Clinical studies have demonstrated that radial head arthroplasty is safe and effective for restoring elbow stability following comminuted unreconstructable radial head fractures associated with complex instability1-4. Biomechanical studies have shown that insertion of a correctly sized metallic radial head implant can reliably restore elbow and forearm kinematics and stability to nearly normal levels5-7. A primary technical goal during radial head arthroplasty is placement of an implant that closely replicates the dimensions of the native radial head. Insertion of an implant that is too thick, which lengthens the radius axially or "overstuffs" the radiocapitellar joint, has been associated with pain, loss of elbow motion, articular cartilage erosions of the capitellum, and posttraumatic arthritis8-11.
Several reliable intraoperative anatomical landmarks have been identified to help surgeons to select a radial head implant of correct length12-14. Knowledge of these guidelines is especially important when the radial head is highly comminuted and cannot be used to approximate implant size as well as during revision arthroplasty, in which the native radial head is absent. Despite the known importance of using a radial head implant of correct length, insertion of an incorrectly sized prosthesis is not uncommon15. Currently, there are no reliable radiographic criteria with which to diagnose overlengthening by a radial head implant13,16,17, which makes the management of patients with a painful radial head prosthesis challenging15.
The purpose of this study was to determine if radiographs of the contralateral, normal elbow could be used to diagnose and quantify the magnitude of overlengthening due to a radial head implant. The first part of the study was performed to establish the side-to-side concordance of joint dimensions to determine if radiographs of the contralateral elbow could be used as surrogates to calculate radial head implant length. In part II, we sought to validate the radiographic measurement technique as a method for diagnosing and calculating the magnitude of radial head implant overlengthening in a controlled cadaveric experiment.
Part I: Side-to-Side Concordance of Radiographic Elbow Anatomy
Patients
A series of 100 matched bilateral elbow radiographs was assembled from a cohort of fifty patients with no known history of elbow injury and no clinical symptoms related to the elbow. The radiographs were obtained with use of a standardized imaging protocol. The mean age of the patients (twenty-five men and twenty-five women) was forty-seven years (range, eighteen to eighty years). This study was approved by our institutional review board.
Radiographic Methods
All radiographs were obtained with use of a commercial digital imaging system (GE Healthcare, Piscataway, New Jersey). Commercially available imaging software (Centricity; General Electric, Burlington, Vermont) was used to magnify images 2×. The radiographic joint-space dimensions in the coronal plane were determined by measuring the distance between the ulna and the humerus at four locations, as described by Rowland et al.16. These four locations (from medial to lateral) included the medial side of the medial ulnohumeral facet (M-MUF), the lateral side of the medial ulnohumeral facet (L-MUF), the medial side of the lateral ulnohumeral facet (M-LUF), and the lateral side of the lateral ulnohumeral facet (L-LUF) (Fig. 1). Once these measurements were obtained, a side-to-side statistical comparison was done and intrarater reliability and interrater reliability were calculated (see Statistical Methods).
Part II: Cadaver Study to Diagnose Overlengthening Due to the Radial Head Implant
Calculation of Radial Head Implant Length
A formula was developed to calculate the length of a radial head implant on the basis of contralateral elbow radiographs. The formula is based on a previous study of radial head implant overlengthening by Frank et al., who demonstrated that thicker implants resulted in increased varus angulation of the elbow with hinging open of the lateral ulnohumeral joint13. The method requires an anteroposterior radiograph of the contralateral, normal elbow to calculate the virtual length of the native radial head (L-NRH). The virtual length of the radial head implant (L-RHI) is then determined on the post-arthroplasty anteroposterior radiograph of the ipsilateral elbow. Next, L-NRH is subtracted from L-RHI, equaling the magnitude of implant overlengthening. The calculation method requires six steps, as is outlined in Figure 2.
Step I: On an anteroposterior radiograph of the contralateral, normal elbow, mark the lateral edge of the lateral ulnohumeral joint on both the humerus (h) and the ulna (u).
Step II: Create an angle, with the apex starting at the most medial aspect of the coronoid, by drawing lines bisecting points h and u.
Step III: Draw a line in the coronal plane that bisects the native radial head. This line starts at a point in the center of the radial neck (cRN) and bisects a point at the center of the radial head (cRH).
Step IV: The portion of this line that lies between the limbs of the angle is the virtual length of the native radial head (L-NRH).
Step V: On the anteroposterior radiograph of the elbow with the radial head implant, all of the same angles and lines are used. Points h and u are marked, and this is followed by creation of an angle with its apex at the most medial extent of the coronoid. A line is drawn that bisects the radial head implant. This line starts at the distal center point of the radial implant stem (cRS) and bisects a point at the center of the radial head implant (cRHI). The portion of this line that lies between the limbs of the angle is the virtual length of the radial head implant (L-RHI).
Step VI: The amount of implant overlengthening is calculated with the formula: (L-RHI) — (L-NRH) = magnitude of overlengthening in millimeters.
Validation of Method: Specimen Preparation and Radiographs
Eight fresh-frozen human cadaveric upper extremities (four pairs) were thawed overnight at room temperature. Prior to their inclusion, the elbows were assessed with radiographs to look for pathological changes and the range of motion and stability of the elbows were evaluated. The average age of the donors at the time of death was sixty-seven years (range, sixty to eighty-one years). The specimens were transected at the midpart of the humerus with the elbow, forearm, and hand left intact. Skin adjacent to the elbow was removed, and the biceps, brachialis, and triceps tendons were identified and were sutured with number-2 high-strength suture (Hi-Fi; Conmed Linvatec, Utica, New York) in preparation for loading to simulate muscle tone. The specimens were tested with the muscles loaded to simulate resting muscle tone, with 10 N applied to the biceps, 10 N applied to the brachialis, and 20 N applied to the triceps13.
Each specimen was mounted on a custom radiolucent rig to control elbow flexion/extension and forearm rotation (Fig. 3). A lateral approach to the elbow was performed between the anconeus and the extensor carpi ulnaris (the Kocher interval) with release of the anterior and posterior aspects of the capsule and preservation of the medial collateral ligament. The lateral collateral ligament was released from its origin on the lateral epicondyle to simulate traumatic instability.
The radial head prostheses were then implanted. These prostheses included an implant of the correct length and implants that resulted in 2, 4, 6, and 8 mm of overlengthening, as determined by measuring the maximum and minimum diameters of the native radial head in situ with digital calipers (Digimatic; Mitutoyo, Aurora, Illinois) with an accuracy of 0.03 mm. The radial head implant (Evolve Proline; Wright Medical Technology, Arlington, Tennessee) was selected on the basis of the minimum diameter. The thickness of the implant was measured with digital calipers, and an equal thickness of native radial head was resected with use of a microsagittal saw, with the cut thickness incorporating the thickness of the saw blade. The canal of the radius was reamed and prepared according to the guidelines in the implant manufacturer's technical manual. A radial head implant of correct length was then inserted, and the lateral collateral ligament was repaired. The lateral collateral ligament was repaired by creating two divergent bone tunnels starting at the isometric point on the lateral humeral epicondyle and exiting along the posterior aspect of the lateral humeral condyle. The lateral collateral ligament was sutured with a number-2 braided suture (Hi-Fi) in a locking fashion, and the suture ends were then passed through the bone tunnels and tensioned at 20 N with the elbow placed in 90° of flexion and the forearm in neutral rotation. The lateral collateral ligament sutures were then clamped to the rig to maintain constant ligament length. Use of the clamp mechanism made it possible to easily release the sutures to allow access to the joint for implantation of radial head implants of various length followed by retensioning of the lateral collateral ligament.
After implantation of a radial head prosthesis, anteroposterior radiographs relative to the forearm were obtained with the elbow in 45° of flexion and the forearm in neutral rotation. In summary, anteroposterior radiographs were made of the native elbow; after insertion of an implant of correct length; and after insertion of implants that resulted in 2, 4, 6, and 8 mm of overlengthening. A total of forty-eight radiographs were made for the eight specimens; these included eight radiographs of the native specimen and forty with an implant in place (eight radiographs for each of five testing conditions [correct length and overlengthening of 2, 4, 6, and 8 mm]).
Validation of Method: Testing the Formula
The method for calculating radial head implant overlengthening with use of a radiograph of the contralateral elbow was tested in a blinded fashion by two experienced elbow surgeons. The forty radiographs of the implants with varying degrees of overlengthening and the corresponding radiographs of the contralateral, normal elbow were blinded, and the testing formula was applied to calculate the magnitude of overlengthening. Examiner 1 (G.S.A.) evaluated the set of forty implant radiographs twice (eighty observations), and Examiner 2 (G.J.W.K.) evaluated the set once (forty observations) for a total of 120 observations. The derived values, based on measurements obtained from the radiographs of the contralateral, normal elbow, were then compared with the known values of the amount of overlengthening by an independent investigator with expertise in statistical methods (J.C.M.).
Statistical Methods
Descriptive statistics including the mean and variability of radiographic measures were calculated. A Pearson correlation test was used to compare the dimensions of the right elbow with those of the left elbow at each of the four joint-space measurement sites. Correlation coefficients between 0.6 and 0.75 were regarded as good; those between 0.751 and 0.89, very good; and those of 0.90 and over, excellent.
In order to assess intrarater and interrater reliability in part I of the study, each radiographic joint-space measurement was conducted twice, two weeks apart, by Examiner 3 (D.M.R.) and a single time by Examiner 1 (G.S.A.). In order to assess intrarater and interrater reliability in part II, each radiographic calculation measurement was conducted twice, two weeks apart, by Examiner 1 (G.S.A.) and a single time by Examiner 2 (G.J.W.K.). Intraclass correlation coefficients (ICCs [2,1])18 were calculated to assess the agreement between the different raters and between the measurements made at the different intervals by the same rater to assess reliability, and between the calculated and actual implant sizes to assess the validity of the measurement method. The 95% confidence intervals (CIs) were calculated around the ICCs. A Bland and Altman19 technique that allows for graphical representation of the differences between the calculated and actual implant sizes across the spectrum of implants was conducted, and the mean error plus the two standard deviation limits around that error were portrayed. A standard contingency table was used to calculate the sensitivity, specificity, and likelihood ratios for the accuracy of the measurement method for diagnosing radial overlengthening.
Source of Funding
This research was financially supported by the Alexandra Kirkley Young Investigator Award from the Canadian Orthopaedic Foundation. Equipment support was provided by Wright Medical Technology (Mississauga, Ontario).
Part I
No significant differences (p > 0.2) in the average radiographic dimensions at the four ulnohumeral joint-space measurement sites were identified between left and right elbows. The right-to-left correlation (Pearson correlation coefficient) was very good to excellent for the lateral facet (M-LUF and L-LUF), with coefficients between 0.82 and 0.90 for both observers. The correlation was good for the medial facet (M-MUF and L-MUF), with coefficients between 0.61 and 0.74.
The intraobserver reliability was excellent for M-LUF (ICC = 0.93; 95% CI = 0.88 to 0.96), L-LUF (ICC = 0.91; 95% CI = 0.84 to 0.95), M-MUF (ICC = 0.94; 95% CI = 0.89 to 0.96), and very good for L-MUF (ICC = 0.85; 95% CI = 0.73 to 0.92). The inter-item correlation matrix reported ICCs between 0.73 and 0.93, representing good-to-excellent interobserver reliability. We concluded, therefore, that measurements on radiographs of the contralateral, normal elbow, and specifically those of the lateral ulnohumeral facet, could function as effective surrogates for joint-space measurements and for use in the radiographic measurement method.
Part II
The radiographic measurement method was used successfully to calculate the magnitude of radial head implant overlengthening within ±1 mm in 104 (87%) of the 120 scenarios tested (Fig. 4 and Appendix). The method was effective for calculating implant size within ±2 mm in all (100%) of the scenarios tested. The overall diagnostic accuracy of the test—i.e., its ability to correctly identify overlengthening, within 1 mm, when it was present—was excellent (Table I). The ability of the test to precisely identify the magnitude of overlengthening was 92% when the amount of overlengthening was set at 2 mm (n = 24), 96% when it was set at 4 mm (n = 24), 87% when was set at 6 mm (n = 24), and 54% when it was set at 8 mm (n = 24).
The reliability of the radiographic measurements, based on repeated measurements performed by a single blinded orthopaedic surgeon on separate occasions or based on separate measurements performed by two different orthopaedic surgeons, was excellent (ICC > 0.95). The overall agreement between the actual measurement and the calculated measurement was high (ICC = 0.98). The mean error between the actual and calculated measurements was 0.4; when we added a two-standard-deviation safety margin around the mean error, the difference fell to between −1.4 to 1.0 mm.
There is no standard nomenclature to describe an incorrectly sized radial head implant. The insertion of an implant that is too thick in the axial plane has been described as "overstuffing of the radiocapitellar joint" or "overlengthening of the radius."9,12-14,17 When an implant that is too thick is inserted, the overall length of the radius increases13,17; therefore, we prefer to use the term "overlengthening." It has been proposed that insertion of an implant that leads to radial overlengthening accelerates the development of posttraumatic radiocapitellar arthritis, restricts the range of elbow motion, and alters elbow kinematics8,9,11. Van Glabbeek et al.9, in a biomechanical study, reported substantial increases in radiocapitellar contact pressures with as little as 2.5 mm of overlengthening. Therefore, the ability to prevent and diagnose small degrees of radial overlengthening is clinically important.
Several intraoperative anatomical features have been described as assisting with the selection of a correctly sized radial head implant4,12-14,17. Doornberg et al.12 performed a computed tomography (CT) imaging study on seventeen patients to examine the relationship between the lateral edge of the coronoid process and the length of the radial head; they reported that, on average, the radial head was 0.9 mm more proximal than the lateral edge of the coronoid. Unfortunately, this anatomical relationship demonstrated a high degree of variability, with measurements ranging from −2.7 to +0.4 mm. Doornberg et al. recommended placement of the implant's articular surface even with or just slightly more proximal than the lateral edge of the coronoid. A limitation of this CT-based study was that cartilage thickness was not accounted for and forearm rotation was not controlled; both variables may have contributed to the high degree of variability found in the results.
Van Riet et al.14 also examined the relationship of the lateral edge of the coronoid to the radial head. In the eight cadaver specimens that they examined, with the forearm in neutral rotation, the distance from the stump of the fractured radial neck to the lateral edge of the coronoid equaled the thickness of the native radial head. Van Riet et al. concluded that this measure could be used reliably in cases in which the radial head was absent.
In another cadaver study, Frank et al.13 examined several intraoperative parameters to determine which best correlated with correct implant length. Visual assessment of the lateral ulnohumeral joint was found to be the best way to identify overlengthening intraoperatively. With the native radial head and with a radial head implant of the correct length, the lateral ulnohumeral joint surfaces apposed each other without a visual gap. In cases with ≥2 mm of overlengthening, a visible gap occurred at the lateral ulnohumeral joint, which was a significant predictor of overlengthening (p < 0.05).
As discussed above, several intraoperative parameters have been identified as assisting with the selection of a correctly sized radial head implant. Unfortunately, radiographic parameters that aid in the diagnosis of overlengthening postoperatively remain elusive. Shors et al.17 conducted a cadaver study to determine if radiographs could be used to diagnose overlengthening due to a radial head implant. They examined specimens under four different conditions: medial and lateral collateral ligaments intact, only the medial collateral ligament disrupted, only the lateral collateral ligament disrupted, and the medial and lateral collateral ligaments disrupted. Their findings indicated that alteration of radial implant length by −2 to +4 mm did not result in a substantial change in the radiographic parameters measured.
Frank and colleagues13 also conducted an imaging study to determine if standard radiographs could be used to diagnose overlengthening by an implant postoperatively. They examined parallelism of the medial ulnohumeral joint on anteroposterior radiographs16 and hypothesized that overlengthening would lead to loss of the normal parallelism of apposing joint surfaces. Their results demonstrated that asymmetry of the medial ulnohumeral joint seen on radiographs became significant (p < 0.05) only when there was at least 6 mm of overlengthening, and the authors concluded that medial joint asymmetry was a relatively insensitive radiographic sign.
Moon et al.20 examined the ability of an overlay device to measure native radial head and neck lengths on anteroposterior elbow radiographs. They compared the use of the overlay device with measurement techniques relying on the bicipital tuberosity. They found the overlay device to be a reliable tool for assessing radial head and neck length; however, use of the technique was examined only in intact, native elbows.
To our knowledge, no one has reported radiographic criteria that can be used to diagnose radial head implant overlengthening accurately. In the present study, we defined and validated a measurement technique based on radiographs of the contralateral elbow that is sufficiently accurate for diagnosing and calculating the magnitude of radial head implant overlengthening and that can be used reliably by different orthopaedic surgeons. The technique requires bilateral anteroposterior elbow radiographs made orthogonal to the forearm and a commercially available image viewing program with a calibrated length tool. The technique is accurate when radiographs of the contralateral, normal elbow are available. When we studied the side-to-side concordance of elbow joint measurements, the lateral joint measurements demonstrated better correlations than did the medial joint measurements. Thus, the measurement technique utilizes lateral elbow joint landmarks to decrease potential errors introduced by variability in medial elbow joint measurements.
The limitations of this study include those that are inherent to any cadaver study, the small sample of specimens, and the fact that only the overlengthened state was examined. Also, radiographs were made in a controlled manner to limit variability, which may be difficult in a busy clinical setting. In this study, radiographs were obtained at 45° of elbow flexion because flexion contractures are clinically common after radial head arthroplasty and 45° of flexion should be attainable by most patients. Similarly, the forearm was imaged in neutral rotation as some degree of supination and pronation loss typically occurs clinically. For the described technique to work, the contralateral elbow must be normal, without osseous deformity or arthritis, to function as a surrogate. In addition, the technique led to overestimations of radial head size at higher degrees of overlengthening. The technique was effective for correctly predicting the size of the implants when there was 2, 4, or 6 mm of overlengthening. With 8 mm of overlengthening, however, the ulnohumeral joint started to subluxate and translate, which resulted in overestimation of implant size by a mean of 0.97 mm. This overestimation, however, has little clinical relevance as the technique still appropriately identified the radial head as being overlengthened. Finally, this method is effective in identifying overlengthening on radiographs, but we cannot comment on the magnitude of overlengthening that is clinically relevant or make recommendations about the indications for revision surgery.
The primary goal of this study was to identify a method for diagnosing and estimating the magnitude of radial head implant overlengthening postoperatively, as previous methods were found to be ineffective. The decision to revise a radial head implant, however, should not be based on the radiographic diagnosis of overlengthening; it must be based on clinical findings such as pain and stiffness. The described measurement technique, based on radiographs of the contralateral, normal elbow, was found to be accurate in predicting radial head implant length. Although this method was effective for diagnosing implant overlengthening in the postoperative situation, the focus should remain on preventing this problem by inserting an implant of correct length at the time of the primary surgery. Surgeons responsible for the treatment of patients with elbow fractures should be familiar with the described intraoperative criteria12-14 that assist with determination of the correct radial head implant length.