journal of

Electronic supplementary material:

The online version of this article contains supplementary material.
global health

Prevalence of rheumatoid arthritis in

VIEWPOINTS
low– and middle–income countries: A

PAPERS
systematic review and analysis

Igor Rudan1*, Simrita Sidhu1*,

Angeliki Papana2, Shi–Jiao Meng3, Background Rheumatoid arthritis (RA) is an autoimmune disorder
that affects the small joints of the body. It is one of the leading causes
Yu Xin–Wei3, Wei Wang3,4, Ruth M.
of chronic morbidity in high–income countries, but little is known
Campbell–Page5, Alessandro Rhyll
about the burden of this disease in low– and middle–income coun-
Demaio6,7, Harish Nair1, Devi Sridhar1, tries (LMIC).
Evropi Theodoratou1, Ben Dowman1,
Methods The aim of this study was to estimate the prevalence of RA
Davies Adeloye1, Azeem Majeed8,
in six of the World Health Organization's (WHO) regions that har-
Josip Car8,9, Harry Campbell1, Wei
bour LMIC by identifying all relevant studies in those regions. To
Wang 3,10, Kit Yee Chan1,11; Global accomplish this aim various bibliographic databases were searched:
Health Epidemiology Reference PubMed, EMBASE, Global Health, LILACS and the Chinese data-
Group (GHERG) bases CNKI and WanFang. Studies were selected based on pre–de-
fined inclusion criteria, including a definition of RA based on the
*
Joint first authorship 1987 revision of the American College of Rheumatology (ACR) def-
1
The University of Edinburgh Medical School,
Edinburgh, Scotland, UK
inition.
2
University of Macedonia, Thessaloniki, Greece Results Meta–estimates of regional RA prevalence rates for countries
3
School of Public Health, Capital Medical of low or middle income were 0.40% (95% CI: 0.23–0.57%) for
University, Beijing, China
4
Department of Integrated Early Childhood and Southeast Asian, 0.37% (95% CI: 0.23–0.51%) for Eastern Mediter-
Development, Capital Institute of Paediatrics, ranean, 0.62% (95% CI: 0.47–0.77%) for European, 1.25% (95%
Beijing, China CI: 0.64–1.86%) for American and 0.42% (95% CI: 0.30–0.53%)
5
The University of Toronto, Toronto, Canada for Western Pacific regions. A formal meta–analysis could not be per-
6
Copenhagen School of Global Health,
formed for the sub–Saharan African region due to limited data. Male
University of Copenhagen, Denmark
7
Harvard Global Equity Initiative, Harvard prevalence of RA in LMIC was 0.16% (95% CI: 0.11–0.20%) while
Medical School, Cambridge, USA the prevalence in women reached 0.75% (95% CI: 0.60–0.90%).
8
School of Public Health, Imperial College This difference between males and females was statistically signifi-
London, London, UK cant (P < 0.0001). The prevalence of RA did not differ significantly
9
Lee Kong Chian School of Medicine, Nanyang
Technological University, Singapore between urban and rural settings (P = 0.353). These prevalence esti-
10
Edith Cowan University, Perth, Australia mates represent 2.60 (95% CI: 1.85–3.34%) million male sufferers
11
School of Public Health, Peking University and 12.21 (95% CI: 9.78–14.67%) million female sufferers in LMIC
Health Science Centre, Beijing, China in the year 2000, and 3.16 (95% CI: 2.25–4.05%) million affected
males and 14.87 (95% CI: 11.91–17.86%) million affected females
in LMIC in the year 2010.
Correspondence to: Conclusion Given that majority of the world’s population resides in
Dr Kit Yee Chan LMIC, the number of affected people is substantial, with a projection
Centre for Population Health Sciences to increase in the coming years. Therefore, policy makers and health–
The University of Edinburgh Medical School
care providers need to plan to address a significant disease burden
Teviot Place, Edinburgh EH8 9AG
Scotland, UK both socially and economically.
kchan16@gmail.com

www.jogh.org • doi: 10.7189/jogh.05.010409 1 June 2015 • Vol. 5 No. 1 • 010409

 Rudan, Sidhu et al.

In recent years there has been a shift in diseases and health lation from a few studies is problematic given that there is
related challenges that the world is facing. Non–commu- ample evidence that RA is a variable disease in time and place
nicable diseases (NCD) have emerged as the leading cause [11]. Moreover, the burden of NCD has increased over the
of death worldwide, accounting for two–thirds of all deaths past decade in LMIC, while it has decreased in high–income
and deaths are projected to increase in the coming years countries [11]. RA also has a substantial economic impact,
[1]. Contrary to popular belief, these diseases are not lim- which can be quantified as direct (cost of medication, hos-
pital stay and visits, care–givers and helpers); indirect (loss
VIEWPOINTS

ited to the developed world; they are increasingly prevalent

in low– and middle–income countries (LMIC), which are of productivity from absenteeism or early retirement); and
intangible costs that are measured by the impact on quality
Papers

facing the double burden of both communicable and non–

of life [12,13]. In the United States, the direct cost of RA was
communicable diseases [2]. In LMIC, constrained health
approximately US$ 13 500 per affected person per year, and
care facilities, lack of resources and funds at individual and
indirect costs could range between US$ 1000 and US$
national level lead to limited treatment and support for
33 000 per affected person per year [14]. However, not much
NCD, which mainly affect the working age population with
is known about costs in the developing world [13].
a negative impact on household incomes and equity. The
high burden of NCD poses additional problems for LMIC, This paper aims to provide an estimate for the global and
creating a vicious cycle by worsening poverty that in turn regional burden of rheumatoid arthritis by systematically
results in a further rise of NCD [1]. Acknowledgment of reviewing relevant literature in both English databases and
the serious implications of the global burden of NCD has those in other languages; to study the variation in the prev-
alence of rheumatoid arthritis by gender, region and setting
led to an international response. The United Nations (UN)
(urban/rural); and to discuss the significance of these prev-
High–level meeting on NCD in 2011 addressed these is-
alence estimates and their implications for public health
sues and has paved the way for tackling them, by provid-
policy.
ing guidance on how to strengthen national capacities to
address NCD and integrate prevention and control activi-
ties across sectors and at all levels of governance and METHODS
health–care provision in LMIC [1,2].
Definition of population under study and
While there is wide recognition of the four main NCD with literature search
a major contribution to the global burden – cardiovascular
diseases, cancers, diabetes mellitus and chronic respiratory The World Bank database was referenced to compile a list
illnesses [2] – there are a large number of other NCD that of all the LMIC in the world [15]. Thereafter, all LMIC were
grouped into regions in accordance to the World Health
cause extensive morbidity, but are neglected as they do not
Organization (WHO) regions [16]. WHO divides the world
significantly contribute to mortality. One such disease is
into six regions; Southeast Asian Region (SEAR), Eastern
rheumatoid arthritis (RA), the most common type of in-
Mediterranean Region (EMR), Western Pacific Region
flammatory musculoskeletal disorder [3,4], in which the
(WPR), Europe (EUR), The Americas (AMR) and Sub–Sa-
quality of life has been reported to be lower than in patients
haran Africa (AFR) [17–19]. A systematic literature search
suffering from most of the other NCD [5,6]. It is a chronic
was conducted separately for each region to find popula-
systematic autoimmune inflammatory disease, character-
tion based studies that documented the prevalence of RA.
ised by a symmetrical persistent synovitis of the joints of
Medline (1946 – July week 1, 2013), EMBASE (1976–
the hands, wrist, feet and knee resulting in tender swelling
2013 week 26) and Global Health (GH) (1973 – 2013
of joints, pain, limitation in motion and morning stiffness.
week 26) were searched using the OVID search engine
Its systematic features include fatigue, generalised weak-
(search terms available in Online Supplementary Docu-
ness, loss of weight and low grade fever [7]. As the disease
ment). Both Medical Subject Headings (MeSH terms) and
advances, irreversible tissue damage occurs, with destruc-
keywords were used in OVID. Other online databases such
tion of bone and cartilage leading to joint deformity, mus-
as PubMed, Web of Knowledge (WoK) and databases se-
cle atrophy, and progression that may involve all joints of
lective to regions, such as LILACS for Latin America, CNKI
the body [8]. For the purpose of clinical trials, RA is diag-
and Wan Fang for China, and IndMed for India, were also
nosed using the American College of Rheumatology (ACR)
thoroughly searched. PubMed was searched for all regions,
criteria, formerly known as the American Rheumatology
as it proved to be broader and more sensitive in picking up
Association (ARA) criteria [7,9].
studies. Grey literature was also searched for all low–mid-
The prevalence of RA in the western world is 1–2% [10], dle income countries using SIGLE (OpenGrey), Google
and is believed to be 1% worldwide [11]. However, this Scholar and Global Health library. The search of grey lit-
global estimate is based on a few sporadic studies over dif- erature resulted in 10 and 149 studies, respectively, none
ferent time periods, in a limited number of LMIC. Extrapo- of which were relevant to this analysis.

June 2015 • Vol. 5 No. 1 • 010409 2 www.jogh.org • doi: 10.7189/jogh.05.010409

 Prevalence of rheumatoid arthritis in low– and middle–income countries

Inclusion and exclusion criteria for study evant studies. Full texts of selected studies were analysed
selection and inclusion and exclusion criteria were applied. Data
After the initial screen, inclusion and exclusion criteria from all relevant studies was extracted into an Excel spread-
were applied to retain only the studies that were free of any sheet, where sample size (age, sex–specific, mean age),
apparent bias. We included studies conducted in LMIC methodology, criteria used for diagnosing RA, study loca-
from all WHO regions that were population based or com- tion (urban or rural) and prevalence rate were documented

VIEWPOINTS
munity based, studies conducted after 1987 that used ARA/ for each study.
ACR diagnostic criteria (see Table 1), focused on adult
Adjustment of prevalence rates

PAPERS
populations (typically 15+ or 18+ years, with exclusion of
juvenile forms in the former studies) and reported the prev- Once the final set of studies was retained (Figure 1), crude
alence rate of RA. We excluded review articles with second- prevalence rates, sex–specific prevalence rates, urban and
ary data only (with the exception of sub–Saharan Africa, rural prevalence rates and male–to–female ratio of RA cas-
where the amount of data was particularly scarce), hospi- es were adjusted to the same measurement unit and ex-
tal–based studies (for lack of representativeness of the gen- pressed as a percentage. Data extracted from each study is
eral population), studies conducted prior to or during 1987 shown in the Online Supplementary Document. Checks
(for inconsistent case definition), studies on other types of for internal consistency of the data were made and possible
arthritis in adults, studies on juvenile forms of arthritis and significant correlations between prevalence rate and the
studies that used other diagnostic criteria to measure the sample size, year of publication of the study, sex and resi-
prevalence of RA in the population. dency were made.
We retained studies that clearly presented the method of
diagnosing RA, beginning with how the sample population Statistical analyses
was recruited and evaluated, along with the criteria used
All statistical analyses are shown in the Online Supple-
for diagnosis of RA. We expected that trained personnel or
mentary Document. We first tested the distribution of the
specialists be involved in the field work, and we excluded
reported prevalence of RA across all identified studies for
the studies where self–reporting was the primary method
normality using the one–sample Kolmogorov-Smirnov test.
of case ascertainment. Specialists (doctors, rheumatolo-
We concluded that the results did not indicate normal dis-
gists) needed to be involved in the next step of confirma-
tribution, presumably because of substantial heterogeneity
tion. Any study where there was no direct contact between
the assessors and sample population, such as telephone in the included studies (Z = 1.831, P = 0.002). We then per-
surveys, were excluded. There is a high probability of mis- formed a meta–analysis of all identified studies in all LMIC
classification and oversight of cases by untrained or inad- using the DerSimonian–Laird method, to determine the
equately trained personnel, or through indirect contact. “LMIC” prevalence rate (Online Supplementary Docu-
ment).
Figure 1 summarises the process of study selection for all
six WHO regions. First, duplicates were excluded and titles We then displayed mean and median prevalences in each
and abstracts of the retained papers were evaluated for rel- of the six WHO regions (Figure 2). The Kruskal-Wallis

Table 1. The criteria of the American College of Rheumatology (ACR) established in 1987 to assist clinical diagnosis of rheumatoid
arthritis*
1. Morning stiffness Morning stiffness in and around the joints, lasting at least 1 hour before maximal improvement
2. Arthritis of 3 or more joints At least 3 joint areas simultaneously have had soft tissue swelling or fluid (not bony overgrowth alone) observed
by a physician. The 14 possible areas are right or left PIP, MCP, wrist, elbow, knee, ankle, and MTP joints
3. Arthritis of hand joints At least 1 area swollen (as defined above) in a wrist, MCP, or PIP joint
4. Symmetric arthritis Simultaneous involvement of the same joint areas (as defined in 2) on both sides of the body (bilateral involve-
ment of PIPs, MCPs, or MTPs is acceptable without absolute symmetry
5. Rheumatoid nodules Subcutaneous nodules, over bony prominences, or extensor surfaces, or in juxta–articular regions, observed by a
physician
6. Serum rheumatoid factor Demonstration of abnormal amounts of serum rheumatoid factor by any method for which the result has been
positive in <5% of normal control subjects
7. Radiographic changes Radiographic changes typical of rheumatoid arthritis on postero–anterior hand and wrist radiographs, which must
include erosions or unequivocal bony decalcification localized in or most marked adjacent to the involved joints
(osteoarthritis changes alone do not qualify

PIP – proximal interphalangeal; MCP – metacarpophalangeal; MTP – metatarsophalangeal

*Four out of the seven criteria need to be met in order to establish the diagnosis of RA, with criteria 1–4 required to be present for at least 6 weeks. Pa-
tients with two clinical diagnoses are not excluded [9].

www.jogh.org • doi: 10.7189/jogh.05.010409 3 June 2015 • Vol. 5 No. 1 • 010409

 Rudan, Sidhu et al.
VIEWPOINTS
Papers

Figure 1. Flowchart presenting the literature search and the process of study selection (WoK = Web of Knowledge;
G.H. = global health).

An important possible confounding effect was differences in

the mean age of the sample between the studies. This is be-
cause, although all samples were defined as “adult popula-
tion” (usually 15 years of age or older), the relative contribu-
tion of elderly population varied in different countries
because of difference in sampling strategies and overall life
expectancy. We explored the association between the preva-
lence of RA and the mean age using linear correlation coef-
ficients (Pearson, Kendall’s tau, and Spearman), and also the
generalised dependence measure mutual information to ex-
clude the potential effect of age distribution on generalisabil-
ity of the results (Online Supplementary Document).
Association between gender and prevalence of RA, where
a considerable difference between sexes was expected, was
Figure 2. The relationship between mean and median preva- explored using the paired samples t–test. We then conduct-
lence of rheumatoid arthritis in low and middle–income ed a gender–specific meta–analysis to estimate prevalence
countries in six WHO regions of the world. and confidence intervals in men and women, using the
DerSimonian–Laird method. In addition, box–and–whis-
one–way analysis of variance by ranks examined whether kers plots of regional prevalence by gender were also pre-
samples originated from the same distribution. After this, sented (Online Supplementary Document). We also ex-
we conducted a series of region–specific meta–analyses to amined the difference in prevalence of RA between urban
estimate regional prevalence and confidence intervals, us- and rural populations. Since the RA prevalence is not nor-
ing the DerSimonian–Laird method (Online Supplemen- mally distributed, we performed the non–parametric
tary Document). Mann-Whitney U test to test the null–hypothesis.

June 2015 • Vol. 5 No. 1 • 010409 4 www.jogh.org • doi: 10.7189/jogh.05.010409

 Prevalence of rheumatoid arthritis in low– and middle–income countries

RESULTS
In our study, the majority of studies were from mainland
China, with additional studies from Taiwan and Hong
Kong. Mexico, Turkey, Iran, India, Pakistan, Philippines
and Russia were also represented through multiple studies.
The median year of publication was 2004, making the es-

VIEWPOINTS
timate useful for application to both the world population
in 2000 and 2010. Twenty–one studies used the ARA cri-

PAPERS
teria, and all the remaining studies used the 1987 revised
ACR criteria. Some studies had multiple cohorts. Each co-
hort was recognised separately during analysis, so that the
final 48 studies resulted in 60 cohorts. In case of the Afri-
can region, only two studies were found from the entire
region that fulfilled the criteria for inclusion. One of the
Figure 3. Regional median, minimum and maximum observed
studies had a very small sample size and did not find a sin- value and inter–quartile range for the prevalence of rheumatoid
gle case of RA, so it was excluded as uninformative. The arthritis in low and middle–income countries in six WHO
other study had a crude prevalence rate of 1%, but we felt regions of the world.
that it would be inadequate to base an entire regions' prev-
alence rate on a single study. Therefore, we decided to dis-
and the generalised dependence measure mutual informa-
card both of those studies and replace them by Bowman’s
tion did not show an effect of mean age of the sample on
systematic analysis in 2012 [20]. Bowman included all the
the reported prevalence of RA across the studies (P = 0.0599;
studies ever conducted in Africa in his estimate, irrespec-
P > 0.05), implying that the differences in age structure of
tive of the year of study, and the prevalence rate from his
samples in different studies were not the main determinant
study was then applied to the population statistics of the
of the observed heterogeneity.
region in 2000 and 2010.
When all studies were analysed in one large meta–analysis, An investigation into differences in prevalence by gender
irrespective of their heterogeneity, this resulted in an using paired samples t–test indicated that the male and fe-
“LMIC” estimate for the prevalence of RA of 0.53% (95% male RA prevalence differed significantly (P < 0.0001),
CI: 0.45–0.61%). Analysis of heterogeneity confirmed that which was expected. We therefore conducted a separate
the data were highly heterogeneous (I2 = 96%) (Online meta–analysis of the RA prevalence in LMIC countries for
Supplementary Document). We then studied the mean men and women. Male prevalence was 0.16% (95% CI:
and median prevalence in each of the six WHO regions 0.11–0.20%) (Figure 4) while the prevalence in women
(Figure 2) and presented box–and–whiskers plot of the was five times higher, amounting to 0.75% (95% CI: 0.60–
results from studies in each region (Figure 3). The Kruskal- 0.90%) (Figure 5). Data seemed to be less heterogeneous
Wallis one–way analysis of variance by ranks test showed for men (I2 = 49.6%) than for women (I2 = 83.7%).
that the prevalence in at least one of the WHO regions was We also examined the difference in prevalence of RA be-
statistically different from the others (P = 0.029). tween urban and rural populations, wherever information
A series of region–specific meta–analyses were conducted was available to allow for comparison. Since we established
to estimate regional prevalence of RA. The meta–analysis that the RA prevalence was not normally distributed, we
estimates of regional RA prevalence rates were 0.40% (95% performed the non–parametric Mann-Whitney U test to
CI: 0.23–0.57%) for Southeast Asia, 0.37% (95% CI: 0.23– test the null hypothesis. The significance of the test was
0.51%) for Eastern Mediterranean, 0.62% (95% CI: 0.47– P = 0.353, indicating that the prevalence in the urban and
0.77%) for European LMIC countries, 1.25% (95% CI: rural settings do not differ significantly (Figure 6).
0.64–1.86%) for American LMIC countries and 0.42%
After all the previous analyses, a strategy was needed for
(95% CI: 0.30–0.53%) for Western Pacific LMIC countries,
estimating the number of persons living with RA in LMIC
respectively. This analysis could not be performed for Af-
in the years 2000 and 2010. Possible approaches were: (i)
rica due to limited data. The data sets were heterogeneous
to apply the meta–analysis of the crude prevalence from all
in all the regions (I2 varied from 74.2% to 97.3%).
identified studies to the total number of persons 15 years
We then studied whether the mean age of the sample con- or older in LMIC; (ii) to use the estimate of prevalence for
tributed to the observed prevalence rates. Linear correla- males and for females that resulted from the meta–analyses
tion coefficients (Pearson, Kendall’s tau, and Spearman) of all studies that reported the rates separately by gender;

www.jogh.org • doi: 10.7189/jogh.05.010409 5 June 2015 • Vol. 5 No. 1 • 010409

 Rudan, Sidhu et al.
VIEWPOINTS
Papers

Figure 4. Meta–analysis of
rheumatoid arthritis
prevalence in men, based on
all available information from
low and middle–income
countries in six WHO regions
of the world.

then, to apply those two estimates of prevalence to male tion. However, the number of studies typically available for
and female populations in LMIC; (iii) to use regional me- different regions was simply too small to be sure whether
dians or the estimates based on regional meta–analysis and the observed differences between regions were real, or just
apply them to the regional populations aged 15 years or stochastic. The same applies to approach (iv). Therefore,
older; (iv) finally, to use sex–specific regional estimates and approach (ii) was used, because it accounted for the most
apply it to male and female population aged 15 years or important confounding variable – gender – and because it
older in respective regions. provided a lot of information for meta–analysis in each gen-
der. This allowed an assumption that the rates considered
Given the quantity and quality of the information that was
representative for all males and all females in LMIC were
obtained through this systematic review, the most appro-
more likely to be accurate than region–specific rates. More-
priate (and robust) approach was to use gender–specific
over, the observed heterogeneity of the underlying data was
estimates of prevalence for the whole LMIC region and ap-
the lowest in gender–specific meta–analysis across LMIC.
ply them to male and female populations in LMIC. There
are a number of reasons why other approaches were not This gives an estimate of male prevalence of 0.156% (95%
preferred and we will list them here. Although the quan- CI: 0.11–0.20%) (Figure 4) that needs to be applied to the
tity of information was the largest for the approach (i) male population aged 15 years or more in LMIC in 2000 and
above, there is uncertainty in some studies over the com- 2010. In females, the prevalence of 0.747% (95% CI: 0.60–
position of sample by gender, and whether it is representa- 0.90%) is used (Figure 5). The UN Population Division's
tive of the underlying population. Given that gender is an estimates for the number of males aged 15 years or older in
extremely important determinant of prevalence, the ap- LMIC in the year 2000 is 1.667 billion, and in the year 2010
proach (i) would suffer from a possible confounding effect it is 2.206 billion [17]. For women, the corresponding fig-
of the gender composition of the sample. The strength of ures are 1.634 billion for the year 2000 and 1.991 billion in
approach (iii) was that it could account for regional varia- 2010 [17]. This translates into 2.60 (95% CI: 1.85–3.34%)

June 2015 • Vol. 5 No. 1 • 010409 6 www.jogh.org • doi: 10.7189/jogh.05.010409

 Prevalence of rheumatoid arthritis in low– and middle–income countries

VIEWPOINTS
PAPERS
Figure 5. Meta–analysis of
rheumatoid arthritis preva-
lence in women, based on all
available information from low
and middle–income countries
in six WHO regions of the
world.

million male sufferers and 12.21 (95% CI: 9.78–14.67%)

million female sufferers in the year 2000, and 3.16 (95% CI:
2.25–4.05%) million affected males and 14.87 (95% CI:
11.91–17.86%) million affected females in the year 2010 in
the countries of low and middle income.

DISCUSSION
We presented a robust estimate of the number of individu-
als suffering from RA in low and middle income countries
in 2000 and 2010. There have already been several at-
tempts to estimate the prevalence of RA at the global, re-
gional and national level and also in LMIC [10,11,14,18,19].
In comparison to previous estimates that presented both
higher and lower estimates than our study, our estimate is
based primarily on a substantial amount of evidence from
LMIC on sex–specific prevalence. We demonstrated that
gender is a principal determinant of RA in LMIC and that
age distribution of the population and being an urban
Figure 6. A scatterplot of observed prevalence rates of rheuma-
toid arthritis in six WHO regions based on urban or rural
dweller do not contribute significantly to disease develop-
residency of the examinees. No statistical differences were noted ment. This is different from some other diseases, such as
(see Online Supplementary Document for further detail). dementia and cancer, where age seems to be the main de-

www.jogh.org • doi: 10.7189/jogh.05.010409 7 June 2015 • Vol. 5 No. 1 • 010409

 Rudan, Sidhu et al.

terminant, or schizophrenia, where being an urban dwell- ies towards a single plausible estimate. Besides three stud-
er and family history seem to be the main driver of the dis- ies, in which we adjusted the estimate, all others determined
ease occurrence [1,2]. Therefore, we believe that our the prevalence rate using the same age cut–off (15 years or
strategy for deriving the estimate was more appropriate older), again leading to comparable estimates underlying
than used in some previous studies. Moreover, we provide each regional and the overall prevalence rate.
the full data set used to develop the estimates in Online Although the best quality of epidemiological work on RA
Supplementary Document and all our methods are trans-
VIEWPOINTS

in LMIC has been conducted by WHO–ILAR–COPCORD

parent and replicable by other groups. [23], this program covers a limited number of countries
Papers

Still, there are limitations in all estimates of the current and this current review includes a larger number of studies
global burden of RA. The criteria of defining the disease conducted elsewhere, by different researchers, leading to
have changed over time and the estimates that don't take wider coverage. Still, nearly all of the retained studies close-
this into account will be internally inconsistent. Moreover, ly followed the three–step methodology set by WHO–
a mixture of studies, both hospital and population based ILAR–COPCORD and used similar questionnaires, thereby
studies, could be considered, although the former will pres- decreasing the methodological variability in the studies
ent the more severe end of disease spectrum and bias the [23]. The questionnaires were translated in local languages
results. A major strength of our study is that it made use of and tested before the start of the studies in almost all cases.
all literature available in all languages, including two major Nevertheless, variation remained even within the selected
Chinese databases and the database with grey literature. studies that share the same three–step methodology. The
We therefore believe that we are presenting the most ad- assessors at each stage were different between the studies.
vanced estimate of RA burden to date. However, limitations In some studies, trained nurses administered the question-
are still large: there are very few data points (particularly in naire, while in others this was done by trained volunteers
Africa) and most LMIC countries do not have a single pub- and students. At the second and third stage most studies
lished epidemiological study. Moreover, most of the studies involved rheumatologists, but a handful of studies had gen-
used for this estimate are quite small and they are unlikely eral medical doctors or internists evaluate the potential cas-
to be nationally representative. Also, this study uses a wide es. Moreover, among the studies conducted by COPCORD,
range of years to provide estimates for 2000 and 2010, there was a slight regional variation in the questionnaire
which is a limitation given that prevalence may be chang- used, given that it was modified over time, decreasing the
ing over time and that the time trend reported here arises comparability of studies. Although the number of partici-
from demographic changes, rather than our understanding pants at each step is given, the reason of non–participation
of the epidemiological situation. is not stated in all studies. This may have led to non–re-
Our paper also aimed to explore whether other major co- spondent bias. Although this problem cannot be easily con-
variates, besides gender, affect the frequency of the occur- trolled, it still needs to be acknowledged, as there may be
rence of RA. We were unable to demonstrate significant ef- a difference in characteristics of those who participate and
fects for either urban / rural living or age structure of the those who do not.
study sample. Comparing all urban, rural and mixed stud- Research in developed countries does not seem to suggest
ies amongst each other, we were unable to demonstrate sig- a growing trend in the prevalence of the disease. However,
nificant differences between prevalence rates in urban or the total number of cases grew considerably between 2000
rural areas. This is contrary to some previous reports that and 2010 because the population of LMIC older than 15
suggested that the prevalence might be higher in urban ar- years has grown in this period [24]. Even a slight increase
eas [21]. Moreover, previous reports suggested that the in the prevalence rate, eg, an additional prevalence of 0.1%,
prevalence rate in LMIC is lower than in the developed would translate to an increment of 4 million affected per-
countries [22], which our study seems to generally support. sons. Given the fact that RA affects the working age popu-
One of the major strengths of our study is that it involved lation and most of the employment in these countries is
a systematic search of ten large databases, resulting in the still for manual labour, it greatly decreases the productivity
identification of 10 599 studies initially and 48 studies se- of countries as a whole [25]. Additionally, the high costs of
lected for inclusion. Native speakers translated studies in a treatment, borne by individuals themselves in the most
language other than English, specifically Chinese and LMIC, counters other efforts to decrease poverty and im-
Spanish. This has greatly increased the pool of studies avail- prove living standards.
able for analysis, as it led to the inclusion of studies other- As RA is an important condition with significant morbid-
wise excluded due to language barriers. All of the studies ity and economic impact, it should be at the forefront in
used the same definition to identify RA in patients: the health care policy. Despite this, RA as part of a larger group
1987 revised ACR criteria (previously known as ARA cri- of NCD receives less than 3% of annual development as-
teria). This enabled comparison and convergence of stud- sistance for health to low and middle income countries.

June 2015 • Vol. 5 No. 1 • 010409 8 www.jogh.org • doi: 10.7189/jogh.05.010409

 Prevalence of rheumatoid arthritis in low– and middle–income countries

The neglect of NCD on the global stage can be explained agencies already working towards generating information,
not only by the gaps in estimates on burden of disease but such as WHO–ILAR [23], should be undertaken and their
also from a lack of strategic communication about the ur- efforts should be supported. In Africa, the African League
gency of the problem [26]. of Associations of Rheumatology (AFLAR) already exists,
In this paper, estimates of RA morbidity aim to take a first but little has been done in terms of research and surveil-
step in raising awareness of policy makers and health care lance of rheumatologic diseases [20]. Such associations

VIEWPOINTS
workers, as previously they have had to rely on the evi- should be supported, encouraged and pressurised by gov-
dence that was generated mainly in the developed world. ernments to carry on more work in this area. The WHO–

PAPERS
The lack of specialists to diagnose and treat this condition ILAR–COPCORD program was developed to identify all
should also be addressed. In certain African nations there types of musculoskeletal disorders, and not specifically de-
is only one rheumatologist for the entire population [21]. signed for rheumatoid arthritis nor as an epidemiological
Therefore, an increase in the number of specialists in this
study program [23]. Therefore, further research should be
area is urgently needed in LMIC [27]. As this takes time, the
specifically orientated towards rheumatoid arthritis, with
existing doctors in the community should be offered spe-
greater attention on the methodology. The lack of informa-
cific education on RA, including newer treatment regimens
tion from more than 100 LMIC countries should be ad-
and the management of the associated comorbid conditions.
As most of the population resides in rural areas, incorporat- dressed and gaps filled. Studies should also include age
ing identification and treatment of the disease in commu- groups of those with the disorder, thereby providing more
nity health care system is crucial to reach all those in need information on who bears the greatest burden and allow-
of diagnosis and treatment. Funding should be targeted at ing age–standardised comparison. Simultaneously, infor-
increasing efficacy of treatments in LMIC. Efforts should be mation about risk factors should be obtained and incorpo-
made to increase the availability of treatment – both anti– rated in study designs.
inflammatories and the newer biological agents that have
The estimates presented in this paper provide a building
proven to be greatly beneficial – at affordable costs. The
block for future epidemiological studies by suggesting the
newer biological agents are very expensive and it is unlike-
way forward in disease assessment in LMIC context and
ly that many LMIC could afford to supply them [25]. More-
the methodology that could be deployed. It could also be
over, RA is associated with an increased risk of other dis-
eases (such as cardiovascular diseases) and the management used to create awareness among health–care workers and
of these comorbid conditions is also important. All doctors education of people about the disease and encourage
should be made aware of treatment protocols already in use health–seeking behaviour for provision of available treat-
by high–income countries with emphasis on early treat- ment, which can decrease the burden associated with dis-
ment, to slow disease progression and elimination of pain. ability and bring about a decrease in morbidity. It is also
This should lead to improvements in quality of life and de- noteworthy to point out the reoccurring theme of lack of
crease the occurrence of co–morbid conditions, such as de- data from the poorest countries. Policy makers from these
pression. The set–up of supportive treatment, such as phys- countries should show more dedication and step up their
iotherapy, should be encouraged [11–18]. efforts towards research in the health care sector, as gener-
Infrastructure for research in areas where it is currently un- ating information about the burden of disease is the first
available should be set up. Allegiance with international step in decreasing its prevalence.

Acknowledgements: The work was partly completed during the corresponding author's employment as a
Senior Research Fellow at The University of Melbourne.
Funding: Part of this work was funded by an internal grant from the University of Melbourne.
Ethical approval: Not required.
Authorship declaration: KYC and IR conceptualised the study. SS and IR conducted the literature review
for all databases apart from Chinese and Latin American. KYC, ARD, SJM, YXW and WW (China) conduct-
ed the review of the Chinese literature. RMC conducted the review of Latin American literature. BD and DA
contributed the data and analyses for sub–Saharan Africa. AP performed all statistical analyses. SS, IR and
KYC drafted the paper. ARD, HN, DS, ET, JC, AM, HC and WW (China-Australia) contributed to writing
of the final version of the paper and checked the paper for important intellectual content.
Competing interests: All authors have completed the Unified Competing Interest form at www.icmje.org/
coi_disclosure.pdf (available on request from the corresponding author) and declare no conflict of interest.

www.jogh.org • doi: 10.7189/jogh.05.010409 9 June 2015 • Vol. 5 No. 1 • 010409

 Rudan, Sidhu et al.

1 United Nations. Prevention and control of non–communicable diseases: Report of the Secretary–General. New
York: United Nations, 2011. Available: http://www.un.org/ga/search/view_doc.asp?symbol=A/66/83&referer=/

REFERENCES
english/&Lang=E. Accessed: 2 May 2014.
2 World Health Organization. UN High Level Meeting on NCD: Summary report of the discussion at the round
tables. Geneva: World Health Organization, 2011. Available: http://www.who.int/nmh/events/moscow_
NCD_2011/round_tables_summary.pdf. Accessed: 2 May 2014.
3 Badley EM, Tennant A. Disablement associated with rheumatic disorders in a British population: problems with
activities of daily living and level of support. Br J Rheumatol. 1993;32:601-8. Medline:8339133 doi:10.1093/
VIEWPOINTS

rheumatology/32.7.601
4 Woolf AD, Pfleger B. Burden of major musculoskeletal conditions. Bull World Health Organ. 2003;81:646-56.
Papers

Medline:14710506
5 Haroon N, Aggarwal A, Lawrence A, Agarwal V, Misra R. Impact of rheumatoid arthritis on quality of life. Mod
Rheumatol. 2007;17:290-5. Medline:17694261 doi:10.3109/s10165-007-0604-9
6 Pollard L, Choy EH, Scott DL. The consequences of rheumatoid arthritis: quality of life measures in the indi-
vidual patient. Clin Exp Rheumatol. 2005;23(5 Suppl 39):S43-52. Medline:16273784
7 Creemers MCW, Van de Putte LB. Rheumatoid arthritis – the clinical picture. In: Isenberg DA, Madison PJ, Woo
P, Klars D, Breedveld FC (eds). Oxford textbook of rheumatology. III Edition. Oxford: Oxford University Press,
2004.
8 Currey HLF. Essentials of Rheumatology. 2nd Ed. London: Churchill Livingstone, 1988.
9 Arnett FC, Edworthy SM, Bloch DA, McShane DJ, Fries JF, Cooper NS, et al. The American Rheumatism Asso-
ciation revised 1987 criteria for the classification of rheumatoid arthritis. Arthritis Rheum. 1988;31:315-24.
Medline:3358796 doi:10.1002/art.1780310302
10 Alamanos Y, Voulgari PV, Drosos AA. Incidence and prevalence of rheumatoid arthritis, based on the 1987 Amer-
ican College of Rheumatology criteria: a systematic review. Semin Arthritis Rheum. 2006;36:182-8. Med-
line:17045630 doi:10.1016/j.semarthrit.2006.08.006
11 Chopra A, Abdel–Nasser A. Epidemiology of rheumatic musculoskeletal disorders in the developing world. Best
Pract Res Clin Rheumatol. 2008;22:583-604. Medline:18783739 doi:10.1016/j.berh.2008.07.001
12 Cooper NJ. Economic burden of rheumatoid arthritis: a systematic review. Rheumatology (Oxford). 2000;39:28-
33. Medline:10662870 doi:10.1093/rheumatology/39.1.28
13 Osiri M, Maetzel A, Tugwell P. The economic burden of rheumatoid arthritis in a developing nation: results from
a one–year prospective cohort study in Thailand. J Rheumatol. 2007;34:57-63. Medline:17183621
14 Kvien TK. Epidemiology and burden of illness of rheumatoid arthritis. Pharmacoeconomics. 2004;22(2 Suppl
1):1-12. Medline:15157000
15 The World Bank. Data: Country and Lending Groups. 2013. Available: http://data.worldbank.org/about/
country–classifications/country–and–lending–groups. Accessed: 1 June 2014.
16 World Health Organization. Quantifying environmental health impacts: Country profiles of environmental bur-
den of disease by WHO regions. 2013. Available: www.who.int/quantifying_ehimpacts/national/countryprofile/
regions/en/index.html. Accessed: 1 June 2014.
17 United Nations. World population prospects: The 2012revision. Available: http://esa.un.org/unpd/wpp/unpp/
panel_population.htm. Accessed: 1 June 2014.
18 Mijiyawa M. Epidemiology and semiology of rheumatoid arthritis in third world countries. Rev Rhum Engl Ed.
1995;62:121-6. Medline:7600065
19 Cross M, Smith E, Hoy D, Carmona L, Wolfe F, Vos T. The global burden of rheumatoid arthritis: estimates from
the Global Burden of Disease 2010 study. Ann Rheum Dis. 2014;73:1316-22. Medline:24550173 doi:10.1136/
annrheumdis-2013-204627
20 Dowman B, Campbell RM, Zgaga L, Adeloye D, Chan KY. Estimating the burden of rheumatoid arthritis in Af-
rica: A systematic analysis. J Glob Health. 2012;2:020406. Medline:23289081 doi:10.7189/jogh.02.020406
21 Kalla AA, Tikly M. Rheumatoid arthritis in the developing world. Best Pract Res Clin Rheumatol. 2003;17:863-
75. Medline:12915162 doi:10.1016/S1521-6942(03)00047-0
22 Akhter E, Bilal S, Kiani A, Haque U. Prevalence of arthritis in India and Pakistan: a review. Rheumatol Int.
2011;31:849-55. Medline:21331574 doi:10.1007/s00296-011-1820-3
23 Darmawan J, Muirden KD. WHO–ILAR COPCORD perspectives past, present, and future. J Rheumatol.
2003;30:2312-4. Medline:14677169
24 Cohen B. Urbanization in developing countries: Current trends, future projections, and key challenges for sus-
tainability. Technol Soc. 2006;28:63-80. doi:10.1016/j.techsoc.2005.10.005
25 Hoving JL, van Zwieten MC, van der Meer M, Sluiter JK, Frings–Dresen MH. Work participation and arthritis:
a systematic overview of challenges, adaptations and opportunities for interventions. Rheumatology (Oxford).
2013;52:1254-64. Medline:23472042 doi:10.1093/rheumatology/ket111
26 Maher A, Sridhar D. Political priority in the global fight against non–communicable diseases. J Glob Health.
2012;2:020403. Medline:23289078 doi:10.7189/jogh.02.020403
27 Halabi H, Alarfaj A, Alawneh K, Alballa S, Alsaeid K, Badsha H, et al. Challenges and opportunities in the early
diagnosis and optimal management of rheumatoid arthritis in Africa and the Middle East. Int J Rheum Dis. In
press. Medline:24620997

June 2015 • Vol. 5 No. 1 • 010409 10 www.jogh.org • doi: 10.7189/jogh.05.010409