Article Text


Multiple sclerosis in sibling pairs: an analysis of 250 families


OBJECTIVES To assess the potential contribution of genetic factors to clinical phenotype in multiple sclerosis.

METHODS Using a cohort of 262 pairs of coaffected siblings from 250 families with multiple sclerosis, intersibling concordance analysis was used to explore underlying genetic mechanisms in disease pathogenesis by assessing parameters of disease course, clinical presentation, age and year of onset, and measures of disability and handicap.

RESULTS Adjusted intraclass correlation coefficients were not significant for either age of onset or for year of first symptom. One third of sibling pairs were concordant for presenting symptom (81/262), a result that was non-significant. However, course type was identical in 50% of the sibling pairs (κ=0.17 (95% confidence interval (95% CI) 0.08 to 0.26)) indicating a significant result. Severity of the disease at assessment, using the Kurtzke and CAMBS scales, demonstrated that whereas there was no agreement for relapse rate in the previous year within the sibship, there was significant concordance for measures of disability (κ=0.11 (95% CI 0.04 to 0.19)), progression (κ=0.09 (95% CI 0.01 to 0.18)) and handicap (κ=0.08 (95% CI 0.02 to 0.14)).

CONCLUSIONS Within a sibship, the clinical presentation tends to be different. However, once established, concordance is more likely to be seen for the ultimate course, leading in the end to similar disability and handicap scores. These results are consistent with the hypothesis that genes influence both disease susceptibility and evolution in multiple sclerosis.

  • multiple sclerosis
  • sibling pair
  • concordance

Statistics from

The familial tendency of multiple sclerosis is well recognised with relative risks of 100–190 for identical twins1 2; 7–13 for half3 and full siblings4; and 5.5 for the offspring of single affected,4 and 60 for conjugal parents.5. Coaffected siblings can be used as a resource for identifying susceptibility genes and have formed the basis for genome screens in various diseases through linkage analysis including insulin dependent diabetes,6 asthma,7 and inflammatory bowel disease.8 Three original genome screens were carried out in multiple sclerosis,9-11 with an additional one from a smaller genetically isolated Finnish population.12 Current approaches include attempts to identify susceptibility genes within these regions of interest using the positional candidate approach—now accelerated by the first draft of the human genome map; further full genome linkage screens with the aim of meta-analysis; and full linkage disequilibrium genome screening using case-control and family based association methods. Together these reinforce the role of the HLA region13 14; they identify areas of genetic overlap between populations; and document large exclusion zones which encompass 95% of the genome for genes of large effect and 65% for genes of moderate effect.

Sibling pairs also provide an opportunity for examining clinical phenotypic heterogeneity within kindreds; and for determining the relative contribution of genetic and environmental forces in shaping the disease expression. The first set from the United Kingdom coaffected sibling pair cohort has already been described numbering 166 families and contributing 177 sibling pairs.15 This series is now extended with reanalysis of the cumulative total of 262 pairs, primarily using a κ function which is able to incorporate potentially confounding variables and generate 95% confidence intervals (95% CIs) for the results obtained, through a novel bootstrap simulation process.



As part of an ongoing study of the genetic epidemiology of multiple sclerosis notifications were received from members of the Association of British Neurologists, with limited recruitment also occurring through articles distributed by the Multiple Sclerosis Society.

Patients were contacted by either a neurology research registrar or experienced neurology research sister, and the project explained. If the family was interested in taking part, arrangements were made to visit the proband at home; a standard history and examination were performed, 30 ml of blood drawn for extraction of DNA and subsequent genetic analysis, and arrangements made for the full medical notes to be obtained from the supervising clinician. A newsletter was distributed to all those involved in the project to keep them up to date with general progress of the research. Ethical approval was obtained from the local research ethics committee and multicentre research ethics committee.


Case definition followed the Poser criteria16 with each sibling pair including at least one member who was clinically definite (grade A), and the other either laboratory supported (B) or clinically probable (C). Efforts were made to exclude possible phenocopies such as the hereditary parapareses and ataxias. In addition, the entire cohort was screened for the autosomal dominant spinocerebellar ataxia (SCA) triplet expansion mutations,17 all male-male pairs tested for the presence of adrenoleukodystrophy (by assaying fasting concentrations of very long chain fatty acids), and more recently a random sample for mutations in the CADASIL gene. No examples of these disorders were identified.

Sibling 1 was designated as the member who first exhibited symptoms rather than the oldest or the first diagnosed sibling.15Although guided by the patient's own recall in assigning clinical onset of disease, meticulous attention was placed on clinical assessment during the visit, where a full history and examination were taken, supplemented by copies of the case notes. In this way a complete description of the disease presentation and progression was built up. Where there was potential inaccuracy due to cognitive impairment, then the documentary evidence was complemented by discussions with the relevant clinician.

Disease course and character were described in accordance with the European Database for Multiple Sclerosis (EDMUS) protocol,18 with four categories being defined: relapse-remitting (RR), primary progressive (P), secondary progressive (RR-P), and unknown (U). An attempt was made to assign symptoms to one or more neuroanatomical locations. Bilateral limb dysfunction and sphincter involvement were scored as cord disease; unilateral limb dysfunction alone as cortical disease; ataxia/dysarthria with or without cranial nerve/limb dysfunction as brainstem/cerebellar disease (no attempt was made to distinguish these two); whereas optic nerve and cognitive dysfunction were generally unambiguous. Despite limitations, this represents an attempt to demarcate the anatomical syndromes of multiple sclerosis. Disease severity was graded according to the Kurtzke expanded disability status scale (EDSS)19 and CAMBS.20

In the 5% of families with trios affected, three possible pairs AB, AC, and BC, were weighted as two thirds so that each family contributed two pairs to the analysis.


Kappa and intraclass correlation coefficient were used to measure pairwise concordance. The first was used for categorical data and the second for continuous data. These were initially developed for measuring agreement between raters, but their application in sibling pair analysis was suggested by Fleiss.21 Application to multiplex sibships substituted pairs of siblings as pairs of raters. Our analysis assessed whether sibling pairs were more alike than a random pair sampled from the group. Statistics close to 1 indicated that sibling pairs were extremely alike and values of 0 indicated that the sibling pairs were no different from any random pair generated from within the cohort. Any variable could have been tested for this type of concordance between siblings. We used: sibling sex, disease course, presenting symptoms, and site, Kurtzke and CAMBS disability/handicap scales, and age/year of onset, with statistics reported as the point estimate with the corresponding 95% CIs. The bootstrap CIs22 were not symmetric about the point estimate; therefore the construction of approximate CIs under the normal distribution would have been wrong. Results were considered significant at the 5% level if the interval did not contain 0.

As previously noted by Robertson et al, various confounding relations may be present in analysing clinical features of sibling pairs15; however only univariate adjustment was required as the confounders had minimal effect on the data. Therefore, more complex multiple variable adjustments were not performed, and the interpretation of results specifically itemises only the unadjusted data. Adjustment for sibling sex could not be made in most cases because the κ function became undefinable due to the many rare cells in the cross tabulation of outcome and sibling sex. For those situations where adjustment for sex could be made, there was no alteration in the original results; sibling sex was therefore excluded from the list of possible adjustors. Our analysis includes adjusted values for the intraclass correlation coefficient and the κ statistic23 for variables including duration, year of birth, age of first symptoms, and year of first symptoms.



The register of familial multiple sclerosis had received 1013 referrals at the time of this study, of which 456 were affected sibling pairs. 152 were excluded before the visit because of: duplicate referrals (14), not a sibling pair (five), untraceable siblings (19), cosibling dead (48), geographical inaccessibility (11), participation refused (17), medical staff did not recommend contact (seven), monophasic disease (15), and not multiple sclerosis (16). The attrition rate after the visit was 54/304 (18% families visited), with further exclusions due to non-paternity (seven), only one sibling ultimately available (nine), monophasic disease (16), not multiple sclerosis (two), inadequate clinical documentation (10), and possible progressive spastic paraparesis or spinocerebellar ataxia (10).


Table 1 details the characteristics of this cohort in terms of pedigree structure and sex. The total number of pedigrees was 250—238 pairs and 12 trios—equating to an effective number of 262 sibling pairs, an increase of 85 from the previous report. The sex structure of like and unlike sex pairings matches other series, including population based cohorts in that 359/512 people were female (70%).

Table 1

Cohort structure


A total of 395 patients were Poser grade A, 37 grade B, and 80 grade C. (The three with previously described grade D disease were re-evaluated during the study period and their Poser ratings had increased to A.) As expected, there was no agreement within sibships on the degree of diagnostic certainty, as this was influenced by the use of laboratory investigations, which would have been relatively unavailable to the older siblings, and by disease duration which disadvantages clinical status in the second affected.

Table 2 outlines the results which are broken down into the following categories

Table 2

Pairwise concordance for all variables and confounders

Disease course

Identical course type was found in 50% of the sibling pairs if RR and RR-P were considered as separate groups (table 3; κ=0.17 (95% CI 0.08 to 0.26)). If the categories are dichotomised into RR/RR-P (RR*) and P alone, the confidence intervals widened and the result was no longer significant (κ=0.09 (95% CI −0.01 to 0.22)).

Table 3

Disease course within sibships

Presenting symptoms and signs

The presenting symptoms are summarised in table 4. One third of sibling pairs were concordant for presenting symptom (81/262; not significant, κ=0.02 (95% CI −0.07 to 0.15)). Dichotomising the data to limb motor or long tract sensory versus remaining symptoms was also non-significant (κ=0.09 (95% CI −0.02 to 0.21)) and 0.04 (95% CI −0.02 to 0.10)) respectively). There was no significant concordance either for the main presenting site (73/262; κ=0.05 (95% CI −0.01to 0.12; table 5).

Table 4

Symptom presentation

Table 5

Neuroanatomical presentation


Severity of the disease at assessment using the Kurtzke and CAMBS scales demonstrated that whereas there was no similarity of relapse rate in the previous year within sibships, there was significant concordance for measures of disability, progression, and handicap, which remained significant after adjustment.

Age and year of onset

Intraclass correlation coefficients of 0.23 for age of onset and 0.40 for year of first symptom were significant but to a large extent reflected the method of sibling sampling, as previously described in the original cohort analysis (15). Adjusted values were not significant for either parameter (p∼0.06). The median time to diagnosis for the first sibling was 3 years (maximum 31 years, interquartile range 1–10 years), with the second sibling having a median of 4 years (maximum 38 years, interquartile range 2–10 years).


The use of sibling pairs in exploring features of multiple sclerosis dates from the 1930s24 when four probable cases were identified from a pool of 444 siblings. The world literature on validated sibling pairs has gradually increased as has the range of analyses, culminating in 1996 with the publication of three genome screens which utilised a combined total of 599 pairs.9-11Reviewing these published series indicates that various methods of ascertainment ranging from attempted total population coverage (Orkneys25 and Canada26) to national tertiary care referrals27 28 have been used in both the clinically oriented and molecular genetic series.

Two main themes emerge from comparative phenotypic studies: the relative contribution of genetic versus environmental factors (whether there is intersibling clustering for age of onset or year of disease onset); and clinical concordance data, which can also be applied to counselling individual families.

Although mean difference in age of onset was lower than year at onset (1.4 v 4.7 years) in the study of Schapira et al, the difference did not reach significance.29 When sibling pairs and sporadic cases were compared by Doolittle et al,27about double the number of sibling pairs had an age at onset within 5 years of each other compared with sporadic cases, whereas this effect was not found for year at onset: simple comparison of the mean age at onset and year at onset showed that both were similar (∼6 years), demonstrating potential pitfalls in this style of analysis. This was reinforced by the study of Bulman et al 26 which again showed these parameters to be the same (∼3 years); but when the more complex technique of intraclass correlation was used, significance was demonstrated for age at onset. Smaller studies of between seven30 and 22 pairs31 reached no firm conclusions. Our previous series found no correlation for age at onset, but did detect a small difference in year at onset, possibly due to anticipation of symptoms by the second sibling. No significance differences in age or year of onset were found in either the most recent United States or French series.28 32

We found concordance within sibships for disease course and measures of progression, handicap, and disability using both the EDSS and CAMBS measures, but not for presenting symptom and site, or relapse pattern over the preceding year. Comparing these data with the initial set of 166 pairs, the enlarged group reinforces the previous conclusion concerning concordance for disease course, with κ coefficients of ∼0.17 (95% CI ∼ 0.07 to 0.25) regardless of the confounders used; augments them with novel findings for progression and degree of handicap which were previously not measured; demonstrates the positive findings for two measures of disability; but finds that concordance is lost for sex (95% CI −0.09 to 0.13), implying that those original results were false positives due to a smaller sample size. Our disability data support the most recent French cohort which, using the progression index (PI) defined as the ratio of EDSS to disease duration if the duration was more than 5 years, reported a significant concordance in 61 pairs (r=0.234, p=0.030).28

Taken together these results show that, within a sibship, the presentation is more likely to be different than the same. However, once established, concordance is more likely to be seen for the ultimate course, culminating in similar disability and handicap scores. As the disease process becomes established, measures of disease outcome are more likely than not to equilibrate within a sibship. Perhaps this is the expression of the hidden rounds of inflammation, demyelination, and axonal loss, which although initiated in different regions of the neuraxis, subsequently display similar tempos of progression and coalesce with the same outcome in any two siblings.


We are grateful to all those affected with multiple sclerosis and their families for involvement in this research project. The assistance of both the national organisation and local branches of the Multiple Sclerosis Society of Great Britain and Northern Ireland is gratefully acknowledged, coupled with help from many members of the Association of British Neurologists, who identified pedigrees. This work was supported by the Wellcome Trust and the Multiple Sclerosis Society of Great Britain and Northern Ireland.


View Abstract

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.