American Journal of Medical Genetics 34168-172 (1989)

Familial Mediterranean Fever in Armenians: Autosomal Recessive Inheritance With High Gene Frequency David B. Rogers, Mordechai Shohat, Gloria M. Petersen, Joan Bickal, Jane Congleton, Arthur D. Schwabe, and Jerome I. Rotter Division of Medical Genetics, Department of Pediatrics, Harbor-UCLA Medical Center, Torrance, California (D.B.R., J.C.); Division of Medical Genetics, Departments of Medicine and Pediatrics, Cedars-Sinai Medical Center (M.S., G.M.P., J.B., J.I.R .) and Division of Gastroenterology, Department of Medicine, UCLA Medical Center (A.D.S.), Los Angeles

Familial Mediterranean fever (FMF) is a re- current episodic inflammatory disorder of unknown pathogenesis that occurs with high frequency in non-Ashkenazi Jews and Arme- nians. However, there are some differences in the clinical manifestations of FMF in these ethnic groups. FMF has been reported to be an autosomal recessive disease in non-Ash- kenazi Jews, with a male/female ratio of 1.7, indicating reduced penetrance in females. However, the inheritance is less clear for Ar- menians. To resolve this problem, we studied prospectively families of 64 Armenian index cases randomly ascertained at the UCLA FMF clinic. Fifty-three families containing 176 sibs in addition to the probands were an- alyzed by genetic segregation analysis (exclu- sions included: six single-child families, four families in which one of the parents was also affected, and a family with incomplete infor- mation). Upper and lower bounds of the seg regation ratio were estimated, and ranged from .10 f .03 to .18 & .05 when only definitely affected sibs were classified as affected; .17 f .04 to .27 f .05 when considering “possibly affected” sibs as affected; and .19 f .04 to .30 f .05 when incomplete penetrance in females was corrected. A value of .25 is the expected segregation ratio for autosomal recessive in- heritance, and our data are consistent with this mode of inheritance. We can reject auto- soma1 dominant inheritance, where the ex- pected segregation ratio is .5. Using extended pedigree data, we calculated an FMF gene

Fkceived for publication October 28,1988; revision received May 30, 1989.

Address reprint requests to Gloria M. Petersen, Ph.D., Division of Medical Genetics, ASB-372, Cedars-Sinai Medical Center, 8700 Beverly Boulevard, Los Angeles, CA 90048.

0 1989 Alan R. Liss, Inc.

frequency of 0.073 and a carrier rate of 117, which is about four times the frequency in non-Ashkenazi Jews. The parent-to-offspring transmission of FMF reported in Armenians can be explained by its high gene frequency in this ethnic group. We conclude that, despite the clinical differences, FMF is an autosomal recessive disease in all ethnic groups studied to date.

KEY WORDS: inheritance, gene frequency, segregation analysis

INTRODUCTION Familial Mediterranean fever (FMF) is a recurrent

inherited disorder characterized by fever and serosal inflammation manifested by peritonitis, pleuritis, or ar- thritis [Siegel, 1945, 1964; Heller et al., 1958; Sohar et al., 1967; Schwabe and Peters, 1974; Eliakim et al., 19811. Most commonly the disease appears in individ- uals of Mediterranean ancestry, particularly non-Ash- kenazi Jews, Armenians, Arabs, and Turks [Eliakim et al., 19811. The ethnic groups differ clinically in their tendency to develop amyloidosis; this severe complica- tion has been reported to be 60% in selected series of Turkish patients [Ozdemir and Sokmen, 1969; Ozer et al., 19711, 12-29% in non-Ashkenazi Jewish patients, but only around 3% in Armenian patients [Armenian and Khachadurian, 1973; Bakir and Murtadha, 1975; Schwabe and Peters, 19741. As the pathogenesis of FMF is unknown, and there is no specific objective biochemi- cal abnormality, its diagnosis a t present is based en- tirely on clinical criteria [Eliakim et al., 19811.

In non-Ashkenazi Jews, FMF appears to be transmit- ted as a single gene trait with a recessive mode of inheri- tance [Sohar et al., 19611. However, the published data for other ethnic groups are conflicting. Several reported non-Middle Eastern families suggest autosomal domi-

FMF in Armenians 169

sumes single incomplete ascertainment [Fisher, 19341, 2) the a priori method assumes complete ascertainment [Hogben, 19461, and 3) a maximum likelihood estimate of the segregation ratio uses the computer program SEGRAN [Morton, 19691, which also estimates the as- certainment probability. Because segregation analysis is influenced by classification of disease status, it was important to repeat the calculations according to classi- fication criteria that ranged from most conservative (based on clinical observations only) to less conservative but biologically meaningful, based on assumptions about the sex ratio distortion. Thus, segregation anal- ysis was applied to three data sets:

1. Only individuals with clinically apparent FMF were classified as affected.

2. Affected status included clinically apparent FMF but also the possibly affected sibs with recurrent attacks of arthritis or unexplained fever.

3. This classification included affected individuals as defined in “2” above, but a correction was made for incomplete penetrance in females. Since the total af- fected malelfemale ratio (including sibs of probands) was 45/32 = 1.4, it is reasonable to assume that at least 28% of the females carrying the FMF genes were clini- cally unaffected. Therefore, 0.4 x the number of phe- notypically affected female sibs of the proband were randomly converted from phenotypically unaffected to affected status for this last classification.

For estimating the gene frequency of FMF (q), we used information collected on uncles, aunts, and cousins of the probands. Fifty-nine families (excluding those with an affected parent) were used. For an autosomal reces- sive condition, the parents of the probands are obligate carriers, and the uncles and aunts have a 50% chance of being carriers. If we assume that the uncles and aunts are married to an unrelated individual whose chance of carrying the gene in a relatively rare condition is 2q, then the chance for their offspring to be homozygous for the FMF gene is 2q x 112 x 114 = ql4. Summing over the number of affected individuals among all cousins of the probands permits us to calculate the gene frequency (9) and carrier rate (2q). The offspring of affected uncles or aunts were excluded. Similarly, this method could be applied to the aunt/uncle data. Their chance of being affected is ql2.

The gene frequency could then be estimated and used to estimate the expected prevalence of FMF in the Arme- nian population, the expected proportion of parents who would be affected (q in an autosomal recessive condi- tion), the expected proportion of offspring who would be affected (q in an autosomal recessive condition), and the probability of observing transmission of FMF through three consecutive generations (7q2, since 2q is the proba- bility of one of the parents being affected and q is the risk of an offspring of a proband being affected, and there are three possibilities of being a proband over three consecu- tive generations).

RESULTS The 53 probands and their families analyzed by the

segregation analysis included 158 sibs. The 53 probands in these families had 11 definitely clinically affected

nant inheritance, based on transmission of the disease in two or more consecutive generations [Heller et al., 1958; Nilsson and Floderus, 19643. Results of large studies in Arabs and Armenians from Lebanon have been interpreted to suggest both recessive [Khacha- durian and Armenian, 19741, dominant [Reimann et al., 1954; Naffah et al., 19751, and polygenic [Arme- nian, 19821 inheritance. This discrepancy has been ex- plained partially by reduced penetrance in females and the lack of recognition of milder forms of the disease [Heller et al., 1958; Khachadurian and Armenian, 1974; Reimann et al., 1954; Naffah et al., 1975; Armenian, 1982; Goksel and Bankman, 19631.

The aim of this study was to determine the mode of inheritance and the gene frequency in Armenians by taking into account milder forms of the disease and considering the reduced penetrance in females.

MATERIALS AND METHODS We ascertained prospectively 64 Armenian patients

(36 males, 28 females) who visited the UCLA FMF out- patient clinic between April 1,1983 and May 30,1988. Patients are referred to the FMF Clinic mainly by prac- titioners in the Los Angeles area. The diagnosis of FMF was made according to accepted criteria [Heller et al., 1958; Schwabe and Peters, 1974; Eliakim et al., 19811. All patients were interviewed and examined by the au- thors (D.B.R., M.S., with final determination of FMF diagnosis made by A.D.S.). Complete family histories and pedigrees were obtained from each case, including information about aunts, uncles, and cousins of each proband, their ages, parental ethnic origins, and whether they were diagnosed as having FMF. Affected sibs and relatives were defined as those who were diag- nosed as such in the FMF clinic or who were found to suffer from typical symptoms by medical records. A sib- ship was ascertained only once; if a sib of a proband was identified in the FMF clinic, heishe was considered to be an affected sib and not an index case. Among the sibs of the probands we also defined a separate group of “possi- bly affected individuals that included those who could not be diagnosed clearly as having FMF but who suf- fered from recurrent attacks of either arthritis or unex- plained recurrent fever alone.

In order to perform a genetic segregation analysis to test an autosomal recessive hypothesis, it was necessary to develop a data set composed of probands and their sibs, and all sibships had to have unaffected parents. The 64 families consisted of 179 sibs including the pro- bands: nine sibs aged 3-10 years (five of whom were affected), 22 sibs aged 11-20 years (12 affected), and 145 sibs who were older than 21 years. Three infants who were less than two years old were excluded from all calculations. Of the 64 possible families, we excluded six because they were single child families, four because one of the parents was affected, and one because of incom- plete information on the proband’s sibs. The remaining 53 families with 53 probands (32 males and 21 females) included a total of 158 sibs and were analyzed by seg- regation analysis.

Segregation analysis was performed using three dif- ferent methods [Emery, 19861: 1) the “sib” method as-

170 Rogers et al.

sibs (6 males, 5 females), six of whom had been seen in the UCLA FMF clinic. Another seven sibs had a history consistent with our definition of “possibly affected.” Ta- ble I shows the observed numbers of affected offspring in the 53 sibships with unaffected parents, and the ex- pected numbers according to the maximum likelihood method. The segregation ratios found for the three data sets are shown in Table 11. This ratio ranged from 0.10 t 0.03 to 0.18 t 0.05 when only confirmed cases of FMF were included, and 0.17 2 0.04 to 0.27 t 0.05 when the seven “possibly affected sibs were added to the affected group. Following correction for the incomplete pene- trance in the females, two normal females (5,x the number of affected sibs, 4) were randomly converted to affected status, and a segregation ratio of 0.19 ? 0.04 to 0.30 2 0.05 was observed.

The gene frequency estimation was based on 59 pro- bands (excluding those with an affected parent where the carrier rate for uncles1aunts is 2/31. They had 305 uncles and aunts, five of whom were affected; nine of the probands’ 317 cousins were similarly affected. The gene frequency was therefore estimated between 0.114 to 0.033 when calculated through the cousins and aunt1 uncles, respectively. The mean gene frequency was 0.073, which yields a carrier rate of about 117. This predicts an FMF prevalence of 11187 among the Arme- nian population.

These data allow certain predictions. In our study group, the expected number of families with an affected parent would therefore be 1113.7 x 126 = 9.2. The probability of a proband having an affected child is 1113.7 x 33 = 2.4. The odds ofobservingapedigree with affected individuals in three consecutive generations is 1113.7 x 1113.7 x 7 = 1/27.

There were four families among the 63 families with complete information in which one of the probands’ par- ents was affected. In these four families, six ofthe 12 sibs (50%) were affected. The 63 probands had 33 offspring, two of whom were affected. We observed that only one of the 63 pedigrees had an affected individual in each of three generations.

Analysis of the probands’ parental origins demon- strated only one family where one of the parents was not Armenian (Russian Jewish); otherwise all the patients we identified (including cousins of probands) were fully Armenian.

TABLE I. Observed and Expected Number of Affected Individuals in the 53 Studied Families*

Family size

No. of families

Definitely affected

individuals

Probably affected

sibs Unaffected

sibs

2 3 4 5 6 7 Total

24 17 6 2 3 1

53

29 20 8 3 3 1

64

4 1 1

1 -

7

15 30 15 7

14 6

87

Expected

27.43 22.05 8.78 3.28 5.48 2.02

69.04

*Expected No. of affected sibs computed under the hypothesis of a segregation ratio = .25. Excluded six single-child families and four fam- ilies with an affected parent.

TABLE 11. Segregation Ratio Computed From Alternative Methods and Data Sets ( t- S.E.) (see “Materials and

Methods” for Description of Groups Analyzed)

Incomplete ascertainment Complete

(Single ascertainment incomplete (a priori

method) SEGRAN method) Group 1 0.10 k 0.03 0.13 t 0.04 0.18 C 0.05 Group 2 0.17 2 0.04 0.21 ? 0.05 0.27 ? 0.05 Group 3 0.19 k 0.04 0.23 t 0.05 0.30 t 0.05

DISCUSSION The familial nature of FMF has been well recognized,

but its mode of transmission has been subject to some controversy. Although Heller et al. [1958] had inter- preted the data as suggesting dominant inheritance, they were able to show in a later study using a much larger group of cases and a more thorough study of family relationships that FMF in non-Ashkenazi Jews is recessively inherited [Sohar et al., 19611. The disease was shown to be frequent among brothers and sisters and rare among their antecedents. In addition, the rate of consanguinity among parents of the probands was much higher than in the general population. The pub- lished data for other ethnic groups are conflicting (Table 111). Many FMF studies included patients from different ethnic groups in Lebanon (Armenian and Christian or Moslem Arabs) in different proportions, and some pooled data retrospectively, all of which may increase the chance of ascertaining families with many and more severely affected offspring [Reimann et al., 1954; Naffah et al., 1975; Armenian, 19821. Nevertheless, a lower- than-expected number of cases among sibs has been observed in all of these studies, which can only partially be explained on the basis of reduced penetrance in fe- males, onset of the disease in late childhood, and failure to diagnose milder forms of the disease.

In our study, we prospectively ascertained all the pa- tients who were referred to the UCLA FMF Clinic in a 6-year period. While our conclusions may be confined to some degree by this sample size, we were able to make some substantive observations. Since more than 80% of the FMF patients manifest their disease before age 20 years, and only 14 unaffected sibs were under that age at ascertainment, we assume the age factor had no signifi- cant role in reducing the number of observed patients. Six of the 11 affected sibs of the probands were evaluated in the UCLA clinic, suggesting that our ascertainment is partially complete. The use of the complete ascertain- ment method, measured in our study by the a priori formula, gives the upper bound of the segregation ratio, while using the formula for incomplete ascertainment methods gives the lower bound. Therefore, the true seg- regation ratio is between 0.10 i 0.03 to 0.18 t 0.05, comparable to that observed for non-Ashkenazi Jews by Sohar et al. [1961]. Both they and we found fewer-than- expected numbers of affected offspring; this could be explained by reduced penetrance in females and failure to recognize milder forms of FMF. Indeed, a segregation

TAB

LE 1

11.

Sum

mar

y of

Pre

viou

s Se

greg

atio

n A

naly

sis

Stud

ies

in F

MF*

M

ale/

fe

mal

e N

o. o

f T

otal

no.

A

utho

rs

Popu

latio

n A

scer

tain

men

t ra

tio

sibs

hips

of

sib

s

Soha

r et a

l. [

196l

] 95

% N

on-A

shke

nazi

R

etro

spec

tive

1.41

12

3 66

4 Je

ws

anal

ysis

of

two

pool

s of

dat

a

Rei

man

n et

al.

68

% A

rmen

ians

58

of

thei

r 72

cas

es;

1.25

58

N

.G.

[195

41

32%

Ara

bs

outp

atie

nt c

linic

Kha

chad

uria

n an

d 53

% A

rmen

ians

, Sp

ecia

l ou

tpat

ient

2.

5 12

0 46

0 A

rmen

ian

[197

4]

34%

Mos

lem

s,

clin

ic

9% C

hris

tian

s

-

Naf

fah

et a

l. [

1975

1 46

% A

rmen

ians

, Sp

ecia

l out

pati

ent

1.86

26

25

% C

hris

tian

s,

clin

ic

+ 17

25

% M

osle

ms

Ara

bs

outp

atie

nt c

linic

A

rmen

ian

1198

21

Arm

enia

ns a

nd

Ret

rosp

ectiv

e N

.G.

121

558

No.

of

affe

cted

Si

bs

Oth

er fi

ndin

gs

232

N.G

.

130

(exp

ecte

d 16

0)

33

(exp

ecte

d 40

) 16

3 (e

xpec

ted

193)

9 fa

mil

ies2

ge

ner.

, 1

fa

mily

-3

gene

r.

Whe

n pa

rent

af

fect

ed 5

0%

tran

smis

sion

; In

crea

sed

cons

angu

inity

in

FMF

fam

ilies

Si

blin

gs a

ffec

ted

in

9 fa

mili

es;

1

fam

ily f

rom

A

rmen

ian

refu

gee

cam

p w

ith

20 a

ffec

ted

in 5

gen

erat

ions

10

fam

ilies

wit

h a

pare

nt a

ffec

ted;

21

of

thei

r 58

si

bs w

ere

affe

cted

Sh

owed

A.D

. pa

tter

n of

in

heri

tanc

e 14

mor

e fa

mili

es

wit

h a

pare

nt

affe

cted

; 31

/91

sibs

wer

e af

fect

ed.

Fam

ilial

ca

ses

are

mor

e se

vere

ly a

ffec

ted

Aut

hors

’ G

ene

conc

lusi

ons

freq

uenc

y

Aut

osom

al r

eces

sive

1:

52

Aut

osom

al d

omi-

N

.G.

nant

Aut

osom

al r

eces

sive

0.

032

wit

h re

duce

d pe

ne-

tran

ce i

n fe

mal

es

Aut

osom

al d

omi-

N

.G.

nant

Poly

geni

c N

.G.

~

*N.G

. =no

t giv

en; A

.R. =

auto

som

al re

cess

ive;

A.D

. = au

toso

mal

dom

inan

t; G

ener

. = ge

nera

tion.

172 Rogers et al.

ratio of 0.19 & 0.04 to 0.30 0.05 was found after correcting for these two parameters.

We also computed a high FMF gene frequency among Armenians. Our minimum estimates of the carrier rate is 1 in 7, which is 3-4 times higher than the carrier rate of 1 in 23 reported for the non-Ashkenazi Jewish popula- tion at risk in Israel [Sohar et al., 19611. In contrast to the high rate of consanguinity observed in the Israeli study, there was no evidence for consanguinity closer than third degree of relationship in our sample, despite careful questioning about family history. The gene fre- quency was independently estimated in our study for cousins (0.114) and uncles/aunts (0.033). The discrepan- cies between them can be explained by the fact that many of the uncles and aunts died at younger ages in the Armenian holocaust in 1914, and the information about their health is partially incomplete when compared to that for the cousins. Another factor that might have reduced our estimate of true gene frequency is the low penetrance in females, and the possibility that one of the parents of some of the cousins were non-Armenian. Nev- ertheless, this high gene frequency predicts a disease rate of 1 in 187 among Armenians, which is consistent with an estimate of 1 in 200-400 for Fresno County, California (based on autopsy data) [Schwabe and Peters, 19741. This agreement of the two independent preva- lence calculations from the same population supports the autosomal recessive inheritance of this condition. Further, the observed numbers of affected parents and offspring of the probands agreed with those expected using the estimated gene frequency, as did the observed and expected incidence of transmission of the disease in three consecutive generations.

Therefore, the parent-to-offspring transmission of FMF reported in Armenians can be explained by its high gene frequency in this ethnic group. Although FMF appears to be an autosomal recessive disease in Armenians and non-Ashkenazi Jews, it may be that they are different at the molecular level, as there are some clinical differences. The use of restriction frag- ment length polymorphisms and linkage analysis will soon be feasible genetic methods to apply to FMF. The particularly high gene frequency in Armenians strongly suggests that the FMF gene has had or even may continue to have a selective advantage in the Medi- terranean area.

ACKNOWLEDGMENTS This research was supported in part by NIH Inflam-

matory Bowel Disease Center grant DK 36200 and a

grant from the Stuart Foundations. We gratefully ac- knowledge the contributions to this study made by F. Lennie Wong and Susan Chao. We thank Jesse Hanifan and Tracey Peppers for preparation of the manuscript.

REFERENCES Armenian HK (1982): Genetic and environmental factors in the aetiol-

ogy of familial paroxysmal polyserositis. An analysis of 150 cases from Lebanon. Trop Geogr Med 34:183-187.

Armenian HK, Khachadurian AK (19731: Familial paroxysmal poly- serositis. Clinical and laboratory findings in 120 cases. Leb Med J

Bakir F, Murtadha M (19751: Periodic peritonitis. Report of 41 cases without amyloidosis. Trans R SOC P o p Med Hyg 69:111-117.

EIiakim M, Levy M, Ehrenfeld M (1981): “Recurrent Polyserositis (Fa- milial Mediterranean Fever, Periodic Disease).” Amsterdam: ElsevieriNorth Holland Biomedical Press.

Emery AEH (19861: “Methodology in Medical Genetics.” Edinburgh: Churchill Livingstone.

Fisher RA (1934): The effect of methods of ascertainment upon the estimation of frequencies. Ann Eugen 6:13-25.

Goksel V, Bankman NH (1963): La serosite periodique dans la popula- tion autochtone turque. Rev Med Moyen Orient 20:497-503.

Heller H, Sohar E, Sherf L (1958): Familial Mediterranean fever. Arch Intern Med 10250-71.

Hogben L (1946): “An Introduction to Mathematical Genetics.” New York: Norton.

Khachadurian AK, Armenian HK (1974): Familial paroxysmal poly- serositis (familial Mediterranean fever) incidence of amyloidosis and mode of inheritance. Birth Defects 10:62-66.

Morton NE (1969): Segregation analysis. In Morton NE (ed): “Com- puter Applications in Genetics.” Honolulu: University of Hawaii Press, pp 129-139.

Naffah J , Bitar E, Nasr W, Khouri K (1975): Etude genetique de la polyserite paroxystique familiale, a propos de 72 cas. Nouv Presse Med 4:1031-1037.

Nilsson SE, Floderus S (1964): Nine cases of hereditary andnonheredi- tary periodic disease. Acta Med Scand 175341-346.

Ozdemir AI, Sokmen C (1969): Familial Mediterranean fever among the Turkish people. Am J Gastroenterol 51:311-316.

Ozer FL, Kaplaman E, Zileli S (1971): Familial Mediterranean fever in Turkey. A report of twenty cases. Am J Med 50:336-339.

Reimann HA, Moadie J , Semerdjian S, Sahyoun P (1954): Periodic peritonitis-heredity and pathology. Report of 72 cases. JAMA

Schwabe AD. Peters RS (1974): Familial Mediterranean fever in Ar-

Siegal S (1945): Benign paroxysmal peritonitis. Ann Intern Med 23:l-21.

Siegal S (1964): Familial paroxysmal polyserositis. Analysis of 50 cases. Am J Med 365393-918.

Sohar E, Gafni J, Pras M, Heller H (1967): Familial Mediterranean fever. A survey of 470 cases and review of the literature. Am J Med

Sohar E, Pras M, Heller J , Heller H (1961): Genetics of familial Medi-

26:605-614.

154:1254-1259.

menians: Analysis of 100 cases. Medicine 53:53-46.

43227-253.

terranean fever. Arch Intern Med 107529-538.