1 Introduction

Osteochondritis dissecans of the humeral capitellum (OCD) is an injury characterized by a defect in the subchondral bone and its overlying cartilage.¹ While the exact etiology is unknown, the pathogenesis of OCD has been hypothesized to be a combination of repetitive microtrauma to an area of the capitellum with tenuous vascular supply.^1,2 As such, OCD is most commonly seen in adolescent overhead athletes in which the elbow is repetitively exposed to valgus or shear forces, with the prevalence being 3.4% in baseball players between ages 12 and 18.²

These patients generally present with worsening activity-related pain and stiffness in their dominant arm.³ As with several pathologies, with low grade lesions the symptoms may be minimal. However, higher grade lesions cause significant pain and mechanical symptoms such as catching, clicking and locking, and restricted range of motion (ROM).² Current treatments for OCD include conservative modalities, arthroscopic/open debridement, microfracture, and osteochondral autograft transplantation.² The choice of treatment depends on size and stability of the lesion. Stability of OCD lesions can be classified intraoperatively into four stages by the International Cartilage Repair Society (ICRS) classification.⁴ ICRS Stage I indicates a stable lesion with continuous, softened area covered by intact cartilage; ICRS Stage II indicates a stable lesion with partial discontinuity; ICRS Stage III indicates a lesion with complete discontinuity, but not displaced; ICRS Stage IV indicates a displaced or loose fragment in the bed.

Although there are several imaging modalities and classifications used to evaluate OCD of the capitellum, there is a lack of agreement in terms of determining stability with different stages of each classification. The aim of this study was to determine which imaging modality and classification system most accurately predicts the intraoperative findings of stability.

2 Methods

Following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, a literature search was conducted using a boolean of the following terms: “osteochondritis”, “capitellum”, “capitellar”, and “elbow”. Studies that met the following criteria were included: (1) graded OCD lesions preoperatively based on radiographic criteria (2) provided number of patients in each OCD classification grade (3) written in English. Case reports, surgical techniques, and reviews were excluded. This was performed by evaluating the PubMed, Embase, and Ovid libraries until July 2020. The search was performed by two authors in an independent fashion and once completed, the results of the included studies were cross tabulated with a perfect agreement. Once all the studies were identified, the quality of these studies was evaluated utilizing the MINORS criteria, and studies with a criteria of less than 5 points were excluded (Fig. 1).⁵

Close

Fig. 1 Study flow diagram. Of the 230 studies identified using specific search strings 44 studies ultimately met final inclusion criteria for this report.

Export to PPT

The following data was extracted from qualifying studies: level of evidence, number of patients, patient demographics, capitellar physis status, ROM, ICRS classification as the gold standard, preoperative classification system used, and the division of patients within the system. All data was inputted into an electronic spreadsheet (Microsoft Excel, Microsoft Office, Redmond, Washington), by one of the authors (AP) and then this was verified by a second author (JJ). Due to the available data, a meta-analysis was not performed, however, all descriptive analyses were performed with statistical analyses included averages wixth the same electronic spreadsheet.

In studies that provided preoperative classification and ICRS classification for each patient, contingency tables were constructed on an electronic spreadsheet (Microsoft Excel, Microsoft Office, Redmond, Washington), and the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of each classification grade for its corresponding ICRS classification were evaluated. Additionally, the sensitivity, specificity, PPV, and NPV of each classification system for the diagnosis of OCD lesion stability were also evaluated.

3 Results

Following the aforementioned criteria, 44 articles were selected. A total of 2,038 patients were evaluated with a mean age of 12.1 years (range = 8.9–34 years) at presentation. Of the 1455 patients whose gender was documented, 85% (n = 1,235 patients) were male. Of the 155 patients whose capitellar epiphyseal status were evaluated, 67% (n = 104 patients) presented with a closed capitellar epiphyseal plate. These patients had a mean pre-operative extension of −6° and a mean pre-operative flexion of 111° with a total arch of 117° in the affected elbow. Of the 406 elbows that were classified intraoperatively according to the ICRS classification, 18 (4%) were stage I, 61 (15%) were stage II, 144 (35%) were stage III, and 183 (45%) were stage IV.

There was a total of 19 classification systems observed. Two systems were used for computed tomography (CT) scans,^6,7 five systems were used for magnetic resonance imaging (MRI),^8–12 eight systems were used for radiographs,^13–20 two systems were used for ultrasound,^13,21 and two system used both radiograph and MRI findings.^22,23 Eight classification systems evaluated lesion morphology, 5 evaluated the interface between the lesion and the subchondral bone, 7 evaluated the underlying capitellar and cartilage morphology, 2 evaluated the size of the lesion and 1 evaluated the physeal plate status. Four systems evaluate more than one of these criteria. The results are summarized in Table 1.

Two classification systems have been assessed for their intraobserver reliability, and six classification systems have been assessed for their interobserver reliability.^9,12,24 Intraobserver and interobserver reliability was evaluated either with the Siegel and Castellan multirater κ or the intraclass coefficient (ICC). Both variables are interpreted as follows: 0–0.2, slight agreement; 0.21–0.4, fair agreement; 0.41–0.6, moderate agreement; 0.61–0.8, substantial agreement; 0.81–1, almost perfect agreement, with 0 indicating no agreement beyond chance alone and 1 indicating perfect agreement.^25,26 One study found the Minami classification to have an interobserver κ of 0.27 and Ferkel and Sgaglion classification to have an interobserver κ of 0.22.²⁴ Another study found the Nelson and Dipaola classification to have an interobserver κ of 0.19.²⁷ The Kohyama classification was reported to have an average intraobserver ICC of 0.925 and an average interobserver ICC of 0.927.¹² The Itsubo classification was reported to have an average intraobserver ICC of 0.88 and an average interobserver ICC of 0.86.⁹ Another study found the Itsubo classification to have an interobserver κ of 0.18.²⁷

Four classification systems were compared directly with intraoperative ICRS staging.^8,9,12,17 The classification grade with the highest predictive value for determining that lesions are ICRS grade I was the Nelson grade 2, with a sensitivity, specificity, PPV and NPV of 100%.²⁸ The classification grade with the highest predictive value to determine lesions that are ICRS grade II was the Kohyama grade 2, with a sensitivity of 71.43%, specificity of 97.01%, PPV of 83.3%, and NPV of 94.2%.¹² The classification grade with the highest predictive value for determining lesions are ICRS grade III was the Kohyama grade 3, with a sensitivity of 96.97%, specificity of 87.5%, PPV of 84.2%, NPV of 97.7%.¹² The classification grade with the highest predictive value for determining lesions are ICRS grade IV are Itsubo grade 5 and Minami grade III. Lesions classified as Itsubo grade 5 have a sensitivity, specificity, PPV, and NPV of 100% when predicting if a lesion is ICRS grade IV.⁹ One study found that lesions classified as Minami grade III also have a sensitivity, specificity, PPV, and NPV of 100% when predicting if a lesion was ICRS grade IV.⁹ However, Mirzayan et al., found Minami grade III lesions to have a sensitivity of 80%, specificity of 100%, PPV of 100%, and NPV of 80% when determining if lesions were ICRS grade IV.²⁹ Satake et al., found Minami grade III lesions to have a sensitivity of 64.7%, specificity of 81.8%, PPV of 64.7%, and NPV of 81.8% when determining if lesions were ICRS grade IV.⁴ A study by Iwasaki et al., showed Minami grade III lesions to have a sensitivity of 100%, specificity of 0%, and PPV of 63.2% when determining if lesions are ICRS grade IV.³⁰ The results at summarized in Table 2.

Six classification systems have been assessed for their sensitivity in predicting unstable lesions.^{6,8,9,11,12,17} The classification system with the best predictive value for stability is the Itsubo classification, with a sensitivity of 100%, specificity of 80%, PPV of 93%, and NPV of 100%.⁹ Another study showed the Itsubo classification had a sensitivity of 65%, specificity of 20%, PPV of 76.47%, and NPV of 12.50%.³¹ The Kohyama classification had the second best predictive value for stability, with a sensitivity of 98.4%, sensitivity of 84.2%, PPV of 95.3%, and NPV of 94.1%.¹² However, since findings of stability on imaging are often treated nonoperatively and are therefore not graded intraoperatively, the NPV is subject to prevalence bias. The results are summarized in Table 3.

4 Discussion

OCD of the capitellum is a leading cause of permanent elbow disability in adolescent overhead athletes.¹³ Untimely or inadequate treatment can lead to radiocapitellar joint degeneration, which can lead to functional limitations and long-term pain.³² Preoperative imaging is the initial step to determine appropriate treatment for OCD. Agreement of radiographic findings with intraoperative findings would greatly guide treatment decisions and prognosis. We found that none of the current classification systems for OCD were effective in providing stages that predicted each ICRS grade with sensitivity, specificity, PPV, and NPV above 80%. However, we found that current MRI-based classification systems had better predictive value of ICRS lesion grade than radiographic classification systems. Additionally, we found that the Itsubo and Kohyama classifications, which utilize MRI, were most effective in determining OCD lesion stability. The Kohyama classification also had the highest interobserver agreement of all evaluated classification systems, although this may be skewed due to the evaluating group's familiarity with their own classification system.

The most popular classification method is the Minami classification, which was originally characterized for the elbow anteroposterior radiographs. Claessen et al., compared the Minami criteria with other radiographic and CT-based classification systems and found that it had the most reliable interobserver agreement between surgeons and radiologists specializing in orthopedic elbow pathology.²⁴ However, the agreement was only fair, and they were not able to establish correlation between classification and intraoperative findings.²⁴ Itsubo et al., also found that the Minami criteria did not correspond well with ICRS classification.⁹ A study by Satake et al., compared radiographs with intra-operative findings, finding that Minami criteria correlated with intraoperative diagnosis of instability in 44% of their cases.⁹ However, a study by Takahara et al., suggested Minami grade II and III lesions are unstable and proposed a guideline advising that these lesions should undergo surgical intervention.¹⁷ Thus, the Minami criteria provides accessible and reliable staging that can help decide between a nonoperative and operative approach to initial treatment.

Takahara et al. developed another radiograph-based classification system that classified OCD as “early” or “advanced” depending on the displacement of the lesion.¹⁸ This system however was insufficient in predicting results for nonoperation with only 55% of “early” lesions healing with nonoperative treatment.¹⁸ The Matsuura classification similarly classified lesions based on displacement of the lesion, however Matsuura stage I and II corresponded with varying levels of Takahara “early” lesions. With this differentiation, Matsuura et al., found that 90.5% of stage I and 52.9% of stage II lesions successfully healed with conservative treatment.³³ Kida et al., further expanded on the Matsuura classification by separating a stage for spontaneously healed lesions and surgically healed lesions to better match the natural progression and healing process of the disease.¹⁴ Difficulties in characterizing cartilage and tissue contrast on radiographs are cited as limitations of the Matsuura classification.³³ Two other radiograph-based classifications, the Iwase and Mitsunaga classifications, have also been proposed, but their effectiveness in predicting prognosis and intraoperative findings have not been well characterized.^15,20

MRI has been widely used to assess size, location, loose bodies, and underlying cartilage of OCD lesions. The Nelson classification system is the most commonly used MRI grading system and the second overall most popular grading system, which was originally described by Nelson et al., to grade OCD of the talus and knee.⁸ While this method was originally found to have good agreement with intraoperative arthroscopic grading of talar and knee OCD, two pitfalls identified were (1) the high false negative rate to detect loose bodies and (2) difficulties differentiating between severe grade 2 and grade 3 lesions.⁸ Wu et al., found that Nelson grade 4 lesions were significantly more associated with concomitant radial head lesions than Nelson grade 3 lesions.³⁴ Yamamoto et al., also found that osteochondral autograft transfer showed satisfactory results in Nelson grade 3 and 4 elbows in reducing pain postoperatively short term.³⁵

Another popular and validated MRI-based classification is the Itsubo classification. Itsubo et al., found this classification to have a sensitivity and specificity of 100% and 80% in diagnosing lesion instability in one study.⁹ However, van de Ende et al., found the Itsubo classification to have a sensitivity and specificity of 65% and 20% in determining lesion instability, respectively.³¹ Differences in these findings may be attributable to small cohort sizes and difficulty differentiating synovial fluid flow and bone marrow edema.³¹

Other proposed MRI based classification include the Dipaola, Hefti, and Kohyama classifications. The Dipaola classification was originally developed to grade OCD of the talus with 100% interobserver agreement and good correlation with arthroscopic grading of the talus.¹¹ However, Iwasaki et al., found the Dipaola classification to have a correlation of 41% with intraoperative staging of capitellar OCD.³⁰ In a study utilizing the Hefti staging, 64.3% stage I and 31.4% stage II lesions were found to heal with nonoperative treatment.³⁶ The Kohyama classification was shown to have sensitivity and specificity of 98.4% and 84.2% in identifying lesion instability and 88.9% accuracy in matching ICRS classification; however there have not been any other studies assessing the Kohyama classification's correlation with intraoperative findings.¹² Limitations of MRI include potential error in identifying loose bodies and limited diagnostic accuracy of subchondral bone information.^11,12 Because OCD patients are generally younger and more active, another limitation includes the longer acquisition time.

Currently CT is not widely used clinically to detect OCD lesions. This could be due to hazards of radiation exposure in immature elbows as well as the absence of a validated classification system for the capitellum.³¹ However, the Ferkel and Sgaglione classification and Clanton and DeLee classification have recently been adapted from the talus and knee, respectively, to evaluate OCD of the capitellum. The Ferkel and Sgaglione classification was shown to have fair interobserver agreement but did not provide any prognostic factors in terms of subchondral bone healing.^24,37 In contrast, van de Ende et al., found that the Clanton and DeLee classification yields a sensitivity of 100% in detecting OCD and 90% in determining lesion instability.³¹ A recent study comparing CT and MRI of OCD lesions showed that CT was more sensitive for indications of instability, such as fragmentation and tilt of the subchondral bone, loose bodies, and osteophytes.³⁸ These findings, along with quick acquisition time, make CT a potential modality to better characterize OCD lesions, especially when MRI findings are inconclusive.^9,38

Ultrasonography has been identified as a low cost, convenient, and safe diagnostic method to detect OCD in the field before the injury advances.^14,39 Ultrasonography can characterize subchondral bone and overlying cartilage in addition to identifying nondisplaced fragments.^9,40 Currently only Ishizaki et al., and Yang et al., have documented an OCD classification system using ultrasonography.^13,21 With the Ishizaki criteria, 90% of grade 2 and 3 abnormal findings were confirmed as OCD radiographically compared to 30.8% of grade 1a and 1b, suggesting utility only in diagnosing more progressed forms of OCD.¹³ Yang et al., also found their classification scheme to have a PPV of 100% when comparing later stage OCD with MRI findings, compared to an overall PPV of 66.7%.²¹ Neither classification scheme was compared to intraoperative findings. Nonetheless, ultrasonography can serve as an early screen for OCD.

Some systems use both radiographs and MRI to classify OCD lesions. Bradley and Petrie proposed a classification scheme that utilized MRI imaging to further classify early lesions with essentially normal radiographs.²² Similarly, Koasaka et al., utilized MRI to further classify Minami stage II into an early and late lesion.²³ Agreement with intraoperative findings have not been well characterized with either of these classification systems.

Two radiograph-based classification systems incorporated capitellar defect size as their criteria. Takahara et al., originally reported that larger osteochondral defects potentially facilitate degenerative changes and require more complex procedures than removal of the fragment.¹⁹ A study by Miyaki et al. further found large lesions with open proximal radial physis to be correlated with radial head enlargement and early osteoarthritis of the radiohumeral joint following arthroscopic debridement.¹⁶ Evaluation of defect size and physis status can therefore direct the choice of surgical intervention. Additionally, while lesion size could potentially improve the accuracy of radiographic classification, this has not been explored in the literature.

Based on these studies, we propose the following sequence of steps to determine OCD lesion stability. A patient with elbow pain or a positive screen by ultrasound should first obtain an elbow radiograph and be classified under the Minami criteria. Minami grade I lesions with an open growth plate, unrestricted motion, and no radial head enlargement are considered stable.^17,41 Minami grade III lesions are considered unstable.¹⁷ Minami grade I lesions with closed growth plate, restricted motion, or radial head enlargement and all Minami grade II lesions should obtain an elbow MRI and be graded by the Kohyama criteria for further classification. Kohyama stage 1 lesions are classified as stable, and Kohyama stage 4 lesions are unstable. Kohyama stage 2 and 3 lesions should obtain a CT scan. Lesions presenting with subchondral fragmentation and tilt, loose bodies, or osteophytes should be classified as unstable, and lesions without these features should be classified as stable.³⁸ This is summarized in Fig. 2.

Close

Fig. 2 Imaging flowchart to determine stability.

Export to PPT

This study had several limitations. As with every systematic review, we are limited by the available data, hence the limitations of the included studies are also our limitations. First, most studies have relatively small sample size and are retrospective in nature. Also, radiographic findings of stability generally lead to nonoperative treatment, but the gold standard of identifying stability is operation. This leaves the NPV of radiographic classification systems for instability subject to prevalence bias. Additionally, while our imaging flowchart suggests that all Kohyama stage 3 lesions receive a CT for further classification of the subchondral bone, Kohyama stage 3 lesions can also present with cartilage displacement, which would render this further classification redundant and the CT unnecesary. Finally, we did not determine the clinical implications in terms of patient reported outcomes or functional outcomes in patients who had a specific grading for each criteria. However, this is the most comprehensive review on radiographic criteria and can aid further studies in determining the best classification, or possibly create a new classification.

In conclusion, OCD can present in adolescents who participate in overhead, weightbearing sports. A combination of radiograph, MRI, and CT is most likely to yield the best characterization of an OCD lesion, which can guide treatment decisions and prognosis. However, with the information provided in our study, further prospective studies can be performed with the ultimate goal of determining the criteria that can more accurately classify these lesions and aid in managing patients with this potentially complex problem.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Disclaimer

None.

CRediT authorship contribution statement

Alex Pu: Conceptualization, Data curation, Formal analysis, Writing – original draft, Writing – review & editing. Julio J. Jauregui: Conceptualization, Data curation, Formal analysis, Writing – original draft, Writing – review & editing. Harold I. Salmons: Data curation, Writing – review & editing. Tristan B. Weir: Conceptualization, Writing – review & editing. Joshua M. Abzug: Conceptualization, Supervision, Writing – review & editing. Mohit N. Gilotra: Conceptualization, Supervision, Writing – review & editing.