THE INHERITANCE OF HIP OSTEOARTHRITIS IN ICELAND

Transcription

1 ARTHRITIS & RHEUMATISM Vol. 43, No. 12, December 2000, pp , American College of Rheumatology 2785 THE INHERITANCE OF HIP OSTEOARTHRITIS IN ICELAND THORVALDUR INGVARSSON, STEFÁN EINAR STEFÁNSSON, INGILEIF B. HALLGRÍMSDÓTTIR, MICHAEL L. FRIGGE, HALLDÓR JÓNSSON, JR., JEFF GULCHER, HELGI JÓNSSON, JÓN INGVAR RAGNARSSON, L. STEFAN LOHMANDER, and KÁRI STEFÁNSSON Objective. To assess, in a population-wide study in Iceland, the genetic contribution to hip osteoarthritis (OA) leading to total hip replacement (THR). Methods. Information from 2 population-based databases in Iceland was combined: a national registry of all THRs performed between 1972 and 1996, and a genealogy database of all available Icelandic genealogy records for the last 11 centuries. A genetic contribution to THR for OA was assessed by 1) identifying familial clusters of OA patients with THR, 2) applying the minimum founder test (MFT) to estimate the minimum number of ancestors ( founders ) that would account for the genealogy of all 2,713 patients with THR for OA, compared with the average number of founders for control lists, 3) calculating an average pairwise kinship coefficient (KC) for the patient list and control lists, and 4) estimating the relative risk (RR) for THR among relatives of OA patients who have undergone the procedure. One thousand matched control lists, each the same size as the patient list, were created using the genealogy database. Results. A large number of familial clusters of patients with THR for OA were identified. The MFT showed that OA patients descended from fewer founders than did subjects in the control groups (P < 0.001). The Supported by decode Genetics Iceland, the Swedish Medical Research Council, Lund Medical Faculty and University Hospital, the Swedish Rheumatism Association, the King Gustav V 80-year Fund, and Kock Foundations. Thorvaldur Ingvarsson, MD: Central Hospital, Akureyri, Iceland, and University Hospital, Lund, Sweden; Stefán Einar Stefánsson, PhD, Ingileif B. Hallgrímsdóttir, MS, Michael L. Frigge, PhD, Jeff Gulcher, MD, PhD, Kári Stefánsson, MD, PhD: decode Genetics, Reykjavik, Iceland; Halldór Jónsson, Jr., MD, PhD, Helgi Jónsson, MD, PhD, Jón Ingvar Ragnarsson, MD, PhD: University Hospital, Reykjavik, Iceland; L. Stefan Lohmander, MD, PhD: University Hospital, Lund, Sweden. Address reprint requests to Thorvaldur Ingvarsson, General Hospital, 600 Akureyri, Iceland; or to Kári Stefánsson, decode Genetics, Lynghálsi 1, 110 Reykjavik, Iceland. Submitted for publication May 25, 2000; accepted in revised form August 21, average pairwise KC among patients with OA was greater than in the control population (P < 0.001). The RR for THR among siblings of OA patients was 3.05 (95% confidence interval ). Conclusion. This population-based study shows that Icelandic patients with hip replacement for OA are significantly more related to each other than are matched controls drawn from the Icelandic population. These findings support a significant genetic contribution to a common form of OA and encourage the search for genes conferring an increased susceptibility to OA. Osteoarthritis (OA) is a complex disorder with a high prevalence in the population, which increases with age. An imprecise distinction between diagnostic and nondiagnostic disease confounds all studies on OA. Multiple pathogenetic mechanisms are implicated in the development of OA, in which a complex genetic background appears to interact with environmental factors to initiate the disease and drive its progression. Common forms of OA lack a clear Mendelian pattern of transmission. Early studies (1,2) illustrated that Heberden s nodes of the fingers are hereditary, and genetic associations with other manifestations of OA have subsequently been confirmed (3,4). Other work has further strengthened the case for the influence of heredity on common forms of OA (5 12). Additional evidence for a genetic predisposition to OA is given by a number of rare subtypes of OA that have a genetic basis and are associated with the onset of OA at an early age (for review, see ref. 13). The spectrum of these forms of hereditary OA is quite varied, encompassing mild disorders that do not become clinically apparent until late in adult life to severe forms that manifest during childhood. The severe forms of OA share the characteristic of being caused by mutations in genes encoding macromolecules expressed in cartilage. Examples of mutated

2 2786 INGVARSSON ET AL genes that may be responsible for these conditions include genes encoding for cartilage-specific collagens and cartilage oligomeric matrix protein (13 22). Additional reports have presented evidence either for or against the association of OA with yet other genes encoding molecules related to cartilage function (23 26). Finally, linkage studies involving large numbers of families and sibling pairs with OA, coupled with novel techniques for genome-wide scans, are now gathering evidence for several other predisposing loci for OA (27 29). The prevalence of hip OA and of total hip replacement (THR) for OA in Iceland is considerably higher than in other Scandinavian countries (30,31). We have previously identified a large Icelandic family with a very high prevalence of hip OA, suggesting the presence of a hereditary risk factor (32,33). Iceland, an island in the North Atlantic Ocean, was settled around AD 900. The current Icelandic population of 275,000 descends from a few founders, and immigration has always been minimal. The population has experienced several catastrophic decimations due to disease and famine, sometimes in connection with natural disasters. Written and oral records maintain a detailed family history for the Icelandic population since the time of settlement more than a millennium ago. The isolation of the population and the unique genealogic information, coupled with high-quality health service records, make the Icelandic population ideal for genetic research. In this study, we have used information obtained from an Icelandic national registry of patients who have undergone joint replacement for hip OA, as well as data from a newly developed genealogy database. With these unique population-based resources, we identified numerous, large familial clusters of patients with THR for OA and performed several population-based tests to assess the genetic contribution to hip OA. PATIENTS AND METHODS Patient population. For this study, the criterion of THR for the diagnosis of hip OA was used to identify cases. This allowed the generation of a population-based list of patients with hip OA. We have previously shown that the THR procedure as performed for hip OA in Iceland is consistently preceded by the existence of severe disease and the presence of advanced radiologic OA (30,31,34). The radiologic pattern of hip OA in Iceland is similar to that reported in other studies (30,34 37), both with regard to the proportion of bilateral cases and the radiologic appearance (lateral, mixed, medial). The mean age at the time of THR surgery in Iceland during the years was 69 years, both for women and for men (31), in concordance with that reported for other Scandinavian countries (38,39). Icelandic hip arthroplasty registry. We performed a computer-aided search of hospital records from all 6 orthopedic clinics in Iceland. This generated information on all patients who had undergone THR for primary OA of the hip from the time that this procedure was generally introduced in Iceland in 1972 up to, and including, Information on identity, sex, age at operation, diagnosis, and type of prosthesis was registered. A total of 3,887 patients who received THR surgery were identified. One of the authors (TI) reviewed all of the patients records in 3 of the 6 hospitals to ascertain a correct diagnosis. Thus, 2,500 of the 3,887 cases were verified against the original patient records. The proportion of incorrect diagnoses was 2%. Diagnoses such as fracture or rheumatoid arthritis, for example, were excluded from further analysis, and therefore a total of 2,713 cases of THR for hip OA (1,383 among women) were used for further analysis in this study. We thus developed a national registry of all THRs done in Iceland from the beginning of these operations in 1972 until Icelandic genealogy database. Investigators at decode Genetics have entered all available Icelandic genealogy records from the last 11 centuries into a computerized database. It is estimated that between 800,000 and 1,000,000 people have lived in Iceland since the original settlement (40). Approximately 600,000 of these individuals are now included in the genealogy database, including the entire current population of 275,000 and most of their ancestors back to the ninth century. Each individual is given a unique personal identifier number (PN) and, in the database, this is connected to the corresponding PNs of the father and the mother. Maternal connections in the genealogy database have been estimated to be 99.3% accurate by examining the mitochondrial DNA sequences of maternally related individuals (40). By examining the genotypes of more than 20,000 Icelanders, it was estimated that the sum of the laboratory error rate and the nonpaternity rate is 1.5%. Data encryption. All patient lists used at decode Genetics are encrypted by the Data Protection Commission of Iceland, before arriving at the laboratory (for details, see decode Genetics Web site at protection/index.htm). Encrypted versions of the databases were used to examine the familial relationships in this study. Control population. In order to assess whether the observed familial clustering of THR patients was significantly different than what could be expected within the population of Iceland, we generated 1,000 independently drawn, matched control sets using the national genealogy database. These control sets were used for the minimum founder test (MFT) and also in producing P values and 95% confidence intervals for the kinship coefficients (KC) and relative risks (RR) for THR patients. Each patient had a matched control in every control set. Thus, each control set was the same size as the patient list. The matched control for each patient was drawn so that the control had the same year of birth as the patient, which avoids biases due to different clustering patterns in different birth cohorts. We also made sure that the matched control had the same number of ancestors going back 5 generations in the national genealogy as did the patient. This ensures that

3 HIP OA IN ICELAND 2787 the control lists are as connected to the genealogy as the patient list. Identification of familial clustering of THR for OA. Familial clusters were generated by combining information in the genealogy database with the patient list of 2,713 patients with THR for OA and the control lists. Familial clusters at several different meiotic distances were thus constructed both for the THR patient list and the control lists. A cluster at a given meiotic distance, e.g., 4, was constructed by first selecting 1 individual at random from the list. All individuals who were on the list and were related to this individual at a meiotic distance of 4 or less were then added to the cluster. The process was repeated for all individuals thus added to the cluster by identifying all individuals in the list who were related to them at a meiotic distance of 4 or less, and they were added to the same cluster. When no more individuals were found in the genealogy database who connected at 4 meioses, the cluster was closed and the next cluster created by picking at random an individual who was not in the first cluster and then reiterating the process. It should be noted that clusters tend to grow laterally because of marriage, as well as vertically, and that an individual in a cluster is not necessarily related to all of the other individuals in the cluster. Cluster size is therefore not appropriate as a measure of a familial component, but still gives an indication of the degree of familial clustering. Finally, the size of the clusters obtained from the patient list were compared with the mean size of the clusters obtained from the control lists. MFT analysis. The MFT was used to test whether the individuals on the list of patients with THR for OA were more closely related than were the individuals on the control lists, which would indicate a familial component of THR for OA (40). The extensive genealogic information in the database was used to find all ancestors of the individuals on the patient and control lists. Reaching a given number of generations back, we identified the minimum set of ancestors ( founders ) such that everybody on the list was a descendant of one of these founders. The more related the people on a list were, the fewer founders were needed to account for all individuals on the list. To generate complete data for a given list of individuals, the genealogy database was used to go back to a certain year and find the minimum number of founders who were born in that year, or later, for the relevant list. This was done for a number of years, both for the patient list and for the control lists. The results are presented on a graph that shows the minimum number of founders needed to account for the patient list as a function of year of founder birth, in comparison with the mean (plus 2 SD above and below the mean) of the minimum number of founders for the control lists. Calculation of the KC. The KC is a measure of the relationship of 2 relatives. It is defined as the probability that a randomly selected allele from each of a pair of individuals is inherited from a common ancestor, i.e., that the alleles are identical by descent (IBD) (41). For any pair of relatives, the KC is approximately one-half of the expected proportion of their genome shared due to common ancestry. In the case of no consanguinity, the KC is one-fourth for first-degree relatives, one-eighth for second-degree relatives, one-sixteenth for third-degree relatives, etc. For this study, the pairwise KC distribution for the patient list was calculated by calculating the KCs of every possible pair of patients and generating the average KC for THR for OA (40). The result was compared with the average pairwise KC for the 1,000 matched control lists. Due to the large size of the pedigree sets, Monte Carlo simulations were used to estimate the average pairwise KC. For each set of individuals, the ancestors 12 generations back were identified. Each ancestor was attributed a set of unique alleles and Mendelian inheritance analysis was used to simulate the transmission of the alleles through the pedigrees. Allele sharing for every pair could thus be calculated. Determination of the RR. The RR for THR among siblings of affected individuals is equal to the risk of THR for OA in siblings divided by the risk of THR for OA in the general population. The RR can be calculated for different types of relatives, e.g., siblings, cousins, or spouses. Since THR for OA is not common among young people, we restricted our calculations to individuals born between 1900 and 1935, for calculations both of the population risk and of the risk in the relatives. Mathematically, the RR for relatives of type R ( R )is estimated by the following equation: ˆ R Aff n a Aff N a /N where N a /N is an estimate of the population risk, N a is the total number affected, and N is the total number of individuals in the population (born between 1900 and 1935). The risk in relatives of type R is estimated by Aff n a / Aff n, where Aff n a is the sum of all affected relatives of individuals on the list and Aff n is the sum of all relatives of type R individuals on the list. Note that we do not have to rely on a control group to estimate the population risk since we have a population-based list of patients with THR for OA and we also know the total number of individuals in the population born in the period of interest. The R was 1 for the control lists since they were a random sample from the general population. We used the estimates of the control s to derive the error distribution of and construct confidence intervals for which can be converted back to a confidence interval for. The square-root scale was used because it is the appropriate variance-stabilizing transform in this setting. Testing for differences by sex. To assess the presence of a possible sex difference in the inheritance of OA associated with THR, the patient list was divided according to sex. For both lists, 1,000 matched control lists were generated, matched for sex as well as for age and ancestral connections in the genealogy database. The average pairwise KC was calculated and the MFT was applied on both lists. Ethics. All studies were approved by the Bioethics Committee of the Icelandic Health Ministry and the Data Protection Commission of Iceland. RESULTS Familial clustering of THR for hip OA. At a meiotic distance of 5, 2 of every 3 Icelandic patients with THR for OA were identified as belonging to a cluster (Table 1). At this meiotic distance, 384 clusters n,

4 2788 INGVARSSON ET AL Table 1. Meiotic distance, group Familial clustering of total hip replacement for hip osteoarthritis* Total number of individuals in clusters Total number of clusters Mean size of clusters Size of largest cluster 2 Patients Controls Patients 1, Controls Patients 1, Controls 1, Patients 1, Controls 1, Patients 2, ,430 Controls 1, , Patients 2, ,954 Controls 2, , * Data on 2,713 cases of total hip replacement for osteoarthritis were examined for relationship in the national genealogy database, and the number and size of clusters was calculated at each given meiotic distance. For comparison, the cluster sizes for the 1,000 similar-sized and matched control lists were calculated. Data for controls are given as the mean SD. varying in size from 190 to 2 patients were identified. Even at a smaller meiotic distance of 2, 1 of every 3 THR cases belonged to a cluster, varying in size from 8 to 2 patients. Clusters were also created for the 1,000 control lists. The most informative number is the total number of individuals belonging to clusters. At 7 meioses, only 1 of the 1,000 control lists had more individuals in clusters than did the patient list. At 6 meioses, the corresponding number was 2 of 1,000 control lists, at 5 meioses, 1 of 1,000, and at 4 meioses, 3 of 1,000, whereas Figure 1. An example of a pedigree identified through the computerized search in the combined national total hip replacement (THR) and genealogy databases. In this pedigree of 25 individuals containing 8 THR cases at a meiotic distance of 5, patients with THR for osteoarthritis (OA) are denoted by solid symbols (square symbols, males; round symbols, females) and open symbols denote individuals without THR for OA. The diagonal line denotes individuals who are dead. at 2 meioses, none of the 1,000 control lists had more individuals in clusters compared with the patients. An example of a pedigree identified through the computerized search in the combined national THR and genealogy databases is shown in Figure 1. This pedigree contains 8 cases of THR for OA, at a meiotic distance of 5 or less. Results of the MFT. We used the MFT to examine the relationships among the OA hip replacement patients, and compared these relationships with those for the individuals in the control lists (Figure 2). The minimum number of founders needed to account for the genealogy of the individuals in each list was calculated for a number of years, going back to It should be noted that in the current generation, each individual is a founder to him- or herself, and therefore the minimum number of founders is the same as the size of the list for this generation. As the calculations go back in time, more and more connections are found among the individuals and the number of founders decreases. At the end of the nineteenth century, the curves begin to diverge, and the minimum number of founders needed for the list of patients with THR for OA were fewer at all times than the average minimum number of founders needed for the 1,000 control lists. The greatest difference was seen in the year 1860, when the number of founders for the THR patient list diverged 11.7 SD units from the mean of the control lists (P 0.001).

5 HIP OA IN ICELAND 2789 Table 2. controls* Average pairwise kinship coefficients for patients and Kinship coefficient, 10 5 Patient list Patients Controls P THR OA patients All [1st] [1st, 2nd] THR OA female patients All [1st] [1st, 2nd] THR OA male patients All [1st] [1st, 2nd] * Data on 2,713 cases of total hip replacement for osteoarthritis (THR OA) were used to calculate average pairwise kinship coefficients, and compared with average pairwise kinship coefficients calculated from 1,000 similar-sized control lists drawn from the Icelandic population and matched for year of birth, sex, and the accumulated connections 5 generations back. Data for controls are given as the mean SD. [1st] first-degree relatives subtracted; [1st, 2nd] first- and second-degree relatives subtracted. Figure 2. Results of the minimum founder test for the complete patient list of 2,713 cases of total hip replacement (THR) due to osteoarthritis in Iceland, as well as for the separate sex-stratified lists. The mean values for the 1,000 similar-sized control lists are shown along with 2 SD units (solid lines surrounded by dotted lines), while the THR patient data are denoted with a solid line and solid circles. When the MFT was applied to the separate male and female patient lists, the greatest difference in the minimum number of founders was seen in the middle of the nineteenth century for both sexes. The number of founders for male patients with THR for OA deviated 9.2 SD units from the mean of their matched control groups (P 0.001) and the value for female patients with THR for OA deviated 6.3 SD units from that of their controls (P 0.001). KC values. The average pairwise KC for the THR for OA patient list, and for the lists of female and male THR for OA patients, were calculated, as well as the average pairwise KC for the control lists (Table 2). To counteract a possible ascertainment bias, we also calculated the KC after removing first-degree relatives and after removing both first- and second-degree relatives. The average pairwise KC of the patients with a THR for OA was significantly higher than that for the controls, and the difference was maintained after removing firstand second-degree relatives. The difference was maintained for the sex-stratified lists. Findings of RR. We estimated the prevalence of THR due to OA in the Icelandic population to be 2.7% for individuals born between 1900 and For indi- Table 3. Relative risks of having total hip replacement for osteoarthritis for siblings, cousins, and spouses of patients* Relative ˆR 95% CI for ˆR ˆR 1 P Siblings Cousins Spouses * 95% CI 95% confidence interval.

6 2790 INGVARSSON ET AL viduals born between 1915 and 1920, it was even higher, at a prevalence of 4%. Siblings and cousins showed a significantly increased RR for THR for OA (Table 3). We did not observe an increased risk of THR for OA among spouses of affected individuals. DISCUSSION The availability of a genealogic database comprising the entire Icelandic population and dating back to the original settlement, combined with a national registry for hip arthroplasty, provided us with the opportunity to study the genetic contribution of hip OA associated with THR. The use of a complete populationbased patient registry eliminated possible bias in the data collection, such as oversampling of familial material as opposed to sporadic cases, and problems in proband identification. The universal access to the health care system provided in Iceland would further tend to minimize possible phenotypic differences associated with differences in socioeconomic status. Indications for THR for OA are similar in all orthopedic centers in Iceland, as is the average age of patients at the time of THR (31,38,39). The radiologic characteristics of the hip OA leading to THR studied here are similar to those reported in other investigations (34 37). A broad range of algorithms was applied to make the best use of the databases. This included algorithms that allowed us to identify, for each person on a list, all relatives of a specific type, e.g., siblings or cousins. Recursive pedigree algorithms were also used that identified, for each person on a list, all ancestors within a given number of generations back. Having identified the ancestors, we were then able to search for ancestors common to any 2 or more members of the input list and thereby create clusters of related individuals and ancestral pedigrees of individuals related to the same founder. The combined use of the databases enabled the identification of the largest familial clusters of patients with total joint replacement for hip OA described to date. The clustering of patients in pedigrees that extend beyond the nuclear family suggests that the pathogenesis of OA has a familial component. Our results extend those of previous reports that have demonstrated the influence of genetic factors in THR due to OA (7). It is possible that the large pedigrees containing multiple individuals that we have identified using this novel method may represent unusual cases of extensive familial clustering rather than the general relationship pattern among THR for OA patients. However, the significant genetic component in hip OA leading to THR was further supported by the examination of familial relationships by application of 3 different statistical methods. As a method to examine the relatedness of the entire group of patients with THR for OA, the MFT was used to estimate the fewest possible founders for a given list of patients or controls. The results showed a significantly lower number of ancestors for the patient list, as compared with the matched control lists. The largest difference between the patients and the controls was observed in the middle of the nineteenth century. Calculation of the KC was used to compare the degree of IBD sharing within the patient and control lists, and the results again showed a significantly greater familial element for patients with THR for OA than for the controls. This, however, does not rule out a possible presence of environmental components shared by the nuclear family or sampling bias by an influential family member. To analyze whether environmental factors affected familial clusters, and thereby the familial components, the KC calculations were also done with removal of all first-degree relatives, and then again after removal of all first- and second-degree relatives. Predictably, the difference in the KC between patients and controls was reduced but still significant. This supports the notion that the difference observed is due to genetic factors. Finally, a more conventional measure of relatedness was used by calculating the overall risk of having a THR operation for OA. For individuals born in Iceland between the years 1900 and 1935, the overall risk was 2.7%. Siblings of affected individuals were 3.05 times more likely to have a THR for OA than were controls. Cousins of affected individuals showed a modest, but significant, increase in the risk of THR operation, while spouses did not show a significant increase in risk. The possible contribution of family-related environmental factors was not specifically addressed in the present study. However, we found no increase in the RR for THR in spouses, contrary to that in the first- and second-degree relatives of the affected patients, pointing to a limited impact of general environmental factors contributing to hip OA among family members in Iceland. To determine genetic factors in multifactorial disease, controls used in sibling studies should ideally resemble siblings with respect to environmental risk factors, but differ from them with regard to possible genetic determinants. Furthermore, they should be representative of the general population in terms of their susceptibility to disease. Spouses of patients fulfill both these criteria by their longstanding proximity to the

7 HIP OA IN ICELAND 2791 patient, and have previously been used as controls in family studies (7). Spouses and siblings would be expected to share a common environment, and to show similar positive or negative biases toward THR for OA. In addition, any significant impact of environmental factors that could bias this study is unlikely, since we have shown that the familial component in hip OA extends for multiple generations beyond the nuclear family. Accordingly, it was argued that the chance that observed associations are related to shared environmental factors decreases for relatives beyond the nuclear family (42). Historically, Iceland has been relatively uniform with respect to socioeconomic status, with little difference between the wealthiest and poorest families, further decreasing the likelihood of bias. In conclusion, our findings significantly extend previous observations in this area and support a substantial genetic contribution to a common form of OA (43). Our study further encourages the search for the genes responsible for increased susceptibility to OA in the Icelandic population. ACKNOWLEDGMENTS We thank Thorgeir Thorgeirsson and Hjörtur H. Jónsson for reviewing the manuscript. REFERENCES 1. Stecher RM, Hersh AH. Heberden s nodes: the mechanism of inheritance in hypertrophic arthritis of the fingers. J Clin Invest 1954;23: Stecher RM. Heberden s nodes: heredity in hypertrophic arthritis of the finger joints. Am J Med Sci 1941;201: Marks JS, Stewart IM, Hardinge K. Primary osteoarthritis of the hip and Heberden s nodes. Ann Rheum Dis 1979;38: Hirsch R, Lethbridge-Cejku M, Scott WW, Reichle R, Plato CC, Tobin J, et al. Association of hand and knee osteoarthritis: evidence for a polyarticular disease subset. Ann Rheum Dis 1996;55: Kellgren JH, Lawrence JS, Bier F. Genetics factors in generalised osteoarthritis. Ann Rheum Dis 1963;22: Anderson JJ, Felson DT. Factors associated with osteoarthritis of the knee in the First National Health and Nutrition Examination Survey (NHANES I). Am J Epidemiol 1988;128: Chitnavis J, Sinsheimer JS, Clipsham K, Loughlin J, Sykes B, Burge PD, et al. Genetic influences in end-stage osteoarthritis: sibling risks of hip and knee replacement for idiopathic osteoarthritis. J Bone Joint Surg Br 1997;79: Spector TD, Cicuttini F, Baker J, Loughlin J, Hart DJ. Genetic influences on osteoarthritis: a twin study. BMJ 1996;312: Felson DT, Couropmitree NN, Chaisson CE, Hannan MT, Zhang Y, McAlindon TE, et al. Evidence for a Mendelian gene in a segregation analysis of generalized radiographic osteoarthritis: the Framingham Study. Arthritis Rheum 1998;41: Hirsch R, Lethbridge-Cejku M, Hanson R, Scott WW Jr, Reichle R, Plato CC, et al. Familial aggregation of osteoarthritis: data from the Baltimore Longitudinal Study on Aging. Arthritis Rheum 1998;41: Holderbaum D, Haqqi TM, Moskowitz RW. Genetics and osteoarthritis: exposing the iceberg. Arthritis Rheum 1999;42: Cicuttini MF, Spector DM. The genetics of osteoarthritis. J Clin Pathol 1996;49: Jimenez SA, Williams CJ, Karasick D. Hereditary osteoarthritis. In: Brandt KD, Doherty M, Lohmander LS, editors. Osteoarthritis. Oxford (UK): Oxford University Press; p Francomano CA, Liberfarb RM, Hirose T, Maumenee IH, Streeten EA, Meyers DA, et al. The Stickler syndrome: evidence for the close linkage to the structural gene for type II collagen. Genomics 1987;1: Palotie A, Vaisanen P, Ott J, Ryhanen L, Elima K, Vikkula M, et al. Predisposition to familial osteoarthritis linked to type II collagen gene. Lancet 1989;I: Ala-Kokko L, Baldwin CT, Moskowitz RW, Prockop DJ. A single base mutation in the type II procollagen gene (COL2AI) as cause of primary osteoarthritis associated with mild chondrodysplasia. Proc Natl Acad Sci U S A 1990;87: Ritvaniemi P, Körkkö J, Bonaventure J, Vikkula M, Hyland J, Paassilta P, et al. Identification of COL2A1 gene mutations in patients with chondrodysplasias and familial osteoarthritis. Arthritis Rheum 1995;38: Pun YL, Moskowitz RW, Lie S, Sundstrom WR, Block SR, McEwen C, et al. Clinical correlations of osteoarthritis associated with a single-base mutation (arginine 519 to cysteine) in type II procollagen gene: a newly defined pathogenesis. Arthritis Rheum 1994;37: Bleasel JF, Bisagnifaure A, Holderbaum D, Vacherlavenu MC, Haqqi TM, Moskowitz RW, et al. Type II procollagen gene (COL2A1) mutation in exon 11 associated with spondyloepiphyseal dysplasia; tall stature and precocious osteoarthritis. J Rheumatol 1995;22: Bleasel JF, Holderbaum D, Mallock V, Haqqi TM, Williams HJ, Moskowitz RW. Hereditary osteoarthritis with mild spondyloepiphyseal dysplasia are there hot spots on COL2A1? J Rheumatol 1996;23: Bleasel JF, Holderbaum D, Brancolini V, Moskowitz RW, Considine EL, Prockop DJ, et al. Five families with arginine 519- cysteine mutation in COL2A1: evidence for three distinct founders. Hum Mutat 1998;12: Briggs MD, Hoffman SMG, King LM, Olsen AS, Mohrenweiser H, Leroy JG, et al. Pseudoachondroplasia and multiple epiphyseal dysplasia due to mutations in the cartilage oligomeric matrix protein gene. Nat Genet 1995;10: Meulenbelt I, Bijkerk C, de Wildt SCM, Miedema HS, Valkenburg HA, Breedveld FC, et al. Investigation of the association of the CRTM and CRTL1 genes with radiographically evident osteoarthritis in subjects from the Rotterdam study. Arthritis Rheum 1997;40: Meulenbelt I, Bijkerk C, Miedema HS, Breedveld FC, Hofman A, Valkenburg HA, et al. A genetic association study of the IGF-1 gene and radiological osteoarthritis in a population-based cohort study (the Rotterdam study). Ann Rheum Dis 1998;57: Horton WE Jr, Lethbridge-Cejku M, Hochberg MC, Balakir R, Precht P, Plato CC, et al. An association between an aggrecan polymorphic allele and bilateral hand osteoarthritis in elderly white men: data from the Baltimore Longitudinal Study of Aging (BLSA). Osteoarthritis Cartilage 1998;6: Loughlin J, Irven C, Ferguson C, Sykes B. Sibling pair analysis shows no linkage of generalized osteoarthritis to the loci encoding type II collagen, cartilage link protein or cartilage matrix protein. Br J Rheumatol 1994;33: Chapman K, Mustafa Z, Irven C, Carr AJ, Clipsham K, Smith A, et al. Osteoarthritis-susceptibility locus on chromosome 11 q, detected by linkage. Am J Hum Genet 1999;65:

8 2792 INGVARSSON ET AL 28. Leppävuori J, Kujala U, Kinnunen J, Kaprio J, Nissilä M, Heliövaara M, et al. Genome scan for predisposing loci for distal interphalangeal joint osteoarthritis: evidence for a locus on 2q. Am J Hum Genet 1999;65: Loughlin J, Mustafa Z, Irven C, Smith A, Carr AJ, Sykes B, et al. Stratification analysis of an osteoarthritis genome screensuggestive linkage to chromosomes 4, 6, and 16. Am J Hum Genet 1999;65: Ingvarsson T, Hägglund G, Lohmander LS. Prevalence of hip osteoarthritis in Iceland. Ann Rheum Dis 1999;58: Ingvarsson T, Hägglund G, Jónsson H Jr, Lohmander LS. Incidence of total hip replacement for primary osteoarthrosis in Iceland Acta Orthop Scand 1999;70: Ingvarsson T. Seventeen siblings with coxarthrosis [abstract]. J Bone Joint Surg Br 1991;73-B Suppl 2: Ingvarsson T, Lohmander LS. An Icelandic family with hip osteoarthrosis [abstract]. Ann Rheum Dis 1996;55: Ingvarsson T, Hägglund G, Lindberg H, Lohmander LS. Assessment of primary hip osteoarthritis: a comparison of two radiographic methods using colon radiographs. Ann Rheum Dis 2000; 59: Ledingham J, Dawson S, Preston B, Milligan G, Doherty M. Radiographic patterns and associations of osteoarthritis of the hip. Ann Rheum Dis 1992;51: Danielson L, Lindberg H, Nilsson B. Prevalence of coxarthrosis. Clin Orthop 1984;191: Forsberg K, Nilsson B. Coxarthrosis on the island of Gotland: increased prevalence in a rural population. Acta Orthop Scand 1992;63: Herberts P, Malchau H. Long-term registration has improved the quality of hip replacement. Acta Orthop Scand 2000;71: Havelin LI, Espehaug B, Vollset SE, Engesaeter LB, Langeland N. The Norwegian arthroplasty register: a survey of 17,444 hip replacements Acta Orthop Scand 1993;64: Gudmundsson H, Gudjartsson DF, Kong A, Gudbjartsson H, Frigge M, Gulcher JR, et al. The inheritance of human longevity in Iceland. Eur J Hum Genet 2000;8: Lange K. Statistics for biology and health: mathematical and statistical methods for genetics analysis. New York: Springer; Sadovnick AD, Macleod PM. The familial nature of multiple sclerosis: empiric recurrence risks for first-, second-, and thirddegree relatives of patients. Neurology 1981;8: Lindberg H. Prevalence of primary coxarthrosis in siblings of patients with primary coxarthrosis. Clin Orthop 1986;203:273 5.