Study Design
This study is an institutional review board-approved, single-center, nonblinded, prospective randomized clinical trial (10-582, protocol) and is registered with Clinicaltrials.gov (#NCT01430468). All subjects provided written informed consent prior to enrollment and randomization.
All consecutive patients who underwent primary anatomic total shoulder arthroplasty and agreed to obtain follow-up computed tomographic (CT) imaging within the first four months after surgery were considered to be eligible and were asked to participate. Consent was obtained by the surgeons (J.P.I., J.J.B., and P.J.E.). Forty-nine patients were eligible and five declined to participate because of the requirement to obtain postoperative imaging, resulting in forty-four patients enrolled in the study. Three experienced attending shoulder surgeons from the Cleveland Clinic (J.P.I, J.J.B., and P.J.E) participated. All three surgeons were familiar with use of ArthroPlan software (Cleveland Clinic, Cleveland, Ohio) and participated in a preclinical study on the Sawbones model (Pacific Research Laboratories, Vashon Island, Washington) (unpublished data).
Randomization
Randomization occurred four weeks prior to surgery with use of a block-of-five randomization method. Neither the consenting physician (M.D.H.) nor the patients were aware of the block size, order of randomization, or allocation at the time of consent. No patient refused to participate or was reassigned after randomization. Randomization bias by assignment was eliminated with use of a non-surgeon investigator (M.D.H.).
Sample Size
Previous experience with this new technology and standard instrumentation11 demonstrated that a patient population of thirty (fifteen patients in each group) would be sufficient to provide 80% power for the study parameters.
Surgical Methods and Preoperative Planning
Forty-four patients were enrolled in the study. Of these, thirteen were omitted from this analysis: ten did not undergo surgery for medical or personal reasons, and three underwent reverse total shoulder arthroplasty. The remaining thirty-one patients were randomly assigned into one of two groups: (1) the glenoid positioning system group, which used patient-specific instruments for the placement of the guidewire, the novel surgical simulator for preoperative planning13,14 (ArthroPlan), and standard preoperative anteroposterior (Grashey view) and axillary radiographs; or (2) the standard surgical group, which used the instruments provided by the manufacturer for placement of the guidewire, preoperative two-dimensional CT reformatted to be perpendicular to the plane of the scapula, and the same preoperative radiographs for preoperative planning. It has been shown that most two-dimensional CT scans result in errors in measuring actual version or inclination, because the plane of image acquisition (gantry angle) can deviate from being perpendicular to the plane of the scapula15,16. To avoid this as a source of error in the standard surgical group, the study protocol required the original two-dimensional images, in the DICOM (Digital Imaging and Communications in Medicine) format, to be reformatted into three dimensions along the plane of the scapula with use of ArthroPlan. The plane of the scapula was defined by three points (the center of the glenoid fossa, the scapula trigonum, and the inferior angle of the scapula) (Fig. 1). Reformatted two-dimensional images were defined to be perpendicular to the plane of the scapula for axial and sagittal images and parallel to the plane of the scapula for coronal images. For the standard surgical group, the surgeons (J.P.I., J.J.B., and P.J.E.) measured the glenoid version on the mid-axial, two-dimensional image as described by Friedman et al.17 and used the same image for selection of the glenoid size and type. Glenoid inclination was measured on the mid-coronal image as the angle between a line perpendicular to the glenoid center line (defined as the line from the center of the glenoid to the os trigonum) and a line from the superior to inferior rim of the glenoid (Fig. 1).
The available implants were the same for all surgeons in both groups. Either a multiple-peg, all-polyethylene glenoid with a uniform thickness backside (standard poly component) or a multiple-peg, all-polyethylene glenoid with a 3-mm, 5-mm, or 7-mm augmentation on one side of the backside (augmented poly component) was used for patients judged to have asymmetric bone loss. In the standard surgical group, the manufacturer (DePuy, Warsaw, Indiana) provided a guide-pin insertion guide that used a freehand method of pin placement that was expected to result in a component that was perpendicular to the plane of the scapula in version and inclination. This freehand method involved placing the glenoid sizing disc on the glenoid and marking the estimated center of the glenoid vault with a cautery. This trajectory of the guide pin was selected because it is the average position of the normal glenoid17-20 and is a commonly accepted standard for glenoid component positioning21-24.
After completion of the entire study, the surgeons (J.P.I., J.J.B., and P.J.E.) performing surgery on the standard surgical group used ArthroPlan to virtually place the glenoid component that they used at the time of surgery into the zero version and zero inclination position. The placement of the glenoid component was done after surgery in the standard surgical group to prevent surgeon bias with regard to use of the software prior to surgery in this patient group. In addition, the surgeons were asked to retrospectively decide whether they would use the standard or augmented poly component.
Before surgery the surgeon individualized the component type and its position to the patient anatomy for the glenoid positioning system surgical group (Fig. 2). Premorbid glenoid anatomy was determined in the pathologic condition with use of the vault model method13,14. The shape and quantification of the normal glenoid vault have been previously described25. An averaged, volumetric reconstruction was used to create a standardized, three-dimensional glenoid vault model, which is a highly conserved shape across normal individuals and can be used to estimate native glenoid version. On the basis of the vault location, the glenoid component was placed in the patient’s native version and inclination. The selected position and type of component used in the simulator were compared with the final component position by means of the postoperative three-dimensional CT scan.
For the glenoid positioning system group, patient-specific instruments were created within ArthroPlan. The patient-specific instruments were based on the selected glenoid component as well as its position in and surface anatomy of the pathologic glenoid (Fig. 3). The patient-specific instruments engaged a defined position on the surface of the glenoid to insert two parallel guide pins: a center guide pin on the glenoid fossa, which was used to ream the glenoid in the desired version and inclination angle, and a second guide pin placed in the base of the coracoid process (Figs. 3-E and 3-F). Three to five versions of the first patient-specific instrument were provided in order for the surgeon to select the best-fit option at the time of surgery. The second patient-specific instrument was placed into the peripheral drill guide and was used to control rotation by rotating the guide so that the arm of the second patient-specific instrument contacted the coracoid pin (not shown). Therefore, by using the first and second patient-specific instruments in sequence, patient and implant-specific information for five degrees of freedom for component location was exactly prescribed (the sixth degree of freedom, reaming depth, was not controlled in this study).
Patient-Specific Instrument Design, Fabrication, and Sterilization
After preoperative planning and patient-specific instrument design was complete in ArthroPlan, the patient-specific instruments were manufactured as stereolithography devices (Astro Manufacturing & Design, Eastlake, Ohio) and were sterilized for surgery.
Operative Technique
The technique for surgical exposure and the use of instruments for bone preparation over a guidewire were identical between the glenoid positioning system group and the standard surgical group. The routine surgical and postoperative care was also the same between the two groups.
The surgical steps using the patient-specific instruments in a glenoid positioning system surgery on the right total shoulder for an anatomic poly component are shown in Figure 3. In all cases, the surgeons (J.P.I., J.J.B., and P.J.E.) used the patient-specific instruments and observed an accurate fit between the bone and the patient-specific instruments.
The Software Measurement of Glenoid Parameters and the Preoperative Planning Stage
In both surgical groups, a postoperative CT scan of the shoulder was performed within two weeks after surgery. Both the preoperative and postoperative CT scans were taken with the patient lying supine with the arm at the side and the scapula imaged by means of a 64-detector CT scanner (Siemens Sensation 64; Siemens Medical Solutions USA, Malvern, Pennsylvania). The acquisition parameters were 120 kV, 140 mAs, 0.6-mm collimation, 512 × 512 matrix, no gantry tilt, and 50-cm field of view. The field of view of each scan included the entire scapula. Images were reconstructed with use of a semi-smooth algorithm, B40, at 0.6-mm increments in the axial plane. Postoperative CT scans were made by means of the metallic artifact reduction technique to allow for more precise measurements of the final implant placement in the registration software.
Outcomes Measured and Data Analysis
In both surgical groups, after surgery the actual location of the component was compared with the desired location by the following method. The postoperative CT was loaded into custom software, allowing the postoperative scapula to be registered to the preoperative scapula with use of mutual information image registration26,27, and was placed back into ArthroPlan. A non-surgeon investigator (E.J.R. or J.A.B.) who had been blinded to the randomization group then superimposed the virtual implant over the postoperatively imaged implant in a best-fit fashion, allowing differences between the planned and actual implant positions to be measured by the position of the pegs of the anatomic glenoid. The total offset is the displacement of the implant centroid between the planned and actual implant positions. The medial-lateral, superoinferior, and anteroposterior offsets are the components of the total offset in each of the coronal, sagittal, and transverse planes, respectively. Inclination and version differences were measured as the angle of the implant axis defined by the center peg of the implant between the planned and actual implant orientation, projected to the coronal and transverse planes, respectively (Fig. 4). These calculated differences match the anatomic definition of version and inclination for the glenoid. Differences in anteroposterior, superoinferior, and medial-lateral offset were expressed in millimeters, and those in roll, inclination, and version were expressed in degrees. This method of volumetric registration with and without metal implants has been validated in patients, Sawbones models (Pacific Research Laboratories), and whole-body cadaver shoulders implanted with both anatomic and reverse shoulder components. The registration and repeated measurements demonstrate accuracy and reproducibility to be <3° and <1 mm.
Adverse Events
Clinical complications of any type that were related or were not related to surgery and those of component malposition that required revision surgery were reported.
Statistics
Data were statistically analyzed with use of JMP version 9.0.0 software (SAS Institute, Cary, North Carolina). The absolute difference between the actual outcome and the planned outcome was compared between the standard surgical group and the glenoid positioning system group with use of a Student t test for continuous data and a Fisher exact test for categorical data. Results were considered to be significant at p < 0.05.
Source of Funding
This study was supported by a State of Ohio Third Frontier technology development grant TECH 09-073. The funds were used to pay all non-standard-of-care imaging, statistical analysis, the salaries for two of the authors (J.A.B. and E.J.R.), and a stipend for one author (M.D.H.). Dr. Iannotti is an inventor of the ArthroPlan software and the patient-specific instrument technology. The intellectual property is owned by the Cleveland Clinic. The Cleveland Clinic licensed the ArthroPlan software to Zimmer after completion of this study and Dr. Iannotti receives royalty income through the Cleveland Clinic. There was no financial support from Zimmer for the performance of this study or for the preparation of this manuscript.
Study Population
Overall, there were no significant differences in patient demographic characteristics between the two groups.
For all study patients, the average preoperative retroversion was −13.0° ± 11.8° and the average inclination was 8.3° ± 5.9°. In the standard surgical group, the average preoperative glenoid retroversion (and standard deviation) was −11.3° ± 13.3° (range, −38.9° to 16.6°). In the glenoid positioning system group, the average preoperative glenoid retroversion (and standard deviation) was −14.8° ± 10.1° (range, −27.4° to 6.7°) (see Appendix).
Planned Versus Actual Component Placement
Table I depicts the average deviation (planned versus actual) from the ideal preoperative planned component placement for the standard surgical and glenoid positioning system groups for all outcome measures. The use of glenoid positioning system technology resulted in a significant improvement in the accuracy of glenoid component placement for inclination and medial-lateral offset (p < 0.05). The glenoid positioning system technology also resulted in improved correction of version (p = 0.11) when compared with the standard surgical group technology. The average deviation in version from the preoperative plan (and standard deviation) was 6.9° ± 4.4° (range, −14° to 11°) in the standard surgical group and 4.3° ± 4.5° (range, −8° to 12°) in the glenoid positioning system group. The average deviation in inclination (and standard deviation) was 11.6° ± 7.0° (range, 1.0° to 24°) in the standard surgical group and 2.9° ± 3.4° (range, −13° to 6°) in the glenoid positioning system group (Table I).
Differences in the accuracy of glenoid component placement were also compared among the three participating shoulder surgeons (J.P.I., J.J.B., and P.J.E.). There was some variability among surgeons, with the most experienced surgeon (J.P.I.) gaining the greatest benefit from the technology (data not shown).
The results of each individual patient were plotted on the basis of the severity of disease (preoperative retroversion) (Fig. 5). As the severity of retroversion increased, there was a trend in both groups toward a less accurately placed glenoid in version (Fig. 5-A) and inclination (Fig. 5-B). However, there was a noticeable difference in the accuracy of placement between the standard surgical group and the glenoid positioning system group as the severity of preoperative retroversion increased. A subset analysis of the most retroverted and the least retroverted glenoids (Table II) in both groups demonstrated a significant improvement in glenoid placement for the glenoid positioning system group in the most retroverted group (see Appendix).
There was a significant difference in the selection of the optimal implant type between the standard surgical group and the glenoid positioning system groups. In ten patients (59%) in the standard surgical group, the surgeons (J.P.I., J.J.B., and P.J.E.) postoperatively selected a different implant from the one that had been selected preoperatively and used at surgery. The choices in implant were between the use of a standard glenoid and that of an augmented glenoid. This selection of a different implant occurred only in one patient (7%) in the glenoid positioning system group (Table III).
On the basis of previously established criteria, a component was considered malpositioned if the component position deviated in excess of 10° from the planned optimal position24,28. With use of these criteria, the use of glenoid positioning system technology significantly reduced the number of malpositioned implants in version and/or inclination from 75% (twelve cases) of the standard surgical group to 27% (four cases) of the glenoid positioning system group (Table III).
Adverse Events
One patient (5.9%) in the standard surgical group experienced a transient partial axillary nerve injury. No patient in the glenoid positioning system group had an adverse event.
This is a prospective, randomized clinical trial designed to test the hypothesis that the use of novel three-dimensional preoperative planning surgical simulation software and patient-specific instrumentation is superior to standard preoperative planning and generic surgical instrumentation for the placement of glenoid components. We could not define if the differences between the glenoid placement of the standard surgical group and that of the glenoid positioning system group were clinically relevant, and this study was not designed to define clinical importance. We did define the ability of this technology to improve component position, particularly in cases in which there was more severe deformity. Further study would be required to define if improved clinical outcomes are associated with improved component positioning.
Although similar technology is commercially available for knee replacement, it has not undergone clinical testing with a control group in a randomized trial. We believe that this is the first Level-I clinical trial to test this type of technology in any joint replacement. Iannotti et al. demonstrated more accurate component placement with use of standard instrumentation and preoperative planning when there was <10° of preoperative glenoid retroversion11. Our data supported the use of glenoid positioning system technology for glenoid component placement, particularly in patients who had more severe bone deformity. When the standard surgical group was compared with the glenoid positioning system group, no differences were observed among patients who had <7° of retroversion, but a significant improvement was observed in the accuracy of component placement in patients who had bone deformity and retroversion in excess of 16°.
With use of previously established criteria for a malpositioned glenoid24,28, glenoid positioning system technology reduced the number of cases with deviations in glenoid positioning >10° in the plane of version or inclination. Malpositioned glenoid components in the glenoid positioning system group can be related to several factors that could limit the precision provided by the technology. First, the software must accurately define the patient’s osseous anatomy. It is necessary to accurately separate the humeral head from the glenoid surface to render an accurate patient-specific instrument. Defining and separating are difficult with loss of cartilage and severe bone deformity. This problem was best managed by means of software algorithms capable of achieving an accurate separation and therefore an accurate patient-specific instrument. The ability to render accurate models of the bones is also dependent on the quality of the primary data acquisition and obtaining images of the entire scapula. Second, the surgeon must remove all soft tissues from the glenoid to represent what was imaged by the software. Segmentation of the bone in the software can remove or not remove calcified or ossified portions of the glenoid rim or labrum, and these same structures may or may not be removed by the surgeon. Third, the surgeon must provide sufficient exposure of the glenoid to allow for proper seating of the patient-specific instrument. Lastly, the surgeon must ream the glenoid to the proper depth and precisely in line with the central guidewire to avoid angulation of the reamer, which can still occur with a 2.5-mm guidewire. In every glenoid positioning system surgical case in this study, the surgeons (J.P.I., J.J.B., and P.J.E.) felt confident that the bone was exposed correctly, the patient-specific instrument fit very well on the bone, and they reamed in line with the guidewire. Despite these claims, we cannot rule out the influence of some or all of the above factors in the four glenoid positioning system cases that had >10° of variation in version or inclination. All cases that deviated >10° in the glenoid positioning system group were from the surgeons having the least experience with this technology. The accuracy of the technology is limited by these factors.
In this study, for the standard surgical group, we planned to place the glenoid in 0° of version and inclination with respect to the plane of the scapula. This position was selected because it is commonly used in surgical practice, and in this study, it was required that we establish a component position preoperatively for the standard surgical group. In the normal population, glenoid version varies widely from 5° of anteversion to 15° of retroversion, with an average preoperative retroversion range from 1° to 6° depending on the source materials and methods of measurement17,20. It is not currently a standard of care to place the glenoid component in the patient’s native premorbid version if that information is obtainable. We have shown in a few prior studies that there is a wide range of native version as determined by the vault model method in both the normal and pathologic condition13,14,25,29. This technology allows the surgeon to more accurately define the patient’s premorbid version from the pathologic state to allow placement of the glenoid component in a position individualized to the patient’s native normal anatomy. The vault model method was the method used to position the glenoid in patients in the glenoid positioning system surgical group. Using the patient-specific version would result in less frequent overcorrection of the patient retroversion, less over-reaming and medialization of the glenoid joint line, and fewer occurrences of peg perforation of the glenoid vault. In cases of extreme posterior bone loss, surgeons were given the option to correct asymmetric posterior glenoid wear with the use of an augmented posterior glenoid component for both groups of patients. It would have been difficult, if not impossible, to correct more severe bone loss without use of the augmented component. In this study, the same surgeons (J.P.I., J.J.B., and P.J.E.) were less likely to select an augmented glenoid for correction of increased glenoid retroversion in the standard surgical group. We believe that this is a result of the inability of the same surgeons to define the pathology and need for this type of glenoid when using standard two-dimensional CT scan imaging when compared with three-dimensional imaging and preoperative planning software with use of the vault model technology.
This study used a freehand method for the placement of a guidewire in patients in the standard surgical instrument group. Recently, some implant manufacturers have supplied fixed or variable angle guides to assist the surgeon in placement of a guidewire. Each of these different methods of placing the guidewire may show improvements in implant placement over the freehand method but have not been reported in the literature.
Computer-assisted surgical techniques are becoming increasingly common for hip8,30-33 and knee30,34-40 arthroplasty, but remain limited in shoulder arthroplasty10,41. To our knowledge, there have been limited instances of the use of patient-specific instruments in the spine42 and knee40,43,44, even fewer in the hip, and none in the shoulder.
In conclusion, this study demonstrated that the use of preoperative planning software and patient-specific instrumentation is not only feasible but also adds accuracy to glenoid component placement. The accuracy of the new technology is dependent on imaging, the surgeon, and the associated pathology.
Disclosure: One or more of the authors received payments or services, either directly or indirectly (i.e., via his or her institution), from a third party in support of an aspect of this work. In addition, one or more of the authors, or his or her institution, has had a financial relationship, in the thirty-six months prior to submission of this work, with an entity in the biomedical arena that could be perceived to influence or have the potential to influence what is written in this work. Also, one or more of the authors has had another relationship, or has engaged in another activity, that could be perceived to influence or have the potential to influence what is written in this work. The complete Disclosures of Potential Conflicts of Interest submitted by authors are always provided with the online version of the article.