doi: 10.1046/j.1529-8817.2003.00062.x
An Analysis of Consanguinity and Social Structure
Within the UK Asian Population Using
Microsatellite Data
A. D. J. Overall1,‡,∗ , M. Ahmad1 , M. G. Thomas2 and R. A. Nichols1
1
School of Biological Sciences, Queen Mary, University of London, London E1 4NS, UK
2
The Centre for Genetic Anthropology, University College London, London WC1E 6BT, UK
Summary
We analysed microsatellite genotypes sampled from the Pakistani and Indian communities in Nottingham, UK, to
investigate the genetic consequences of substructuring mediated by traditional marriage customs. The application of
a recently developed likelihood approach identified significant levels of population substructure within the Pakistani
community as a whole, as well as within the finer divisions of castes and biradheri. In addition, high levels of cryptic
or unacknowledged consanguinity were detected within subgroups of this community, including biradheri. The
Indian sample showed no significant evidence of either substructure or consanguinity. We demonstrate that estimates
of disease gene frequencies can be inaccurate unless they are made jointly with estimates of population substructure
and consanguinity ((θ ≡ F ST ) and C). The magnitude of these estimates also highlights the importance of accounting
for the finer scale of social structuring when making decisions regarding the risk of recessive disorders in offspring.
Keywords: Consanguinity, substructure, inbreeding, microsatellites, UK Asian population.
Introduction
The UK Asian population, now numbering over a million individuals, was largely founded by migrants from
the Indian sub-continent during the 1950s. This particular episode of migration was motivated by numerous factors, including large dam building projects in
Kashmir and the booming textile industry throughout many of the UK Midland cities at this time.
The majority of these early founders originated from
just a few locations throughout the Sub-continent,
namely the Punjab, Gujarat and to a smaller extent
Bangladesh. Each of these regions is quite distinctive in their religions and culture which, in turn,
∗
Correspondence to: Dr. Andrew D. J. Overall, Institute of Cell,
Animal and Population Biology, Ashworth Laboratories, West
Mains Road, University of Edinburgh, Edinburgh EH9 3JT, UK.
Tel: +44 (0)0131 651 3047. Fax: +44 (0)0131 650 5455. E-mail:
andy.overall@ed.ac.uk
‡
Present address: Institute of Cell, Animal and Population Biology, Ashworth Laboratories, West Mains Road, University of
Edinburgh, Edinburgh EH9 3JT, UK.
C
University College London 2003
have influenced the pattern of settlement of family
members and the subsequent development of communities (Ballard, 1994). Although marriage traditions were not abandoned on arrival, the distance between Britain and the sub-continent has modified the
traditional relationships between couples, particularly
within cultures known to have quite elaborate marriage practices, such as consanguineous unions within
Pakistani Muslim communities. An earlier study (Darr
& Modell, 1988) identified an increase in consanguinity within a British-Pakistani Muslim community and
suggested that this increase resulted from the direct effects of migration. Generally, however, studies of social
structure have focused largely on the people of the subcontinent itself. For example, consanguinity has been
estimated to range between 38 – 49% within the Punjab
(Bittles et al. 1992) and differentiation between 8 Indian
Hindu castes has been estimated to be high (GST = 0.04
(Mukherjee et al. 1999)). How these studies relate to
the fine-scale social structure of the present Asian population of the UK is largely unknown. For this study,
Annals of Human Genetics (2003) 67,525–537
525
A. D. J. Overall et al.
genetic and sociological data were collected from the
Asian community of Nottingham, UK, to measure the
degree of consanguinity and substructure within this
community and to identify any correlation between social networks, such as caste and biradheri (extended family networks) and patterns of multilocus homozygosity.
Consanguinity and substructure within populations
both have their consequences for health. Consanguinity, usually defined as marriage between second cousins
or closer, can lead to an increased birth prevalence of
recessive disorders through inheritance of a copy of
a recessive allele from each parent (Devi et al. 1987;
Chitty & Winter, 1989; Özlap et al. 1990; Bundey &
Alam, 1993; Zlotogora, 1997; Hutchesson et al. 1998;
Modell & Darr, 2002). There is also evidence that consanguinity can lead to an increased likelihood of spontaneous abortion (Hussain, 1998) and it has been estimated that first cousin progeny experience 4.4% more
pre-reproductive mortalities than non-consanguineous
offspring (Bittles & Neel, 1994). There is also evidence
that consanguineous couples are more likely to have had
consanguineous parents (Darr & Modell, 1988; Hussain,
1998), resulting in an accumulation of identity of alleles
at loci above that expected simply through the consideration of parental relatedness. A detailed review of the
outcomes of consanguineous offspring and its relevance
to clinical genetics is given by Bittles (2001). Substructure within populations can also lead to a localised increase in the frequency of a recessive allele, and hence
disorder (Heinisch et al. 1995; Bittles, 2001), but there
are important differences with the case of consanguinity (Overall et al. 2002). In a subdivided population,
genetic drift and founder events can elevate the local
frequency of deleterious alleles, especially in societies
that have strongly endogamous kinship groups. Because
the increased frequency is due to chance effects, different alleles are likely to benefit in different subpopulations. For example, certain inherited disorders have
been found to occur almost exclusively within individual tribal groups in Oman (Rajib & Patton, 1999). On
the other hand, polygenic disorders due to epistatic interactions between alleles at multiple genes are more
likely in populations with consanguineous marriage,
as consanguineous offspring have an elevated probability of identity by descent (IBD) across the whole
genome.
526
Annals of Human Genetics (2003) 67,525–537
Despite these differences in pattern of disease within
populations, differentiating consanguinity from population substructure using genetic information is not practicable using conventional genetic analyses. The alternative of asking about marriage patterns has its own
difficulties. In part, this is because of the sensitive nature
of kin-relationships; there is often difficulty in obtaining reliable information orally. Furthermore, there may
be unrecognised patterns due to the interplay between
traditional marriage practices and other factors such as
small population size or elevated endogamy in poorly
defined subgroups. Small population size in some UK
Asian populations, for example, may lead to cryptic consanguinity within traditionally caste endogamous communities. In addition, substructure may occur within
castes through the formation of extended families, or
biradheri.
It could be important to recognise this finer scale of
social structuring within the UK Asian communities,
particularly when offering advice relating to the potential risks of genetic disorders within future offspring.
This study addresses these issues by quantifying the relative contributions of consanguinity and substructure, at
the caste and biradheri level, using a recently developed
method that distinguishes between the two, by applying
genotypic information (Overall & Nichols, 2001).
Materials and Methods
Study Populations
A research worker of Punjabi descent and fluent in
Punjabi collected information using a simple questionnaire. The location of the subject’s family origin was
noted, where the origin refers to the previous home
within the sub-continent of the subject’s family. Information regarding caste and parental relatedness was
also collected. For the questionnaire, both caste and biradheri were recorded. Pakistani biradheri have been
treated in the literature as equivalent traditional social/occupational groups to the Indian castes (Shami
et al. 1994; Wang et al. 2000). Biradheri has also been
used to identify a sub-caste (Vertovec, 1994) or a localised endogamous kin-network, within which endogamous marriages are arranged (Ballard, 1990; Shaw,
1994). By this definition, individuals belonging to a
caste, such as Rajput, can belong to different and
C
University College London 2003
Consanguinity and Substructure in UK Asian Populations
relatively unconnected biradheri within this caste. We
attempted to identify individuals belonging to a tightknit kin-network within otherwise potentially substructured castes. A combination of both substructure and
consanguinity may therefore be expected within the UK
Asian communities.
The data was obtained on a voluntary and anonymous basis at a GP surgery in Nottingham, in accordance with ethical committee guidelines (ELCHA Research Ethics Committee). Staff ensured that samples
were not obtained from close relatives, although more
distantly related individuals such as these belonging to
the same biradheri, could not be easily identified. In
total, 188 individuals contributed to the questionnaire
and had buccal-scrape samples collected for subsequent
genetic analysis.
Genetic Data
DNA was extracted from buccal swabs using the
Chelex procedure (Walsh et al. 1991) and amplified at
ten microsatellite loci: D2S1338, D3S1358, D8S1179,
D16S539, D18S51, D19S433, D21S11, FGA, THO1
and vWA, using SGMPlus fluoro-labelled primers
(P.E. Biosystems, 1999). The PCR products were run
through a 5% polyacrylamide gel on an ABI 377 sequencer and the fragments were analysed using GeneScan and Genotyper software.
Genetic Analysis
The heterozygosity for each of the loci is given
in Table 2a, along with the heterozygosity for two
broadly sampled populations for comparison: US AfroCaribbean (n = 195) and US Caucasian (n = 200); data
obtained from P.E. Biosystems User Manual (1999). Differences in allele frequency distributions were quantified
as F ST values estimated for each locus using the Arlequin
software package (Schneider et al. 2000), and results are
given in Table 2b.
Genetic identity was treated in a traditional hierarchical fashion that requires identification of alleles within
individuals, subpopulations and total populations (Weir
& Cockerham, 1984). For the subsets of individuals who
could be identified as belonging to a specific group,
such as a caste or ancestral country of origin, F-statistics
were calculated using the Arlequin software package
C
University College London 2003
(Schneider et al. 2000). Estimates of F ST values for
each identified Asian location represented by approximately 20 individuals (India, Pakistan, Jullundur,
Mirpur and Kashmir) are given in Table 2c. With the
Indian castes, only two qualified (Jat-Sikh and HinduKhatri), whereas for the Pakistani castes only Rajput
numbered above 20 individuals. The remainder of the
castes were pooled and are referred to as ‘non-Rajput’.
It is believed by some historians that the Rajput now
represents the amalgamation of 36 Hindu warrior castes
on conversion to Islam, as well as comprising individuals
self-identifying with this group (Wang et al. 2000). Nevertheless, inter-caste differentiation may still be recognisable due to ancient isolation, albeit with an expected
reduction in power through comparing heterogeneous
populations. Estimates of F IS were calculated for the
same groups and are given in Table 3. This hierarchical
F-statistics approach is possible when individuals can
be reliably allocated into a series of subgroups. Using
AMOVA (Arlequin 2000; Schneider et al. 2000), for
example, would identify which level of subdivision results in the most differentiation. However, this approach
requires that individuals be allocated to specific groups,
and this allocation is difficult for the UK Asian population as social boundaries, rather than geographic barriers, may be influencing reproductive isolation. Unfortunately, social barriers within any community are difficult
to delineate. Identifying castes within Pakistani communities, therefore, unlike Indian communities, may not
accurately reflect endogamous groups. Incorrectly pooling differentiated subgroups will result in an increase in
homozygosity over Hardy-Weinberg expectations, the
Wahlund effect, which might be falsely attributed to
consanguinity. In those cases where there was unexplained excess homozygosity we used the method of
Overall & Nichols (2001) to analyse the genetic data.
This approach estimates the relative contributions of
substructure and consanguinity to excess homozygosity without prior assignment of specific subgroups.
Maximum Likelihood Estimation
of Consanguinity and Substructure
The logic underpinning the method to distinguish consanguinity and substructure can be clarified by considering the elevated allelic homozygosity that can be
Annals of Human Genetics (2003) 67,525–537
527
A. D. J. Overall et al.
observed within subpopulations, which is quantified by
F IS (Wright, 1921; Cockerham, 1973). In a large randomly mating subpopulation, the number of closely
related couples is small. A pattern of marriage that
favours higher frequencies of related pairings would
elevate the probability of genes being IBD. Consequently homozygosity in individuals (I) is inflated above
random expectations based on the subpopulation (S)
frequencies (hence the subscripts I and S). The expected effect on F IS estimates is essentially given by
F̂ I S = C kg =1 c g Rg . Here, c g is the proportion of the
consanguines (C) inbred to degree Rg (e.g., c 1 is the
proportion of the population inbred to degree R1 ,
where R1 = 1 /16 for offspring of first cousins. c 2
could be the proportion inbred to degree R2 where
R = 1 /8 for offspring of half sibs, and so on, where
2k
g =1 c g = 1). Generally, the effect on individuals with
R < 1 /32 is negligible, so practical calculations need only
consider the first five categories of inbreeding (up to
k = 5).
If subpopulations comprising the total population
have different allele frequencies, then there will be an
excess of homozygosity over expectation for an undivided panmictic population, the Wahlund effect. Some
alleles will be locally common in a particular subpopulation (S), and pairs of such alleles drawn from the same
subpopulation will occur at a higher rate than pairs of
alleles drawn from the total population, predicted from
the allele frequencies in the total population (T). This
correlation between alleles from the same subpopulation
is measured by F ST , or its equivalent θ (Cockerham,
1973).
It follows that both population substructure, or consanguinity, or some combination of the two could explain the excess homozygosity observed in a population.
Consider the case where the excess homozygosity over
HW expectations is F = 0.03.
Where there is both population substructure
and consanguinity this excess implies C kg =1
(c g [Rg + (1 − Rg )θ]) + (1 − C)θ = 0.03. Notice that
in the extreme case of no consanguineous pairings (C =
0) this reduces to θ = 0.03, so that the excess is explained entirely by differentiation between allele frequencies between the subpopulations, in accordance
with Wright’s island model (Wright, 1931). Conversely,
if there is no population substructure (θ = 0) we obtain
528
Annals of Human Genetics (2003) 67,525–537
C kg =1 c g Rg = 0.03; so that the effect is accounted for
by consanguinity.
In the case of substructure alone, it is important to
appreciate that θ relates to the increased probability of
IBD at each locus within every individual. The effect of
consanguinity is quite different. For example, if the excess homozygosity has come about through first cousin
unions, such that C × 1 /16 = 0.03, then, clearly, only
a proportion of the population is contributing to this
excess, about 50% (C = 0.5). It is only these individuals that are expected to have an increased probability
of alleles being IBD at each of their loci. The remaining 50% of the sample are expected to have genotypes
corresponding to HW expectations. For this reason, the
distribution of the number of homozygous loci, within
an individual, is different for each of these two scenarios.
It is these differences in the distribution of homozygous
loci within individuals that allows the relative contributions of consanguinity and substructure to be estimated
(Overall & Nichols, 2001).
The parameters we are interested in estimating are C,
the proportion of the population inbred through consanguineous parents and θ, the magnitude of substructure. Because we expect consanguinity to be largely between first-cousin unions, as observed by Darr & Modell
(1988), we set from our general treatment above c g =
1 and R = 1 /16 . The θ and C parameters are estimated
using , the likelihood of the sample:
p i (θ + (1 − θ) p j ) i = j
(1 − C )
=
Loci 2 p i p j (1 − θ )
i = j
Individuals
p i (R + (1 − R)(θ + (1 − θ ) p j )) i = j
+C
Loci 2 p i p j (1 − R)(1 − θ )
i = j
(1)
where p i is the allele frequency for allele i (Overall
& Nichols, 2001). This method was applied to 10,000
combinations of values for C and θ , resulting in a grid of
likelihood values. Figure 1 illustrates the results of equation (1) for sub-samples of the data that gave positive
estimates of F IS (table 3).
Results
The Asian population of Nottingham broadly consists
of individuals originating from two principal regions of
C
University College London 2003
Consanguinity and Substructure in UK Asian Populations
Table 1 Sample data (numbers of individuals (I), 1st cousin offspring (1C), 2nd cousin offspring (2C) and individuals belonging to
biradheri (B) grouped into country of origin, location within country and caste.)
N
N
Country of
origin
I
1C
2C
B
Location
I
India
87
1
0
2
Jullundur
Ludihana
Punjab
Amristar
Hoshiapur
Phagare
Delhi
Gujurat
Jammu
Khushab
Purawara
India
Mirpur
Kashmir
Lahore
Jhelum
Rawalpindi
Islamabad
Sargodar
Chawaal
Faislabad
Gujaranwala
Sailkot
Sial
44
9
9
8
7
3
1
1
1
1
1
2
39
28
11
6
6
4
2
1
1
1
1
1
Pakistan
101
21
10
27
the Punjab: Mirpur and Jullundur. Mirpur is located
within the Kashmir area of Pakistan and Jullundur in
the northwest of India, approximately 150 km Southeast of Mirpur. Although these two locations are the
most common regions of origin in our sample, a further 19 locations of variable specificity were identified
through a questionnaire (Table 1). The sampled individual’s caste is also presented in Table 1, along with the
number of individuals identified as being offspring of
1st and 2nd cousin unions and the number belonging to
a biradheri. Multiple origins were not observed in our
sample (i.e., the recent ancestry of the volunteer was
reported to have originated in the same location in India or Pakistan through both parents). The age structure
of the volunteers was broad, between 20 and 65 years
(data not shown), and included migrants as well as offspring and grandchildren of the migrant generation. The
specificity of the locations recorded varies; for example,
nine individuals gave the Punjab as a location and 28
gave Kashmir, both of which cover large areas. Other
C
University College London 2003
N
1C
2C
B
Caste
I
1
Jat-Sikh
Hindu-Khatri
Hindu-Brahmin
Hindu-Pandit
Hindu-Rai
Sikh-Ramgara
Sikh-Sein
Muslim-Gujur
Muslim-Jat
None stated
56
19
4
1
1
1
1
2
1
1
Rajput
Bains
Gujar
Kashmiri
Moghul
Jat
Sheikh
Bhatt
Chaudri
Syed
None stated
41
13
6
6
4
2
2
1
1
1
24
1
1
13
7
1
6
4
10
9
1
1
3
1
1
1
1C
2C
1
8
5
1
2
1
B
1
1
4
3
1
1
14
5
1
1
2
1
1
3
1
1
3
individuals specified cities. However, the majority of
the individuals’ ancestries could be traced back to just
three locations: Jullundur (51% of Indians), Mirpur and
Kashmir (38% and 27% of Pakistanis respectively). These
proportions are typical of those observed throughout
British Pakistanis and Indians (Ballard, 1994). Castes are
also unevenly represented, with Jat-Sikh being the most
prevalent of the Indian castes (at 65%), and Rajput the
most prevalent of the Pakistani castes (56%). The prevalence of the Rajput caste might be indicative to some
degree of the self-identification of non-Rajput individuals that occurred on their conversion to Islam (Wang
et al. 2000).
Of the 188 individuals scored for 10 loci, 96% of
the genotypes gave reliable results with 5 samples failing
to amplify at any loci. Table 2b indicates that both the
Indian and Pakistani samples are significantly different
from both Afro-Caribbean and US Caucasian over all
loci, albeit with low F ST values. Table 2c gives the values of F ST for those groups that are numerically well
Annals of Human Genetics (2003) 67,525–537
529
A. D. J. Overall et al.
Table 2a Expected heterozygosities for the 10 SGM Plus loci (AC = Afro-Caribbean, USC = US Caucasian from P.E. Biosystems
(1999))
India
Pakistan
AC
USC
D2
D3
D8
D16
D18
D19
D21
FGA
THO1
vWA
0.885
0.875
0.893
0.885
0.732
0.614
0.754
0.787
0.793
0.792
0.788
0.796
0.838
0.707
0.804
0.750
0.855
0.716
0.876
0.872
0.768
0.772
0.847
0.782
0.813
0.890
0.863
0.838
0.885
0.928
0.861
0.861
0.840
0.850
0.744
0.761
0.780
0.750
0.817
0.806
Table 2b F ST values (AC = Afro-Caribbean, USC = US Caucasian from P.E. Biosystems (1999))
India-AC
Pakistan-AC
India-USC
Pakistan-USC
AC-USC
D3
D8
D16
D18
D19
D21
FGA
THO1
vWA
Total
0.006∗
0.010∗
0.000
0.001
0.006∗
0.000
0.001
0.002
0.000
0.003
0.006∗
0.006∗
0.002
0.003
0.009∗
0.000
0.000
0.001
0.004
0.005∗
0.003
0.000
0.004
0.005∗
0.007∗
0.003
0.002
0.003
0.003
0.001∗
0.005
0.001
0.001
0.002
0.003
0.000
0.000
0.004
0.003
0.004∗
0.015∗
0.019∗
0.004
0.013∗
0.027∗
0.003
0.007∗
0.003
0.003
0.008∗
0.003∗
0.003∗
0.002∗
0.003∗
0.007∗
0.08
0.10
F ST significantly greater than 0 (p < 0.05).
Table 2c F ST estimates between identifiable subgroups of Asian
sample
Group
F ST
India
Pakistan
Jullundur
Mirpur
Kashmir
Jat-Sikh
Hindu-Khatri
Rajput
Non-Rajput
0.0006
A
1.0
0.8
0.0000
0.6
0.0000
0.0058
C
∗
D2
0.4
0.2
0.1
9
0.
0.02
0.5
Annals of Human Genetics (2003) 67,525–537
0.0
0.00
0.5
530
0.1
represented (N ≈ 20). Whether countries (India and
Pakistan), regions (Jullundur, Mirpur and Kashmir) or
castes (Jat-Sikh, Hindu-Khatri and Rajput) were compared with each other, significant differences calculated
as F ST were not observed. Artificial clusters such as
Jullundur vs non-Jullundur and Mirpur/Kashmir vs
non-Mirpur/Kashmir also gave non-significant values
of F ST (results not shown). The largest value of F ST observed was that between Rajput and the remainder of
the Pakistani caste members (non-Rajput).
Table 3 gives estimates of F IS for the same groups.
Small, but non-significant values were generally observed within the Indian groups, with a negative value
for the Hindu-Khatri caste. The Pakistani groups, on
the other hand, all gave significant positive values
of F IS .
0.04
0.06
θ
Figure 1 Joint estimates of the degree of population
substructure (θ) and the proportion of consanguineous unions
(C). (A) Likelihood surface for Pakistani sample with 10%, 50%
and 90% CLV envelopes. N = 101. (B) Likelihood surface for
Mirpuri sample with 10%, 50% and 90% CLV envelopes. N =
39. (C) Likelihood surface for Pakistani biradheri with 10%,
50% and 90% CLV envelopes. N = 28.
Figures 1 (A-C) show the likelihood plots for the
Pakistani, Mirpuri and biradheri samples respectively.
The axes are C, the estimated proportion of the
sample with consanguineous parents (R = 1 /16 ), and θ,
the estimated magnitude of substructure. The envelopes
C
University College London 2003
Consanguinity and Substructure in UK Asian Populations
Discussion
B
1.0
0.8
C
0.6
0.9
0.4
0.2
0.5
0.1
0.5
0.0
0.00
0.02
0.04
0.06
0.08
0.10
θ
C
1.0
0.8
C
0.6
0.9
0.4
0.2
0.1
0.0
0.00
0.02
0.5
0.1
0.5
0.04
0.06
0.08
0.10
θ
Figure 1 (continued)
enclose the 10, 50 and 90% critical likelihood values
(CLV). These values are found by taking the cumulative
sum of the likelihood values such that the corresponding
cumulative sum is just less than or equal to 10, 50 or 90%
of the total cumulative sum.
Table 3 F IS estimates for each of the
identifiable subgroups of the Asian sample.
Country
F IS
Region
F IS
Caste
F IS
India
0.00844
Jullundur
0.01081
Pakistan
0.04378∗∗
Mirpur
Kashmir
0.04971∗
0.08416∗∗
Jat-Sikh
Hindu-Khatri
Rajput
Non-Rajput
0.00887
− 0.03352
0.04257∗
0.05733∗∗
∗
C
University College London 2003
Relative to most European societies, the social structure of the UK Asian population is complex, being
largely a result of marriage traditions. Hindu and Sikh
communities, for example, are often caste-endogamous.
Strict rules of exogamy can also prevail so that marriage into the father’s, mother’s, and both parents
mother’s descent group (got) is prohibited in addition
to the reciprocal exchange of women between families
(Ballard, 1990). It may be that the negative F IS value
observed for the Hindu Khatri caste (Table 3) is reflecting caste endogamy, but got exogamy. Marriage within
Pakistani Muslim societies, however, has fewer exclusions: first order relatives; parents, sibs and parent’s sibs
(Ballard, 1990). Amongst Punjabi Muslims, an active
preference for first cousin marriage is also common
(Bittles et al. 1991) and there is evidence that such arrangements are increasing in prevalence (Darr & Modell,
1988). One other important aspect of Pakistani
Muslim societies is the extent of family networks that
develop through biradheri endogamy, where marriages
within these large kin-networks often occur between related individuals (Darr & Modell, 1988; Ballard, 1990).
Despite the recognition of such groups, it is often not
possible to identify a simple nested hierarchy of population divisions within which individuals can be placed. It
is more often the case that identifiable divisions overlap,
making population substructure difficult to quantify.
The F ST values presented in Table 2c show that no
significant deviations were found between the more
obvious divisions of the Asian sample. The F IS estimates in Table 3, however, vary widely between the
divisions with significant positive values estimated for
the major Pakistani groups (caste and locations) as
well as the total Pakistani sample itself. These samples
were inspected further using the likelihood approach
(equation (1)).
p < 0.05,∗∗ p < 0.01
Annals of Human Genetics (2003) 67,525–537
531
A. D. J. Overall et al.
The excessive homozygosity observed for the
Pakistani sample appears to be contributed to by substructure (Figure 1(A)). The maximum likelihood values are for θ = 0.029 and C = 0.1. With this magnitude of variation within the Pakistani subgroup the
F ST value presented in Table 2c is likely to be an underestimate. The questionnaire identified 20% of the
Pakistani individuals as having parents related as first
cousins. This proportion appears to be low, relative to
other studies on UK Pakistani communities (Darr &
Modell, 1988; Qureshi, 2003), but is corroborated by
the genetic data. The contour of Figure 1(A) shows that
the top 10% most likely parameter combinations exclude values above C > 0.3.
The locations with significant F IS values are Mirpur
and Kashmir. The Mirpuri sample (Figure 1(B)) is of
interest as, similar to a previous study on the Mirpuri
(Overall & Nichols, 2001), it shows high levels of consanguinity, the most likely parameter combination being
39% consanguines (R = 1 /16 ) and θ = 0.015. With N =
39, the CLV envelopes are broad, though the 10% CLV
excludes zero for both parameters. Again, Figure 1(B)
is in accordance with the questionnaire, which identifies 1 /3 of the Mirpuris as having parents related as
first cousins. The estimates of C for the Mirpuris are in
keeping with other studies of UK Pakistani communities where estimates are around 55% first cousin marriages (Darr & Modell, 1988; Modell, 1991), a value
not excluded by the 10% most likely parameter values.
This result suggests some variation in the proportion
of consanguinity in this Pakistani sample. The majority of Pakistani immigrants trace their origins to rural
Mirpur and, as other studies have observed (Rao & Inbaraj, 1977; Bittles, et al. 1991; Bittles, 1994), the highest
rates of consanguinity are to be found in rural communities. It appears that this trend has continued in their
immigrant descendants. The fact that the Pakistani sample as a whole has a lower value for C might reflect a
large proportion of non-Mirpuri individuals descending
from more urban locations on the subcontinent.
The likelihood method was unable to distinguish the
most likely parameter combination for the Kashmiri
sample, most probably due to the small sample size
(N = 28), the method being sensitive to sample size
when either C is close to 1 or when the magnitude
of θ is equivalent to the R value being considered, as
532
Annals of Human Genetics (2003) 67,525–537
the value in Table 3 suggests is the case. Under these
conditions the distribution of homozygous genotypes is
expected to be similar for both complete consanguinity
and substructure, and the two cannot be distinguished.
The limitations of the method are detailed in Overall
& Nichols (2001). In addition to increasing sample size,
the resolution of the method is also greatly improved by
the addition of more polymorphic loci.
An issue not generally considered in previous studies of UK Pakistanis is the heterogeneity that can exist within communities through more complex social
structuring, which is in contrast to the detailed investigations conducted on the sub-continent (e.g., Hussain,
1998; Wang et al. 2000) and elsewhere around the world
(Bittles, 2001). Both of the Pakistani caste groupings
gave significant positive F IS estimates. The non-Rajput
group (figure not shown), as expected given the numerous castes and biradheri that it comprises, shows
evidence of substructure with the most likely parameter combination of θ = 0.028 and C = 0. The single
Rajput caste (figure not shown) also has a maximum
likelihood of θ = 0.037 and C = 0, supporting the
Y-chromosome haplotype diversity observed by Wang
et al. (2000) in showing evidence of substructure within
the Rajput.
Figure 1(C) shows the likelihood contour for individuals whose parents were identified as belonging to
the same ‘biradheri’. Individuals who gave this response
to the questionnaire did so as an alternative to consanguinity and it is unclear to how related the parents were.
The excess homozygosity observed within this group is
high, and corresponds to both substructure and consanguinity, with a maximum likelihood value of consanguinity equivalent to 47% first cousin offspring and
θ = 0.033. Because the estimation of C requires a prior
estimate of R, we cannot be certain that the magnitude
of excess homozygosity refers to first cousin offspring.
If an R-value corresponding to second cousin offspring
is inputed into equation (1) (R = 0.03125), a maximum likelihood estimate of C of around 100% results.
It is possible, due to the sensitivity of the questionnaire,
that respondents may have preferred to claim that their
parents were from the same biradheri rather than acknowledge first cousin consanguinity. Table 1 reveals
that, unlike consanguinity which is largely confined to
the rural Mirpur and Kashmir, individuals whose parents
C
University College London 2003
Consanguinity and Substructure in UK Asian Populations
belong to the same biradheri are far more widespread,
suggesting that a proportion of those questioned gave an
accurate response. Whatever the cause, this magnitude
of consanguinity is clearly relevant to the genetic health
of the UK Pakistani population, as relationships other
than acknowledged first cousin offspring have received
little attention.
The maximum likelihood estimate for substructure
(θ = 0.033) is indicative of the isolation expected to
exist between large kin-networks. Such social structuring has been compared with the tribes of the Middle East
(Modell & Darr, 2002), with the implication that inherited disorders may be unevenly distributed as a result of
restricted intermarriage. The evidence of substructure
within the UK Pakistani biradheri group certainly suggests this as a possibility. The magnitude of substructure
and consanguinity within our biradheri grouping is consistent with that found by Wang et al. (2000), where
biradheri were identified as groups equivalent to the Indian caste. Wang et al. (2000) identified three biradheri
in a study of communities throughout the Punjab: the
Awan, Khattar and Rajpoot, and found significant differentiation between them and inbreeding within them
at autosomal loci. The result of our likelihood analysis
points to a similar conclusion, although it is important
to note that the likelihood estimate is an average of the
different degrees of endogamy expected within different biradheri, as well as between the individual families within biradheri; differences that were observed in
the Pakistani population by Wang et al. (2000). Because
only a proportion of those individuals identifying with
a caste claimed to belong to a biradheri in our survey, numbers were too few to make meaningful comparisons between biradheri from different caste groups,
which would have been a more appropriate comparison
with the Wang et al. (2000) study. However these groups
are stratified into endogamous units, the Rajput example does provide an indication that social groupings can
contain significant diversity, possibly through past social
and geographic isolation.
A further comparison was used to test the link with
reported consanguinity: estimates from Pakistani individuals who report that their parents are unrelated and
not from biradheri. The likelihood surface (not shown)
for this group, as would be expected if the parent’s re-
C
University College London 2003
lationship was accurately reported, gave the most likely
parameter combination of zero for both C and θ.
The Indian groups did not show any significant deviations from Hardy-Weinberg equilibrium. This result
at first appears surprising, given that caste endogamy
may be expected to generate substructure in the total
population, as is seen on the sub-continent with Hindu
caste and tribal groups (Mukherjee et al. 1999). This
may reflect the small sample sizes for all but the JatSikh caste in our data set. In addition, Sikh caste endogamy is not expected to be adhered to in marriage
arrangements as strongly as within Hindu communities (Ballard, 1994). The Jat-Sikh caste (N = 56) was
represented by individuals who could trace their ancestry to five different Indian locations. Unlike the Pakistani Rajput caste, which was similarly composed of
individuals from various origins, the Jat-Sikh did not
show any evidence of substructure. This possibly reflects the fact that the majority (34/56) of the Jat-Sikhs
trace their origins to Jullundur, a prosperous and well
connected region of the Punjab likely to have experienced significant influxes from surrounding areas, particularly during the partition (Ballard, 1990). Although
many Indian Sikhs are also known to be twice migrants,
having first migrated to East Africa before moving on
to the UK from the 1960s onwards, these predominantly belong to the Ramgara (Ramgarhia) caste, with
the Jat-Sikhs mostly migrating to the UK directly from
the Indian Sub-continent (Bhachu, 1986). Another possible cause of this result is that the traditional caste endogamy that may still be present on the sub-continent
has been disrupted by a combination of restrictive migration policies, settlement, and a community adapting and assimilating into a western society (Ballard,
1990). There is thus little evidence from the Jat-Sikhs
that caste endogamy plays an important role within this
UK community.
In a previous study on the Jullundur community of
Nottingham (Overall & Nichols, 2001), excess homozygosity was observed where the most likely explanation
was substructure, and it was suggested that this could be
through caste endogamy. However, the Jullundur subsample obtained for this study, with care being taken not
to re-sample the same individuals, showed no significant
excess of homozygosity (Table 3).
Annals of Human Genetics (2003) 67,525–537
533
A. D. J. Overall et al.
Relevance of Consanguinity and Substructure
to Genetic Health
D = C [q (R + (1 − R) (θ + (1 − θ) q ))]
+ (1 − C) [q + (θ + (1 − θ) q )] ,
where it then follows that
q =
(C R (1 − θ ) + θ ) −
0.008
0.006
q
0.004
0.4
0.2
0.04
0
0.02
θ
0.8
0.000
0.6
1
0.002
C
Figure 2 The change in frequency estimates for a
recessive allele (q) with C and θ with R = 0.0625 and
observed incidence of 1:10000.
when either C = 0.614 or θ = 0.048 and when D =
1:10000 when either C = 0.223 or θ = 0.015.
The magnitude of substructure detected in our
survey can have appreciable effects. As an illustration,
(C R (1 − θ) + θ )2 − 4 (C R (1 − θ) + θ − 1) D
. (2)
2 (C R (1 − θ) + θ − 1)
Figure 2 illustrates the effect of substructure and consanguinity on the estimated frequency, q, where the coefficient of consanguinity is R = 0.0625 and where the
observed incidence is 1:10000. This Figure shows that,
as expected, more homozygotes are expected for a given
allele frequency with increasing consanguinity through
the addition of autozygous homozygotes. For a given
incidence, therefore, the estimated frequency of the disease allele declines with increasing consanguinity. This
effect is more marked with rarer disorders, reflecting the
fact that carriers of rare mutations are unlikely to mate
with carriers of the same disorder unless they are related
(Modell & Darr, 2002). Substructure has a similar effect on this estimation. In these examples, a halving of
the estimated gene frequency when D = 1:1000 occurs
534
0.010
0
One important implication of our results is that simple methods of estimating disease gene frequencies may
need to be modified. Some studies (Hutchesson et al.
1998; Hall et al. 1999) use an equation derived from
Dahlberg (1930) and Edwards (1989) to estimate the
frequency of specific autosomal recessive inborn errors
of metabolism, q, from estimates of D, the incidence of
the disorder, C, the proportion of the population from
consanguineous unions and R, the inbreeding coefficient where D = C(q(R + (1 − R)q)) + (1 − C)q2 .
Our results suggest that this approach needs additional
modification to include a parameter for substructure.
The Pakistani group, for example, showed evidence of
substructure where the frequency of a disease gene could
vary between subpopulations; estimating q from such a
substructured population would lead to an overestimate
due to a Wahlund effect. In a case such as this a more
complete formulation of disease incidence would be:
D=1:10000
Annals of Human Genetics (2003) 67,525–537
equation (2) was used to modify the results of Hutcheson et al. (1998) on the estimation of 10 autosomal
recessive gene frequencies. Figures 3(A) and 3(B) show,
respectively, the disease frequencies and gene frequencies
estimated from a North-Western European and a Pakistani group from the West Midlands, UK. Data were
obtained from the West Midlands neonatal screening
programme and the National Census. Figure 3(A) shows
that five out of the ten disorders have significantly different disease frequencies (higher in the Pakistani sample).
Figure 3(B) shows the corresponding gene frequency
estimates. One set takes into account only consanguinity (filled circles, C = 0.7 and R = 0.0686 according
to Hutchesson et al. (1998)) and our new estimates allow for both consanguinity and substructure. One uses
C
University College London 2003
Consanguinity and Substructure in UK Asian Populations
A
0.0008
Disease frequency
0.0007
0.0006
0.0005
0.0004
0.0003
0.0002
B
Niemann Pick
disease type C
Non-ketotic
hyperglycinaemia
San Fillipo
diseaseType C
Mucopolysaccharidosis type 1
MCADD
Hyperoxaluria
type1
Cystinosis
Galactosaemia
Tyrosinaemia type 1
0
Phenylketonuria
0.0001
0.016
Gene frequency
0.014
0.012
0.010
0.008
0.006
0.004
Niemann Pick
disease type C
Non-ketotic
hyperglycinaemia
San Fillipo
diseaseType C
Mucopolysaccharidosis type 1
MCADD
Hyperoxaluria
type1
Cystinosis
Galactosaemia
Tyrosinaemia type 1
0.000
Phenylketonuria
0.002
Figure 3 (A) Incidence frequencies of ten autosomal recessive inborn errors of metabolism In
NW European (open squares) and Pakistani (filled circles) populations of the West Midlands,
UK. Data from Hutchesson et al. (1998). The error bars represent 95% confidence intervals
where the observed cases of each disorder are considered to be Poisson distributed, according to
the method of Hutchesson et al. (1998). (B) Estimated gene frequencies of ten autosomal
recessive inborn errors of metabolism. Error bars as in figure 3(A). Open squares: NW
√
European; estimated as (incidence). Filled circles: Pakistani; estimated using eq. (2) with C =
0.7, R = 0.0686, θ = 0 as in Hutchesson et al. (1998). Open circles: Pakistani; estimated using
eq. (2) with C = 0.4, R = 0.0625, θ = 0.01. Filled triangle: Pakistani; estimated using eq. (2)
with C = 0.5, R = 0.0625, θ = 0.03.
the parameter estimates from the Mirpuri (open circles,
C = 0.4, R = 0.0625 and θ = 0.01); another from
the biradheri (closed triangles, C = 0.5, R = 0.0625
and θ = 0.03) samples. The general trend is as noted
by Hutchesson et al. (1998), with only PKU showing
C
University College London 2003
any significant differentiation between the two ethnic
groups for all three parameter combinations considered.
In general, the modified gene frequency estimates incorporating the Mirpuri parameters tend to be higher
than those from Hutcheson et al. (1998), while those
Annals of Human Genetics (2003) 67,525–537
535
A. D. J. Overall et al.
incorporating the biradheri parameters tend to be lower.
The biradheri type parameters decrease the gene frequency estimate for tyrosinaemia type 1 such that the
differences between the two ethnic groups are no longer
significant. The Mirpuri type parameters increase the
gene frequency estimate for MCADD (medium chain
acyl-CoA dehydrogenase deficiency) such that the differences for this disorder are no longer significant.
Higher disease incidence in Pakistani communities
relative to NW European populations has generally been
attributed to consanguinity. The results of this study indicate that the magnitude of consanguinity practiced
amongst certain divisions of the population varies. For
example, individuals whose family origins trace back
to rural Pakistan, such as Mirpur, are more likely to
be consanguines. The Hutchesson et al. (1998) model
of inferring gene frequency is appropriate for such
cases. A higher incidence may also be contributed to
by substructure, such as we found with the Nottingham
Pakistani sample. If the subpopulations could be identified, then greater accuracy would be obtained by treating them separately. However, our results provide evidence of cryptic population structure. Equation (2) is
therefore of use to estimate the average frequency of the
recessive allele within a substructured population, and
takes into account any differing incidence of disease between the subpopulations.
The UK Pakistani population has been studied in detail, in particular because of the clear implications a traditional consanguineous society has for recessive disorders (Terry et al. 1985; Darr & Modell, 1988; Chitty &
Winter, 1989; Bundey et al. 1991). The results of this
study suggest that simply identifying broad estimates of
the proportion of consanguineous unions, along with
an average coefficient of consanguinity, may provide too
vague a picture of how disease incidence relates to the
risk of congenital disorders. The substructure identified
by our likelihood analysis indicates the potential for heterogeneity in disease gene frequencies across the various
Pakistani subpopulations. It would be helpful to those
offering health advice to be able to identify whether
such heterogeneity is accounted for by caste/biradheri
differentiation, or differentiation due to some other
identifiable subgroup such as the Mirpuris or Kashmiris.
The Indian community, on the other hand, shows
no significant evidence of substructuring, despite indi536
Annals of Human Genetics (2003) 67,525–537
cations of a Wahlund effect in a previous study on the
Jullundur community (Overall & Nichols, 2001) and the
expectation of caste endogamy. This is perhaps not so
surprising considering how successful the Indian community has been in settling and establishing itself within
the UK society (Ballard, 1990). In contrast to Indian
customs, the Pakistani marriage strategies, in particular those of the Mirpuris, remain very much as they
were during the early immigration period and continue
to involve a high proportion of trans-continental arrangements (Ballard, 1990). Traditional marriage practices will continue to generate the social structuring of
the Pakistani population for some time, and for this reason the influence of both consanguinity and substructure on the assessment of risk of congenital recessive
disorders needs to be pursued.
Acknowledgements
This study was supported by The Sir Jules Thorn Charitable
Trust grant 98/28A. We are grateful for the comments made
by two anonymous referees, which improved the manuscript
considerably. Also thanks to Roger Ballard (Centre for Applied South Asian Studies) for guidance in setting up the sampling procedure, and the communities of Nottingham who
participated in the project.
References
Bhachu, P. (1986) Twice Migrants. London: Tavistock.
Ballard, R. (1990) Migration and Kinship: the differential effect of marriage rules on the process of Punjabi migration
into Britain. In: Migration and Ethnicity (eds C. Clarke & C.
Peach), Cambridge University Press.
Ballard, R. (1994) The emergence of Desh Pardesh. In: Desh
Pardesh: The South Asian presence in Britain (Ed. R. Ballard).
London: Oxford University Press.
Bittles, A.H. (2001) Consanguinity and its relevance to clinical
genetics. Clin Genet 60, 89–98.
Bittles, A.H., Mason, W.M., Greene, J. & Appaji, Rao,
N. (1991) Reproductive behavior and health in consanguineous marriages. Science 252,789–794.
Bittles, A.H. & Neel, J.V. (1994) The costs of human inbreeding and their implications for variations at the DNA level.
Nature Genet 8, 117–121.
Bittles, A.H., Shami, S.A. & Appaji Rao, N. (1992) Consanguineous marriage in South Asia: incidence, causes and
effects. In: Minority Populations: Genetics, Demography and
Health (eds A.H. Bittles & D.F. Roberts), pp. 102–117. London: Macmillan.
C
University College London 2003
Consanguinity and Substructure in UK Asian Populations
Bittles, A.H. (1994) The role and significance of consanguinity as a demographic variable. Pop Dev Rev 20, 561–584.
Bundey, S. & Alam, H. A five year prospective study of the
health of children in different ethnic groups, with particular
reference to the effect of inbreeding. Eur J Hum Genet 1,
206–219.
Bundey, S., Alam, H., Kaur, A., Mir, S. & Lancashire, R.
(1991) Why do Pakistani babies have high perinatal and
neonatal mortality rates? Paediatr Per Epidemiol 5, 101–114.
Chitty, L.S. & Winter, R.M. (1989) Perinatal mortality indifferent ethnic groups. Arch Dis Child 64, 1036–1041.
Cockerham, C.C. (1973) Analysis of gene frequencies.
Genetics 74, 679–700.
Dahlberg, G. (1930) Inzucht bei Polyhybridat beim Menschen. Hereditas 14, 83–96.
Darr, A. & Modell, B. (1988) The frequency of consanguineous marriage amoung British Pakistanis. J Med Genet
25, 186–190.
Devi, A. R. R., Appaji Rao, N. & Bittles, A.H. (1987) Inbreeding and the incidence of childhood genetic disease in
Karnataka, Sounth India. J Med Genet 24, 362–365.
Edwards, J.H. (1989) Familiarity, recessivity and germline mosaicism. Ann Hum Genet 53, 33–47.
Hall, S.K., Hutchesson, A.C.J. & Kirk, J.M. (1999) Congenital
hypothyroidism, seasonality and consanguinity in the West
Midlands, England. Acta Paediatr 88, 212–215.
Heinisch, U., Zlotogora, J., Kafert, S. & Gieselmann, V. (1995)
Multiple mutations are responsible for the high frequency of
metachromatic leukodystrophy in a small geographic area.
Am J Hum Genet 56, 51–57.
Hussain, R. (1998) The role of consanguinity and inbreeding as a determinant of spontaneous abortion in Karachi,
Pakistan. Ann Hum Genet 62, 147–157.
Hutchesson, A.C.J., Bundey, S., Preece, M.A., Hall, S.K. &
Green, A. (1998) A comparison of disease and gene frequencies of inborn errors of metabolism among different
ethnic groups in the West Midlands, UK. J Med Genet 35,
366–370.
Modell, B. (1991) Social and genetic implications of customary consanguineous marriage among British Pakistanis. Report of a meeting held at the Ciba Foundation on 15 January
1991. J Med Genet 28, 720–723.
Modell, B. & Darr, A. (2002) Genetic counselling and customary consanguineous marriage. Nat Rev Genet 3, 225–
229.
Mukherjee, N., Majumder, P.P., Roy, B., Roy, M., Dey, B.,
Chakraborty, M. & Banerjee, S. (1999) Variation at 4 short
tandem repeat loci in 8 population groups of India. Human
Biology 71, 439–446.
Overall, A.D.J. & Nichols, R.A. (2001) A method for distinguishing consanguinity and population substructure using multilocus genotype data. Mol Biol Evol 18, 2048–
2056.
C
University College London 2003
Overall, A.D.J., Ahmad, M. & Nichols, R.A. (2002) The effect of reproductive compensation on recessive disorders
within consanguineous human populations. Heredity 88,
474–479.
Özlap, I., Coskun, T., Tokol, S., Demircin, G. & Mönch,
E. (1990) Inherited metabolic disorders in Turkey. J Inher
Metab Dis 13, 732–738.
Biosystems, P.E. (1999) AmpF STR SGM Plus user manual.
California: Perkin-Elmer.
Qureshi, N., Gilbert, P. & Raeburn, J.A. (2003) Consanguinity and genetic morbidity in a British primary care setting:
a pilot study with trained linkworkers. Ann Hum Biol 30,
140–147.
Rajib, A. & Patton, M.A. (1999) Analysis of the population
structure in Oman. Commun Genet 2, 23–26.
Rao, P.S.S. & Inbaraj, S.G. (1977) Inbreeding in Tamil Nadu,
South India. Soc Biol 24, 281–288.
Schneider, S., Roessli, D. & Excoffier, L. (2000) Arlequin ver.
2000: A software for population genetics analysis. University of Geneva, Switzerland: Genetics and Biometry Laboratory.
Shami, S.A., Grant, J.C. & Bittles, A.H. (1994) Consanguineous marriage within social/occupational class boundaries in Pakistan. J Biosoc Sci 26, 91–96.
Shaw, A. (1994) The Pakistani community in Oxford. In:
Desh Pardesh: The South Asian presence in Britain (Ed. R.
Ballard). London: Oxford University Press.
Terry, P.B., Bissenden, J.G., Condie, R.G. & Mathew, P.M.
(1985) Ethnic differences in congenital malformations. Arch
Dis Child 62, 866–879.
Vertovec, S. (1994) Caught in an ethnic quandary: IndoCaribbean Hindus in London In: Desh Pardesh: The South
Asian presence in Britain (Ed. R. Ballard). London: Oxford
University Press.
Walsh, P.S., Metzgar, D.A. & Higuchi, R. (1991) Chelex(R)
100 as a medium for the simple extraction of DNA for
PCR based typing from forensic material. Bio Techniques 1,
91–98.
Wang, W., Sullivan, S.G., Ahmed, S., Chandler, D.,
Zhivotovsky, L.A. & Bittles, A.H. (2000) A genome-based
study of consanguinity in three co-resident endogamous
Pakistan communities. Ann Hum Genet 64, 41–49.
Weir, B.S. & Cockerham, C.C. (1984) Estimating F statistics
for the analysis of population structure. Evolution 38, 1358–
1370.
Wright, S. (1921) Systems of mating. Genetics 6, 111–178.
Wright, S. (1931) Evolution in Mendelian populations.
Genetics 16, 97–159.
Zlotogora, J. (1997) Genetic disorders among Palestinean
Arabs. Am J Med Genet 68, 472–475.
Received: 19 March 2003
Accepted: 11 July 2003
Annals of Human Genetics (2003) 67,525–537
537