You are on page 1of 9

REVIEWS

Family history of rheumatoid arthritis:


an old concept with new developments
Thomas Frisell1, Saedis Saevarsdottir2,3 and Johan Askling1,3
Abstract | Family history of rheumatoid arthritis (RA) is a proxy for an individual’s genetic and, in
part, environmental risk of developing RA, and is a well-recognized predictor of disease onset.
Although family history of RA is an old concept, the degree of familial aggregation of RA, whether
it differs by age, sex, or serology, and what value it has for clinical decisions once a diagnosis of RA
has been made remain unclear. New data have been emerging in parallel to substantial progress
made in genetic association studies. In this Review, we describe the various ways that familial
aggregation has been measured, and how the findings from these studies, whether they are
based on twins, cohorts of first-degree relatives, or genetic data, correspond to each other and
aid understanding of the aetiology of RA. In addition, we review the potential usefulness of family
history of RA from a clinical point of view, demonstrating that, in the era of big data and genomics,
family history still has a role in directing clinical decision-making and research.

Rheumatoid arthritis (RA) clusters in families, mean‑ In this Review we provide an overview of the his‑
ing that relatives of patients with RA are at increased tory and the current state of research on the familial
risk of developing the disease. Indeed, family history aggregation of RA, including the different study designs
has long been recognized as one of the strongest risk and metrics used in this field (BOX 1; FIGS 1,2), and dis‑
factors for developing RA1,2 and, as such, it is routinely cuss what can be learned about the disease. We also
assessed as part of the clinical work‑up of individuals explore whether this information might have value in
with inflammatory arthritis. the clinical setting.
The study of familial aggregation of disease is moti‑
vated by potential clinical usefulness and by the con‑ Familial risk of RA
cern of patients, who worry about what risks they might A brief historical overview
have inherited or could be passing on to the next gen‑ Although a heritable component to RA was suspected as
eration in terms of developing RA, the clinical course early as 1810 (REF. 2), collection of data sets of sufficient
and treatment. Accurate estimation of the strength and quality to prove the presence of familial aggregation, let
characteristics of familial aggregation is difficult. Ideally, alone to calculate precisely its magnitude and character‑
1
Clinical Epidemiology Unit, it requires large, representative populations of families istics, has proven difficult. Inspired by early case reports
Department of Medicine
with valid diagnoses of RA. Data based on registries and of familial disease1–3, several studies in the 1950s com‑
Solna, Karolinska Institutet,
T2 Karolinska University total-population studies are providing increasingly pre‑ pared self-reported family history of RA among clini­cally
Hospital, SE‑171 76 cise calculations of the predictive value of family history verified cases with that among control subjects. Cases
Stockholm, Sweden. of RA for disease onset. reported two to six times higher prevalence of RA in their
2
Institute of Environmental Familial clustering is caused by genetic and environ‑ first-degree relatives (FDRs)4–6. One estimate suggested
Medicine, Karolinska
Institutet, BOX 210, SE‑171
mental factors shared within families, and, therefore, its that risk was raised by as much as 15-fold7. Self-reported
77 Stockholm, Sweden. magnitude and pattern of distribution provides infor‑ family history, however, is recognized as being likely to
3
Rheumatology Unit, mation on the aetiology of RA. In particular, observed result in substantial misclassification unless diagnosis in
Department of Medicine familial risks can be used to estimate the heritability of relatives can be verified5. Self-reporting might also lead to
Solna, Karolinska Institutet,
RA, which is essential to contextualize other aetiological overestimation, for example if patients with RA are more
Karolinska University
Hospital, SE‑171 76 studies, especially genetic studies. Several studies have aware of the disease history of relatives with RA than are
Stockholm, Sweden. also addressed the usefulness of family history in the control individuals.
Correspondence to J.A. clinic, with the hypothesis that such information might To overcome these limitations, several ambitious
johan.askling@ki.se predict clinical course and prognosis if familial factors projects in the late 1950s and the 1960s involved clini‑
doi:10.1038/nrrheum.2016.52 influencing disease onset also influence disease severity, cal examinations of hundreds of relatives of individuals
Published online 21 Apr 2016 or to test if treatment response is itself hereditary. with RA and of matched controls. The results largely

NATURE REVIEWS | RHEUMATOLOGY ADVANCE ONLINE PUBLICATION | 1


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

Key points Over the following decades, several influential stud‑


ies demonstrated higher RA concordance in monozy‑
• Family history of rheumatoid arthritis (RA) is one of the strongest known risk factors gotic than in dizygotic twins17–19. Together with the
for developing RA, conferring twofold to fourfold increased risk in accumulating evidence for a substantial contribution
first-degree relatives of the HLA region to the risk of RA20, the findings in
• The heritability of RA seems to be ~40%, and is higher for seropositive than for twins settled the point of whether RA clustered in fam‑
seronegative RA ilies, but left unanswered the question of the strength of
• Familial risk does not differ by sex, which suggests that genetic effects do not explain aggregation. In the early 1990s, opinions on the impor‑
why RA is more common in women than men tance of familial aggregation of RA ranged from it being
• Family history remains a strong independent predictor of RA onset, despite advances very weak21 to it being one the strongest risk factors for
made in identifying genetic risk alleles during the past decade RA22. This issue was not resolved in contemporary stud‑
• Data so far suggest familial aggregation for treatment response but not for ies. Several small studies reported high rates of RA in
clinical characteristics relatives of index patients but did not include compari‑
son groups23–25. Two prospective cohort studies in rela‑
tives of patients treated at RA clinics found that risk was
corroborated the earlier findings2,8–10. Critics, however, only roughly doubled in FDRs26,27. A case–control study
claimed that the studies used unrepresentative samples reported about a fourfold increase in risk28, but the
and non-validated definitions of RA11. In particular, the power was limited and the generalizability questionable.
lack of association reported in total-population sur‑
veys performed in Native American populations12–15 Registries and total-population studies
was interpreted as evidence for bias in earlier studies. To avoid potentially problematic selection bias in
If some weak familial aggregation of RA did exist, it hospital-based populations but to retain sufficient sta‑
was argued at the time, it did not fit a simple Mendelian tistical precision, complete records of RA and familial
inheritance pattern, making heritable factors unlikely relationships in large unselected populations are needed.
to be important for RA 10–12. Total-population sur‑ Several countries have national registries, which have been
veys did have the advantage of avoiding the potential used for studies of familial aggregation (TABLE 1). Although
selection bias of hospital-based surveys, but the low few in number, these studies have been important for
prevalence of RA resulted in low power, and the lack resolving several issues about familial clustering.
of statistical significance was interpreted as absence of One study used complete pedigrees of about 270,000
familial aggregation12,16. living Icelanders29, obtained from a data set established as
part of the deCODE project30. The study identified more
than 1,000 living patients with RA and found signifi­
Box 1 | Measurement of familial aggregation cant familial aggregation of RA in FDRs (λ ~4) and
The translation between measures of familial clustering is dependent on disease second-degree relatives (λ ~2); definitions of λ are pro‑
prevalence. The two most common methods to measure clustering use relative risks: vided in BOX 1 and FIGURE 1. Of note, the study was esti‑
mated to cover ~60% of all patients with RA in Iceland,
Risk in relative of affected individual and attendance at the participating rheumatologists’
λ=
Risk in the general population clinics might itself have been familial.
Hemminki et al.31 identified almost 50,000 Swedish
and patients hospitalized for RA up to 2004. They analysed
NatureRisk
Reviews | Rheumatology
in relative of affected individual data in national registries of hospitalizations and famil‑
Familial risk =
Risk in relative of unaffected individual ial relationships, and found familial risks (standardized
incidence ratios) of 3.0 in parents and 4.6 in siblings.
The λ, a traditional measure, only requires background data on disease incidence. In Denmark, the risk for RA among parents of ~9,000
NatureorReviews
The use of ordinary case–control | Rheumatology
cohort designs, yielding familial relative risks, has RA patients was compared with that in parents of more
become more common. Each pair of relatives can be analysed separately or exposure
than 3,000,000 RA‑free individuals, yielding hazard ratios
can be defined as any affected first-degree relative. If family size does not depend on
disease status, with these two definitions the familial risk will be roughly equal. of 2.6–2.9 (REF. 32).
Twin studies often estimate concordance rates. Probandwise concordance is the A comprehensive analysis of the national Swedish
proportion of affected twins whose co‑twins are also affected, and pairwise registries, with longer follow‑up than the other registry
concordance is the proportion of all twin pairs where both twins are affected. studies, and, importantly, including patients receiving
The probandwise rate is the numerator of the λ, and might, for example, be compared non-primary outpatient care, identified about 90,000
with the expected population risk. The pairwise rate is difficult to compare to any other patients with RA33. Familial risks were in line with the
rate, and its use is discouraged94. The sample concordance will differ from the previous total-population estimates, with an odds ratio
population concordance unless affected twins were independently and completely of 3.2 (95% CI 3.0–3.3) for FDRs and of 1.9 (1.7–2.2)
ascertained, although ascertainment correction is possible42,43,94. for second-degree relatives. Of note, the estimates
The same familial risk would imply stronger familial aggregation for a common than
remained similar when restricting analysis to the ~12,000
for a rare disease. Thus, an alternative metric is the tetrachoric correlation, which is the
correlation in relatives’ liability to develop RA under the liability-threshold model. patients with recent-onset RA in the clinical Swedish
These correlations are directly related to heritability, and for a given disease prevalence Rheumatology Quality of Care Register and ~2,800 in
heritability gives a specific familial risk, although other factors shared by relatives might the EIRA case–control study (http://www.eirasweden.se/
increase familial risk further. index1.htm)33, where the validity of diagnosis and serology
was better substantiated than in the full Swedish register.

2 | ADVANCE ONLINE PUBLICATION www.nature.com/nrrheum


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

16 0.6 Since familial clustering might also be expected to arise


λ, FDR from shared non-genetic factors, upper limits to her‑
Familial risk, FDR 0.5 itability of RA were set in these studies. Attempts have
Tetrachoric correlation
8 also been made to tease apart genetic and non-genetic

Tetrachoric correlation
0.4 familial factors.
Relative risk

4 0.3
Twin studies
The classic method of separating nature from nurture
is to contrast familial risks between relatives who have
0.2
different degrees of genetic and environmental relation‑
2
ships, such as adoptees and twins. Twins are particularly
0.1 useful, since, under some assumptions39, the difference
in correlation between monozygotic and dizygotic twins
1 0 would be directly proportional to the heritability.
0 10 20 30 40 50 60 70 80 90 100 In what is perhaps the most influential study of famil‑
Heritability (%) ial aggregation in RA, MacGregor and colleagues40 used
data from two previously published twin studies to esti‑
Figure 1 | Inter-relations of familial risk measures. Assuming a disease prevalence
of 1%, the λ and the familial risk are virtually inseparable,Nature
whereas the tetrachoric
mate heritability. The first of these had studied same-sex
Reviews | Rheumatology
correlation is always half the heritability and is unaffected by prevalence. If factors other Finnish twins with RA, identified through the national
than additive genetics contribute to familial clustering, all measures would be increased Sickness Insurance Register 17, and reported pairwise
further. Conversely, if familial risk measures are used to estimate heritability in the concordance rates of 12% in monozygotic and 3% in
presence of shared environmental or non-additive genetic effects, heritability will be dizygotic pairs (BOX 1). The second study had recruited
overestimated. FDR, first-degree relative. 203 affected twin pairs through participating rheumatol‑
ogists and a media campaign in the UK, and reported a
probandwise concordance rate of 15% in monozygotic
Several early studies speculated that the familial and 4% in dizygotic twin pairs18. MacGregor et al. esti‑
aggregation was modified by sex or age at onset, but mated the heritability to be 65% in the Finnish study
results were inconclusive23,28,34–36. The registry-based group and 53% in the UK sample. At the time, many
studies ruled out any sex difference in familial risk31–33, researchers (including those of the two twin studies) had
which is interesting given the marked female predom‑ interpreted the familial risks of 2–5 and the probandwise
inance of RA, but show that early age at onset confers concordance rates in monozygotic twins of 15–20% to be
increased familial risk33. This finding might reconcile evidence for a weak effect of genes17,18,21. On the herita‑
some of the differences observed between studies and bility scale, however, it became clear that these data were
types of relatives37. actually consistent with a substantial genetic influence.
Overall, the risk of RA in FDRs of affected individu‑ However, MacGregor and colleagues did not separate
als is increased roughly threefold, and in second-degree shared genetic and environmental factors, since models
relatives approximately twofold, with the association explicitly assumed that twin correlation was due com‑
being modified by age at onset but not by sex. The pletely to additive genetics. Despite the analysis being
largest total-population surveys so far have all been done after finding that the effect of shared environment
Scandinavian and analyses in other ethnic groups would was not significantly different from zero, power for this
be valuable. test was very low 40. Unfortunately, low power is seen in
all twin studies of RA. Thus, although assessment of
Heritability of RA these populations could in theory shed light on the rela‑
The driving force behind most studies of the familial tive importance of genetic versus shared environmental
aggregation of RA, rather than being to improve predic‑ factors in RA, the scarcity of suitable participants has
tion of disease onset, has been interest in how impor‑ so far made these studies noninformative on this point.
tant genetic factors are in the development of RA. The Two other twin studies presented concordance rates
role of genetic factors is often quantified as heritability, consistent with substantial heritability and limited (or no)
that is, the proportion of liability to develop a disease influence of shared environment, but low power made
that is due to genetic factors (BOX 2). Heritability esti‑ them inconclusive19,41. By contrast, nationwide Danish
mates are inherently based on several assumptions, and studies have reported low monozygotic twin concord‑
the method used should be taken into account when ance and low heritability 42 but substantial influence of
interpreting the values reported for RA (TABLE 2). shared environment 43, and have concluded that genes
The first heritability estimates for RA, reported by seem less important than other factors for the develop‑
Lawrence in 1970 (REF. 2), were 60–70% for seropositive ment of RA. Of note, though, the confidence limits were
RA. The sample, however, was not well described and broad and, therefore, the apparent discrepancy might be
estimates were based on the assumption that familial due to chance variation across studies44,45 (TABLE 2).
aggregation was due entirely to additive genetic effects38. The most precise estimate of the heritability of RA
Under the same assumption, we used data for FDRs in comes from a large study in FDRs rather than twins,
the Swedish Multigeneration Register and the Swedish but could be an overestimate if family environment
Patient Register and estimated that heritability was 40%33. contributes to the risk of RA. Yet, although the roles of

NATURE REVIEWS | RHEUMATOLOGY ADVANCE ONLINE PUBLICATION | 3


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

A good illustration is the weak contribution of the


FDR, prevalence 0.5% shared epitope alleles, which, despite high prevalence,
128
MZ, prevalence 0.5%
strong effect on RA, and obvious familial aggregation,
FDR, prevalence 1%
64
MZ, prevalence 1% has been algebraically shown to yield a familial risk of
FDR, prevalence 2% about 1.2 (REF. 27). For shared environment to explain
32
Relative risk

MZ, prevalence 2% any substantial part of familial aggregation, an unknown


16 factor of extremely strong influence or a multitude of
factors of weaker importance would be required, as is
8 the case for genetic risk factors52,53. A special case of a
4
non-genetic factor would be if patients with mild RA
were diagnosed more often if they had a family history of
2 RA. This pattern would seem likely, but since the familial
risk remains at least equally strong for more easily diag‑
1 nosed severe RA, such factors cannot explain more than
0 10 20 30 40 50 60 70 80 90 100
a small part of the familial aggregation.
Heritability (%)
Methods to directly estimate heritability through
Figure 2 | Familial risk as a function of heritability and prevalence. The relative measurement of genetic relatedness, estimated from
risk of developing rheumatoid arthritis (RA) in first degree relatives (FDRs) and genome-wide genetic markers, have developed rapidly 54.
Nature Reviews | Rheumatology
monozygotic twins (MZ) of patients with RA, compared with that in relatives of non‑RA In contrast to estimates based on family history, which,
controls, as a function of the heritability is shown. For a given disease prevalence, the as described earlier, might overestimate heritability,
heritability gives a specific familial risk. The range should cover plausible values for RA,
genome-based methods underestimate it because they
and in most populations is 0.5–1.0%. Although the translation between heritability and
can only estimate the proportion tagged by the meas‑
familial risk is dependent on prevalence, the dependence is slight for FDRs at moderate
levels of heritability. ured markers and miss effects from rare and non-tagged
variants. The latter estimates should be free from envi‑
ronmental and non-additive influences, however, and
several environmental risk factors have been confirmed, the comparison with pedigree-based estimates should
including smoking 46 and alcohol use47,48, some evidence be instructive. Estimates so far have varied considerably,
suggests that any overestimation would be limited. however. For instance those from Stahl and colleagues53
of 45% are close to pedigree-based heritability, whereas
Nature, not nurture? Lee et al.55 calculated heritability of anti-citrullinated pep‑
Although familial clustering might seem to be expected tide antibody (ACPA)-positive RA to be 19% (TABLE 2).
from shared environmental factors, a study that com‑ Clearly, the latter estimate is lower than most of those
bined information on family history with established from FDR and twin studies, which, if this estimate is
environmental and genetic RA risk factors showed no correct, suggests there must be plenty of rare variants,
part of the familial aggregation was explained by various non-additive genetic effects, or shared environmental
non-genetic risk factors. By contrast, the shared epitope influences on RA to explain the difference to heritability
and 76 single-nucleotide polymorphisms explained estimated from familial aggregation.
about 20% of the familial risk49. These findings might
seem surprising, but are in line with the general find‑ Seropositive RA
ing from twin studies that shared family environment Two-thirds of patients with RA are seropositive for rheu‑
is of negligible importance for most studied diseases50. matoid factor (RF) or ACPAs. Recent genetic studies have
Indeed, an individual risk factor would need an extreme shown differences and similarities in the genetic risk
association with RA and extreme familial aggregation to factors for patients with seropositive and seronegative
yield more than a modest effect on familial aggregation51. RA52,56,57, which supports the view that these phenotypes

Table 1 | Total-population studies of familial aggregation of rheumatoid arthritis


Authors Country Ascertainment of RA Number Estimate (95% CI)
of patients
Grant et al., Iceland Patients attending an RA 1,412 λ for siblings 4.4 (3.3–5.7), for
2001 (REF. 29) outpatient clinic, covering about parents 3.9 (2.6–5.5) and for offspring
60% of all RA patients in Iceland 2.8 (1.6–4.2)*
Hemminki et al., Sweden Hospitalized patients 47,361 IRR for siblings 4.6 (3.0–7.2) and for
2009 (REF. 31) parents 3.0 (2.8–3.2)
Somers et al., Denmark Hospitalizations and 9,118 HR for mother–daughter 2.7 (2.4–3.1)
2013 (REF. 32) outpatient visits and for father–son 2.9 (2.2–3.8)
Frisell et al., Sweden Hospitalizations and 90,372 OR for FDRs 3.2 (3.0–3.3), for siblings
2013 (REF. 33) outpatient visits 3.6 (3.1–4.0), for parents 3.1 (2.9–3.3)
and for offspring 3.2 (3.0–3.5)
FDR, first-degree relative; HR, hazard ratio; IRR, incidence rate ratio; OR, odds ratio; RA, rheumatoid arthritis. *λ denotes risk of
disease for a type of relative to an affected individuals divided by the risk in the general population.

4 | ADVANCE ONLINE PUBLICATION www.nature.com/nrrheum


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

Box 2 | Heritability 3.95 [95% CI 3.72–4.19] in FDRs) compared with that


for seronegative RA (2.47 [2.24–2.72])33,66. These risks
Heritability of a trait is defined as the proportion of its variance attributable to additive translate to heritability of 40–50% for seropositive and
genetic effects (that is, the sum of the individual average effects of all alleles across the 20–30% for seronegative RA, and the degree of familial
genome). Heritability is sometimes referred to as narrow-sense heritability to avoid aggregation overlaps substantially 33,49,66,67. Although the
mistaking it for the larger proportion that is due to all genetic effects (broad-sense
validity of the serology may be questioned in diagnoses
heritability), which would also contain effects due to dominance and interactions
between alleles at different loci. The concept of heritability, although alternatively based purely on registry information, the pattern seemed
named, was introduced by Fisher in 1918 (REFS 95,96), who showed that the expected to be strengthened when assessed in subsets of patients
correlation between relatives’ phenotypes could be calculated. In reverse, observed with more-robust diagnoses33 and more-contemporary
phenotypic correlations between relatives can be used to estimate heritability38,39. patients66, and in the EIRA study population, where
Fisher’s model, which was extended by Falconer38,39, assumes that risk of disease is the all participants had been tested for RF and ACPA49.
sum of and interactions between genetic and environmental factors. The tetrachoric Nevertheless, these studies used different samples from
correlation between relatives, therefore, would be the sum of contributions from these within the same population, and further studies done in
components, and the relative importance of genetic and environmental factors settings outside Sweden would be valuable.
calculated by solving equation systems of relatives with different relatedness. In recent analyses of data from the Nurses’ Health
In practice, simplifications are needed, and heritability is often estimated from twins,
Study, family history of RA and/or systemic lupus erythe­
assuming contributions only from additive genetics and shared environmental factors:
matosus was equally associated with seropositive and
• Correlation for monozygotic twins = 1 × heritability + 1 × shared environment
seronegative RA, but family history was self-reported,
• Correlation for dizygotic twins = 0.5 × heritability + 1 × shared environment frequently many years after disease onset. The high prev‑
Alternatively, hereditability is estimated from first-degree relatives (FDRs), assuming alence (~30% of RA patients reported a family history of
only contributions from additive genetics: RA and/or SLE) suggests that recall bias or diagnostic
• Correlation for FDRs = 0.5 × heritability uncertainty might have affected results68,69. Heritability
Heritability estimates have practical applications in animal breeding, for instance to estimated from observed genetic relatedness is consistent
identify traits responsive to selection. Heritability is used in humans to try to answer the with a considerable difference between sero­positive  and
question of nature versus nurture or to set an upper limit for how much all risk alleles seronegative RA, being 19% for ACPA-positive RA
explain a given disease, which acts as a benchmark against which to compare currently and zero for ACPA-negative RA55.
identified genes. Family studies have also addressed the nature of the
genetic overlap between seropositive and seronegative
RA. Part of the difference in heritability is probably caused
are distinct but related. By focusing on genome-wide sig‑ by the stronger associations between the HLA region and
nificant alleles, these studies have limited ability to assess seropositive RA than seronegative RA56. Nevertheless,
the magnitude of the difference between phenotypes, the strongest known genetic risk factor, the HLA‑DR4
and whether the lower number of associations with sero­ shared epitope, explains only a minor part of the higher
negative RA reflects lower heritability or is merely due to familial risk49, meaning that other factors must also be
limited power remains unclear. involved. The familial co‑aggregation of seropositive and
Family studies might provide the answer. Studies sero­negative RA is substantial, with family history of one
published in the 1950s and 1960s consistently reported being a strong predictor of onset of the other and both
a stronger familial aggregation for seropositive than sero­ having similar familial co‑aggregation with other sys‑
negative RA2,9,58. Of note, however, these early studies were temic inflammatory diseases66. This finding suggests that
performed before the current classification criteria for RA the alleles associated with other chronic inflammatory
were established. The seronegative RA group, therefore, diseases are prim­arily shared by the RA subsets. Familial
probably contained patients who would now be diagnosed co‑aggregation would also be expected if a proportion of
as having other arthritis-related diseases, which might people with seronegative RA should be classified as sero‑
have diluted the true familial aggregation. Among later positive. This likelihood is supported by studies in which
studies, only a few reported a stronger familial aggregation substantial proportions of patients negative for anti-cyclic
for seropositive RA34 and most reported no significant citrullinated protein antibodies carried other specific
difference between phenotypes, although power was gen‑ ACPAs, anti-mutated citrullinated vimentin antibodies,
erally limited and point estimates favoured seropositive or anti-carbamylated protein  antibodies70,71.
RA17,27,28,59–62. A reanalysis of the data from twins reported
by Silman and colleagues18 has been widely referenced as Missing heritability
evidence for equal heritability (~67% in ACPA-positive Genome-wide association studies in collaborative pro‑
and ACPA-negative RA)63,64, but the estimates were jects that have pooled several large RA cohorts have
based on only two and zero ACPA-negative concordant identified more than 100 risk loci associated with
monozygotic and dizygotic pairs, respectively 65. Two RA52,72,73. The proportion of heritability not yet explained
other twin studies have presented estimates separated by by identified risk alleles is sometimes referred to as the
ACPA serostatus, but with too low power for the findings missing heritability. It provides important contextual‑
to be informative41,43. Alternative designs are needed to ization for further genetic studies of RA, indicating the
explore the role of seropositive RA further. amount of genetic effects still to be identified. From
In an analysis of the data of the Swedish national reg‑ the perspective of familial risk, this quantity is interest‑
istries and the EIRA case–control study, familial aggrega‑ ing because it gives an idea of when (or if) family history
tion of seropositive RA was clearly increased (odds ratio could be replaced by genetic factors as predictors of RA.

NATURE REVIEWS | RHEUMATOLOGY ADVANCE ONLINE PUBLICATION | 5


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

Table 2 | Studies estimating the heritability of rheumatoid arthritis


Study Sample Relationship Heritability estimate (95% CI) Main assumptions
Overall Seropositive Seronegative
Lawrence, 1970 Manchester family survey* FDRs — 60% (36–84)‡ — Only additive genetic effects
(REF. 2)
Lawrence, 1970 380 twin pairs from ARC Twins — MZ, 70%§; — Only additive genetic effects
(REF. 2) twin survey DZ, 60%§
MacGregor et al., 27,000 Finnish twins, linked Twins 65% (50–77) — — Only additive genetic
2000 (REF. 40) to sickness insurance register, effects after finding family
from Aho et al.17 environment not significantly
different from zero
MacGregor et al., 203 UK twin pairs recruited Twins 53% (40–65)|| — — Only additive genetic
2000 (REF. 40) through clinics (~33%) and a (proband‑wise effects after finding family
media campaign (~67%), from selection) environment not significantly
Silman et al.18 different from zero
van der Woude 148 UK twin pairs (the subset Twins 66% (44–75)|| 68% (55–79) 66% (21–82) Only additive genetic effects
et al., 2009 from Silman et al.18 with serum
(REF. 65) samples)
Stahl et al., 2012 5,485 patients with RA from Bayesian — 36%¶ — Only additive genetic effects
(REF. 53) Stahl et al.73 and Wellcome polygenic
Trust study98 model
Stahl et al., 2012 5,485 patients with RA from GCTA — 45%¶ — Only additive effects of
(REF. 53) Stahl et al.73 and Wellcome tagged variants
Trust study98
Svendsen et al., 45,000 Danish twins, screened Twins 12% (0–76). — — Only additive genetic
2013 (REF. 43) for RA C: 50% (0–72) effects and shared family
environment
Frisell et al., 2013 90,000 patients with RA from FDRs 41% (39–42)# 45% (42–47)# 28% (23–33)# Only additive genetic effects
(REF. 33) Swedish national registries
Frisell et al., 2013 2,871 Swedish patients with FDRs 37% (32–44)# 52% (42–61)# 21% (0–44)# Only additive genetic effects
(REF. 33) RA from the EIRA study
Haj Hensvold ~7,000 Swedish twins, from RA Twins 39% (0–66), 41% (0–74), — Only additive genetic
et al., 2015 patient registry C: 0 (0–44) C: 2 (0–54) effects and shared family
(REF. 41) environment
Lee et al., 2015 9,638 patients with RA from GCTA — 19.4% –1% (−5 to 5)‡ Only additive effects of
(REF. 55) Stahl et al.73, Okada et al.52, 18–20.8)‡ tagged variants
and ACPA-negative cases
from EIRA study
ACPA, anti-citrullinated protein antibody; C, common family environment; DZ, dizygotic twins; FDR, first-degree relative; GCTA, genome-wide complex trait
analysis; MZ, monozygotic twins. *No further information about this sample is available. ‡Confidence intervals calculated from ±1.96 × SE. §Standard errors not
reported. ||Difference in estimates depend on assumptions about sampling scheme: MacGregor et al. assumed that each pair was recruited through affected index
twin (λMZ = 15) and van der Woude et al. used a model assuming that pairs were symmetrical99, which effectively counts concordant pairs twice (λMZ = 27). ¶Estimate
derived from publication by summing reported estimates for MHC and non-MHC. #CIs not included in original publication.

The proportion of the heritability explained by loci heritability we have reported of ~50% for ACPA-positive
identified so far has been claimed to be quite high, RA and ~20% for ACPA-negative RA on the basis of
~50%64,72,74 or even higher 63 in some reports. However, familial aggregation33, the missing heritability would be
these estimates have been based on different measures at least 50–60% and 65–75%, respectively.
of heritability and the degree of how much is explained In line with these estimates, a study showed that the
by the HLA region has varied73,75–77. In a study that used shared epitope and leading single-nucleotide polymor‑
molecular data, variants identified in the HLA region phisms from 76 identified loci only explained about
would explain 12.7% of the liability to develop ACPA- 20% of the familial aggregation for ACPA-positive RA
positive RA78, but the proportion was much lower for and virtually nothing for ACPA-negative RA49. Another
ACPA-negative RA. Non-HLA loci explain far less, with study found that self-reported familial RA was associ‑
100 explaining only 5.5% of RA heritability in European ated with substantially increased prediction of RA when
populations and 4.7% in an Asian population, although added to a model that included a literature-based genetic
no reference was made to what value the heritability was risk score for the disease69. These reports suggest that
assumed to have52. All identified loci, therefore, seem many alleles remain unidentified, and the predictive
to explain <20% of liability for ACPA-positive RA and value of the family history of disease is not likely to be
<5% for ACPA-negative RA. Thus, compared to the replaced by genotyping any time soon.

6 | ADVANCE ONLINE PUBLICATION www.nature.com/nrrheum


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

Family history in the clinical setting


Individuals referred to rheumatology clinics are rou‑ General population
FDRs of RA patients Healthy RA
tinely asked about family history of chronic inflam‑
matory diseases as part of the diagnostic work‑up.

Population proportion
This practice seems sensible considering the familial
aggregation of RA, including cross-disease aggregation
with other inflammatory arthritis66, but the usefulness
of such information is not always clear. Of note, the
increased risks described in this Review are related to
the general population, and would probably be quite dif‑
ferent among individuals who present with symptoms of
arthralgia or systemic inflammation. Indeed, some stud‑
ies show that self-reported family history of RA can be
a predictor for not having RA in the clinical setting 79,80, Liability to develop RA
probably because self-reporting is unreliable and subject
to recall bias. Studies have suggested that the positive Figure 3 | The liability-threshold model. The distribution
predictive value of self-report is 25–75%, depending of risk of rheumatoid arthritis (RA) is shown in the general
Nature Reviews | Rheumatology
on the phrasing of the question23,59. Familial aggrega‑ population and first-degree relatives (FDRs) of patients
tion of RA seems to be mostly specific, although there with RA according to the standard model of disease
is some familial co‑aggregation with connective-tissue development in quantitative genetics38,97. Under this
model, everyone is assumed to have a value on a liability
and autoimmune diseases31,66,81 but little to none with
continuum related to risk of developing RA, but only those
osteoarthritis or unspecified arthralgia66. If family his‑
with values above a certain threshold actually develop the
tory of disease is to help in making a diagnosis, care disease. An individual’s liability is a sum of his or her risk
must be taken that the individual’s report is as accurate factors, such as individual risk alleles, and lifestyle factors
as possible. like smoking, with each additional risk factors moving the
Whether family history of RA is of any value for person closer to the threshold. In populations at increased
clinical care and prediction beyond diagnosis is risk compared with the general population, the proportion
unclear. The liability-threshold model (FIG. 3) suggests past the threshold is higher. Heritability estimates
that it would be, since patients with RA and a fam‑ correspond to the proportion of the variance in liability
ily history of RA would on average have higher lia‑ that is explained by additive genetic effects, where
additive refers to the average effects summing without
bility than RA patients without such a family history.
interactions on the liability scale.
Assuming that the risk of developing RA displays some
correlation with disease activity, we could hypothesize
that patients with familial RA should have more severe
disease, and possibly be more resistant to treatment. As with disease severity in general, there is limited evi‑
So far, however, there is little evidence to support dence to support such predictive value. In one prospec‑
this hypothesis. tive study, family history of RA was linked to increased
risk of radiographic progression90, whereas in another
Disease severity no association was found61. In one study, while family
Early studies generally found increased familial risks history of RA did not predict response to methotrex‑
for erosive and active RA2, but since these studies pre‑ ate or anti-TNF therapy 67, there was some indication
dated modern RA classification criteria, the findings that 1‑year drug survival in FDRs taking TNF inhib‑
might reflect diagnoses with high specificity rather itors was predictive of the patient’s own drug survival
than the relevant underlying pathogenic mechanisms. on these drugs67. This finding is tentatively supported
By contrast, later studies showed, with few exceptions82, by several genotype-based heritability estimates. In
no substantial differences in clinical characteristics a small Icelandic study, heritability of joint destruc‑
(beyond seropositive status and age at onset) between tion was estimated to be 60%91, and in a Dutch study
familial and sporadic RA34,61,62,67,83–85. Later studies also the heritability for change in swollen and tender joint
demonstrate no familial aggregation of RA disease char‑ counts in patients taking TNF inhibitors was 60–80%92.
acteristics per se86,87, although, again, with the exception By contrast, a UK study found no significant heritabil‑
of seropositive status and age at onset 33,88,89 and in one ity of response to any TNF inhibitors by any measure,
study rheumatic nodules88, which otherwise would have although point estimates for heritability were possibly
been expected if these disease characteristics identified slightly increased in the subset treated with anti-TNF
aetiologically distinct subsets. monoclonal antibodies93. These studies were limited by
low power and need to be replicated, but so far is seems
Treatment response that family history of specific treatment responses could
Since a substantial proportion of the identified RA risk be predictive of response, whereas overall family history
loci tag pathways targeted by current therapies or that of RA is not. The clinical usefulness of this information
are involved in general inflammatory response52, it is unknown, but it could mean that attention should
seems reasonable that family history of RA might be pre‑ be paid to family history of responsiveness to different
dictive of disease progression and treatment response. treatment approaches.

NATURE REVIEWS | RHEUMATOLOGY ADVANCE ONLINE PUBLICATION | 7


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

Conclusions antirheumatic therapies. Current evidence clearly sup‑


Family history of RA is an old concept, but fascinat‑ ports family history remaining an integral part of the
ing new data continue to emerge. Registry-based work‑up for RA, but indicates that care should also be
total-population studies have validated findings from ear‑ taken to ensure that relatives do indeed have RA, rather
lier studies in smaller and selected samples, and it is now than osteoarthritis or other arthralgia that confer no
beyond doubt that family history is one of the strongest increase in RA risk among FDRs. Asking about the
risk factors for developing RA, conferring a twofold to antirheumatic treatments that affected relatives receive
fourfold increase in risk among FDRs. Despite the pro‑ might be one way to discriminate between non‑RA
gress made in genetic association studies, this predictive joint complaints and frank RA. Additionally, the risk
value is unlikely to be replaced by genetic markers any associated with family history decreases with increas‑
time soon. ing distance of relatives and, therefore, not searching
From an aetiological perspective, studies in FDRs further than second-degree relatives seems reasonable.
support heritability of about 40% for RA, which is lower Limited evidence so far suggests family aggregation of
than the often-cited 60% that was based on twin studies. treatment response but not disease course, but there are
Heritability is not modified by sex, but evidence suggests no clear-cut findings. Yet, in situations where the choice
that seropositive RA (for RF or ACPA) is more familial between two treatments is neutral in all other respects,
than seronegative RA. However, there seems to be little a family history of lack of response to one of them may
difference between seropositive and seronegative RA strengthen the choice for the other.
in genetic and familial risk factors that are shared with Finally, looking ahead, since the largest total-population
other arthritis-related diseases. studies so far have used Scandinavian registers, repli‑
From a clinical point of view, information on fam‑ cations in other populations would be desirable. More
ily history might inform the work‑up for RA diagno‑ research is also needed on family history as a predictive
sis and predictions for disease course and response to factor for prognosis.

1. Blumberg, B. S. Genetics and rheumatoid arthritis. 16. Hellgren, L. Inheritance of rheumatoid arthritis. 31. Hemminki, K., Li, X., Sundquist, J. & Sundquist, K.
Arthritis Rheum. 3, 178–185 (1960). Acta Rheumatol. Scand. 16, 211–216 (1970). Familial associations of rheumatoid arthritis with
2. Lawrence, J. S. Heberden Oration, 1969. Rheumatoid 17. Aho, K., Koskenvuo, M., Tuominen, J. & Kaprio, J. autoimmune diseases and related conditions.
arthritis — nature or nurture? Ann. Rheum. Dis. 29, Occurrence of rheumatoid arthritis in a nationwide Arthritis Rheum. 60, 661–668 (2009).
357–379 (1970). series of twins. J. Rheumatol. 13, 899–902 (1986). 32. Somers, E. C., Antonsen, S., Pedersen, L.
3. Batterman, R. C., Grossman, A. J. & Leifer, P. 18. Silman, A. J. et al. Twin concordance rates for & Sorensen, H. T. Parental history of lupus and
The occurrence of rheumatoid arthritis in twins. rheumatoid arthritis: results from a nationwide study. rheumatoid arthritis and risk in offspring in a
Acta Rheumatol. Scand. 2, 161–169 (1956). Br. J. Rheumatol. 32, 903–907 (1993). nationwide cohort study: does sex matter?
4. Empire Rheumatism Council Scientific Advisory 19. Bellamy, N., Duffy, D., Martin, N. & Mathews, J. Ann. Rheum. Dis. 72, 525–529 (2013).
Committee & Lewis-Faning, E. Report on an Enquiry Rheumatoid arthritis in twins: a study of 33. Frisell, T. et al. Familial risks and heritability of
Into the Aetiological Factors Associated with aetiopathogenesis based on the Australian Twin rheumatoid arthritis: role of rheumatoid factor/anti-
Rheumatoid Arthritis (British Medical Association, Registry. Ann. Rheum. Dis. 51, 588–593 (1992). citrullinated protein antibody status, number and type
1950). 20. Gregersen, P. K., Silver, J. & Winchester, R. J. of affected relatives, sex, and age. Arthritis Rheum.
5. Miall, W. E. Rheumatoid arthritis in males; an The shared epitope hypothesis. An approach to 65, 2773–2782 (2013).
epidemiological study of a Welsh mining community. understanding the molecular genetics of susceptibility to 34. Barrera, P., Radstake, T. R., Albers, J. M., van Riel, P. L.
Ann. Rheum. Dis. 14, 150–158 (1955). rheumatoid arthritis. Arthritis Rheum. 30, 1205–1213 & van de Putte, L. B. Familial aggregation of
6. Stecher, R. M. Heredity in joint diseases: the genetics (1987). rheumatoid arthritis in The Netherlands: a cross-
of Heberden’s nodes, rheumatoid arthritis, ankylosing 21. Deighton, C. M. & Walker, D. J. The familial nature of sectional hospital-based survey. European Consortium
spondylitis, rheumatic fever, gout, and osteoarthritis rheumatoid arthritis. Ann. Rheum. Dis. 50, 62–65 on Rheumatoid Arthritis families (ECRAF).
of the hip. Doc. Rheumatol. (1956) 19, 5–74 (1957). (1991). Rheumatology (Oxford) 38, 415–422 (1999).
7. Neri Serneri, G. G. & Bartoli, V. Heredopathology of 22. Wordsworth, P. & Bell, J. Polygenic susceptibility in 35. Deighton, C. M., Wentzel, J., Cavanagh, G.,
reactive mesenchymal disease (so‑called collagen rheumatoid arthritis. Ann. Rheum. Dis. 50, 343–346 Roberts, D. F. & Walker, D. J. Contribution of inherited
disease). I. Research on the hereditary factors in (1991). factors to rheumatoid arthritis. Ann. Rheum. Dis. 51,
acute primary rheumatism. Acta Genet. Med. 23. Wasmuth, A. G., Veale, A. M., Palmer, D. G. 182–185 (1992).
Gemellol. (Roma) 5, 155–189 (in Italian) (1956). & Highton, T. C. Prevalence of rheumatoid arthritis in 36. Pritchard, M. H. Evidence for a hypothetical non-HLA
8. de Blécourt, J. J., Polman, A., de Blécourt- families. Ann. Rheum. Dis. 31, 85–91 (1972). susceptibility gene in rheumatoid arthritis.
Meindersma, T., Erlee, T. J. D. & Drion, E. F. Hereditary 24. Thomas, D. J., Young, A., Gorsuch, A. N., Br. J. Rheumatol. 33, 475–479 (1994).
factors in rheumatoid arthritis and ankylosing Bottazzo, G. F. & Cudworth, A. G. Evidence for an 37. Leu, M., Czene, K. & Reilly, M. Bias correction of
spondylitis. Ann. Rheum. Dis. 20, 215–223 (1961). association between rheumatoid arthritis and estimates of familial risk from population-based cohort
9. Lawrence, J. S. & Ball, J. Genetic studies on autoimmune endocrine disease. Ann. Rheum. Dis. 42, studies. Int. J. Epidemiol. 39, 80–88 (2010).
rheumatoid arthritis. Ann. Rheum. Dis. 17, 160–168 297–300 (1983). 38. Falconer, D. S. Inheritance of liability to certain
(1958). 25. Silman, A. J., Hennessy, E. & Ollier, B. Incidence of diseases estimated from incidence among relatives.
10. Masi, A. T. & Shulman, L. E. Familial aggregation and rheumatoid arthritis in a genetically predisposed Ann. Hum. Genet. 29, 51–76 (1965).
rheumatoid disease. Arthritis Rheum. 8, 418–425 population. Br. J. Rheumatol. 31, 365–368 (1992). 39. Falconer, D. S. & Mackay, T. F. C. Introdution to
(1965). 26. Jones, M. A., Silman, A. J., Whiting, S., Barrett, E. M. Quantitative Genetics (Pearson Education Ltd, 1996).
11. O’Brien, W. M. The genetics of rheumatoid arthritis. & Symmons, D. P. Occurrence of rheumatoid arthritis 40. MacGregor, A. J. et al. Characterizing the
Clin. Exp. Immunol. 2 (Suppl.), 785–802 (1967). is not increased in the first degree relatives of a quantitative genetic contribution to rheumatoid
12. Bennett, H. & Burch, T. A. The distribution of population based inception cohort of inflammatory arthritis using data from twins. Arthritis Rheum. 43,
rheumatoid factor and rheumatoid arthritis in the polyarthritis. Ann. Rheum. Dis. 55, 89–93 (1996). 30–37 (2000).
families of Blackfeet and Pima Indians. 27. del Junco, D., Luthra, H. S., Annegers, J. F., 41. Haj Hensvold, A. et al. Environmental and genetic
Arthritis Rheum. 11, 546–553 (1968). Worthington, J. W. & Kurland, L. T. The familial factors in the development of anticitrullinated protein
13. Burch, T. A., O’Brien, W. M. & Bunim, J. J. Family and aggregation of rheumatoid arthritis and its antibodies (ACPAs) and ACPA-positive rheumatoid
genetic studies of rheumatoid arthritis and relationship to the HLA‑DR4 association. arthritis: an epidemiological investigation in twins.
rheumatoid factor in Blackfeet Indians. Am. J. Public Am. J. Epidemiol. 119, 813–829 (1984). Ann. Rheum. Dis. http://dx.doi.org/10.1136/
Health Nations Health 54, 1184–1190 (1964). 28. Koumantaki, Y. et al. Family history as a risk factor annrheumdis-2013-203947 (2013).
14. O’Brien, W. M., Bennett, P. H., Burch, T. A. & Bunim, for rheumatoid arthritis: a case–control study. 42. Svendsen, A. J., Holm, N. V., Kyvik, K., Petersen, P. H.
J. J. A genetic study of rheumatoid arthritis and J. Rheumatol. 24, 1522–1526 (1997). & Junker, P. Relative importance of genetic effects in
rheumatoid factor in Blackfeet and Pima Indians. 29. Grant, S. F. et al. The inheritance of rheumatoid rheumatoid arthritis: historical cohort study of Danish
Arthritis Rheum. 10, 163–179 (1967). arthritis in Iceland. Arthritis Rheum. 44, 2247–2254 nationwide twin population. BMJ 324, 264–266
15. Bunim, J. J., Burch, T. A. & O’Brien, W. M. Influence of (2001). (2002).
genetic and environmental factors on the occurence 30. Gulcher, J. & Stefansson, K. Population genomics: 43. Svendsen, A. J. et al. On the origin of rheumatoid
of rheumatoid arthritis and rheumatoid factor in laying the groundwork for genetic disease modeling arthritis: the impact of environment and genes —
American Indians. Bull. Rheum. Dis. 15, 349–350 and targeting. Clin. Chem. Lab. Med. 36, 523–527 a population based twin study. PLoS ONE 8, e57304
(1964). (1998). (2013).

8 | ADVANCE ONLINE PUBLICATION www.nature.com/nrrheum


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

44. Silman, A. J. Commentary: do genes or environment 64. Kurko, J. et al. Genetics of rheumatoid arthritis — a comparison of the demographic and clinical
influence development of rheumatoid arthritis? BMJ a comprehensive review. Clin. Rev. Allergy Immunol. characteristics of 956 patients. J. Rheumatol. 15,
324, 264 (2002). (2013). 400–404 (1988).
45. MacGregor, A. J., Lanchbury, J., Rigby, A. S., Kaprio, J. 65. van der Woude, D. et al. Quantitative heritability of 84. Silman, A. J., Ollier, W. E. & Currey, H. L. Failure
& Snieder, H. Using twin studies to label disease as anti-citrullinated protein antibody-positive and anti- to find disease similarity in sibling pairs with
genetic or environmental is inappropriate. BMJ 324, citrullinated protein antibody-negative rheumatoid rheumatoid arthritis. Ann. Rheum. Dis. 46, 135–138
1100 (2002). arthritis. Arthritis Rheum. 60, 916–923 (2009). (1987).
46. Sugiyama, D. et al. Impact of smoking as a risk factor 66. Frisell, T. et al. Familial aggregation of arthritis- 85. Deighton, C. M. & Walker, D. J. What factors
for developing rheumatoid arthritis: a meta-analysis of related diseases in seropositive and seronegative distinguish probands from multicase rheumatoid
observational studies. Ann. Rheum. Dis. 69, 70–81 rheumatoid arthritis: a register-based case–control arthritis same sex sibships from sporadic disease?
(2010). study in Sweden. Ann. Rheum. Dis. 75, 183–189 J. Rheumatol. 19, 237–241 (1992).
47. Di Giuseppe, D., Alfredsson, L., Bottai, M., Askling, J. (2016). 86. Balsa, A. et al. Clinical and immunogenetic
& Wolk, A. Long term alcohol intake and risk of 67. Frisell, T., Saevarsdottir, S. & Askling, J. Does a family characteristics of European multicase rheumatoid
rheumatoid arthritis in women: a population based history of RA influence the clinical presentation and arthritis families. Ann. Rheum. Dis. 60, 573–576
cohort study. BMJ 345, e4230 (2012). treatment response in RA? Ann. Rheum. Dis. http:// (2001).
48. Kallberg, H. et al. Alcohol consumption is associated dx.doi.org/10.1136/annrheumdis-2015-207670 87. Jarvinen, P., Koskenvuo, M., Koskimies, S.,
with decreased risk of rheumatoid arthritis: results (2015). Kotaniemi, K. & Aho, K. Rheumatoid arthritis in
from two Scandinavian case–control studies. 68. Sparks, J. A. et al. Contributions of familial rheumatoid identical twins: a clinical and immunogenetic study
Ann. Rheum. Dis. 68, 222–227 (2009). arthritis or lupus and environmental factors to risk of of eight concordant pairs derived from a nationwide
49. Jiang, X. et al. To what extent is the familial risk of rheumatoid arthritis in women: a prospective cohort twin panel. Scand. J. Rheumatol. 20, 159–164
rheumatoid arthritis explained by established study. Arthritis Care Res. (Hoboken) 66, 1438–1446 (1991).
rheumatoid arthritis risk factors? Arthritis Rheumatol. (2014). 88. Jawaheer, D., Lum, R. F., Amos, C. I., Gregersen, P. K.
67, 352–362 (2015). 69. Sparks, J. A. et al. Improved performance of & Criswell, L. A. Clustering of disease features
50. Polderman, T. J. et al. Meta-analysis of the heritability epidemiologic and genetic risk models for rheumatoid within 512 multicase rheumatoid arthritis families.
of human traits based on fifty years of twin studies. arthritis serologic phenotypes using family history. Arthritis Rheum. 50, 736–741 (2004).
Nat. Genet. 47, 702–709 (2015). Ann. Rheum. Dis. 74, 1522–1529 (2014). 89. Cui, J. et al. The influence of polygenic risk scores on
51. Khoury, M. J., Beaty, T. H. & Liang, K. Y. Can familial 70. van Beers, J. J. et al. ACPA fine-specificity profiles in heritability of anti-CCP level in RA. Genes Immun. 15,
aggregation of disease be explained by familial early rheumatoid arthritis patients do not correlate 107–114 (2014).
aggregation of environmental risk factors? with clinical features at baseline or with disease 90. Rojas-Villarraga, A. et al. Familial disease, the
Am. J. Epidemiol. 127, 674–683 (1988). progression. Arthritis Res. Ther. 15, R140 (2013). HLA‑DRB1 shared epitope and anti-CCP antibodies
52. Okada, Y. et al. Genetics of rheumatoid arthritis 71. Jiang, X. et al. Anti-CarP antibodies in two large influence time at appearance of substantial joint
contributes to biology and drug discovery. Nature cohorts of patients with rheumatoid arthritis and their damage in rheumatoid arthritis. J. Autoimmun. 32,
506, 376–381 (2014). relationship to genetic risk factors, cigarette smoking 64–69 (2009).
53. Stahl, E. A. et al. Bayesian inference analyses of the and other autoantibodies. Ann. Rheum. Dis. 73, 91. Knevel, R. et al. Genetic predisposition of the
polygenic architecture of rheumatoid arthritis. 1761–1768 (2014). severity of joint destruction in rheumatoid arthritis:
Nat. Genet. 44, 483–489 (2012). 72. Eyre, S. et al. High-density genetic mapping identifies a population-based study. Ann. Rheum. Dis. 71,
54. Yang, J., Lee, S. H., Goddard, M. E. & Visscher, P. M. new susceptibility loci for rheumatoid arthritis. 707–709 (2012).
GCTA: a tool for genome-wide complex trait analysis. Nat. Genet. 44, 1336–1340 (2012). 92. Umicevic Mirkov, M. et al. Estimation of heritability
Am. J. Hum. Genet. 88, 76–82 (2011). 73. Stahl, E. A. et al. Genome-wide association study meta- of different outcomes for genetic studies of TNFi
55. Lee, S. H. et al. New data and an old puzzle: the analysis identifies seven new rheumatoid arthritis risk response in patients with rheumatoid arthritis.
negative association between schizophrenia and loci. Nat. Genet. 42, 508–514 (2010). Ann. Rheum. Dis. 74, 2183–2187 (2014).
rheumatoid arthritis. Int. J. Epidemiol. 44, 1706–1721 74. Bowes, J. & Barton, A. Recent advances in the genetics of 93. Plant, D., Wilson, A. G. & Barton, A. Genetic and
(2015). RA susceptibility. Rheumatology (Oxford) 47, 399–402 epigenetic predictors of responsiveness to
56. Han, B. et al. Fine mapping seronegative and (2008). treatment in RA. Nat. Rev. Rheumatol. 10, 329–337
seropositive rheumatoid arthritis to shared and distinct 75. Raychaudhuri, S. et al. Common variants at CD40 and (2014).
HLA alleles by adjusting for the effects of heterogeneity. other loci confer risk of rheumatoid arthritis. 94. McGue, M. When assessing twin concordance, use
Am. J. Hum. Genet. 94, 522–532 (2014). Nat. Genet. 40, 1216–1223 (2008). the probandwise not the pairwise rate. Schizophr. Bull.
57. Padyukov, L. et al. A genome-wide association study 76. Deighton, C. M., Walker, D. J., Griffiths, I. D. 18, 171–176 (1992).
suggests contrasting associations in ACPA-positive & Roberts, D. F. The contribution of HLA to rheumatoid 95. Fischer, R. The correlation between relatives on the
versus ACPA-negative rheumatoid arthritis. arthritis. Clin. Genet. 36, 178–182 (1989). supposition of Mendelian inheritance.Trans. R.
Ann. Rheum. Dis. 70, 259–265 (2011). 77. Hasstedt, S. J., Clegg, D. O., Ingles, L. & Ward, R. H. Soc. Edinb. 52, 399–433 (1918).
58. de Blécourt, J. J., Boerma, F. W. & Vorenkamp, E. O. HLA-linked rheumatoid arthritis. Am. J. Hum. Genet. 96. Visscher, P. M., Hill, W. G. & Wray, N. R. Heritability
Rheumatoid arthritis (R.A.) factor in near relatives of 55, 738–746 (1994). in the genomics era — concepts and misconceptions.
sero-positive and sero-negative patients with 78. Raychaudhuri, S. et al. Five amino acids in three HLA Nat. Rev. Genet. 9, 255–266 (2008).
rheumatoid arthritis. Ann. Rheum. Dis. 21, 339–341 proteins explain most of the association between MHC 97. Todorov, A. A. & Suarez, B. K. Genetic liability model.
(1962). and seropositive rheumatoid arthritis. Nat. Genet. 44, Encyclopedia of Biostatistics (John Wiley & Sons,
59. Wolfe, F., Kleinheksel, S. M. & Khan, M. A. Prevalence 291–296 (2012). 2005).
of familial occurrence in patients with rheumatoid 79. Thompson, A. E. et al. Comprehensive arthritis referral 98. Wellcome Trust Case Control Consortium. Genome-
arthritis. Br. J. Rheumatol. 27 (Suppl. 2), 150–152 study — phase 2: analysis of the comprehensive wide association study of 14,000 cases of seven
(1988). arthritis referral tool. J. Rheumatol. 41, 1980–1989 common diseases and 3,000 shared controls. Nature
60. Laivoranta-Nyman, S. et al. Immunogenetic (2014). 447, 661–678 (2007).
differences between patients with familial and non- 80. Larkin, J. G. Family history of rheumatoid arthritis — 99. Tsonaka, R., van der Woude, D. & Houwing-
familial rheumatoid arthritis. Ann. Rheum. Dis. 59, a non-predictor of inflammatory disease? Duistermaat, J. Marginal genetic effects estimation
173–177 (2000). Rheumatology (Oxford) 49, 608–609 (2010). in family and twin studies using random-effects
61. Radstake, T. R. et al. Familial versus sporadic 81. Cardenas-Roldan, J., Rojas-Villarraga, A. models. Biometrics 71, 1130–1138 (2015).
rheumatoid arthritis (RA). A prospective study in an & Anaya, J. M. How do autoimmune diseases cluster
early RA inception cohort. Rheumatology (Oxford) 39, in families? A systematic review and meta-analysis. Author contributions
267–273 (2000). BMC Med. 11, 73 (2013). T.F. researched data for the article, J.A. and T.F. wrote the
62. Radstake, T. R. et al. Genetic anticipation in 82. Deighton, C. M., Roberts, D. F. & Walker, D. J. Effect of article, and all authors made substantial contributions to dis-
rheumatoid arthritis in Europe. European Consortium disease severity on rheumatoid arthritis concordance cussions of the content and review/editing of the manuscript
on Rheumatoid Arthritis Families. J. Rheumatol. 28, in same sexed siblings. Ann. Rheum. Dis. 51, 943–945 before submission.
962–967 (2001). (1992).
63. de Vries, R. Genetics of rheumatoid arthritis: time for 83. Wolfe, F., Kleinheksel, S. M. & Khan, M. A. Competing interests statement
a change! Curr. Opin. Rheumatol. 23, 227–232 (2011). Familial versus sporadic rheumatoid arthritis: The authors declare no competing interests.

NATURE REVIEWS | RHEUMATOLOGY ADVANCE ONLINE PUBLICATION | 9


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.

You might also like