sowers MF, Jamadar D, et al. The natural history of emergent
osteoarthritis of the knee in women. Osteoarthritis Cartilage 2002;10:849–54.
[1] Felson DT, Lawrence RC, Dieppe PA, et al. Osteoarthritis: new insights. Part 1: [33] Dawson J, Juszczak E, Thorogood M, et al. An investigation of risk factors for
the disease and its risk factors. Ann Intern Med 2000;133:635–46. symptomatic osteoarthritis of the knee in women using a life course approach.
[2] Davis MA, Ettinger WH, Neuhaus JM. The role of metabolic factors and blood J Epidemiol Community Health 2003;57:823–30.
pressure in the association of obesity with osteoarthritis of the knee. J Rheuma- [34] Hootman JM, Macera CA, Helmick CG, et al. Influence of physical activity-related
tol 1988;15:1827–32. joint stress on the risk of self-reported hip/knee osteoarthritis: a new method
[3] WHO. See (http://www.whoint/mediacentre/factsheets/fs311/en/indexhtml). to quantify physical activity. Prev Med 2003;36:636–44.
[4] Felson DT, Anderson JJ, Naimark A, et al. Obesity and knee osteoarthritis. The [35] Manek NJ, Hart D, Spector TD, et al. The association of body mass index and
Framingham Study. Ann Intern Med 1988;109:18–24. osteoarthritis of the knee joint: an examination of genetic and environmental
[5] Davis MA, Ettinger WH, Neuhaus JM. Obesity and osteoarthritis of the knee: influences. Arthritis Rheum 2003;48:1024–9.
evidence from the National Health and Nutrition Examination Survey (NHANES [36] Sayer AA, Poole J, Cox V, et al. Weight from birth to 53 years: a longitu-
I). Semin Arthritis Rheum 1990;20:34–41. dinal study of the influence on clinical hand osteoarthritis. Arthritis Rheum
[6] Bagge E, Bjelle A, Eden S, et al. Factors associated with radiographic osteoarthri- 2003;48:1030–3.
tis: results from the population study 70-year-old people in Goteborg. J [37] Yoshimura N, Nishioka S, Kinoshita H, et al. Risk factors for knee osteoarthritis
Rheumatol 1991;18:1218–22. in Japanese women: heavy weight, previous joint injuries, and occupational
[7] Schouten JS, van den Ouweland FA, Valkenburg HA. A 12 year follow up study in activities. J Rheumatol 2004;31:157–62.
the general population on prognostic factors of cartilage loss in osteoarthritis [38] Szoeke CE, Cicuttini FM, Guthrie JR, et al. Factors affecting the prevalence of
of the knee. Ann Rheum Dis 1992;51:932–7. osteoarthritis in healthy middle-aged women: data from the longitudinal Mel-
[8] Hart DJ, Spector TD. The relationship of obesity, fat distribution and osteoarthri- bourne Women’s Midlife Health Project. Bone 2006;39:1149–55.
tis in women in the general population: the Chingford Study. J Rheumatol [39] Tangtrakulwanich B, Geater AF, Chongsuvivatwong V. Prevalence, patterns,
1993;20:331–5. and risk factors of knee osteoarthritis in Thai monks. J Orthop Sci 2006;11:
[9] Felson DT, Zhang Y, Hannan MT, et al. Risk factors for incident radiographic 439–45.
knee osteoarthritis in the elderly: the Framingham Study. Arthritis Rheum [40] Thelin N, Holmberg S, Thelin A. Knee injuries account for the sports-related
1997;40:728–33. increased risk of knee osteoarthritis. Scand J Med Sci Sports 2006;16:
[10] Felson DT, Lawrence RC, Hochberg MC, et al. Osteoarthritis: new insights. Part 329–33.
2: treatment Approaches. Ann Intern Med 2000;133:726–37. [41] Yoshimura N, Kinoshita H, Hori N, et al. Risk factors for knee osteoarthritis in
[11] Hart DJ, Doyle DV, Spector TD. Incidence and risk factors for radiographic knee Japanese men: a case-control study. Mod Rheumatol 2006;16:24–9.
osteoarthritis in middle-aged women: the Chingford Study. Arthritis Rheum [42] Harms S, Larson R, Sahmoun AE, et al. Obesity increases the likelihood of total
1999;42:17–24. joint replacement surgery among younger adults. Int Orthop 2007;31:23–6.
[12] Hochberg MC, Lethbridge-Cejku M, Tobin JD. Bone mineral density and [43] Manninen P, Riihimaki H, Heliovaara M, et al. Overweight, gender and knee
osteoarthritis: data from the Baltimore Longitudinal Study of Aging. osteoarthritis. Int J Obes Relat Metab Disord 1996;20:595–7.
Osteoarthritis Cartilage 2004;12:S45–8. [44] Shiozaki H, Kogab Y, Omori G, et al. Obesity and osteoarthritis of the knee
[13] Wang Y, Simpson JA, Wluka AE, et al. Relationship between body adiposity in women: results from the Matsudai Knee Osteoarthritis Survey. The Knee
measures and risk of primary knee and hip replacement for osteoarthritis: a 1999;6:189–92.
prospective cohort study. Arthritis Res Ther 2009;11:R31. [45] Cooper C, Snow S, McAlindon TE, et al. Risk factors for the inci-
[14] Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of observational studies in dence and progression of radiographic knee osteoarthritis. Arthritis Rheum
epidemiology: a proposal for reporting. Meta-analysis Of Observational Studies 2000;43:995–1000.
in Epidemiology (MOOSE) group. JAMA 2000;283:2008–12. [46] Jarvholm B, Lewold S, Malchau H, et al. Age, bodyweight, smoking habits and
[15] Orsini N, Bellocco RSG. Generalized least squares for trend estimation of sum- the risk of severe osteoarthritis in the hip and knee in men. Eur J Epidemiol
marized dose-response data. Stata J 2006;6:40–57. 2005;20:537–42.
[16] Greenland S, Longnecker MP. Methods for trend estimation from summa- [47] Liu B, Balkwill A, Banks E, et al. Relationship of height, weight and body
rized dose-response data, with applications to meta-analysis. Am J Epidemiol mass index to the risk of hip and knee replacements in middle-aged women.
1992;135:1301–9. Rheumatology (Oxford) 2007;46:861–7.
[17] Haftenberger M, Lahmann PH, Panico S, et al. Overweight, obesity and fat [48] Reijman M, Pols HA, Bergink AP, et al. Body mass index associated with onset
distribution in 50- to 64-year-old participants in the European Prospec- and progression of osteoarthritis of the knee but not of the hip: the Rotterdam
tive Investigation into Cancer and Nutrition (EPIC). Public Health Nutr Study. Ann Rheum Dis 2007;66:158–62.
2002;5:1147–62. [49] Grotle M, Hagen KB, Natvig B, et al. Obesity and osteoarthritis in knee, hip
[18] Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat and/or hand: an epidemiological study in the general population with 10 years
Med 2002;21:1539–58. follow-up. BMC Musculoskelet Disord 2008;9:132.
[19] Egger M, Davey Smith G, Schneider M, et al. Bias in meta-analysis detected by [50] Lohmander LS, Gerhardsson de Verdier M, Rollof J, et al. Incidence of severe
a simple, graphical test. BMJ 1997;315:629–34. knee and hip osteoarthritis in relation to different measures of body mass: a
[20] Tobias A. Assessing the influence of a single study in meta-analysis. Stata Tech population-based prospective cohort study. Ann Rheum Dis 2009;68:490–6.
Bull 1999;47:15–7. [51] Niu J, Zhang YQ, Torner J, et al. Is obesity a risk factor for progressive radio-
[21] Anderson JJ, Felson DT. Factors associated with osteoarthritis of the knee in the graphic knee osteoarthritis? Arthritis Rheum 2009;61:329–35.
first national Health and Nutrition Examination Survey (HANES I). Evidence [52] Toivanen AT, Heliovaara M, Impivaara O, et al. Obesity, physically demanding
for an association with overweight, race, and physical demands of work. Am J work and traumatic knee injury are major risk factors for knee osteoarthritis –
Epidemiol 1988;128:179–89. a population-based study with a follow-up of 22 years. Rheumatology (Oxford)
[22] Ettinger WH, Davis MA, Neuhaus JM, et al. Long-term physical functioning in 2010;49:308–14.
persons with knee osteoarthritis from NHANES. I: Effects of comorbid medical [53] Nishimura A, Hasegawa M, Kato K, et al. Risk factors for the incidence and
conditions. J Clin Epidemiol 1994;47:809–15. progression of radiographic osteoarthritis of the knee among Japanese. Inter-
[23] Spector TD, Hart DJ, Doyle DV. Incidence and progression of osteoarthritis in national Orthopaedics 2010.
women with unilateral knee disease in the general population: the effect of [54] Oliveria SA, Felson DT, Cirillo PA, et al. Body weight, body mass index, and
obesity. Ann Rheum Dis 1994;53:565–8. incident symptomatic osteoarthritis of the hand, hip, and knee. Epidemiology
[24] Hochberg MC, Lethbridge-Cejku M, Scott Jr WW, et al. The association of 1999;10:161–6.
body weight, body fatness and body fat distribution with osteoarthritis of [55] Sandmark H, Hogstedt C, Lewold S, et al. Osteoarthrosis of the knee in men and
the knee: data from the Baltimore Longitudinal Study of Aging. J Rheumatol women in association with overweight, smoking, and hormone therapy. Ann
1995;22:488–93. Rheum Dis 1999;58:151–5.
[25] Kujala UM, Kettunen J, Paananen H, et al. Knee osteoarthritis in former runners, [56] Coggon D, Reading I, Croft P, et al. Knee osteoarthritis and obesity. Int J Obes
soccer players, weight lifters, and shooters. Arthritis Rheum 1995;38:539–46. Relat Metab Disord 2001;25:622–7.
 L. Jiang et al. / Joint Bone Spine 79 (2012) 291–297 297

[57] Manninen P, Heliovaara M, Riihimaki H, et al. Physical workload and the [59] Franklin J, Ingvarsson T, Englund M, et al. Sex differences in the association
risk of severe knee osteoarthritis. Scand J Work Environ Health 2002;28: between body mass index and total hip or knee joint replacement resulting
25–32. from osteoarthritis. Ann Rheum Dis 2009;68:536–40.
[58] Holmberg S, Thelin A, Thelin N. Knee osteoarthritis and body mass [60] Blagojevic M, Jinks C, Jeffery A, et al. Risk factors for onset of osteoarthritis of
index: a population-based case-control study. Scand J Rheumatol 2005;34: the knee in older adults: a systematic review and meta-analysis. Osteoarthritis
59–64. Cartilage 2010;18:24–33.