Thirty-six paired human cadaveric scapulae (age: 60.9 ± 7.5 years; bone mineral density: 0.476 ± 0.163 kg/m3; female, n = 19 - male, n = 17) were cleaned of all soft tissues, and allocated into three groups relative to their size. For this purpose, the scapulae were inspected by the senior surgeon (ADM) and allocated into the three groups according to the best fitting glenoid baseplate guide (small, n = 12; medium, n = 12; large, n = 12). Bone mineral density was measured (dual-energy x-ray absorptiometry) for all scapulae at a 1x1cm region (anterior to posterior) of the inferior third of the glenoid. There was no difference in bone mineral density between glenoid size (small, medium, large) or glenoid positioning groups (inferior vs. superior, during load-to-failure testing). All specimens were thawed overnight and experiments were performed at room temperature within the next 24 h. Specimens were obtained from Medcure Inc. (Portland, OR).
A current reverse arthroplasty system (Univers Reverse II, Arthrex, Naples, FL), used by the senior author in his clinical practice, was used for this study. The glenoid baseplate of this system features an anatomic baseplate shape, as opposed to a circular design, a central peg and screw (6.5 mm) combination, and two (inferior and superior) variable angle screws (4.5 mm). Baseplates of three sizes (small, medium, large; height: 32 mm–38 mm, width: 24.4.mm-30.4 mm, thickness: 4 mm) were implanted according to glenoid size in the three groups. In each group (n = 12), baseplates were either implanted in an inferior position (n = 6), flush with the inferior glenoid rim, or were placed in a superior position (n = 6), defined as 5 mm superior offset from the inferior glenoid rim. The offset was measured and marked using a digital caliper with 0.01 mm resolution (Mitutoyo, Japan).
All surgeries were performed by an experienced, fellowship trained shoulder surgeon (A.D.M.) to reduce performance bias. Similar to the actual in-vivo surgical technique, a guide wire was placed first to ensure correct positioning of the glenoid implants. The goal was to place the baseplate in 0 to 10° retroversion and neutral inclination according to the surgeon’s judgement and clinical practice. Following over-drilling the guide wire to fit the central peg (12.1 mm diameter), the glenoid surface was reamed down beyond the sclerotic zone, the baseplate was aligned with glenoid rotation, impacted, and fixed with a 6.5 mm diameter central screw (15 mm, 20 mm, 25 mm lengths). Two additional 4.5 mm locking screws (24 mm–48 mm length with 4 mm increments) were used at the superior and inferior position. The maximally possible screw lengths to avoid any bone perforation were chosen by measuring with a depth gauge. Specimens were trimmed and potted in custom boxes with the baseplate-surface parallel to the floor in all planes. Scapulae were then mounted in a servo-hydraulic mechanical testing apparatus (MTS, Eden Prairie, MN) to simulate a 60° glenohumeral angle between the glenoid and actuator holding the humeral components (neutral metal cup and 3 mm polyethylene liner) in a custom fixture, which were available in three different sizes (36 mm, 39 mm, 42 mm) corresponding to available glenospheres. The angle for testing was chosen according to previous studies, showing peak shear forces occurring at 60° of abduction [11, 12].
Four stacked 0–45-90° rosette 350 Ohm strain gauges (Vishay, Raleigh, NC) were placed on the scapula 5 mm medial to the glenoid surface at the corresponding 12, 3, 6, and 9 o’clock positions. Strain gauges were fixed to the bone with a thin layer of cyanoacrylate adhesive (Loctite® 411, Henkel, Düsseldorf, Germany) and covered with tape for 24 h before testing. (Fig. 1) A 12-channel strain data recorder (Model 5100 Scanner) was used to log readings of all rosettes simultaneously at 20 Hz (StrainSmart 4.31 Software). Principal strain values were calculated for each rosette. The strain gauges were re-calibrated between each round of testing.
Three pilot studies were performed to verify the accuracy of the strain gauges and get an understanding of the interaction between strain and load values. First, the accuracy and reproducibility of the strain gauges on a homogeneous surface were tested. Two strain gauges were placed on a 1″ polyvinylchloride pipe and a compressive load of 750 N was applied repeatedly (five times) in axial direction on a mechanical testing system (MTS, Eden Prairie, MN), where the strain gauges provided a consistent strain output (sensor 1: 219 ± 4.4, sensor 2: 220 ± 2.2). Average error between the sensors was 0.99 (0.45%) in reading strain under the repeatedly applied 750 N load. Second, to ensure that sequential measurements with different glenospheres were not influenced by the manual exchange of components, a pilot series of repeated strain measurements was performed. Between each trial, the glenosphere was removed and repositioned on the baseplate. The strain gauges were re-calibrated between each repeated measurement and strain values were compared between trials. This process revealed consistent strain readings with an average error of 0.50 (1.16%) across repeated trials and all strain gauge positions.
Third, in an effort to make the concept of strain easier to understand and put into context, strains during load-to-failure were evaluated in order to compare strains encountered during activities of daily living with what failure strain is. A load-to-failure test (n = 12) was conducted by applying a constant compressive force at a rate of 0.2 mm/sec and logging the peak load. Peak loads at failure ranged from 1537.65 N to 4150.30 N, while principal strains at failure were 6087.25 ± 2189.19 in this small pilot cohort. This last pilot study revealed that increasing strain values were consistent with increasing macroscopically visible osseous deformity and location of failure. In combination, these results suggested that the strain gauges were able to adequately register supraphysiological bone strains indicative of component or bone failure.
Testing sequence
Three different glenosphere sizes (36 mm, 39 mm, 42 mm diameter) were available with three different designs each (standard, 2.4 mm inferior offset, 4 mm lateralized). The sequence for testing was chosen at random for different glenosphere sizes as well as designs to avoid any possible bias from micromotions caused by specific baseplate-glenosphere combinations. The glenosphere was placed on the baseplate and secured with a single hammer tap using a dedicated impactor. A compressive load of 100 N was applied for 60 s to center the glenosphere under the humeral component. Then a load of 750 N, which has been used in various prior studies [2, 11,12,13,14,15], was applied for 10 s to register bone strains. The glenospheres were then removed, re-inserted, and the testing sequence was repeated for a total of three times. Mean values were calculated for further statistical analysis. After completion of three test runs, the next glenosphere design was chosen randomly and the process was repeated until all glenosphere types were tested. Then the humeral components were removed and the next size of glenosphere-humeral cup combination was mounted for further testing. According to the possible glenospheres that could be used with the various baseplate sizes, this resulted in a total of 27 test runs for small, 18 for medium, and 9 for large baseplates, respectively (Fig. 2.). The incidence of abutment of the humeral component at the scapular neck was documented for all baseplate-glenosphere-positioning combinations. In case of abutment, the subsequently obtained strain values were excluded from statistical analysis under the assumption that the inferior strain gauge becomes damaged by the engaging humeral component. Following strain testing, a lateralized glenosphere (to avoid abutment as mode of failure) was mounted in the size of the baseplate and a load-to-failure test was performed by applying a constant compressive force at a rate of 0.2 mm/sec while documenting peak load at bone or component failure as well as the mode of failure. (Testworks 4, Eden Prairie, MN). The biomechanical testing setup is depicted in Fig. 3.
Statistical analysis
Descriptive statistics were used to present demographical data of specimens. One-way analysis of variance (ANOVA) with Tukey post-hoc test was used to evaluate differences between three or more independent groups. Linear mixed effects regression was used to generate mean values and 95% confidence intervals for principal strain across conditions (baseplate size, glenosphere size, glenosphere design, and baseplate position), with a random effect at the specimen level. This approach accounted for the correlation between shoulders as specimens were used in multiple conditions. A custom matrix was created to illustrate the incidence of humeral component abutment based on the various baseplate-glenosphere combinations with different baseplate positions. The Chi2-test was used to compare the incidence of abutment between baseplate positioning and baseplate-glenosphere combinations. Confidence intervals of proportions or counts were calculated according to the modified Wald method [16]. The alpha level for all statistics was set at 0.05. All statistical analysis was performed using Stata 14 (StataCorp. 2015. Stata Statistical Software: Release 14. College Station, TX: StataCorp LP).
