Improving Clinical Recognition of Marfan Syndrome

Sponseller, Paul D. MD; Erkula, Gurkan MD; Skolasky, Richard L. MA, ScD; Venuti, Kristen D. CRNP; Dietz, Harry C. III MD

Journal of Bone & Joint Surgery - American Volume:
doi: 10.2106/JBJS.I.00892
Scientific Articles
Abstract

Background: Marfan syndrome is a potentially fatal disorder with cardiovascular, skeletal, and other manifestations that may also be seen in individuals without Marfan syndrome, making diagnosis difficult. Our goals were (1) to examine the ways in which patients have been recognized as having Marfan syndrome, (2) to examine the prevalence of current diagnostic findings, and (3) to determine which physically evident features are most sensitive and specific for referral to confirm a diagnosis of Marfan syndrome.

Methods: Between 2005 and 2007, we prospectively studied 183 consecutive patients with identified Marfan syndrome (Marfan group) and 1257 orthopaedic patients and family members (non-Marfan group). For the Marfan group, we recorded age at the time of recognition and the methods by which the syndrome was recognized; we used Ghent criteria to identify physically and radiographically evident features. For the non-Marfan group, we examined for Ghent criteria that could be noted on the basis of a routine history, physical examination, or radiographs. We used means, odds ratios, and frequencies to analyze the diagnostic use of each finding (α = 0.05).

Results: According to the Ghent criteria, 27% of patients in the Marfan group (mean age at the time of diagnosis, 7.3 years) had major skeletal involvement whereas 19% had zero or one skeletal feature. The most common physical features were craniofacial characteristics, high-arched palate, positive thumb and wrist signs, and scoliosis. In the non-Marfan group, 83% had one skeletal feature, 13% had two skeletal features, and 4% had three skeletal features or more. The physical features with the highest diagnostic yield were craniofacial characteristics, thumb and wrist signs, pectus excavatum, and severe hindfoot valgus.

Conclusions: Musculoskeletal clinicians should be aware of the diagnostic features of Marfan syndrome. Patients with three to four physically evident features, or two highly specific features (e.g., thumb and wrist signs, craniofacial features, dural ectasia, or protrusio), should be carefully reexamined and possibly referred for an echocardiogram or a genetics consultation.

Level of Evidence: Diagnostic Level II. See Instructions to Authors for a complete description of levels of evidence.

Author Information

1c/o Elaine P. Henze, BJ, ELS, Medical Editor and Director, Editorial Services, Department of Orthopaedic Surgery, Johns Hopkins Bayview Medical Center, 4940 Eastern Avenue, #A665, Baltimore, MD 21224-2780. E-mail address for E.P. Henze: ehenze1@jhmi.edu

Article Outline

Marfan syndrome, the result of a fibrillin-1 deficiency, has an estimated prevalence of one to two cases per 10,000 people1. If left untreated, it may decrease life expectancy2, although modern cardiovascular therapy can extend the mean life expectancy of patients with Marfan syndrome from forty-seven to seventy-five years3,4.

Early diagnosis and treatment of this syndrome would be advantageous5-7, but early diagnosis is complicated by two facts: (1) the pleomorphic nature and rarity of the Marfan phenotype pose challenges8, and (2) most of the associated features (e.g., aortic dilatation, aortic dissection, aortic valve prolapse, and ocular lens dislocation) can be revealed only with additional examination or diagnostic tests. Although definitive confirmation of the Marfan syndrome should be made by genetic experts, recognition for this referral comes from many sources. Because the skeletal features are the most visible, orthopaedic surgeons are often the first specialists to assess these patients, which provides a potential opportunity for diagnosis9-11. Ideally, musculoskeletal clinicians would accurately identify patients at risk for Marfan syndrome for referral to geneticists or cardiologists.

The goals of the present study were (1) to examine ways in which patients have been recognized as having Marfan syndrome, (2) to examine the prevalence of current diagnostic findings, and (3) to determine which physically evident features are most sensitive and specific for diagnostic referral.

Back to Top | Article Outline

Materials and Methods

The present study was approved by our institutional review board, and all patients in the study consented to participate.

From July 2005 through January 2007, we conducted a prospective study of two populations: (1) patients who were known to have Marfan syndrome as confirmed by a geneticist (Marfan group) and (2) patients who attended our pediatric orthopaedics and pediatric sports medicine clinics and their first-degree relatives (non-Marfan group). All patients who were identified as having Marfan syndrome had the diagnosis confirmed by a geneticist with use of clinical (Ghent) criteria12,13. Of the 195 potential participants in the Marfan group, twelve declined to participate. Of the 2010 potential participants in the non-Marfan group, 110 declined to participate and 643 were excluded because of the presence of another skeletal syndrome or neurologic disorder. The remaining 183 patients in the Marfan group and 1257 individuals in the non-Marfan group (633 patients from the pediatric orthopaedic clinic, 165 patients from the pediatric sports medicine clinic, and 459 family members) formed our study population (Table I). The mean age (and standard deviation) was 20.3 ± 15.1 years (range, one to sixty-four years) for the Marfan group and 26.5 ± 15.5 years (range, three to eighty-four years) for the non-Marfan group. Males comprised 46.4% of the patients in the Marfan group and 39.3% of the individuals in the non-Marfan group (p = 0.053).

For the patients in the Marfan group, we studied the process of recognition and identified the age at which Marfan syndrome was first diagnosed, the individual who first suspected the diagnosis, and the features that led to this recognition.

Individuals in both groups were examined for the major and minor Ghent criteria that could be determined from the routine history and physical examination. Height and arm span were measured for all subjects. Physical findings included skeletal features used for the diagnosis of Marfan syndrome: scoliosis of >20°, spondylolisthesis, pectus carinatum or severe excavatum, severe hindfoot valgus with medial displacement of the medial malleolus, an arm-span-to-height ratio of >1.05, elbow flexion contractures, and positive thumb and wrist signs, as described elsewhere14 (Fig. 1). The thumb (Steinberg) sign is considered to be positive when the entire distal phalanx protrudes beyond the ulnar border of the clenched fist15. The wrist (Walker-Murdoch) sign is considered positive if the thumb can cover the entire nail of the fifth finger when wrapped around the contralateral wrist16. Both of these tests are manifestations of long, slender limbs and joint laxity. When radiographs were not available, scoliosis was considered to be present if the angle of trunk rotation measured ≥6° on a scoliometer with forward bend17. If scoliosis was present, height was corrected with use of the formula of Bjure et al.18. Craniofacial features were also noted, including slender cranium (dolichocephaly), malar hypoplasia, retrognathia, and enophthalmos with down-slanting palpebral fissures (Fig. 2). Minor physical features that were evaluated included skin striae, high-arched palate, myopia (as evidenced by corrective eyewear), mild pectus excavatum, and recurrent inguinal or incisional hernia. Striae were not included if they occurred in association with marked weight gain or were present in the abdominal area of a female who had been pregnant. We also reviewed any radiographs that had been made during the routine care of those who were our patients. For example, a pelvic radiograph was considered to be positive for acetabular protrusio if the center-edge angle was >50° or if the medial acetabular wall protruded medial to the Kohler line. A magnetic resonance image of the spine, if made, was evaluated for dural ectasia.

Back to Top | Article Outline
Statistical Analysis

The goals of the current study were (1) to examine the ways in which patients have been recognized as having Marfan syndrome, (2) to examine the prevalence of current diagnostic findings, and (3) to determine which physically evident features are most sensitive and specific for genetics referral.

To achieve the first and second goals, we characterized the individuals with Marfan syndrome who were seen in our cohort study. Means and standard deviations were calculated for continuous measures, and numbers and percentages were calculated for categorical measures. Comparisons were made against the individuals without Marfan syndrome with use of the Student t test and the chi-square test when appropriate. The level of alpha was set at 0.05.

To achieve the third goal, logistic regression analysis was used to model the likelihood of a patient having Marfan syndrome, given the presence of a physically evident feature. Simple logistic regression was used to model the impact of a single feature without controlling for additional features. The discriminatory ability of each of these features was assessed through examination of the sensitivity and specificity of each physically evident feature. In addition, the positive and negative predictive values were computed. Physically evident features that showed a significant impact were retained for further analysis. Multivariate stepwise logistic regression was used to model the impact of a single feature while controlling for other features. Thresholds for entry into and removal from the model were 5% and 10%, respectively. A final model with all significant discriminators was computed. A receiver-operator characteristic curve analysis was used to take into account the trade-off between sensitivity and specificity. Accuracy was defined with use of the area under the curve. An area under the curve of 0.50 indicates purely random performance, and a value of 1.00 indicates the maximum value possible.

Back to Top | Article Outline
Source of Funding

There was no external source of funding for this study.

Back to Top | Article Outline

Results

Patients with Marfan syndrome had been diagnosed at a mean age of 7.3 ± 10 years. However, 10% of patients had not been diagnosed until after the age of twenty-eight years. In five patients, recognition came only after a relative's death or aortic dissection. Recognition of a possible diagnosis of Marfan syndrome was made by a pediatrician in 26% of the cases, a family member in 21%, an ophthalmologist in 14%, a family practitioner in 8.4%, an orthopaedist in 7.7%, and another contact in 23%.

Patients in the Marfan and non-Marfan groups had a mean of 2.8 and 0.3 major Ghent diagnostic criteria, respectively. Of those in the Marfan group, 27% had four or more major skeletal diagnostic criteria, which is the Ghent definition of major skeletal involvement (Table I). Conversely, 19% had zero or only one Ghent skeletal diagnostic feature and therefore could be considered to have “silent” Marfan syndrome. In the non-Marfan group, twenty-four individuals were referred for an echocardiogram or genetics consultation, or both; 83% had one skeletal feature, 13% had two skeletal features, and 4% had three skeletal features or more. None were diagnosed with Marfan syndrome, although some were followed because of this possible diagnosis.

Back to Top | Article Outline
Physically Evident Features

The average arm-span-to-height ratio was 1.04 ± 0.04 and 1.01 ± 0.04 in the Marfan and non-Marfan groups, respectively; this difference was significant (p < 0.001). The prevalence of positive physically evident findings in patients with and without Marfan syndrome was 34.4% and 11.5%, respectively, for an arm-span-to-height ratio of >1.05; 59.2% and 21.3% for scoliosis; 43.1% and 1.1% for pectus excavatum; 16.4% and 1.1% for pectus carinatum; 46.5% and 2.0% for severe hindfoot valgus with medial displacement of the medial malleolus; 18.8% and 2.8% for elbow flexion contracture; 79.3% and 4.7% for a positive thumb sign; 74.0% and 7.7% for a positive wrist sign; 66.0% and 2.0% for positive thumb and wrist signs (combined); 57.0% and 6.1% for skin striae; 77.9% and 15.2% for a high-arched palate; 89.5% and 2.7% for craniofacial features; 62.9% and 0.3% for acetabular protrusio; 27.5% and 0% for dural ectasia; and 63.2% and 12.4% for flat feet.

Back to Top | Article Outline
Sensitivity and Specificity

The diagnostic accuracy of each physically evident feature (based on a simple logistic regression model) is presented in Table II. The presence of craniofacial features was the most sensitive discriminator between individuals with and without Marfan syndrome (sensitivity, 89.5%; 95% confidence interval, 84.9% to 94.1%), with a diagnostic accuracy (area under the curve) of 0.934. A positive thumb sign was also a sensitive discriminator between those with and without Marfan syndrome (sensitivity, 79.3% [95% confidence interval, 73.3% to 85.3%]; specificity, 95.3% [95% confidence interval, 94.1% to 96.5%]), with a diagnostic accuracy of 0.873.

Dural ectasia was the most specific discriminator between those with and without Marfan syndrome (specificity, 100%), although it was not sensitive (sensitivity, 27.5%; 95% confidence interval, 18.3% to 36.6%), with a diagnostic accuracy of 0.637. Similarly, acetabular protrusio was a specific (specificity, 99.7%; 95% confidence interval, 99.4% to 100%) although not a sensitive (sensitivity, 37.1%; 95% confidence interval, 27.9% to 46.4%) discriminator between those with and without Marfan syndrome, with a diagnostic accuracy of 0.684.

Based on stepwise multivariate logistic regression, the combination of the presence of craniofacial features and a positive thumb sign provided the highest diagnostic accuracy (area under the curve, 0.997). The presence of craniofacial features and/or a positive thumb sign had a high sensitivity (93.2%) and specificity (93.2%) for the discrimination between individuals with and without Marfan syndrome.

Back to Top | Article Outline
Scoliosis Subset Analysis

Scoliosis was identified in 103 patients in the Marfan group and in 255 individuals in the non-Marfan group. Among patients with scoliosis, the presence of craniofacial features was the most sensitive (sensitivity, 92.3%; 95% confidence interval, 87.2% to 97.4%) as well as a specific (specificity, 92.9%; 95% confidence interval, 89.8% to 96.0%) discriminator between those with and without Marfan syndrome. A positive thumb sign was a sensitive (sensitivity, 83.8%; 95% confidence interval, 76.8% to 90.9%) and specific (specificity, 89.4%; 95% confidence interval, 85.7% to 93.1%) discriminator.

Back to Top | Article Outline

Discussion

The recognition of Marfan syndrome may be life-saving. However, recognition often depends on physically evident features such as those involving the skeleton. The fact that many of these features are present with some frequency in the general population creates a challenge. The present study showed that patients with Marfan syndrome usually were diagnosed in childhood but that >10% were diagnosed as adults. In the latter group, the opportunity to use medical treatment to modify aortic enlargement was lost19. Also, 3% of patients were diagnosed only after a cardiovascular emergency in a relative and many individuals (21%) were first recognized as having the syndrome by family members rather than by medical professionals. Even though many patients with Marfan syndrome present first to an orthopaedic surgeon, fewer than 10% of those in our study were first recognized by an orthopaedic surgeon. The low frequency of Marfan syndrome in the population (0.01% to 0.02%) means that most of the relevant physical findings in isolation will be false-positive findings. The challenge is to recognize and to refer the potential patients with Marfan syndrome while minimizing over-referral (Table III, Fig. 3).

The physical findings with the highest diagnostic yield were the thumb and wrist signs, craniofacial features, and pectus excavatum. Craniofacial features were the most sensitive, and combined thumb and wrist signs were the most specific. The value of craniofacial features needs to be studied further because currently these features are only subjectively defined. The criterion of severe hindfoot valgus (defined as medial deviation of the medial malleolus12,13) is also quite subjective and therefore less useful. If interpreted loosely as “flat foot,” it is not specific at all. The criterion of scoliosis of >20° also contributes only slightly to diagnostic yield (odds ratio, 5.6) because of the high prevalence of this condition in the general United States population. In isolation, none of these physical findings constitutes an indication for referral.

Some of the Ghent criteria also involve skeletal imaging. The presence of dural ectasia is highly specific, as is acetabular protrusio; these criteria should be used for assessment when available. One of the difficulties, however, is that a magnetic resonance imaging study may not be available for many patients.

There are several entities that overlap with Marfan syndrome. Loeys-Dietz syndrome may be characterized by hyperextensible joints, aneurysms of the aorta and other large vessels, scoliosis, and foot deformities20. Hypertelorism is a common feature and is caused by a mutation in the gene of the transforming growth factor receptor proteins. MASS syndrome refers to mitral valve prolapse, mild aortic enlargement, striae, and skeletal findings in the absence of true Marfan syndrome21. Ectopia lentis may also be an isolated finding in some cases because of a fibrillin-1 mutation22. Shprintzen-Goldberg syndrome is, in some cases, a result of fibrillin-1 mutation and may be manifested by scoliosis, arachnodactyly, sternal deformity, and developmental delay23. Patients with homocystinuria may have scoliosis, arachnodactyly, lens dislocations, and vascular thromboses; treatment of this metabolic disorder is important24. Patients with some forms of the Ehlers-Danlos syndrome may have joint hyperextensibility, scoliosis, aortic root dilatation, and sternal deformity25.

Tests for mutations in the gene for fibrillin-1 are now readily available but are expensive. However, this test should be ordered by specialists who understand its implications and limitations. There are several other clinical entities that are caused by mutations in the fibrillin-1 gene: MASS phenotype, Shprintzen-Goldberg syndrome, Weill-Marchesani syndrome26, and isolated ectopia lentis are a few of the most relevant conditions. In addition, a small percentage of patients with Marfan syndrome have not been found to have mutations on this test, perhaps because of large gene deletions.

It is not possible to perform a complete examination for Marfan syndrome in every patient. It also is not known what should be the appropriate referral or false-positive rate. Referral of all patients who have any of the Ghent criteria is clearly impractical because >30% of the individuals in the non-Marfan group in the present study had one Ghent criterion or more. However, a referral rate of 1% to 2% may be clinically reasonable. The fact that none of the twenty-four patients whom we referred during the course of this study was diagnosed with Marfan syndrome highlights this concern, although we had recognized other patients before the time of this study. The finding that 19% of patients with Marfan syndrome had zero or one skeletal finding is a clear indication that not all patients can be clinically recognized on the basis of physical observation. Such patients who evade recognition may be thought of as having “silent” Marfan syndrome in terms of their recognizability.

We used our data to try to identify the best parameters that would provide the most reasonable percentage of genetic referrals and a minimal chance of missing patients with Marfan syndrome. Requiring more clinical criteria that are suggestive of Marfan syndrome before making a referral has the effect of lowering the number of referrals that may be false-positive but increasing the odds of missing a patient who might have Marfan syndrome. Referral of patients who have four or more signs from the set of skeletal major and minor Ghent criteria, evidence of a positive family history, and the presence of dural ectasia provides the minimum rate of referral to a geneticist (0.4%) and would miss 17.8% of patients with Marfan syndrome. Choosing a more liberal threshold of three findings causes potential missing of only 6.1% of patients with Marfan syndrome but increases the referral rate to 2.2%. Because 19% of patients with Marfan syndrome are considered to have a silent form of the disease, this threshold may be the closest possible approximation of a guideline for recognition.

On the basis of the findings of the present study, we suggest the following procedure to help to appropriately identify patients with potential Marfan syndrome who need to be referred to a geneticist. During an initial physical examination, the orthopaedic surgeon should at least briefly visualize the entire patient. Ideally, a patient's shoes should be off and the patient should be wearing clothing that allows for the observation of maximal body area, including the sternum. The patient should be asked about a family history of connective-tissue disorder. If the fingers appear long and slender, the clinician should check the thumb and wrist signs. If either of these signs or another relevant feature is seen, the examiner should check for the others on the list. An anteroposterior radiograph of the pelvis may be examined for acetabular protrusio. If three or four characteristic features are seen, especially if two of them are highly informative features (craniofacial features or the thumb or wrist sign), referral for an echocardiogram or a genetic consultation is suggested. Dural ectasia and acetabular protrusio, although rarely seen in the general population, are highly sensitive for Marfan syndrome and should prompt referral. The time, expense, and chance of a “negative” referral need to be balanced against the risk of missing the diagnosis. It is important to know that in very young patients, the full phenotype may take time to evolve, and the geneticist may need to monitor these patients over time.

NOTE: The authors thank Teri M. McCambridge, MD, for her assistance in this study.

Investigation performed at the Departments of Orthopaedic Surgery and Pediatrics, The Johns Hopkins University, Baltimore, and the Howard Hughes Medical Institute, Baltimore, Maryland

Disclosure: The authors did not receive any outside funding or grants in support of their research for or preparation of this work. Neither they nor a member of their immediate families received payments or other benefits or a commitment or agreement to provide such benefits from a commercial entity.

Back to Top | Article Outline

References

1. Dietz HC Cutting GR Pyeritz RE Maslen CL Sakai LY Corson GM Puffenberger EG Hamosh A Nanthakumar EJ Curristin SM Stetten G Meyers DA Francomano CA. Marfan syndrome caused by a recurrent de novo missense mutation in the fibrillin gene. Nature. 1991;352:337–9.
2. Manusov EG Martucci E. The Marfan syndrome. An underdiagnosed killer. Arch Fam Med. 1994;3:822–6.
3. Bethea BT Fitton TP Alejo DE Barreiro CJ Cattaneo SM Dietz HC Spevak PJ Lima JA Gott VL Cameron DE. Results of aortic valve-sparing operations: experience with remodeling and reimplantation procedures in 65 patients. Ann Thorac Surg. 2004;78:767–72.
4. Bhudia SK Troughton R Lam BK Rajeswaran J Mills WR Gillinov AM Griffin BP Blackstone EH Lytle BW Svensson LG. Mitral valve surgery in the adult Marfan syndrome patient. Ann Thorac Surg. 2006;81:843–8.
5. Amado JA Thomas DJ. Early recognition of Marfan's syndrome. J Am Acad Nurse Pract. 2002;14:201–6.
6. Chan KL Callahan JA Seward JB Tajik AJ Gordon H. Marfan syndrome diagnosed in patients 32 years of age or older. Mayo Clin Proc. 1987;62:589–94.
7. Eaton L Meiner SE. Marfan syndrome: identification and management. Medsurg Nurs. 1999;8:113–7.
8. Aoyama T Francke U Dietz HC Furthmayr H. Quantitative differences in biosynthesis and extracellular deposition of fibrillin in cultured fibroblasts distinguish five groups of Marfan syndrome patients and suggest distinct pathogenetic mechanisms. J Clin Invest. 1994;94:130–7.
9. Grahame R Pyeritz RE. The Marfan syndrome: joint and skin manifestations are prevalent and correlated. Br J Rheumatol. 1995;34:126–31.
10. Joseph KN Kane HA Milner RS Steg NL Williamson MB Jr Bowen JR. Orthopedic aspects of the Marfan phenotype. Clin Orthop Relat Res. 1992;277:251–61.
11. Pyeritz RE McKusick VA. The Marfan syndrome: diagnosis and management. N Engl J Med. 1979;300:772–7.
12. Beighton P de Paepe A Danks D Finidori G Gedde-Dahl T Goodman R Hall JG Hollister DW Horton W McKusick VA Opitz JM Pope FM Pyeritz RE Rimoin DL Sillence D Spranger JW Thompson E Tsipouras P Viljoen D Winship I Young I. International Nosology of Heritable Disorders of Connective Tissue, Berlin, 1986. Am J Med Genet. 1988;29:581–94.
13. De Paepe A Devereux RB Dietz HC Hennekam RC Pyeritz RE. Revised diagnostic criteria for the Marfan syndrome. Am J Med Genet. 1996;62:417–26.
14. Judge DP Dietz HC. Marfan's syndrome. Lancet. 2005;366:1965–76.
15. Feingold M. The thumb sign in children. Clin Pediatr (Phila). 1968;7:423–4.
16. Walker BA Murdoch JL. The wrist sign. A useful physical finding in the Marfan syndrome. Arch Intern Med. 1970;126:276–7.
17. Bunnell WP. An objective criterion for scoliosis screening. J Bone Joint Surg Am. 1984;66:1381–7.
18. Bjure J Grimby G Nachemson A. Correction of body height in predicting spirometric values in scoliotic patients. Scand J Clin Lab Invest. 1968;21:191–2.
19. Williams A Davies S Stuart AG Wilson DG Fraser AG. Medical treatment of Marfan syndrome: a time for change. Heart. 2008;94:414–21.
20. Loeys BL Schwarze U Holm T Callewaert BL Thomas GH Pannu H De Backer JF Oswald GL Symoens S Manouvrier S Roberts AE Faravelli F Greco MA Pyeritz RE Milewicz DM Coucke PJ Cameron DE Braverman AC Byers PH De Paepe AM Dietz HC. Aneurysm syndromes caused by mutations in the TGF-beta receptor. N Engl J Med. 2006;355:788–98.
21. Glesby MJ Pyeritz RE. Association of mitral valve prolapse and systemic abnormalities of connective tissue. A phenotypic continuum. JAMA. 1989;262:523–8.
22. Edwards MJ Challinor CJ Colley PW Roberts J Partington MW Hollway GE Kozman HM Mulley JC. Clinical and linkage study of a large family with simple ectopia lentis linked to FBN1. Am J Med Genet. 1994;53:65–71.
23. Shprintzen RJ Goldberg RB. A recurrent pattern syndrome of craniosynostosis associated with arachnodactyly and abdominal hernias. J Craniofac Genet Dev Biol. 1982;2:65–74.
24. Münke M Kraus JP Ohura T Francke U. The gene for cystathionine beta-synthase (CBS) maps to the subtelomeric region on human chromosome 21q and to proximal mouse chromosome 17. Am J Hum Genet. 1988;42:550–9.
25. Farag TI Schimke RN. Ehlers-Danlos syndrome: a new oculo-scoliotic type with associated polyneuropathy? Clin Genet. 1989;35:121–4.
26. Faivre L Gorlin RJ Wirtz MK Godfrey M Dagoneau N Samples JR Le Merrer M Collod-Beroud G Boileau C Munnich A Cormier-Daire V. In frame fibrillin-1 gene deletion in autosomal dominant Weill-Marchesani syndrome. J Med Genet. 2003;40:34–6.
Copyright 2010 by The Journal of Bone and Joint Surgery, Incorporated