Introduction
Damage to the rotator cuff often leads to glenoid osteoarthritis with characteristic erosion patterns. Depending on the location of erosion, the glenoids are subdivided by Favard et al. into five types, ranging from E0 to E4 [1, 2, 3]. According to the typology, E2 types show “glenoid erosion predominantly in the superior pole [4].” Impairment of individual rotator cuff muscles results in changes to the force balance in the glenohumeral joint, which leads to altered humeral head translations and overtime yields distinct erosion patterns [5, 6, 7, 8].
Areas with increased force, according to Wolff's law, undergo higher degrees of bone remodeling and thus exhibit higher bone mineral densities [5, 6, 9, 10, 11, 12, 13, 14].
The severe superior bone loss of type E2 glenoids consequently is responsible for challenges in reverse shoulder arthroplasty (rTSA) as the preferred treatment option [15, 16, 17]. Operative methods such as eccentric reaming, bone grafting and augmented glenoid baseplates address the bone deficiency to decrease postoperative complications and maximize functional results [5, 8, 16, 17, 18, 19, 20, 21]. Studies such as Boileau et al.'s have demonstrated that using bone grafts are effective in correcting large superior deficiencies [17].
However graft resorption remains a potential complication that poses a risks for implant loosening [22, 23, 24, 25, 26]. Consequently, recent literature has increasingly focused on various factors that may influence primary implant stability, aiming to better understand underlying mechanisms and to develop strategies to minimize these complications [5, 13, 27, 28, 29, 30].
According to Spek et al., no single strategy exists for optimal anchoring and “therefore, guidelines should cover multiple surgical options that can achieve adequate baseplate fixation” [31]. For example, literature consistently indicated that the successful anchoring of augmented glenoid implants in dense bone lowers the risk of implant loosening and the need for revisions [27, 32, 33, 34, 35, 36, 37, 38]. The 2025 study by Ritter et al. postulated that a succinct preoperative bone density analysis is necessary for informed decision making regarding the choice of implant and deformity corrections in rTSA [27].
Therefore, exploring parameters, such as bone mineral density as a cause for implant loosening, is particularly relevant for severe superior glenoid deficiencies, as is the case in E2 and E3 types, since they are effectively treated with rTSA [4, 5, 8, 23, 39, 40]. Recent studies have therefore increased their focus on analyzing bone mineral densities in the shoulder joint [5, 6, 27, 30, 35, 36, 37, 38, 39, 41, 42, 43, 44, 45, 46].
For example, Eghbali et al. focused on densities in both healthy [43] as well as pathological glenoids [39]. They evaluated the effect of primary shoulder arthritis and cuff tear arthropathy (CTA) on bone density and how it is influenced by age, sex, and body mass index [39]. On the basis of identifiable density variations between these distinct shoulder pathology groups, they stressed the importance of individualized preoperative planning.
However, their indiscriminate focus on cuff tear arthropathy in general, asks for a differentiated look on various subtypes of CTA.
Letissier et al. and Knowles et al. investigated densities for erosion patterns according to the Walch classification [6, 37]. In 2025, Ritter et al. investigated if bone density and the degree of augmentation influences baseplate stability in reverse shoulder arthroplasty [27]. One of their findings was that lower bone density was associated with higher micromotion and decreased baseplate stability.
They also stated that information on bone density derived from preoperative CT‐scans may guide surgeons whether metal or bone should be used to correct angular deformities. Harada et al. explored glenoid densities of shoulders with symptomatic rotator cuff tears by using a grid that divides the glenoid into seven regions for analyses [47]. This grid comes close to the reported consensus in literature regarding screw placement in rTSA [31].
Their study also showed that shoulders with symptomatic subscapularis and infraspinatus tendon tears had reduced bone densities in central glenoid areas [47]. A study by Anetzberger et al. performed subchondral density analyses of shoulders with rotator cuff damage [48]. However, their study did not differentiate between erosion types nor investigate glenoid regions below the subchondral glenoid.
Each of these studies converge in describing significant density variations across the glenoid of shoulders with rotator cuff damage but mostly lack focus on specific pathology subtypes. Despite the wide variety of approaches to density issues in literature, only a limited amount of studies addressed densities specifically in Favard type glenoids [5, 13].
Mahaffy et al. were among the first to analyze densities of the distinct E2 type erosion by covering multiple layers to a total depth of 10 mm into the glenoid vault [5]. Using a quadrant‐based grid that is compatible with the Favard classification system, they were able to correlate their results to this specific erosion typology. Mahaffy et al. showed significant density differences between the surface areas and deeper parts of the glenoid vault.
Although not a density study per se, but indirectly relevant, is the study by Abdic et al. that quantified erosion orientation and showed that the E2 type erosion pattern did not solely occur in the superior quadrant as previously described but also extended to the neighboring posterior quadrant [8].
To cover the other major Favard subtype, in addition to E2 glenoids, Klarer et al. investigated bone density distributions in E3 types [5, 13]. Their results showed higher densities in the periphery of the superior‐anterior regions for E3 types. These findings suggest revisiting E2 type densities from a perspective that, unlike Mahaffy et al., differentiates between central and peripheral glenoid regions.
Additionally, such an approach would permit to correlate density findings with the recently reported superior‐posterior erosion orientation in E2 types as described by Abdic et al.—results that deviate from the original Favard definition of a distinct superior erosion [1, 3, 4, 8]. Analyzing densities and comparing them to other anatomical parameters, such as glenoid erosion, which the Favard classification is based on, may yield new insights into the pathophysiology of cuff tear arthropathy.
For these reasons, the primary goal of the study was to map the bone density distribution across the glenoid by relying on a multifaceted three‐dimensional grid. The study hypothesized that the densities show significant variations from central to peripheral regions of the glenoid and that these differences, if they exist, correlated with the erosion patterns described in literature for E2 Favard type glenoid. To put the first hypothesis to the test the study relied on statistical analyses to detect significant variations.
The second hypothesis was tested by comparing the significant density variations with the descriptions of E2 type erosion orientation in literature.
Methods
To achieve the goal of the study, i.e., to map the bone density distribution across the glenoid, and to test the ensuing first hypothesis that significant variations exist between central and peripheral regions, we relied on methods largely based on those described by Klarer et al. for the analysis of E3 types [13]. The second hypothesis, that is, if high density areas coincide with erosion characteristic in the pathology definitions, is part of the Discussion section since it only applies and discusses the density data in relation to erosion descriptions in existing literature.
This computational study used retrospective, fully anonymized preoperative imaging data and did not involve direct patient contact or intervention. Informed consent was waived due to the retrospective and anonymized nature of the data, in accordance with institutional regulations. Computed tomography (CT) scans of 27 patients were attributed by an experienced shoulder surgeon (senior author) to the Favard E2 type with an age range between 63 and 88 (mean age, 82.5 ± 5.8 years).
The cohort consisted of 8 male and 19 female patients, including 9 left and 18 right shoulders. The CT scans were obtained by using standard clinical settings (120 to kVp, 512 × 512 resolution) and were saved as DICOM files (Digital Imaging and Communications in Medicine).
Each patient's CT scan was imported into Materialse's Mimics software (Mimics V21.0; Materialize, Leuven, Belgium). Soft tissue, the humerus, and other surrounding bone structures were removed to create an isolated scapula segmentation, with particular attention given to accurately segmenting the glenoid. Specifically, manual segmentation techniques were performed to preserve adequate bone volume, extending medially from the glenoid surface, to ensure complete inclusion of the glenoid vault during the segmentation processes [49].
In this study the term “glenoid surface” is used synonymously for the areas covered by the original articular surface together with its immediate surrounding areas. The glenoid was subsequently reconstructed as a 3D computer model.
For each isolated glenoid model, four anatomical points (i.e., superior, anterior, posterior, and inferior) were identified as the most extreme points of the glenoid rim and used as references for all topographical grids. All anatomical placements were manually identified and positioned by a medical doctor (first author) and checked by a biomechanical engineer (second author), who both had been instructed and trained by an experienced shoulder surgeon (last author). The halfway point along the line between the superior and inferior points served as the glenoid center.
A second line was drawn through the glenoid center at a 90‐degree angle to the superior‐inferior line. To align the ensuing four quadrants according to the Favard classifications, the crosshairs were rotated 45 degrees (see Figure 1).
Ten concentric rings were distributed radially out from the glenoid center. Each patient's individual anatomy determined the specific width of the rings, using one tenth of the overall glenoid radius as the distance between each of the rings. Depending on which line was longer, either the superior‐inferior or the anterior‐posterior diameters were used to calculate the ring radii for each patient.
This guaranteed that the customized ring grid covered the maximum glenoid surface for each patient. The ensuing multifaceted grid consists of 40 distinct sections: Quadrants were split into 10 ring segments, with ring 1 at the center and ring 10 at the periphery of the glenoid. This ensured a detailed mapping of the glenoid surface for the subsequent density analyses.
To specify the bone layer for the density analyses, first, a surface remesh function was carried out in Mimics 3‐matic (Materialize, Leuven, Belgium). In a next manual step, each patient's glenoid surface was isolated and saved as a stereolithographic file (STL) to determine the deepest point of each glenoid surface, i.e., the point with the greatest anatomical depth relative to the glenoid rim. A plane created by the transverse axis (connecting the anterior and posterior points) and the inferior point was translated 18 mm medial to the deepest point of the glenoid surface—this 18 mm distance was sufficient to encompass the volume of the glenoid vault in each sample [49].
The second plane formed the medial limit of the region of interest (ROI), that is, the bone located between this plane and the glenoid surface that was used for all subsequent density analyses.
A point cloud of data points was created in Mimics to export the density information of the ROI in Hounsfield units (HU). Each density point was determined by its individual voxel density in HU and its specific voxel position. All density points were analyzed with respect to their location within the 40 segments of the three‐dimensional ring‐quadrant grid.
To differentiate between bone types, values ranging from 100 to 600 HU were assigned to trabecular, whereas values of 600 HU and above were used to identify cortical bone [5, 13, 50, 51].
IBM SPSS Statistics for Windows, Version 29.0 (IBM Corp., Armonk, NY, USA) was used to perform a mixed repeated‐measures multiple analysis of variance (mixed 3‐way RM‐ANOVA) using quadrant, ring, and bone types (trabecular or cortical) as repeated independent variables. Statistical significance was defined as p < 0.05. Descriptive density results for quadrant and ring regions were represented as mean Hounsfield units (HU) ± 1 standard deviation.
Results
Significant multivariate main effects were observed for the within‐subject factors rings and quadrants (p < 0.001), indicating differences across anatomical locations. There was no evidence of a statistically significant difference in the ring‐quadrant interaction between bone types (trabecular vs. cortical) (p = 0.491). Pairwise comparisons were conducted across combinations of ring, quadrant, and bone type levels.
These analyses revealed significant differences between rings within individual quadrants and bone types.
E2 Bone Densities Across Entire Glenoid Grid
Using the entire glenoid grid, without differentiating between rings and quadrants, the overall mean bone densities were 888 ± 135 (Hounsfield units ± 1 standard deviation) for cortical bone and 377 ± 61 HU for trabecular bone.
E2 Bone Densities Across Quadrants
Cortical bone densities by quadrant—ranging from superior (903 ± 137 (HU), anterior (894 ± 132 HU), posterior (885 ± 138 HU) to inferior (871 ± 136 HU)—showed a statistically insignificant decrease in densities across all quadrants (see Figure 2). Even the largest density difference of 32 HU between the superior and the inferior quadrant had no statistical significance (p = 0.154). For trabecular bone densities, the anterior (402 ± 48 HU) and superior quadrants (395 ± 47 HU) had the highest densities and were significantly denser than posterior (352 ± 64 HU) and inferior (358 ± 67 HU) quadrants (p < 0.001) (see Figure 2); The maximum range difference across all four quadrants was between anterior (402 ± 48 HU) and posterior (352 ± 70 HU) quadrants with statistically significant 50 HU (p = < 0.001).
E2 Bone Densities Across Quadrants and Rings
Comparing E2 Cortical Densities Within Individual Quadrants Across Rings
For cortical densities, ring 7 of the superior quadrant showed the highest cortical density (940 ± 121 HU) and thus was the densest region in E2 type glenoid (see Figures 3 and 4). Compared to the other superior lying regions, ring 7 was significantly denser than ring 1 (877 ± 150 HU) (p = 0.011), ring 2 (880 ± 140 HU) (p = 0.011), and ring 3 (871 ± 119 HU) (p = 0.003). Ring 7 was 60 HU denser than ring 2 and 69 HU denser than ring 3.
In contrast to the superior quadrant, the anterior, posterior, and inferior quadrants showed minimal and insignificant density variation between central and peripheral glenoid regions.
In sum, only the superior quadrant showed significant cortical density variations between glenoid regions, with high‐densities in the mid‐range of the periphery, with a slight dip at the very outer rings, and low‐densities centrally.
Comparing E2 Cortical Densities of Each Individual Ring Across the Four Quadrants
Across the first six rings, (i.e., central regions of the glenoid), cortical bone showed no significant density variation between quadrants (see Figure 3).
In rings 7, 8, and 9, significant density variations between quadrants were detectable: For example, superior ring 7 (940 ± 121 HU) was denser (p = 0.026) than inferior ring 7 (869 ± 142 HU); superior ring 8 (932 ± 151 HU) was denser (p = 0.013) than inferior ring 8 (856 ± 138 HU); superior ring 9 (915 ± 149 HU) was denser than (p = 0.007) inferior ring 9 (841 ± 138 HU). Interestingly, superior ring 10 and inferior ring 10 showed no significant difference (p = 0.675). In sum, cortical densities of E2 type glenoids divided the glenoid into a high‐density superior‐anterior and low‐density posterior‐inferior half.
Statistically significant differences, however, only occurred between the peripheral regions of superior and inferior quadrants, in rings 7 to 9 (see Figure 5).
Comparing E2 Trabecular Densities Within Individual Quadrants Across Rings
Trabecular densities in superior and anterior quadrants showed no significant variation between rings (see Figure 6). The inferior‐posterior glenoid half, however, showed significant differences between central and peripheral areas. In the posterior quadrant, central ring 1 (398 ± 50 HU) was significantly denser (p < 0.001) than posterior rings 6 (351 ± 62 HU), 7 (331 ± 57 HU), 8 (320 ± 59 HU), 9 (312 ± 61 HU), and 10 (291 ± 53 HU) (see Figure 7).
Additionally, the posterior quadrant showed further trends. Trabecular densities did not vary significantly over the first four rings, with ring 1 (398 ± 50 HU) and ring 4 (384 ± 52 HU) only showing a difference of 16 HU. However, starting at ring 5 (363 ± 53 HU) to peripheral ring 10 (291 ± 53 HU), posterior rings showed a steady and statistically significant decrease (p < 0.001) in trabecular bone densities (see Figures 6 and 7).
The inferior quadrant shared similarities with the posterior quadrant, with a few exceptions. In the inferior quadrant, ring 1 (415 ± 73 HU) was not only the densest inferior ring but also the densest trabecular ring section of all quadrants. Going peripherally, it was therefore significantly denser (p = 0.028) than inferior ring 2 (382 ± 44 HU), ring 5 (371 ± 51 HU) (p = 0.044) ring 6 (359 ± 63 HU) (p = 0.004).
Between rings 7 (341 ± 62 HU) and ring 10 (300 ± 60 HU) the differences were highly significant (p < 0.001). The overall difference between inferior ring 1 and ring 10 was 115 HU (p < 0.001). Furthermore, trabecular densities of the inferior quadrant showed hardly any variation between the first four rings, with a maximum variation range of 32 HU between inferior ring 1 (415 ± 73 HU) and ring 2 (382 ± 44 HU).
In the periphery, however, beginning at ring 5 (371 ± 51 HU) to ring 10 (300 ± 60 HU) a steady and significant decrease was observed (p < 0.001). Inferior ring 10 was significantly less dense than all other rings (p = < 0.001).
Along with the major differences mentioned above, even though trabecular densities steadily dropped from center to periphery, both inferior and posterior quadrants showed that at every ring step the difference to either its neighboring ring or adjacent rings were statistically significant. For example, inferior ring 2 (382 ± 44 HU) was denser (p < 0.001) than inferior rings 8 (326 ± 63 HU), 9, and 10; inferior ring 3 was denser (p < 0.001) than rings 8 to 10; ring 4 was denser (p < 0.001) than posterior rings 8 to 10; posterior ring 5 (363 ± 53 HU) was denser (p < 0.001) than posterior ring 9 (312 ± 61 HU) and 10; posterior ring 6 was denser (p < 0.001) than posterior ring 10; posterior ring 8 was less dense (p < 0.001) than rings 1 to 4 and less dense (p = 0.004) than rings 5 and 6; posterior ring 9 was significantly less dense (p < 0.001) than posterior rings 1–5 and less dense (p = 0.002) than ring 6; posterior ring 10 was significantly less dense (p < 0.001) than posterior rings 1–8. A similar trend was observed in the posterior quadrant across rings.
Overall, the inferior and posterior quadrants showed similar trends and the most variation in trabecular densities from central towards peripheral glenoid regions (see Figure 8). Both showed a steady density decrease starting at ring 5 towards ring 10, meaning that inferior‐posterior regions of E2 type glenoids demonstrated high degrees of density variation with overall lower trabecular densities than other quadrants. Superior and anterior quadrants showed minimal fluctuations and behaved similarly (i.e., continuous high trabecular densities), across all rings from center to periphery.
Comparing E2 Trabecular Densities of Each Individual Ring Across the Four Quadrants
E2 trabecular densities of all quadrants across rings showed no differences in rings 1 to 4. Statistically significant difference started at ring 5. Superior ring 5 was denser (p = 0.043) than posterior ring 5; superior ring 6 was denser (p = 0.033) than posterior ring 6; anterior ring 6 was significantly denser (p < 0.001) than posterior ring 6 and inferior ring 6 (p = 0.002); superior ring 7 was denser (p < 0.001) than posterior ring 7 and inferior ring 7 (p = 0.013).
Superior ring 7 (331 ± 57 HU) was denser (p = 0.026) than inferior ring 7 (341 ± 62 HU). Similarly behaved: superior ring 8 and inferior ring 8 (p = 0.013); superior ring 9 and inferior ring 9 (p = 0.007); superior ring 10 and posterior as well as inferior rings 10 (p < 0.001); anterior rings 10 and posterior as well as inferior rings 10 (p < 0.001).
In sum, the biggest E2 trabecular bone density differences were between the low‐density posterior‐inferior half and the high‐density superior‐anterior half of the peripheral glenoid (i.e., ring 5 to 10) (see Figure 6).
Discussion
The main focus of this study was, first, to investigate the bone density variations of E2 type glenoids and to test the hypothesis that significant differences across central and peripheral glenoid regions exist. Based on these results, the study also discusses possible correlations between density distributions and the original (Favard et al.) as well as recently published observations of the E2 erosion pattern [1, 2, 8]. Even though this study employed computational methods similar to those reported previously for E3 types [13], it examined the distinct E2 erosion subtype using an independent patient cohort.
The results showed significant regional density differences between central and peripheral regions of the inferior and posterior glenoid.
As a retrospective study, we had to rely on standard routine shoulder CT scans that had been acquired for preoperative planning and were not originally intended for quantitative bone density assessment. Therefore, we could not take into account individual soft‐tissue and systemic bone density differences. Consequently, this study relied on HU for describing bone density using the above‐mentioned scans.
This methodological choice is reinforced by Dhanjani et al.'s results “that both standard and PC [phantom calibrated] HU measurements of the glenoid and proximal humerus significantly correlate with DEXA T‐scores across multiple anatomical sites.” [52] Their findings support using HU assessments from standard CT scans without phantom‐calibration for analyzing bone density [52, 53, 54, 55, 56, 57]. Therefore, like previous studies, we opted for this methodological choice for clinical purposes in favor of reference‐based methodologies [5, 6, 27, 30, 42, 58].
While cortical and trabecular bone types are typically defined in relation to their morphological configurations—namely, trabecular forming the porous inner network and cortical forming the hard outer bone layer—this study deliberately disregarded their relative spatial distribution and structural classifications. Instead, bone types were based solely on density parameter ranges (100 to 600 HU for trabecular; 600 HU and above for cortical), which have been attributed to these types in previous studies [5, 13, 50, 51]. This means that the term “cortical” bone, which normally is not attributed to bone below the subchondral plate, was nonetheless applied in this study to cortical‐like structures that resemble cortical bone according to its density range, even when present in deeper glenoid regions [5, 6, 13, 59].
When using, in a first step, solely a diagonal‐shaped quadrant grid, the results showed that the superior quadrant had the highest cortical bone density, followed by the anterior, posterior and inferior—each quadrant less dense than the one before. For trabecular bone, the densest quadrants were anterior, followed by superior, then inferior, and the posterior as the least dense. Posterior and inferior quadrants shared almost identical trabecular bone densities.
Overall, the quadrant results for cortical bone resembled the ones produced by Mahaffy et al. [5] However, the trabecular density results differed: Mahaffy's densest trabecular quadrant was the posterior followed by superior, then anterior, and inferior as the least dense one. What might be noteworthy for understanding the partly different results is that Mahaffy et al. stated that trabecular bone densities decreased with increasing glenoid depth. One reason why our quadrant results might be different could be that our study used one single layer with a depth of at least 18 mm.
Mahaffy et al. analyzed densities in several layers up to a maximum depth of 10 mm. Consequently, our layer of analysis included a larger variety of density points, thus most likely yielding slightly different results in specific quadrants. To produce results that are directly comparable to Mahaffy, it would have been necessary to distinguish and analyze the same layers and to specifically focus on the missing layers between the depths of 10 and 18 mm.
However, this study opted to analyze one single layer, in favor of several layers [5], and instead differentiated the quadrant structure by subdividing each quadrant into ten concentric rings from center to periphery.
In a second analytical step, when stratifying by rings and quadrants, cortical densities showed the overall highest densities close to the periphery of the superior quadrant (peaking in ring 7). Conversely, the inferior quadrant had the lowest cortical densities also in the periphery, (i.e., rings 6 to 10). Trabecular bone showed a high‐density superior‐anterior half and a low‐density inferior‐posterior half that had decreased densities towards the periphery.
These bone density results beg the question whether a correlation between density and erosion exists as expected based on Wolff's law [5, 8]. In both quadrant and ring‐quadrant analyses, cortical as well as trabecular densities paralleled Favard's initial E2 type erosion definition: both bone types showed a concentration of high‐density in the peripheral superior regions where erosion has been described to be greatest and thus bone remodeling rates are hypothesized to be equally high. Conversely, lower densities were found in the remaining less eroded inferior‐posterior peripheral regions of the glenoid.
Greater bone densities in the superior‐anterior regions suggested that joint loading in type E2 glenoids occurred preferentially in these areas.
Since this study solely produced data concerning bone density and not for glenoid erosion, it was only possible to relate our results to general descriptions of the Favard typology as presented in pertinent literature. When comparing this study's results to a more recent and quantitative erosion analysis of E2 types, density distribution and erosion orientation only partially overlap (see Figures 5 and 8) [3, 4, 8]. According to Abdic et al., E2 erosion orientation did not solely occur in the superior quadrant but also extended to the neighboring posterior quadrant [8].
Although our density results differed from Abdic et al. with respect to the posterior quadrant, our findings corroborate their results that E2 type erosion is not solely restricted to the superior quadrant but also extends to neighboring ones. This begs the question why our high‐density findings for E2 types in the superior‐anterior quadrants were opposed to the erosion patterns reported by Abdic et al. with the highest erosion in the superior‐posterior glenoid regions [8].
A possible explanation could be that glenoids with underlining cuff pathologies are found to exhibit vast osteophyte formations, especially in anterior and inferior glenoid regions [60]. The idiosyncratic density properties of osteophyte bone could obscure our results [61, 62, 63]. Higher densities in the peripheral anterior area could also be related to the pathological thinness of the anterior glenoid rim containing a small amount of trabecular bone available in the most peripheral rings.
As a result, an increased cortical‐to‐trabecular ratio may be observed in these areas. At any rate, the overall number of trabecular and cortical density points was sufficient for comparisons with more centrally located areas containing more available bone stock.
Another reason for the discrepancy of our results with Abdic et al.'s findings, could be that a percentage of our limited cohort might have suffered from substantial subscapularis damage, which has been linked to higher degree of anterior glenoid erosion [64, 65]. Unfortunately, the density focus of this study did not permit to take into account gradings of rotator cuff conditions such as fatty infiltration [66]. We, like previous studies, neglected qualitative aspects that would differentiate between physiological bone and osteophyte formations [5, 6].
In this respect, our study quantitatively described bone densities, similar to the approach used by Abdic et al. in their bone volume analyses [8].
Furthermore, Abdic et al.'s focus was on erosion volume and not on quantifying regional erosion depth. Since our study analyzed regional densities to a specific depth, our results are not fully comparable. Given that erosion and bone density in the two studies only partially overlap, especially when differentiating between center and peripheral glenoid areas, it would be informative to analyze erosion orientation and depth using the same ring‐quadrant grid.
The application of this grid also for erosion orientation—which would allow to differentiate between center, periphery, depth, and quadrants—could provide more interpretable and comparable results [4, 5, 8]. Frankle et al. corroborated this approach by suggesting the use of 3D models for identifying abnormal rotator cuff arthropathy subtypes [64].
Regarding surgical preparation, the three‐dimensional ring‐quadrant grid also has the potential to be integrated into preoperative planning software. The systematic topographical analyses of bone densities in E2 types can suggest where operative procedures should refrain from excessive removal of bone and which areas should be targeted for implant anchoring, regardless of implant design. The main focus of the study, however, was solely the detailed analysis of bone characteristics in a specific pathology type to complement previous literature [5, 6, 27, 35, 37, 43, 44, 47, 48, 67] and to contribute to a global understanding of the pathophysiology with its interconnected factors of bone erosion due to cuff tear arthropathy.
This study has the following limitations: Although all anatomical landmark placements were checked by a second trained observer, formal intra‐ and inter‐observer reliability was not assessed, which may limit the reproducibility of the measurements. Also, the concentric grid was unable to fully cover the entirety of every patient's glenoid and inevitably neglected some areas of the heavily deformed glenoid. Even the best fit ring structure could only be variable in its dimension.
That means that the diameter in millimeters changed depending on the size of each patient's individual glenoid. The study also neglected void fractions, (i.e., subarticular bone cysts with very low bone density), as well as other scapular landmarks outside the glenoid vault to keep the results comparable to previous studies [5, 6, 13, 36]. In contrast to prior literature that analyzed several bone layers, this study focused on one extensive layer [5, 6, 35, 43].
This kept the results from 40 different segments manageable, especially in their representation for clinical application. We opted to assess more differentiated glenoid regions, (i.e., substantially more surface sections as well as deeper glenoid parts), which are in accordance with previously recommended screw lengths for implant fixation. [17, 30, 31, 68, 69, 70] By relying on one depth layer, we assumed that dense cortical‐like bone was located in regions closest to the glenoid surface and less‐dense trabecular bone in deeper glenoid regions. Mahaffy et al., in their study, stated that cortical‐like bone densities did not differ in deeper glenoid regions [5].
We, therefore, did not identify a specific cut off layer within the 18 mm range for either of the two bone types to include all density points. Density analyses with the ring‐quadrant grid at variable depth increments in future studies would be advisable to address this issue.
Conclusion
The study was able to show significant regional density differences between central and peripheral regions of the inferior and posterior glenoid of E2 types. The highest cortical densities were found in the periphery of the superior quadrant and equally high trabecular densities in central and peripheral regions of superior and anterior quadrants. Regarding surgical procedures, the results suggest refraining from excessive bone removal in these high‐density areas and rather targeting them for (additional) implant fixation.
The discrepancies between the density results when compared to previous density as well as erosion studies on E2 type glenoids ask for a reassessment, using a grid design that differentiates between central and peripheral glenoid areas. Such a grid would also permit a more exact erosion description. Regardless of these possible future approaches, the analyses of E2 type glenoids provided a detailed and comprehensive assessment of topographical bone density variations with indirect relevance for preoperative planning of implant anchoring.
Author Contributions
Moritz Klarer: conception and design, data acquisition, data analysis, data interpretation, data integrity and accuracy confirmation, critical revision, manuscript drafting. Cole T. Fleet: conception and design, data analysis, critical revision.
Joshua W. Giles: critical revision and supervision. James A.
Johnson: conception and design, critical revision and supervision, final approval of the manuscript. George S. Athwal: conception and design, critical revision and supervision, final approval of the manuscript.
Funding
The authors have nothing to report.
Ethics Statement
This computational study did not require Institutional Review Board (IRB) approval as it involved the retrospective analyses (secondary research) of fully anonymized CT imaging data with no identifiable patient information. No direct or indirect patient contact was made during any stages of the study.
Conflicts of Interest
George Athwal is a paid consultant for Stryker and receives royalties from Stryker, Exatech, and Conmed for implants not related to the topic of this research paper. No company had any input into the design or conduction of this study. The remaining authors, their immediate family, and any research foundations with which they are affiliated did not receive any financial payments or other benefits from any commercial entity related to the subject of this article.
Supporting information