Introduction

The unique shape of the trapeziometacarpal, or thumb carpometacarpal (CMC), joint affords a variety of movements in multiple planes. Perhaps as a consequence of its shape and the demands placed on this joint, the CMC joint is prone to the development of osteoarthritis. Precise causes of thumb CMC osteoarthritis remain unclear. Many risk factors have been suggested, including factors such as advanced age, female sex, congenital joint malformation, and genetic predisposition and acquired or environmental factors such as medical conditions, acute trauma, and occupational requirements (long-term low-level stress, eg, repetitive thumb use) [10]. Factors characterizing CMC arthritis include changes in the shape of joint surfaces and deterioration of the ligamentous support system.

Osteoarthritis of the first CMC joint is a common, painful problem, with prevalence reported as high as 15% in adults older than 30 years; as much as 1/3 of all postmenopausal women may be afflicted (based on data from the National Health and Nutrition Survey in 2006) [1, 7, 11, 14]. The condition is diagnosed using history, physical examination, and plain radiographs.

Eaton and Littler [9] introduced the most commonly used classification system for staging the severity of CMC joint arthritis in 1973. Eaton and Glickel [8] adapted this to include degenerative changes to the scaphotrapezial joint in 1987. The Eaton classification remains the most common system employed to determine the stage of the disease.

Reported shortcomings of this system include its limited intra- and interobserver reproducibility; high interobserver reliability is a critical quality measure for any classification system [12]. Perhaps as a consequence of inconsistent reproducibility of radiographic grading techniques, radiographic severity does not predictably correlate with symptoms or treatment recommendations [16]. To our knowledge, the literature on the Eaton classification system has not been systematically reviewed to determine the extent or quality of intra- and interobserver reliability and therefore no major conclusions have been developed on the subject.

Accordingly, we performed a systematic review to assess the degree of intra- and interobserver reliability of the Eaton radiographic classification of CMC joint arthritis.

Search Criteria and Strategy

We performed a systematic review of the following electronic databases: PubMed, Scopus®, and CINAHL (the Cumulative Index to Nursing and Allied Health Literature). Search terms were developed utilizing services provided through Lane Library at Stanford University School of Medicine (Table 1). We supplemented the search for additional relevant studies not identified by the electronic database from the references obtained in the reviewed studies. Search terms included: Eaton, thumb, carpometacarpal joints, metacarpophalangeal joint, hand joints, trapeziometacarpal joint, thumb carpometacarpal joint, CMC, basal joint, observer variation, classification, intrarater, interrater, interobserver, arthritis, osteoarthrit*, spondylarthrit*.

Table 1 Search terms used to retrieve articles assessing intra- and interobserver reliability of the Eaton classification for trapeziometacarpal arthritis
Full size table

Two independent reviewers (AJB, AM) screened retrieved titles and abstracts for articles of potential interest. Disagreement was resolved by consensus. Inclusion criteria were original studies investigating the degree of intra- and interobserver reliability of the Eaton classification for CMC arthritis and published in English language up to and including March 2013. Only studies that used plain radiographs as the mode of imaging were included for analysis. Studies using alternate modes of imaging for assessment of intra- and/or interobserver reliability (eg, CT, MRI, or ultrasonography) were excluded, unless they also included plain radiographs. Furthermore, we excluded editorials, commentaries, and letters to the editor.

The electronic search yielded 274 citations with 32 duplicate studies (Fig. 1). We assessed the potential relevance of the 242 unique citations and excluded 235 based on review of title and abstract. Of the seven remaining articles, three [2, 14, 17] were excluded after full-text review, as they were ultimately deemed not relevant to the study question. This process left four articles [5, 12, 13, 15] that met criteria for final analysis, which together included analysis of 163 patients’ radiographs, and these were critically reviewed. Each study included distributions of mild, moderate, and advanced cases. The observers in each report included hand surgeons, radiologists, and residents.

Fig. 1
figure 1

A flowchart demonstrates the identification of articles included for analysis.

Full size image

Parameters of interest included the level of experience of study participants reviewing the images, number of reviewers involved in the study, number of radiographs per subject/patient reviewed, additional and special views introduced, and limitations discussed in the respective studies. Furthermore, intra- and interobserver reliability calculations (kappa) were extracted. Intra- and interobserver reliability of the Eaton classification were measured using kappa values in all studies except for one [15], which only assessed interobserver reliability.

The level of evidence of the studies included in this analysis was determined using the Oxford Centre for Evidence Based Medicine Levels of Evidence Classification by two independent observers (AJB, AM) [4].

Results

Intraobserver reliability, as measured by the kappa value, ranged from 0.54 to 0.657 (fair to moderate) (Table 2). In the studies reporting on this end point, it was measured at two different time points ranging from 1 week [5, 13] to 6 weeks [12].

Table 2 Studies of intra- and interobserver reliability of the Eaton classification
Full size table

Interobserver reliability, as measured by the kappa value (both within and between groups, as well as being reported overall) ranged from 0.11 to 0.56 (poor to fair) (Table 2).

Based on the Oxford Centre for Evidence Based Medicine Levels of Evidence Classification, all four of the included studies were determined to be Level 3b.

Discussion

Detailed staging of thumb arthritis was first reported in the 1970s. Burton [3] developed a four-stage disease classification system in 1973 (Table 3) that relies on clinical signs, patient symptoms, and radiographs, including involvement of adjacent joints and the thumb metacarpophalangeal joint. Concurrently, Eaton and Littler [9] introduced a purely radiographic staging system for first CMC osteoarthritis (Table 4) that did not include the qualitative clinical findings associated with the Burton classification. Perhaps because of the more objective radiographic nature of the classification, the Eaton classification has been considered useful in clinical practice and has gained widespread acceptance [15]. However, multiple studies demonstrate large variation in the utility of this descriptive classification; furthermore, absence of agreement exists for correlating recommended treatment to CMC radiographic severity [5, 12, 13, 15]. We therefore evaluated the degree of intra- and interobserver reliability of the Eaton system for staging CMC arthritis.

Table 3 Burton classification [3]
Full size table
Table 4 Eaton classification [9, 15]
Full size table

Among the limitations of this study are the small number of studies available for analysis, as well as variations within each study with respect to observer specialty and level of training; hand surgeons, radiologists, and orthopaedic surgery residents were among the most commonly used observer groups.

The Eaton staging system was intended to describe radiographic severity tied to clinically relevant information, yet a disparity exists between clinical symptoms and radiographic disease. The purpose of this study was to perform a systematic review of the available literature assessing reliability and thus a measure of the strength of this system as it relates to consistency. A question then remains whether the radiographic reliability, or lack thereof, contributes to the discrepancy between clinical and radiographic disease. Furthermore, this study does not examine the validity of the Eaton classification as it relates to clinical utility or severity. Specific features that could enhance the clinical utility of the classification as it relates to surgical planning or outcomes are beyond the scope of this examination; this would require an enhancement of the reliability of the Eaton classification as it is currently exists.

Intraobserver reliability of radiographic CMC arthritis staging was fair to moderate, and interobserver reliability was poor to fair. Even when familiarity with the classification system is reported, the kappa values have been poor to moderate [5, 12, 13, 15]. Increasing the number of views and adding three-dimensional views of the CMC joint only modestly improved interobserver reliability of the staging system in two of the studies we evaluated [5, 12]. One of the variables for examination in the Burton [3], Eaton [9], and modified Eaton [2, 8] classifications is subluxation [3] and extent of subluxation [8, 9, 15] of the first metacarpal on the trapezium. Recent studies have confirmed that subluxation is present in asymptomatic subjects, measured with precise stress radiographs [17] and CT [6]. Thus, using degree of subluxation as criteria for staging, rather than a supporting variable, suggests current Eaton stratification is an inaccurate measure of disease severity. A concern with the current system is that observers tend to form their own interpretation of the classification system. This is especially true as it relates to Stage I, which describes joint widening. This is rarely seen in the experience of the senior author (ALL), and a common modified Stage I among hand surgeons describes minimal joint narrowing. Furthermore, since subluxation is probably not relevant, the senior author and colleagues often dismiss the required degree of subluxation in staging advanced disease. This type of informal modification predicts poor interobserver reliability.

Current literature suggests that Eaton stage does not correlate well with clinical findings, and treatment depends on patient response to nonoperative measures, a concept that has been confirmed in a recent Cochrane review evaluating different surgical strategies for various stages of osteoarthritis. That review indicated that patients with Stage I osteoarthritis are likely to benefit from nonoperative interventions, and treatment of those with Stage II to IV osteoarthritis depends on the severity of patient symptoms and their functional demands [16].

Regardless of rater experience level, number of available views, and availability of CT images, radiographic staging of CMC arthritis with the current standard, the Eaton classification, demonstrates intra- and interobserver agreement that is at best moderate and is generally even worse in the middle ranges of disease severity (Stages II and III). Additionally, the choice of treatment strategy tends to vary from surgeon to surgeon, and a single surgeon may recommend different therapies for patients with the same stage of osteoarthritis. This suggests that other qualitative factors, such as reported symptoms, physical examination, and surgeon preference, are more important for choosing treatment modalities than Eaton radiographic staging [15]. Variation in interpretation of the current staging system indicates the need for more accurate and reproducible radiographic measurements that quantify disease severity. Quantified radiographic reference standards may provide a preliminary step for reaching a consensus concerning the treatment of the common thumb CMC arthritis.