Brain (2000), 123, 1102—111 1
`
`Autoimmune disease in first—degree relatives of
`patients with multiple sclerosis
`A UK survey
`
`S. A. Broadley,1 J. Deans,1 S. J. Sawcer,1 D. Clayton2 and D. A. S. Compstonl’3
`
`1University of Cambridge Neurology Unit, Addenbrooke’s
`Hospital, 2MRC Biostatistics Unit, Institute of Public
`Health and 3E. D. Adrian Building, Cambridge, UK
`
`Correspondence to: Alastair Compston, University of
`Cambridge Neurology Unit, Addenbrooke’s Hospital, Hills
`Road, Cambridge C32 2QQ, UK
`E-mail: alastair.compston@medschl.cam.ac.uk
`
`Summary
`Previous studies examining an association with other
`autoimmune diseases have suggested the existence of a
`generalized autoimmune diathesis
`in patients with
`multiple sclerosis. We investigated the prevalence of
`autoimmune disease in first-degree relatives of probands
`with multiple sclerosis using a case—control method. The
`results show an excess of autoimmune disease within these
`
`families, but no significant association was seen with non-
`autoimmune diseases. The higher risk in multiplex than
`simplex families suggests an effect of genetic loading.
`While the increase in risk applies to each autoimmune
`
`disease, autoimmune thyroid disease (and Graves’ disease
`in particular) contributes disproportionately to the excess
`risk. There was no increase in autoimmune disease within
`
`patients with multiple sclerosis themselves when compared
`with the index controls or population data. We conclude
`that autoimmune disease is more common in first-degree
`relatives
`of patients with multiple
`sclerosis
`and
`hypothesize that common genetic susceptibility factors
`for autoimmunity co-exist with additional disease specific
`genetic or environmental factors, which determine clinical
`phenotype in the individual.
`
`Keywords: multiple sclerosis; autoimmune disease; epidemiology; prevalence; familial
`
`Introduction
`
`Autoimmune disease is characterized by humoral or cell
`mediated immune response to self-antigen. This may be
`organ specific or systemic and, given the various overlap
`syndromes and occurrence of more than one autoimmune
`disease in the same patient (Sheehan and Stanton-King,
`1993), the different phenotypes are thought to represent a
`spectrum of immune dysregulation (Gordon and Isenberg,
`1990). Multiple
`sclerosis
`shares many
`clinical
`and
`pathological characteristics of prototypical autoimmune
`diseases (Hafler and Weiner, 1989).
`The majority of autoimmune diseases, including multiple
`sclerosis, are more common in women and show an increasing
`prevalence throughout adult life with peak incidence between
`the ages of 20 and 40 years (Beeson, 1994). A tendency for
`remission during pregnancy, with a transient deterioration or
`increased incidence of onset in the puerperium, has been
`reported for several autoimmune diseases (Mitchell and
`Bebbington, 1992; Tada et al., 1994; Nelson and Ostensen,
`1997), including multiple sclerosis (Confavreux et al., 1998),
`and each shows a variable response to immunosuppressive
`therapy. The presence of autoantibodies, usually associated
`
`© Oxford University Press 2000
`
`Page 1 of 10
`
`with other conditions, is well recognized in multiple sclerosis
`(Spadaro et al., 1999). The significance of this finding is less
`clear since these autoantibodies are not usually associated
`with clinical evidence of disease (Baker et al., 1972). The
`hallmark of multiple sclerosis as an autoimmune disease is
`the perivascular accumulation of autoreactive T cells
`(Stinissen et al., 1998).
`There are occasional case reports linking multiple sclerosis
`and autoimmune disease within families (McCombe et al.,
`1990), and one case—control study looking at the risk of
`chronic inflammatory disease within the immediate family
`members of patients with multiple sclerosis (Midgard et al.,
`1996). This and other studies (Warren and Warren, 1982)
`suggest an association with diabetes, but others have failed
`to show any familial link between these disorders (Alter and
`Sawyer, 1970). However, none of these studies has sufficient
`sample size to be definitive and each fails to make the
`distinction between type 1 and type 2 diabetes.
`As part of ongoing genetic studies in multiple sclerosis,
`we have accumulated a large cohort of simplex and multiplex
`families from throughout the UK. We have conducted a
`
`Biogen Exhibit 2115
`
`Mylan v. Biogen
`IPR 2018-01403
`
`Biogen Exhibit 2115
`Mylan v. Biogen
`IPR 2018-01403
`
`Page 1 of 10
`
`

`

`Autoimmune disease in multiple sclerosis families
`
`1103
`
`Table 1 Population prevalences of autoimmune diseases in the UK
`
`Autoimmune disease
`
`Prevalence (%)
`
`References
`
`Hashimoto’s/hypothyroidism*
`Graves’lhyperthyroidism*
`Rheumatoid arthritis
`Type 1 diabetes mellitus
`Pemicious anaemia
`Systemic lupus erythematosus
`Myasthenia gravis
`Addison’s disease
`Total
`
`0.80
`0.65
`0.55
`0.34
`0.13
`0.027
`0.015
`0.009
`2.52
`
`(Tunbridge et al., 1977)
`(Shank, 1976; Tunbridge et al., 1977)
`(Hochberg, 1990)
`(Gatling et al., 1998)
`(Scott, 1960)
`(Hopkinson et al., 1993; Johnson et al., 1995)
`(Robertson et al., 1998)
`(Willis and Vince, 1997)
`
`Mean figures are given where more than one reference was available. *Adult population figure.
`
`retrospective case—control postal questionnaire survey in
`order to assess whether autoimmune disease is more common
`
`in first-degree relatives of probands with multiple sclerosis.
`
`Methods
`Power calculations indicated that 250 cases and 250 controls
`
`would be required to demonstrate a twofold increase in the
`risk of autoimmune disease, with 0c (the probability of a false
`positive, i.e. type 1 error, result) = 0.05 and B (the probability
`of a false negative, i.e. type 2 error, result) = 0.2, assuming
`a background prevalence of 2.5% (Table 1) and an average
`of three relatives per family. A larger sample size was
`predicted for unequal sample sizes, but this requirement is
`counteracted by the use of simplex and multiplex families in
`order to measure dosage effect. Multiplex families were
`defined as a proband with one or more first-degree relatives
`having multiple sclerosis. Ethical approval was obtained from
`the local Cambridge (CREC) and Oxford and Cambridge
`Regional Ethics Committee (MREC).
`The patients were 773 cases referred by members of the
`Association of British Neurologists from throughout the UK,
`or those volunteering for participation in research through
`the Multiple Sclerosis Society of Great Britain and
`Northern Ireland, and Brunel University. Inclusion criteria
`were a confirmed diagnosis of multiple sclerosis made by a
`consultant neurologist, Caucasian origin and living parents
`born in the UK. Probands were excluded if they failed to
`meet the Poser criteria for a diagnosis of clinically definite
`(Poser
`et al.,
`1983),
`laboratory-supported definite or
`laboratory-supported probable multiple sclerosis (categories
`A—C). All multiple sclerosis probands were visited to establish
`the diagnosis and record clinical details. Confirmation of
`the diagnosis was made using medical records, and where
`applicable, results of MRI, visual evoked potentials and CSF
`examination were obtained.
`
`The autoimmune diseases selected for study are the most
`prevalent of those with recognized associated autoantibodies
`(Patrick, 1993). Their overall population prevalence in the
`UK is 2.5% (Table 1). Psoriasis was also considered as a
`putative T—cell mediated autoimmune disease (Barker, 1998)
`but analysed separately. Crohn’s disease and ulcerative colitis
`
`Page 2 of 10
`
`were investigated because of previous reports linking these
`diseases with individuals having multiple sclerosis (Rang
`et al., 1982) and their families (Minuk and Lewkonia, 1986;
`Sadovnick et al., 1989).
`Non-autoimmune diseases were considered in order to
`
`control for reporting bias. ‘Heart attack’ was included in the
`list of conditions, and age of onset of a disease was requested
`allowing a differentiation to be made between type 1 and
`type 2 diabetes, so that type 2 would provide a further non-
`autoimmune control. The cut-off was made at 30 years,
`recognizing that this is arbitrary and that some cases of later
`onset diabetes are also autoimmune in nature. Atopic asthma
`was included as an ‘immunological’ control condition; there
`are reports linking atopy with multiple sclerosis (Frovig
`et al., 1967), although this is unconfirmed and others have
`found a reduced prevalence in patients with multiple sclerosis
`(Oro et al., 1996).
`Two identical questionnaires, one to be completed by the
`proband and the other by a control, were distributed. A
`reminder letter was sent to non-responders after 2 months.
`Controls were selected by the probands from a choice of
`spouse, partner, carer or friend. The questionnaire requested
`information about the list of specific autoimmune and non-
`autoimmune conditions in parents and siblings. Information
`about offspring was not requested because of ethical issues
`relating to informed consent in those aged under 18 years.
`Since the rate of autoimmune disease in childhood is low,
`this group would be relatively uninformative. Relevant
`information regarding each condition was provided and it
`was requested that all
`living relatives be asked directly
`about these conditions and for information from memory on
`deceased relatives. Where there was any doubt about a
`particular relative, participants were instructed to reply in the
`affirmative, so as to increase the likelihood of including all
`positive diagnoses. Informed consent for participation in the
`study was obtained from all index cases and controls.
`On receipt of completed questionnaires all living relatives
`said to have one of the listed conditions were contacted to
`
`request confirmation from their general practitioner. A repeat
`mail-shot was sent to non-responders after a further 2 months.
`Consent to release of medical information was obtained and
`
`further specific details relating to asthma (atopy, age of onset
`
`Page 2 of 10
`
`

`

`1104
`
`S. A. Broadley et al.
`
`and smoking history) and thyroid disease (previous surgery,
`thyroxine replacement therapy and specific diagnoses) were
`requested. It was therefore possible to distinguish atopic
`asthma from non-atopic asthma and primary hypothyroidism
`and hyperthyroidism from other secondary causes of thyroid
`disease. Atopic asthma was defined as asthma beginning
`before the age of 20 years, asthma with atopic features (hay
`fever or eczema) commencing before the age of 30 years, or
`asthma with atopic features prior to the age of 50 years in
`persons who had never been smokers. For the purposes of
`analysis, primary hypothyroidism and hyperthyroidism were
`assumed to have an autoimmune basis and were included with
`
`Hashimoto’s thyroiditis and Graves’ disease, respectively. The
`majority of cases with primary hyperthyroidism are due to
`Graves’ disease, but the remainder usually also have an
`autoimmune basis (predominantly Hashimoto’s thyroiditis).
`General practitioners were contacted to confirm or refute
`each diagnosis. Three options were given: ‘yes’, the diagnosis
`is correct; ‘no’, this person has never had this disease; or
`‘uncertain’ , it is not possible to confirm or deny this diagnosis
`based on personal knowledge and the available medical
`records. Only affirmative responses have been included in
`the analysis. Demographic details and all diagnostic data
`were entered into a specifically created Microsoft® Access
`database, such that positive diagnoses were entered three
`times: initial notification, relatives’ verification and general
`practitioner
`confirmation. Thus,
`encoding errors were
`minimized.
`
`To explore under-reporting of autoimmune diseases within
`the
`control
`families,
`a
`randomly chosen,
`apparently
`unaffected, member of every third responding family was
`contacted. This individual was asked to confirm their date of
`
`birth, number of relatives in the family and whether they had
`ever been diagnosed with any of the specified conditions.
`Statistical analysis was carried out using the proportion of
`families in whom one or more relatives had the specified
`condition. A x2 test for trend (quend) was used to test for
`dose effect of genetic loading between control, simplex and
`multiplex families with unitary weighting (Fleiss, 1981). The
`x2 test was used for comparisons between two groups and
`the odds ratio, which very closely approximates to the relative
`risk in large samples, used to calculate the sibling risk. The
`age— and sex-adjusted prevalence figure for autoimmune
`disease in control relatives was calculated using the UK mid-
`census estimates for 1997 (Matheson and Pullinger, 1999).
`
`Results
`
`Response rates
`The overall response rates for usable questionnaires were
`375 of 647 (58%) for controls and 571 of 753 (76%) for
`cases. A breakdown of recruitment and reasons for exclusion
`
`are given in Fig. 1. Analysis of the non-responding cases
`revealed that they were twice as likely to have an incomplete
`address (e.g. missing postcode) and were more likely to have
`
`Page 3 0f 10
`
`been originally visited for other aspects of genetic research
`more than 4 years prior to the postal survey. This suggests
`that many of the non-responders may never have received
`the questionnaires due to inadequate postal details or through
`having moved since originally being recruited. Disease
`severity did not influence the likelihood of response and this
`may reflect the option that the questionnaire be completed,
`if necessary, by another relative or carer on behalf of the
`proband. In total,
`the survey included 3439 relatives of
`946 index cases and controls. Thirty-two half-siblings were
`excluded from the analysis. One thousand individuals were
`contacted regarding positive diagnoses, of whom 879 replied.
`
`Diagnostic confirmation
`The number of general practitioners contacted was 799, of
`whom 722 responded. It was therefore possible to confirm
`the diagnosis of autoimmune disease through general
`practitioners in 78% of living affected relatives. The figure
`for all diagnoses was 75%. The positive predictive value of
`diagnoses in living relatives initially reported by the index
`case or control for each condition is listed in Table 2. With
`
`the exception of rheumatoid arthritis, pernicious anaemia and
`ulcerative colitis, reliable reporting figures of 70% or more
`were seen and rates were similar for case and control relatives.
`The confusion of rheumatoid arthritis with osteoarthritis was
`
`predictable and explains the positive predictive value of less
`than 50% in both cases and controls.
`
`Demographics offamilies
`Of the 571 index cases, criteria for Poser category A were
`met in 513 (89.8%), category B in 32 (5.6%) and category
`C in 26 (4.6%). There were 140 males and 431 females (ratio
`1 2 3). Disability scores (from the expanded disability status
`scale) ranged from 0 to 9.5, with 217 having scores less than
`4, 274 from 4 to 7, and 80 greater than 7. A relapsing—
`remitting course was reported in 339 (59.4%), with 180
`(31.5%) having secondary progressive disease and 52 (9.1%)
`primary progressive disease. These figures accord well with
`previously reported cross-sectional surveys of patients with
`multiple sclerosis (Miller et al., 1992).
`The index controls were spouses or partners in 73%. Of
`the control relatives, 22% were no longer living as opposed
`to only 2% of case relatives. Deceased relatives were included
`in the analysis, except where otherwise stated. There were
`375 control, 508 simplex and 63 multiplex families. The
`11% frequency of multiple sclerosis cases with a co-affected
`first-degree relative is consistent with UK population figures
`(Robertson et al., 1996).
`There was a co—affected sibling in 49 of the multiplex
`families. Because of the way in which these families were
`identified (i.e. affected sibling pair), they were predictably
`larger. The 49 co—affected siblings have therefore been
`excluded so as to avoid any potential bias from including
`siblings with multiple sclerosis and concurrent autoimmune
`
`Page 3 of 10
`
`

`

`625
`
`Autoimmune disease in multiple sclerosis families
`
`1105
`
`
`
`Cases contacted
`773
`
`Potential controls
`
` Control replies
`
`Additional controls
`18
`
`484 (77%)
`
`Presumed delivered
`753
`
`Case replies
`612 (81%)
`
`571 (76%)
`
`Multiplex families
`63
`
`,— ______________________
`
`Included families
`
`Case families
`
`I|I IIII I I I
`
`Control families
`
`Simplex families
`508
`375 (60%)
`I
`l. _______________________
`
`Fig. 1 Summary of recruitment and reasons for exclusion of case and control families. Controls were deemed unsuitable if they
`themselves had multiple sclerosis (1), were a blood relative of the index case (30) or non-Caucasian (8). ‘Additional controls’ refers to
`instances where more than one control family was supplied by the same index case.
`
`Table 2 Positive predictive value for each condition
`
`Table 3 Details offamily members
`
`Diagnosis
`
`Cases (%)
`
`Controls (%)
`
`Relative
`
`Control
`
`Simplex
`
`Multiplex
`
`Multiple sclerosis
`Thyroid disease
`Rheumatoid arthritis
`Diabetes (type 1 and 2)
`Pernicious anaemia
`Systemic lupus erythematosus
`Psoriasis
`Crohn’s disease
`Ulcerative colitis
`Asthma
`
`Myocardial infarction
`Overall
`
`87
`85
`32
`94
`45
`67
`55
`80
`55
`74
`
`74
`70
`
`100
`88
`41
`97
`100
`N/A
`78
`100
`64
`87
`
`68
`78
`
`Proportion of notified diagnoses in living relatives confirmed by
`general practitioners. N/A = not applicable.
`
`disease (in fact only one of the 49 had an autoimmune
`disease) and to equilibrate the sibship sizes. The mean ages
`of relatives, the sibship sizes and sex distribution for siblings
`were comparable between the three groups (Table 3). The
`figure of 2.5 for the mean number of offspring per family is
`slightly higher than the population figure for mothers 40
`years ago of just over 2 (Harris, 1997). This difference may
`merely reflect the fact that, by the very nature of this study,
`no families without children were included. When sibling
`numbers for one-third of the control families were checked
`
`with a second family member, five additional siblings were
`identified. When extrapolated to the entire set of control
`families, this only constitutes an error of 1% in the number
`
`Page 4 of 10
`
`Mean age of relatives (range) in years
`Index
`42.8 (19—68)
`40.7 (19—60)
`Mother
`68.4 (28—95)
`67.7 (42—92)
`Father
`67.3 (33—90)
`70.2 (33—93)
`Sister
`42.2 (17—84)
`39.4 (7—61)
`Brother
`41.4 (1—67)
`39.2 (21—56)
`All
`55.6 (1—95)
`55.6 (7—93)
`(excluding index)
`
`40.0 (21—56)
`65.7 (42—85)
`68.6 (48—93)
`36.9 (8—63)
`37.6 (1—52)
`55.8 (1—93)
`
`Sibship size and proportion of sisters
`Sibship size
`1.63
`Sisters (%)
`45
`
`1.65
`51
`
`1.51
`54
`
`Sibship size refers to the mean number of siblings for each family
`type.
`
`of control relatives. Seven control relatives also had a
`
`diagnosis of multiple sclerosis, which was confirmed in all
`six of the living cases. These families remained in the analysis
`as controls.
`
`Autoimmune disease in control relatives
`
`The prevalence data for autoimmune disease in all control
`relatives are summarized by age in Fig. 2. These give an
`age— and sex-adjusted prevalence of 2.3% for the selected
`autoimmune diseases in the control relatives, which compares
`well with the predicted population figure of 2.5%. Ninety-
`one of 117 unaffected control relatives who were contacted
`
`Page 4 of 10
`
`

`

`1106
`
`S. A. Broadley et al.
`_; 0
`
`
`
`Frequency(%)
`
`O-tNO)A01O)\lC0(0 [I
`
`<35
`
`35—44
`
`45—54
`
`55—64
`
`65-74
`
`>74
`
`Age group (years)
`
`Fig. 2 Age-specific prevalence of selected autoimmune diseases in control relatives (11 = 1315). Error
`bars show the standard error.
`
`
`
`'
`
`7
`
`35
`
`30
`
`C a
`
`,
`
`'
`
`
`
`Control
`
`.
`Single
`multiple sclerosis
`
`.
`Multiplex
`
`18
`
`16
`
`14
`
`0:8 12
`5‘ 10
`
`C g
`
`3E
`
`8
`
`64 2 0
`
`r
`A
`m
`
`.
`Simplex
`Family type
`
`.
`Multlplex
`
`Control
`
`A 25
`o\°
`E 20
`)
`g) 15
`\—
`L 10 7%}fl
`5 I0
`
`Fig. 3 Prevalence rates based on one or more relatives having any
`of the selected autoimmune diseases. ‘Confirmed’ (white bars)
`refers to the diagnosis in one or more of these affected relatives
`having been verified by the general practitioner; unconfirmed =
`striped bars. Error bars show the standard error for the combined
`data.
`
`responded. None reported suffering from any of the listed
`autoimmune diseases, confirming that under-reporting was
`not a significant problem.
`
`Autoimmune disease in case and control
`
`families
`The number of families where one or more relatives had one
`
`of the selected autoimmune diseases was 44 of 375 (11.7%)
`in control families, 80 of 508 (15.7%) in simplex families
`and 16 of 63 (27%) in multiplex families (quend = 8.95,
`P = 0.003). Figure 3 shows this result graphically and also
`demonstrates the proportion of families in which at least
`one relative with a diagnosis of autoimmune disease was
`confirmed by the general practitioner. The result for the
`control families is slightly higher than that predicted by the
`population data (3.5 X 2.5% = 8.8%) and reflects the mean
`age of 56 years seen for all relatives. To remove any potential
`bias due to disparity in the sibship size, sibling sex distribution
`or proportion of deceased parents,
`the prevalences of
`
`Page 5 0f 10
`
`Familytype
`
`Fig. 4 Autoimmune disease in living parents. Error bars show the
`standard error. Striped bars = mothers; white bars = fathers.
`
`autoimmune disease in living mothers and living fathers were
`calculated separately (Fig. 4). These data combined also
`show a statistically significant relationship (thrend = 4.06,
`P = 0.04).
`The association with all autoimmune disease was seen
`
`for each individual condition (Fig. 5). The relationships
`for Hashimoto’s disease/autoimmune hypothyroidism and
`Graves’ disease/autoimmune hyperthyroidism reach statist-
`ical significance in their own right, with Graves’ disease/
`primary hyperthyroidism appearing to contribute the largest
`effect. The result for systemic lupus erythematosus just
`fails to reach significance at the 5% level (P = 0.051).
`In the case of rheumatoid arthritis, deceased relatives
`account for half of the unconfirmed control families and
`
`diagnostic confirmation rates are generally low for all family
`types (Fig. 5C). In View of the low positive predictive value
`of around 40% for rheumatoid arthritis seen in living relatives,
`it is reasonable to presume that a fair proportion of these
`unconfirmed rheumatoid arthritis diagnoses are incorrect and
`mistaken for osteoarthritis. If rheumatoid arthritis is excluded
`
`from the analysis, the result becomes more significant (P =
`0.000035) indicating that our quoted P-value is likely to be
`
`Page 5 of 10
`
`

`

`(A)
`
`16
`3 14
`L 12
`a 10
`E
`8
`D'
`3
`5
`9
`4
`LL
`2
`0
`
`Hashimoto’s thyroiditis/hypothyroidism
`
`Control
`
`,
`Simplex
`
`,
`Multiplex
`
`(B)
`
`12
`”a 10
`E
`a 8
`5
`6
`U'
`:i
`4
`92
`Li.
`2
`o
`
`Autoimmune disease in multiple sclerosis families
`
`1107
`
`Graves' disease/Hyperthyroidism
`
`(C)
`
`1O
`
`Rheumatoid arthritis
`
`’a
`8
`at
`a e
`8
`U'
`3
`93
`Li.
`
`4
`2
`
`0
`
`Control
`
`,
`Simplex
`
`,
`Multiplex
`
`Control
`
`
`
`,
`Simplex
`
`.
`Multiplex
`
`fiend = 7.21,
`P = 0.007
`
`Family type
`
`firm, = 1153,
`P: 00007
`
`Family type
`
`12mm = 123,
`P = 0.27
`
`Family type
`
`(D)
`
`3.0
`3 2.5
`g; 2.0
`g 1 5
`0)
`~
`E} 1.0
`LL
`-
`9 o 5
`0.0
`
`Type 1 Diabetes mellitus
`
`Control
`
`_
`Simplex
`
`,
`Multiplex
`
`(E)
`
`3.5
`3 3.0
`g; 2-5
`g 2.0
`02 1.5
`3
`LL
`.
`9 3'2
`0.0
`
`Pernicious anaemia
`
`Control
`
`,
`Simplex
`
`_
`Multiplex
`
`(F)
`
`3.5
`:5 3.0
`9; 2 5
`g 2.0
`w 1.5
`3
`LI.
`.
`2 3'2
`0-0
`
`Systemic lupus erythematosus
`
`Control
`
`Simplex
`
`.
`Multiplex
`
`X2trend = 0014,
`P: 0.9
`
`Family type
`
`xztrend = 0.96,
`P: 0.33
`
`Family type
`
`thrend = 3.81,
`P: 0.051
`
`Family type
`
`Fig. 5 (A—F) Frequency of individual autoimmune diseases. Error bars show the standard error. (C) The proportion of confirmed (white)
`and unconfirmed (striped) cases with rheumatoid arthritis. Error bars and hypothesis test statistic are based on combined data.
`
`a conservative estimate. This result remains significant even
`when the non-responding case families are included on the
`basis that they had no autoimmune disease (xztrend = 8.49,
`P = 0.0036).
`There were no type 1 diabetic relatives in the multiplex
`families where one or two might have been expected. Only
`two cases of myasthenia gravis were reported, one in a
`deceased control relative and the second in a living case
`relative in whom the general practitioner was ‘uncertain’
`about the diagnosis; this individual was excluded. No cases
`of Addison’s disease were reported.
`
`Sibling risk for autoimmune disease
`The number of siblings with one or more autoimmune
`diseases was 10 of 565 (1.8%) for controls and 30 of 982
`(3.1%) for cases. The sex-adjusted sibling (its) risk is 1.65
`(95% confidence interval, 1.0—3.4).
`
`Other conditions in families
`Psoriasis showed a similar trend to the selected antibody
`associated autoimmune diseases (Fig. 6A).
`Inflammatory
`bowel disease showed no association with multiple sclerosis
`(Fig. 6B). The non-autoimmune control diseases, myocardial
`infarction and type 2 diabetes mellitus, and asthma showed
`no familial relationship with multiple sclerosis (Fig. 6C—E).
`
`Autoimmune disease in index cases and
`
`controls
`
`carefully matched for age and sex. However, calculated sex-
`adjusted prevalence rates for autoimmune disease were 2%
`in the index cases and 2.73% in the index controls (Z =
`0.74, P = 0.46). These figures are comparable with a
`population prevalence of 2.5% for these autoimmune diseases
`in the UK. The figures for individual conditions are listed in
`Table 4 and most agree with population data. The numbers
`of cases and controls make interpretation of the apparently
`lower rate of autoimmune disease in cases impossible, but
`do at
`least indicate that there has been no bias towards
`
`autoimmune disease in the cases who responded. There
`were no cases of pernicious
`anaemia,
`systemic
`lupus
`erythematosus, myasthenia gravis or Addison’s disease
`among the index cases and controls.
`
`Discussion
`Here we show an increased risk of other autoimmune diseases
`
`in the relatives of patients with multiple sclerosis. This
`study was large with high response rates and diagnostic
`confirmation, and used a conservative family based method
`of statistical analysis. The sex distributions of the index cases
`and controls are clearly influenced by the use of spouses/
`partners as controls. However,
`these individuals were not
`used for the primary analysis and only their relatives were
`included. Fears that men may be less diligent in reporting
`family history data were not born out, as the frequencies of
`all conditions studied in the control families match published
`figures for the UK population, and random sampling for
`under-reporting of autoimmune disease failed to identify
`missed cases in the control families. In a condition which
`
`An estimate of prevalence for autoimmune disease in index
`cases and controls would require several thousand participants
`
`predominantly affects women, such as multiple sclerosis,
`cases will tend to come from families with a higher proportion
`
`Page 6 0f 10
`
`Page 6 of 10
`
`

`

`Psoriasis
`
`(B)
`
`Inflammatory bowel disease
`
`1108
`
`S. A. Broadley et al.
`
`(A) 18
`A 16
`§ 14
`5. 12
`c 10
`GD
`8
`e 6
`a)
`l.
`4
`
`..
`
`2 I
`
`o
`
`Control
`
`2
`X "end = 3.80,
`p = 0.051
`
`,
`Simplex
`.
`Family type
`
`,
`Multiplex
`
`
`
`6
`q;
`5
`9/
`6 4
`C
`3
`8
`2
`cr
`
`1
`
`O
`
`Control
`
`2
`x trend = 0.0008,
`p = 0.93
`
`.
`Simplex
`.
`Family type
`Present study
`
`Multiplex
`
`Sadovnick
`Minuk and
`and Paty
`Lewkonia
`Previous studies
`
`
`
`(C)
`
`A 3.5
`28 3.0
`>. 2.5
`0 2.0
`=
`g 1.5
`o' 1.0
`9 0.5
`LL 0.0
`
`Myocardial infarction in living relatives
`Normalized to a standard family of two parents,
`one brother and one sister
`
`Control
`
`,
`Simplex
`
`‘
`Multiplex
`
`(D)
`
`3 5
`A 3'0
`'
`o\°
`‘; 2~5
`0 2.0
`8
`3 1-5
`o- 1 0
`9 '
`u. 0.5
`0.0
`
`Type 2 Diabetes mellitus
`
`Control
`
`Simplex
`
`Multiplex
`
`(E)
`
`25
`A
`o\° 20
`I
`o 15
`S
`3 10
`c-
`9.’
`5
`u.
`
`0
`
`Atopic asthma
`
`Control
`
`Simplex
`
`Multiplex
`
`X2trend = 0.13,
`P=0.72
`
`Family type
`
`thrend = 2.72,
`P=O.1
`
`Family type
`
`xzfiend = 0.36,
`P=0.55
`
`Family type
`
`Fig. 6 (A—E) Frequency of other diseases. Error bars show the standard error. (B) The proportion of unconfirmed (striped) and confirmed
`(white) cases for inflammatory bowel disease. Error bars and hypothesis test statistic are based on combined data for present data only.
`Because heart disease was frequently reported as a cause of death in deceased relatives, and could not always be validated only living
`relatives have been analysed for myocardial infarction, and families normalized to a structure of two parents, one brother and one sister.
`
`Table 4 Sex-adjusted prevalences in index cases and controls
`
`Condition
`
`Cases
`(%)
`
`Controls Population Reference
`(%)
`(%)
`
`Hashimoto’s/hypothyroidism
`Graves’/hyperthyroidism
`Rheumatoid arthritis
`Type 1 diabetes mellitus*
`Psoriasis
`Type 2 diabetes mellitus
`Asthma
`Inflammatory bowel disease*
`Myocardial infarction
`
`0.4
`0.5
`0.2
`0.7
`2.8
`0.4
`6.5
`0.5
`0.5
`
`*Not sex-adjusted.
`
`1.2
`0.0
`1.4
`0.3
`3.4
`0.0
`7.1
`1.1
`0.8
`
`0.8
`0.7
`0.6
`0.3
`3.7
`1.5
`6.5
`0.4
`N/A
`
`(Tunbridge et al., 1977)
`(Shank, 1976; Tunbridge et al., 1977)
`(Hochberg, 1990)
`(Gatling et al., 1998)
`(Brandrup and Green, 1981)
`(Gatling et al., 1998)
`(Ertle and London, 1998)
`(Bernstein et al., 1999)
`—
`
`of daughters, and this effect is further increased in affected
`sibling pair families. In addition, the sex distribution of the
`index cases does have a subtle effect on the sibship sex
`distribution, with women tending to have a slightly higher
`proportion of sisters and vice versa for men. However,
`variations in the sibship sex distribution observed in the
`present study would not account for differences in the rates
`of autoimmune disease seen in the three sets of relatives,
`particularly as it is the parents who contribute the largest
`effect. We therefore conclude that the antibody associated
`autoimmune diseases studied are more frequent among first-
`degree relatives of patients with multiple sclerosis than well-
`matched controls, whereas non-autoimmune diseases show
`no such relationship. In addition, the data indicate increasing
`genetic load for autoimmune disease among members of
`multiplex families. However,
`the overall
`risk of other
`autoimmune diseases observed in relatives of our patients
`
`with multiple sclerosis is small, suggesting that genetic
`factors responsible for this apparent autoimmune diathesis
`contribute only a small part
`to genetic susceptibility in
`multiple
`sclerosis. Since ks
`for multiple
`sclerosis
`is
`approximately 20, the figure of 1.65 attributable to the risk
`of non-specific autoimmunity indicates that disease specific
`genetic factors are more significant.
`Autoimmune thyroid disease shows the strongest effect,
`but this may reflect
`the high prevalence compared with
`other autoimmune conditions. Using different methodology,
`without control data, a recent French study found a lower
`overall rate of autoimmune disease in first-degree relatives
`of patients with multiple sclerosis than in our study, but also
`reported a high prevalence of Graves’ disease (Heinzlef et al.,
`1999). This finding is of particular interest since a proportion
`of patients with multiple sclerosis treated with interferon-B
`(Rotondi et al., 1998) and one-third of patients receiving
`
`Page 7 0f 10
`
`Page 7 of 10
`
`

`

`Campath-lH (Coles et al., 1999) develop Graves’ disease.
`This suggests a specific relationship between Graves’ disease
`and multiple sclerosis, although the mechanism is not yet
`understood. The finding of no relatives with type 1 diabetes
`in the multiplex families may have arisen purely by chance,
`but could indicate a specific relationship between type 1
`diabetes and multiple sclerosis: a higher genetic load
`predisposing towards multiple sclerosis may be relatively
`protective against autoimmune diabetes. For example,
`the
`class 2 major histocompatibility complex allele HLA-DR2,
`which is clearly associated with multiple sclerosis (Coraddu
`et al., 1998), is protective in type 1 diabetes mellitus (Noble
`et al., 1996) in northern Europeans.
`Psoriasis shows a similar trend of genetic loading in
`simplex and multiplex families and this confirms
`the
`previously reported increased risk of psoriasis within multiple
`sclerosis families (Midgard et al., 1996). It would therefore
`seem reasonable to include psoriasis in the list of autoimmune
`diseases linked with multiple sclerosis. In contrast, we found
`no association with inflammatory bowel disease. This result
`is in contrast to previous studies (Minuk and Lewkonia,
`1986; Sadovnick et al., 1989) which report a higher than
`expected prevalence of
`inflammatory bowel disease in
`relatives of patients with multiple sclerosis. However, these
`studies calculated the risk of an individual developing both
`conditions from published population prevalence figures, but
`not
`the family risk. The finding of inflammatory bowel
`disease in 1.13% (Minuk and Lewkonia, 1986) and 2.94%
`(Sadovnick et al., 1989) of families, if restricted to parents
`and siblings, is within the confidence intervals for our results
`in both case and control families. Taken together, these results
`provide no evidence supporting an increased prevalence of
`inflammatory bowel disease within first-degree relatives of
`patients with multiple sclerosis.
`Although