Inter-examiner reproducibility of tests and criteria for generalized joint hypermobility and benign joint hypermobility syndrome

Transcription

1 Rheumatology 2007;46: doi: /rheumatology/kem290 Inter-examiner reproducibility of tests and criteria for generalized joint hypermobility and benign joint hypermobility syndrome B. Juul-Kristensen 1, H. Røgind 2, D. V. Jensen 3 and L. Remvig 1 Objective. To test the reproducibility of tests and criteria for generalized joint hypermobility (GJH) and benign joint hypermobility syndrome (BJHS). Methods. A standardized protocol for clinical reproducibility studies was followed using a three-phase study (with a training, an overall agreement and a test phase). An overall agreement of at least 0.80 was required to proceed to the test phase. Phases 1, 2 and 3 used 14 patients (with varying degrees of hypermobility), 20 patients (50% cases) and 40 patients (50% cases), respectively. The inclusion criterion for cases was hypermobility (patients with Ehlers-Danlos Syndrome or BJHS) and for controls, non-hypermobility (patients with shoulder and/or back pain); patients were selected from patients files (phases 1 and 2) or included consecutively from our outpatient clinic (phase 3). Results. The overall agreement in phase 2 was 0.95 for GJH and 0.90 for BJHS. Reproducibility for diagnosing GJH and BJHS in phase 3 showed kappa values of 0.74 and 0.84, respectively. Kappa in the Beighton tests for diagnosing GJH (currently or historically) was generally above 0.80, except for the fifth fingers and elbows (0.60). In the Brighton tests for diagnosing BJHS, kappa was above 0.73, except for the skin signs (0.63). Lowest kappa was found in the Rotès-Quérol tests, where it was 0.57, except for the right shoulder (0.31). Conclusion. We found a good-to-excellent reproducibility of tests and criteria for GJH and BJHS. Future research on the validity of the tests and criteria for joint hypermobility is urgently needed. KEY WORDS: Hypermobility, Beighton tests, Brighton criteria, Rotès-Quérol tests, Standardized protocol. Introduction Joint hypermobility (JH), i.e. increased joint range of motion (ROM), is seen either as a localized condition (LJH) e.g. in a single joint or a more generalized (GJH) one. When seen in parallel with musculoskeletal problems, the total condition is recognized as pathological, called hypermobility syndrome (HMS). A dichotomous principle for judgement of JH was introduced more than 40 yrs ago. Instead of diagnosing hypermobility through ROM in degrees, several authors used fulfilment of a certain extreme ROM as criterion for hypermobility [1 3]. The tests used were revised [4 6], and currently the Beighton tests have gained international acceptance and seem to be the most widely used tests today for diagnosing general joint hypermobility (GJH). Curiously, the Beighton tests have only been described by five photos and a legend [6]. In the Spanish-speaking countries though, the Rote` s-que rol tests have mostly been used [3, 4], supplementing the Beighton tests with tests for the shoulder, the cervical spine, the hip and the toe. Two studies have reported data on inter-examiner reproducibility of hypermobility tests. In one study, kappa for the different Beighton tests varied between 0.79 and 0.93 and for the additional Rote` s-que rol tests between 0.44 and 1.00 [7]. In the other study, the same Beighton tests, except for one (fifth finger), showed kappa between 0.44 and 1.00 [8]. No studies have reported data on intra-examiner reproducibility of these hypermobility tests. The term GJH indicates that ROM in general is increased compared with the mean ROM of the general population. GJH is 1 Department of Medical Orthopaedics and Rehabilitation, Rigshospitalet, University Hospital, Copenhagen, 2 Department of Rheumatology, Frederiksberg Hospital and 3 Department of Rheumatology, Hospital of Nordsjælland, Hørsholm, Denmark. Submitted 15 April 2007; revised version accepted 19 September Correspondence to: B. Juul-Kristensen, Department of Medical Orthopaedics and Rehabilitation, Rigshospitalet, University Hospital of Copenhagen, Blegdamsvej 9, Dk-2100 Copenhagen Ø, Denmark. birgit.juul.kristensen@rh.regionh.dk; bjkprivat@mail.dk claimed to be present in 5 15% of the general population, to be more frequent among children than among adults and to be slightly more frequent among girls/women than among boys/men [9]. However, there is no universally accepted criterion for GJH, and, except for one study that has tested the Beighton tests towards a global index [10], neither the Beighton tests nor the criterion for GJH have been tested for validity [11]. Several different cut points for GJH have been used, e.g. a Beighton score of 6 positive tests out of 9 [12], 5 positive tests [13] and 4 positive tests [14]. Nevertheless, the diagnostic reproducibility has only been reported when using 6 positive tests for GJH with the result: good to excellent [12]. HMS was first described by Kirk et al. [2] as a pathological condition characterized by JH and musculoskeletal complaints. HMS was later renamed benign joint hypermobility syndrome (BJHS), and a set of tests, signs and symptoms the Brighton criteria was suggested and later revised [14]. These criteria include both subjective and objective findings, of which some are also included in the set of criteria for the more serious diseases with hypermobility, such as Ehlers-Danlos syndrome (EDS), Marfan syndrome (MF) and Osteogenesis Imperfecta (OI). The criteria for BJHS include two major and/or eight minor criteria (Appendix 1). The two major criteria, a Beighton score 4 positive tests out of 9 (either currently or historically) and arthralgia in 4 joints for more than 3 months, are more or less given a priori. However, a scientific basis for the minor criteria has only vaguely been described [14, 15]. Neither data on reproducibility nor on the validity of the Brighton criteria for BJHS have been reported. In one of the aforementioned studies, the authors were aware of increasing the overall agreement through a training phase, but the overall agreement was not reported [7]. Neither study [7, 8] took the prevalence of the condition in the sample population into consideration before calculating kappa. The aim of this study was to analyse the inter-examiner reproducibility of first, the criteria for the diagnoses GJH and BJHS; second, the Beighton tests; and third, supplementing Rote` s-que rol tests [4] and the tests for BJHS (the Brighton 1835 ß The Author Published by Oxford University Press on behalf of the British Society for Rheumatology. All rights reserved. For Permissions, please journals.permissions@oxfordjournals.org

2 1836 B. Juul-Kristensen et al. TABLE 1. Kappa values (99% CI) presenting current, and current or historical scores of the Beighton tests: for diagnosing general joint hypermobility (GJH) (phase 3) Name of the Beighton tests and criteria (n ¼ 40) Kappa, currently (99% CI) Kappa, currently or historically (99% CI) Prevalence, currently (%) Prevalence, currently or historically (%) B1-first finger, right 1.00 (1.00, 1.00) 1.00 (1.00, 1.00) B1-first finger, left 0.94 (0.79, 1.09) 1.00 (1.00, 1.00) B2-fifth finger, right 0.61 (0.15, 1.07) 0.60 (0.21, 0.99) B2-fifth finger, left 0.40 a ( 0.01, 0.81) 0.71 (0.40, 1.02) B3-elbow, right 0.64 (0.28, 1.00) 0.62 (0.28, 0.96) B3-elbow, left 0.34 a ( 0.10, 0.78) 0.63 (0.32, 0.94) B4-knee, right 0.90 (0.72, 1.08) 0.90 (0.72, 1.08) B4-knee, left 0.85 (0.64, 1.06) 0.85 (0.64, 1.06) B5-forward bending 0.88 (0.67, 1.09) 0.81 (0.53, 1.09) Beighton score (0.52, 1.04) 0.85 (0.64, 1.06) GJH diagnosis (Beighton score 5) 0.66 (0.30, 1.02) 0.74 (0.46, 1.02) a Values marked indicate a poor kappa. TABLE 2. Kappa values (99% CI) presenting current, and current or historical scores of the tests used in the Brighton criteria for diagnosing Benign Joint Hypermobility Syndrome (BJHS) (phase 3) Name of the Brighton tests and criteria (n ¼ 40) Kappa, currently (99% CI) Kappa, currently or historically (99% CI) Prevalence, currently (%) Prevalence, currently or historically (%) Skin signs (Minor 1) BR-minor 1.1 striae distensae 0.56 (0.22, 0.90) 50 BR-minor 1.2 skin hyperextensibility 0.55 (0.16, 0.94) 28 BR-minor 1.3 thin skin 5 BR-minor 1.4 papyraceous scarring 0.63 (0.27, 0.99) 28 Minor 1 skin signs 0.63 (0.32, 0.94) 63 Eye signs (Minor 2) BR-minor 2.1 drooping eyelids (bilateral ptosis) 3 BR-minor 2.2 myopia 0.84 (0.56, 1.12) 20 BR-minor 2.3 anti-mongoloid slant 0 Minor 2 eye signs 1.00 (1.00, 1.00) 23 Marfanoid habitus signs (Minor 3) BR-minor 3.1 arm span/height ratio > BR-minor 3.2 upper/lower segment ratio < BR-minor 3.3 arachnodactyly (Steinberg sign) 0.83 (0.52, 1.14) 18 BR-minor 3.4 arachnodactyly (wrist sign) 0.63 (0.14, 1.12) 13 Minor 3 marfanoid habitus 0.84 (0.56, 1.12) 20 Beighton score (Minor 4) Beighton score 1 3, (>50 yrs 0 3) 0.85 (0.64, 1.06) 45 Arthralgia (Minor 5) 1 3 joints 0.73 (0.45, 1.01) 38 Dislocation/subluxation (Minor 6) 0.77 (0.49, 1.05) 33 Soft tissue rheumatism (Minor 7) Epicondylitis, tenosynovitis or bursitis 0.83 (0.60, 1.06) 35 Other signs (Minor 8) Varicose veins, hernia, uterine/rectal prolapse 0.78 (0.47, 1.09) 23 Family hypermobility (presence of a first degree relative) 0.89 (0.68, 1.10) 35 Beighton scores 4 (Major 1) 0.78 (0.52, 1.04) 0.85 (0.64, 1.06) Arthralgia 4 joints (Major 2) 0.95 (0.82, 1.08) 43 Major 1 þ (0.61, 1.07) 40 BJHS diagnosis 0.84 (0.61, 1.07) 63 Kappa could not be calculated (almost total agreement, but at the same time a very large probability by chance) since SPSS could not calculate asymmetric cross tabulations. criteria). In addition, the aim was to analyse the agreement or the mutual dependencies between the different Brighton tests and criteria and the final diagnosis of BJHS. Methods Study design and tests In order to optimize the conditions for analysing reproducibility of the tests and diagnoses, a three-phase study was performed as described in a standardized protocol for clinical reproducibility and validity studies [16]. In all the three phases, the included tests were the Beighton tests [6] and the Brighton criteria for BJHS [14] (Tables 1 and 2 and Appendix 1). Performance of the Beighton tests was measured according to photographs [6] and descriptions [14]. Each Beighton test started with a performance of the test by the examiner along with a question. Two of the tests (the first and fifth finger tests) were performed passively by the patients themselves and without muscle contractions. The three other tests (knee, elbow and forward bent back) were performed in a position where the passive structures were challenged mostly (i.e. in extreme positions). Three Rote` s-qu erol tests for hypermobility were included, representing the hip, the shoulder and the cervical spine [4], since these joints are not represented in the Beighton tests (Table 3 and Appendix 1). A standardized clinical protocol was developed (Appendix 1), including the tests for current and historical information. Only if a test was negative currently was the patient asked whether it had been possible to perform a positive test previously (historical information). Each of the positive Beighton tests counted 1 point, giving a maximum of 9 points. Since there is no universally accepted cut-off level for the diagnosis of GJH, we defined a Beighton score of 5 positive tests out of 9 to be the criterion for GJH. Initially, two examiners were trained in order to eliminate the subjective part of the test evaluation as much as possible through a detailed description of the tests, including cut-off points for judging a test as positive or negative the training phase (phase 1). Second, an agreement phase (phase 2) was undertaken,

3 Reproducibility of tests and criteria for GJH and BJHS 1837 TABLE 3. Kappa values (99% CI) presenting current, and current or historical scores of the Rotès-Quérol tests (phase 3) Name of the Rotès-Quérol tests (n ¼ 40) Kappa, currently (99% CI) Kappa, currently or historically (99% CI) Prevalence, currently (%) Prevalence, currently or historically (%) R6-cervical column rotation right 0.35 a ( 0.06 to 0.76) 0.57 (0.23 to 0.91) R6-cervical column rotation left 0.49,a (0.08 to 0.90) 0.57 (0.23 to 0.91) R7-shoulder external rotation right 0.53 a ( 0.09 to 1.15) 0.31 a ( 0.23 to 0.85) R7-shoulder external rotation left 0.79 (0.25 to 1.33) 0.63 (0.11 to 1.15) 8 13 R8-hip abduction right 0.32 a ( 0.07 to 0.71) 0.80 (0.57 to 1.03) R8-hip abduction left 0.39 a ( 0.05 to 0.83) 0.80 (0.54 to 1.06) Indicates that there was a significant difference between the two examiners regarding frequency of positive tests. a Values marked indicate a poor kappa. and for dichotomous data, an observed agreement of 80% for both GJH and BJHS was decided in advance as a requirement for proceeding to the real test phase (phase 3). If the overall agreement was <80%, phases 1 and 2 had to be repeated until the agreement was at least 80%. This would in turn include a new training phase and a revised test description. Inter-examiner reproducibility with nominal data is usually calculated by Cohen s-, which is dependent upon the overall agreement and upon the prevalence [17, 18]. Thus, in phases 2 and 3, the prevalence of the positive tests was designed to be close to 50% to facilitate the interpretation of results by maximizing kappa. Further, to avoid systematic bias, the sequence of the tests was performed in a random order. Patients Patients were aged between 18 and 71 yrs. Cases and controls for phases 1 and 2 were randomly selected from files of patients previously referred to the out-patient clinic of the Department of Medical Orthopaedics and Rehabilitation between 2000 and 2006, and were seen a minimum of 6 months prior to the experiment. Cases and controls in phase 3 were new patients referred to our out-patient clinic and consecutively included in the study. No matching criteria were applied. Phase 1 included 14 cases with the inclusion criterion of 1 9 positive Beighton tests, and phase 2 included 10 cases with the diagnoses BJHS (ICD-10: DM 35.7) or EDS (ICD-10: DQ79.6) and 10 controls consisting of non-hypermobile patients with back or shoulder pain. Phase 3 consisted of 20 patients referred for BJHS or EDS and 20 controls referred for back or shoulder pain. An arbitrary number of 40 patients was selected as a statistical minimum to perform simple reproducibility studies and as a practical choice to make the study easy and cheap to perform by not being too large [16]. The Committee on Biomedical Research Ethics for Copenhagen and Frederiksberg, Denmark, approved the experimental protocol (jnr. KF ). Each patient gave written informed consent according to the Declaration of Helsinki [19]. Procedures All tests were performed in a random order with respect to right and left side and to the test sequence. In phase 1, the examiners (A and B) were unblinded, and the test results were discussed until consensus was reached, and the test description was revised if necessary. In phases 2 and 3, each examiner was blinded with respect to the test results by the other examiner and the diagnosis. Half of the patients (cases as well as controls) were examined first by examiner A and the other half by examiner B. Two observers (C and D) checked the blinding of the examiners, the performance of the tests in a randomized order and in accordance to the test description. If a current test could not be performed (e.g. because of pain) and if a patient could not remember whether a specific condition had been present historically, the test results were coded as a negative answer (0). Data analysis and statistics For the dichotomous data, Cohen s, statistic [presented with 99% confidence intervals (CI)] was used for each of the single tests and criteria for GJH and BJHS [18] and also for the mutual dependencies between the Brighton tests and signs and the final diagnosis of BJHS [16]. For the continuous data, intra-class correlation coefficient [(ICC) model 2.1, two-way random effects model] [20, 21] was calculated for the total scores and signs. McNemar s test was used to test for significant differences between the two examiners regarding frequency of positive tests and Mann Whitney s U-test for differences between cases and controls with respect to age, with P < 0.05 as the level of significance. The statistical program used was SPSS (SPSS Inc., Chicago, IL, USA, version 13.00). Results In phase 1, tests for the knees, the elbows, the cervical spine, the hips and the shoulders needed most training and discussions, including the questions on historical information. In phase 2, the overall agreement for GJH was 0.95 (19 out of 20 patients were classified equally, 7 as positive), and for BJHS it was 0.90 (18 out of 20 patients were classified equally, 11 as positive). In phase 3, cases were significantly younger than controls (34.2 yrs, S.D vs 46.4 yrs, S.D. 17.0, P ¼ 0.02) (Table 4). Using McNemar s test, there was no significant difference between the two examiners, except for left cervical rotation in the Rote` s- Que rol test (currently) (Table 3). Kappa values of the Beighton tests (currently) were between 0.61 and 1.00, except for the left fifth finger and left elbow, where it was 0.40 and 0.34, respectively (Table 1). When including historical data, meaning either a current or historical prevalence, kappa increased to 0.71 and 0.63 for those two tests. For the total Beighton score, either currently or historically, ICC was When calculating kappa for the diagnosis of GJH, it was 0.74, with an overall agreement of 0.88 (35 patients were classified equally, 14 as positive). Kappa values for Rote` s-que rol tests were generally lower, between 0.31 and 0.80, currently as well as either currently or historically (Table 3), and lowest kappa values were right cervical rotation, right shoulder rotation and right hip abduction. ICC for the Rote` s-que rol tests was Kappa for the various Brighton tests were above 0.73 (Table 2), except for abnormal skin signs and marfanoid habitus signs (between 0.55 and 0.66), while ICC for the total number of minor criteria was For the diagnosis of BJHS, kappa was 0.84, with an overall agreement of 0.93 (37 patients were classified equally, 23 as positive). By means of the -statistics, the mutual dependencies between the single Brighton tests and signs and the final BJHS diagnosis were calculated, and thus the frequency by which each test and sign was used for diagnosing BJHS was illustrated for each examiner compared with a random frequency. In Table 5, kappa values indicated that examiners A and B particularly used major criteria 1 and 2, besides the minor criteria 1, 6, 7 and the presence

4 1838 B. Juul-Kristensen et al. of a first degree relative to make the final BJHS diagnosis. Furthermore, the examiners agreed to diagnose BJHS mainly by the combinations of two major or four minor criteria. Discussion In spite of a different cut-off level used for the diagnosis of GJH, the present reproducibility of the criterion ( ¼ 0.74) was at the same level as previously reported (0.78) [12]. For the diagnosis of BJHS, kappa was 0.84 that is usually accepted as a goodto-excellent level of reproducibility. The present reproducibility of the Beighton tests kappa values mostly above 0.80 was at the same level as found in previous studies [7, 8], except for the fifth fingers and elbows (0.60), and also the present higher kappa for the Beighton than the Rote` s-quérol tests is in agreement with previous results [7]. For most of the tests used in the Brighton criteria, kappa was well above 0.73, except for the abnormal skin signs (0.63). There is no other reproducibility study to compare the present results on the Brighton criteria with. All ICC values (the number of Beighton tests, the Rote` s-quérol tests and the number of Brighton criteria) were good to excellent ( ). In none of the two basic test descriptions for the Beighton tests [6, 14] is it stated precisely whether the tests should be performed actively or passively, and naturally this may be a potential source of discrepancy in the interpretation of results from different studies. With this strict and standardized protocol of training and testing, generalizability to the clinical practice may be low, unless the health personnel is relevantly trained. It must be noted that generalizability to, for example, other professional groups is not only depending on the examiners clinical experience in joint mobility measurements, etc. (here it was many years), but also on the extent that the examiners are able to follow the strict test procedures according to a strict and standardized protocol. In one of the two only reproducibility studies [7], kappa was between 0.79 and 0.93 for the Beighton tests and between 0.44 and 1.00 for the Rotès-Quérol tests. In that study, the prevalence TABLE 4. Demographic data for patients (cases and controls) in phases 2 and 3 Phase 2 (n ¼ 20) Phase 3 (n ¼ 40) Females, cases/controls (n) 8/6 17/10 Males, cases/controls (n) 4/2 3/10 Age (yrs) 46.2 (S.D. 3.8, range 21 70) 40.3 (S.D. 16.7, range 18 71) of cases (a score of 5) was 114 out of 173 (66%). In contrast, the second reproducibility study [8] reported lower kappa values on four of the Beighton tests (between 0.44 and 0.82 for two experienced rheumatologists) and even lower kappa when comparing experienced rheumatologists with inexperienced rheumatologists or with the parents of the tested children (kappa between <0.40 and 1.00). In that study, the population was a random selection of 100 children, and the prevalence of positive tests varied between 37 and 50% when tested by the parents and between 13 and 88% when tested by experienced rheumatologists. Taken together, both studies had a prevalence mostly well above or below 50%, influencing kappa negatively [17, 18]. The missing training phase in one of the studies [8], in addition to the missing overall agreement phase in both studies, may have further influenced kappa negatively. A missing randomization of the test sequence in one of the studies [7] may result in a bias towards a higher kappa. In both the studies, the examiners were blinded, but in the present study, where also they were blinded, we cannot exclude that the results were influenced by the number of older participants among cases in phase 3. However, the difference between the groups was only 12 yrs as a mean, and patients ages are not always easy to judge by sight; moreover the number of tests was large (n ¼ 46), and hence this factor is less likely to influence the present results. Some of the reasons for the present low kappa values with respect to current Beighton (left fifth finger and left elbow) and Rote` s-que rol tests (right cervical rotation and right hip abduction) could also be the low prevalence of positive tests (between 18 and 20% prevalence) as seen in previous studies. Including historical information, the prevalence increased remarkably along with the kappa. This is illustrated clearly in the left elbow and the left fifth finger, where the overall agreement increased by 6 and 10% and the prevalence increased by 65% in both the tests. At the same time, the probability by chance only decreased by 16 and 20%. For another of the Rote` s-que rol tests, though (right shoulder external rotation), a low current prevalence was also found. Here, the prevalence only increased a little by including historical presence (from 10 to 13%), while the observed agreement and the probability by chance decreased (from 93 to 85% and from 85 to 77%), and this resulted in a considerably decreasing kappa (from 0.53 to 0.31, equal to a 42% decrease). For all the Rote` s-que rol tests, the generally low kappa values may also be explained by a missing requirement of an overall agreement of at least 80% in phase 2. Although the overall TABLE 5. The mutual dependencies between the major/minor criteria and the final BJHS diagnosis and between the four combinations of diagnostic criteria for BJHS (phase 3) and the final BJHS diagnosis, judged by -statistics (99% CI) for each of the examiners Brighton tests and criteria (n ¼ 40) Examiner A Kappa, currently or historically (99%CI) Examiner B Kappa, currently or historically (99%CI) Major criteria 1. Beighton score (0.26, 0.94) (3) 0.69 (0.38, 1.00) (1) 2. Arthralgia 4 joints for >3 months 0.61 (0.33, 0.89) (2) 0.62 (0.34, 0.90) (2) Minor criteria 1. Skin signs 0.62 (0.28, 0.96) (1) 0.43 (0.07, 0.79) 2. Eye signs 0.30 (0.04, 0.56) 0.32 (0.06, 0.58) 3. Marfanoid habitus 0.23 (0.00, 0.46) 0.32 (0.06, 0.58) 4. Beighton score < ( 0.92, 0.24) 0.64 ( 0.98, 0.30) 5. Arthralgia <4 joints 0.26 ( 0.62, 0.10) 0.29 ( 0.65, 0.07) 6. Dislocation/subluxation 0.49 (0.21, 0.77) (4) 0.35 (0.04, 0.66) 7. Soft tissue 0.44 (0.13, 0.75) 0.44 (0.16, 0.72) (4) 8. Varicose veins, hernia, prolapse 0.07 ( 0.24, 0.38) 0.07 ( 0.19, 0.33) Presence of a first degree relative 0.45 (0.17, 0.73) 0.57 (0.29, 0.85) (3) Combinations of diagnostic criteria 1. Two major 0.53 (0.25, 0.81) (1) 0.62 (0.34, 0.90) (1) 2. One major þ two minor 0.19 ( 0.02, 0.40) 0.17 ( 0.04, 0.38) 3. Four minor 0.49 (0.21, 0.77) (2) 0.36 (0.08, 0.64) (2) 4. Two minor þ one first degree relative 0.16 ( 0.02, 0.34) 0.25 (0.02, 0.48) Numbers in italics are descending rating scores.

5 Reproducibility of tests and criteria for GJH and BJHS 1839 agreement was mostly high (between 0.70 and 0.90), there was actually a significant difference between the examiners in one of the tests (left cervical rotation test), indicating a difficulty in performing, judging and interpreting the patients historical information. This may indicate some unrecognized difficulties with these tests. Generally, the Rote` s-quérol tests were also the tests with the lowest kappa in one of the previous reproducibility studies [7], although they were higher than in the present study (with higher prevalences), but they also had no randomization of the test sequence. In future studies, these tests need to be standardized further or be substituted by other tests for hypermobility, e.g. other joints or planes of movement. A low prevalence of skin hyperextensibility was also seen (28%), which may very well be one of the reasons for this relatively low kappa (0.55). One may argue that the cut point, 3 cm for skin hyperextensibility, was too high as an arbitrary level. However, neither the Villefranche criteria [13] nor the Brighton criteria [14] give any level for abnormal skin extensibility. A level of >4 cm has been used as a pathological level for the EDS diagnosis [22], and as BJHS is considered a less serious condition than EDS, we selected the level 3 cm as the pathological level for BJHS. Generally, it may be argued that missing data were reduced by coding non-performance of a test (because of pain e.g.) and nonremembering of a historical presence as negative answers. Patients could not perform the movement due to pain in only 1.5% of the tests; thus, the influence on kappa is small. Patients could not remember historical presence in 28% of the questions, which is a relatively large percentage (15% corresponding to the Rote` s- Que rol tests, 2% for family hypermobility and 11% for the Beighton scores). Unfortunately, we are not able to give a clear estimate of this influence on kappa, since there was no clear pattern of how it affected the observed probability and the probability by chance. Since a clinician normally will interpret an unclear historical answer as a no, we found it relevant to interpret the answers in the same way. However, this high percentage naturally questions the use of this parameter as a valid criterion. In addition to reproducibility, the criterion validity of the tests and criteria must be known. Although the present study was not designed as a validity study, we analysed the influence of further requirements for diagnosing GJH on the number of patients with GJH, besides the mutual dependencies between the major respective minor criteria and the final BJHS diagnosis. The number of patients diagnosed with GJH, which is claimed to be a general condition and consequently should require positive tests for both upper and lower parts of the body, was reduced in the present study by 2 out of 14 patients (14%) with these requirements. If, additionally, a positive back test was required, 1 out of the 14 patients (7%) was excluded from the diagnosis of GJH. Whether a larger percentage of patients would be excluded when using these additional criteria in a larger population is not known. The criteria for BJHS were primarily proposed for research protocols [23], but were also recommended as a useful clinically diagnostic tool, although no studies support the validity of these tests and criteria. In the present study, some of the tests were not positive at all and thus infrequently used for the diagnosis of BJHS (i.e. bilateral ptosis, anti-mongoloid slant and some of the anthropometric dimensions for marfanoid habitus, Beighton score <4, arthralgia in less than four joints and varicose veins), while others (e.g. myopia) appeared frequently in both cases and controls (Table 5). In total, 14 patients (35%) had a positive family history, but of those only six (15%) patients had two minor criteria in addition, and were given the diagnosis of BJHS. Kappa was only 0.16 and 0.25 (Table 5), showing that the examiners in diagnosing BJHS rarely used this criterion. Neither did the exclusion of all the minor criteria seem to affect the reproducibility of diagnosing BJHS in the present study (also a kappa value of 0.84). The true relevance of these minor criteria must therefore be tested in a large cohort of hypermobile patients compared with a sex- and age-matched control cohort. Further, whether the criteria for BJHS and GJH are valid in predicting, e.g. osteoarthritis and/or other musculoskeletal diseases, also remains to be tested longitudinally. Conclusions Following a strict and standardized protocol for clinical reproducibility studies, the reproducibility of diagnosing GJH and BJHS was found to be high, with kappa values of 0.74 and 0.84, respectively. Additionally, the Beighton tests for GJH (either currently or historically) and the Brighton criteria for BJHS showed a good reproducibility, with kappa values mostly above 0.80 and 0.73, respectively. However, reproducibility of the Rote` s- Que rol tests (currently or historically) for hypermobility showed considerably lower kappa values, mostly between 0.31 and 0.80, and consequently these tests need to be refined in the future. Further research on the validity of tests and criteria for GJH and BJHS is urgently needed. Acknowledgements The authors would like to thank their technical assistant Marianne Ibsen and medical student Marie Wetterslev for their skilful assistance. Consultant Richard Ellis, MD, Southhampton, UK, has kindly revised the English language. Funding: The study was supported by the National Research Fund for Health and Disease, the Research Foundation of the Danish Physiotherapy Association, The National Health Service Research Fund for Promotion of Research in Physiotherapy, Britta Holles Foundation and the sawmill owner Jeppe Juhl and his wife Ovita Juhl s Foundation. Disclosure statement: The authors have declared no conflicts of interest. References Rheumatology key messages Standardized protocols for clinical reproducibility studies are required. Reproducibility for diagnosing GJH and BJHS was good to excellent. Validity of these tests and criteria is lacking. 1 Carter C, Wilkinson J. Persistent joint laxity and congenital dislocation of the hip. J Bone Joint Surg 1964;46B: Kirk JA, Ansell BM, Bywaters EG. The hypermobility syndrome. Musculoskeletal complaints associated with generalized joint hypermobility. Ann Rheum Dis 1967;26: Rotès-Quérol J. [Articular laxity considered as factor of changes of the locomotor apparatus.] Rev Rhum Mal Osteoartic 1957;24: Rotès-Quérol J, Duran J, Subiros R, Pifferer J, Gomez J. La laxité articulaire comme facteur d alterations de l appareil locomoteur (Nouvelle étude 1971). Rhumatologie Mai 1972;24: Beighton P. Articular manifestations of the Ehlers-Danlos syndrome. Semin Arthritis Rheum 1972;1: Beighton P, Solomon L, Soskolne CL. Articular mobility in an African population. Ann Rheum Dis 1973;32: Bulbena A, Duro JC, Porta M, Faus S, Vallescar R, Martin-Santos R. Clinical assessment of hypermobility of joints: assembling criteria. J Rheumatol 1992;19: Hansen A, Damsgaard R, Kristensen JH, Baggers J, Remvig L. Interexaminer reliability of selected tests for hypermobility. J Orth Med 2002;25: Remvig L, Jensen DV, Ward RC. Epidemiology of general joint hypermobility and basis for the proposed criteria for benign joint hypermobility syndrome: review of the literature. J Rheumatol 2007;34: Bird HA, Brodie DA, Wright V. Quantification of joint laxity. Rheumatol Rehabil 1979;18:161 6.

6 1840 B. Juul-Kristensen et al. 11 Remvig L, Jensen DV, Ward RC. Are diagnostic criteria for general joint hypermobility and benign joint hypermobility syndrome based on reproducible and valid tests? A review of the literature. J Rheumatol 2007;34: Mikkelsson M, Salminen JJ, Kautiainen H. Joint hypermobility is not a contributing factor to musculoskeletal pain in pre-adolescents. J Rheumatol 1996;23: Beighton P, De Paepe A, Steinmann B, Tsipouras P, Wenstrup RJ. Ehlers-Danlos syndromes: revised nosology, Villefranche, Ehlers-Danlos National Foundation (USA) and Ehlers-Danlos Support Group (UK). Am J Med Genet 1998;77: Grahame R, Bird HA, Child A. The revised (Brighton 1998) criteria for the diagnosis of benign joint hypermobility syndrome (BJHS). J Rheumatol 2000;27: Bird HA. Special interest group for joint hypermobility. Br J Rheumatol 1993;32: FIMM Academy. Reproducibility and validity, protocol formats for diagnostic procedures in manual/musculoskeletal medicine. Maastricht, The Netherlands: FIMM Academy for Manual/Musculoskeletal Medicine, Fleiss JL. Reliability of measurement. The design and analysis of clinical experiments. New York: John Wiley and Sons, 1986; Altman DG. Inter-rater agreement. In: Altman DG, ed. Practical statistics for medical research. London: Chapman and Hall, 1996; Vollmann J, Winau R. Informed consent in human experimentation before the Nuremberg code. Br Med J 1996;313: Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull 1979;86: Denegar CR, Ball DW. Assessing reliability and precision of measurement: an introduction to intraclass correlation and standard error of mesurement. J Spor Rehabil 1993;2: Holzberg M, Hewan-Lowe KO, Olansky AJ. The Ehlers-Danlos syndrome: recognition, characterization, and importance of a milder variant of the classic form. A preliminary study. J Am Acad Dermatol 1988;19: Grahame R. Hypermobility and hypermobility syndrome. In: Keer R, Grahame R, eds. Hypermobility syndrome - recognition and management for physiotherapists. London: Elsevier Limited, 2004;1 14. Appendix 1. Test protocol for reproducibility study of the Beighton tests and criteria for generalized joint hypermobility and of the Brighton tests and criteria for benign joint hypermobility syndrome (1) Passive apposition of the thumb to the flexor side of the forearm (shoulder 908 flexed, elbow extended and hand pronated), tested on right and left side, is performed by the patient after the following procedure. The examiner performs the test and asks: Can you with a straight arm move your thumb down so it touches the lower part of the forearm? If the test is negative, meaning no touch, the examiner asks: Have you been able to do this previously? (2) Passive dorsiflexion of the little finger >908 (elbow flexed 908, the forearm and hand pronated resting on a table), tested on right and left side, is performed by the patient after the following procedure. The examiner performs the test and asks: Can you with the forearm resting on the table, move your little finger, so it is pointing a little bit backwards? If the test is negative, the examiner asks: Have you been able to do this previously? (3) Passive hyperextension of the elbow >108 (shoulder 908 abducted and hand supinated), tested on right and left side, is performed by the patient after the following procedure. The examiner performs the test and asks: How much are you able to overstretch your elbow in this position (illustrated by the examiner) with your palm pointing towards the roof? If the test is negative, meaning no overstretch, the examiner asks: Have you been able to overstretch the elbow previously? (4) Passive hyperextension of the knee >108 (standing), tested on right and left side, is performed by the patient after the following procedure. The examiner performs the test and asks: How much are you able to overstretch your knee when you are standing straight up? If the test is negative, meaning no overstretch, the examiner asks: Have you been able to overstretch the knee previously? (5) Forward flexion of the trunk, with knees straight, so that the palms of the hands rest easily on the floor, is performed by the patient after the following procedure. The examiner performs the test and asks: Can you with straight knees bend your body forward and place both palms easily on the ground? If the test is negative, meaning no touch on the ground with the whole palm of the hands, the examiner asks: Have you been able to do this previously? The tests (currently and historically) used for BJHS included in the protocol were the following. The two major criteria include: (1) A Beighton score 4/9 or greater (currently or historically, results transferred from the above described test protocol). (2) Arthralgia (3 months) in 4 joints (currently or historically), the examiner asks Have you or have you had constant joint pain for more than three months? If yes: In how many joints at the same time? The eight minor criteria include: (1) Beighton score of 1 3 out of 9, and if aged >50 yrs, a Beighton score of 0 3 (currently or historically, results transferred from the above described test protocol). (2) Arthralgia (3 months) in 1 3 joints, and/or back pain, and/or spondylosis, spondylolysis/spondylolisthesis (currently or historically, results may be transferred from the above, described test protocol). The examiner asks: Have you or have you had constant joint pain for more than three months? If yes, the examiner asks: In how many joints at the same time? Have you or have you had back pain for more than three months, or have you had an X-ray diagnosis of spondylosis, spondylolysis or spondylolisthesis in your back? (Major and minor 1 as well as major and minor 2 criteria are mutually exclusive). (3) Dislocation/subluxation in more than one joint, or in one joint on more than one occasion (historically). The examiner asks: Has one of your joints been dislocated? If yes, the examiner asks: How many different joints? and Has it happened more than once? (4) Soft tissue rheumatism 3 lesions, e.g. epicondylitis, and/or tenosynovitis, and/or bursitis (historically). The examiner asks: Have you had epicondylitis, tenosynovitis or bursitis? If yes, the examiner asks: Have you had three or more than three incidents? (5) Marfanoid habitus, either one of: tall slim, arm span/height ratio >1.03, or upper/lower segment ratio <0.89, or arachnodactyly (positive Steinberg sign/wrist sign) (currently) (6) Abnormal skin, either one of: striae distensae, or skin hyperextensibility (3 cm on the volar aspect of the forearm), or thin skin or papyraceous scarring (>½cm broad papyraceous scars) (currently). The examiner asks: Do you have any stretch marks or broad scars? (7) Eye signs, either one of: drooping eyelids, or myopia or antimongoloid slant (currently). (8) Other signs, either one of: varicose veins, or hernia, or uterine/rectal prolapse (currently or historically). The examiner asks: Have you or have you had varicose veins, hernia or prolapsed uterus/rectum? The examiner question about hypermobility among relatives was: Are your parents, brothers/sisters or children hypermobile? For the diagnosis of BJHS either two major criteria, one major and two minor criteria, four minor criteria or two minor criteria

7 Reproducibility of tests and criteria for GJH and BJHS 1841 plus hypermobility in an unequivocally affected first degree relative. BJHS is excluded by presence of EDS, MF or OI. Tests for hypermobility in the neck, the shoulder and the hip according to Rote` s-quérol (currently or historically) included: (1) Active guided lateral rotation of the head as far as possible without lateral flexion. If the nose of the person points behind the frontal plane (>908rotation), when sitting with straight back and neck, chin retracted, hands placed behind the body to stabilize the shoulder area, the test is judged to be positive. The examiner asks: How far can you rotate your head to the side? If the test is negative, the examiner asks: Have you previously been able to rotate your head more/so much that your nose was pointing backwards? (2) Active guided outward rotation of the shoulder joint as far as possible. If the forearm points behind frontal plane (>908rotation) when sitting, straight back, upper arm along gravity, elbow joint in 908 of flexion and close to the body with forearm in neutral position, the test is judged to be positive. The examiner fixates the coracoid process and the acromion and moves the patient s forearm, and the examiner asks: How far can you rotate your forearm to the side, while you keep your elbow close to your body? If the test is negative, the examiner asks: Have you previously been able to rotate your forearm more/so that the forearm was pointing backwards? (3) Active guided abduction of the hip until the pelvis moves. If the leg can be abducted >458 from the neutral position when lying supine in the anatomically normal position, the test is judged to be positive. The examiner moves the calf with one hand and palpates the anterior superior iliac spine (SIAS) with the other hand, and asks: How far can you with a straight knee move your leg to the side? If the test is negative, the examiner asks: Have you previously been able to move it further out to the side/so far that you were almost able to do the splits?