A Novel Dual‐Energy X‐Ray Absorptiometry Protocol for the Proximal Humerus Reveals a Fatty Infiltration Dependent Reduction in Bone Mineral Density

ABSTRACT

Localized reduction in bone mineral density (BMD) within the proximal humerus may compromise anchor fixation strength and tendon‐to‐bone healing after rotator cuff repair; however, standardized dual‐energy X‐ray absorptiometry (DXA) protocols and diagnostic thresholds are lacking. This study aimed to develop a reproducible proximal humerus DXA protocol, evaluate regional BMD differences between affected and contralateral shoulders, and investigate the relationship between fatty infiltration and regional bone loss. A retrospective analysis was performed on 224 patients who underwent unilateral arthroscopic rotator cuff repair with preoperative bilateral shoulder DXA scans. The proximal humerus was divided into five anatomically defined regions of interest (R₁–R₅), and subgroup analysis was conducted according to supraspinatus fatty infiltration grade. The affected shoulder demonstrated lower BMD across all regions, although significant reductions were observed only in the medial humeral head (R₂, p < 0.001) and averaged humeral head BMD (R₂₃₄, p = 0.011). Patients with mild fatty infiltration (Grades 0–2) showed BMD reduction primarily in periarticular regions, whereas advanced fatty infiltration (Grades 3–4) was associated with diffuse BMD loss across all regions. Femoral neck T‐score was moderately correlated with humeral head BMD (r = 0.66, R ² = 0.488, p < 0.001), and regression analysis identified a proximal humerus osteoporosis threshold of 0.478 g/cm², corresponding to a femoral neck T‐score of −2.17. This protocol enables reproducible regional BMD assessment and demonstrates fatty infiltration–dependent bone loss progression. Femoral neck DXA may serve as a screening tool but is insufficient for definitive diagnosis.

1. Introduction

Rotator cuff tears (RCTs) represent one of the most common degenerative musculoskeletal diseases, leading to significant shoulder pain and restricted function. Most rotator cuff tears involve disruption of the tendon insertion at the proximal humeral enthesis, although tears can also occur at other locations, such as the musculotendinous junction [1]. Following tendon detachment from the insertion site, the loss of mechanical loading at the bone–tendon interface is known to induce subsequent skeletal remodeling. This process aligns with Wolff's law, which posits that bone adapts its structure and density in response to mechanical loading [2, 3]. The lack of appropriate mechanical stimulation following an RCT leads to local bone loss primarily affecting the proximal humerus, where the rotator cuff inserts [4].

In this study, localized osteoporosis refers to region‐specific reduction in bone mineral density (BMD) confined to the proximal humerus, in contrast to systemic osteoporosis, which is defined using dual‐energy X‐ray absorptiometry (DXA) measurements of the lumbar spine and hip to reflect generalized skeletal bone loss. Localized bone loss in the proximal humerus may compromise anchor pull‐out strength during rotator cuff repair, understanding the need for precise assessment to guide the surgical strategy [5]. Moreover, as localized bone loss can adversely influence tendon healing following rotator cuff repair, accurate diagnosis and appropriate pharmacologic management are warranted in patients with diminished local bone quality [6, 7, 8].

Despite the clinical relevance of proximal humerus bone quality, reliably and consistently assessing it remains a significant challenge. DXA is widely considered a cost‐effective modality for diagnosing osteoporosis [9]. However, DXA measurement protocols are primarily standardized for the lumbar spine (LS) and proximal femur. When measuring proximal humeral BMD with DXA, challenges arise, including a lack of standardized positioning protocols to avoid measurement error caused by overlap of complex anatomical structures, such as the bicipital groove [4, 10]. Furthermore, unlike the lumbar spine and hip, no standardized diagnostic cutoff value has been established for BMD of the proximal humerus to guide treatment decisions. Because DXA T‐scores are defined using the mean and standard deviation of a young, healthy reference population, interpretation of proximal humeral DXA measurements is inherently limited by the absence of an appropriate anatomical reference database [11, 12]. Consequently, true T‐scores for the proximal humerus cannot currently be calculated, and BMD measurements in this region must be interpreted cautiously in a relative or region‐specific context.

Several existing studies haNve attempted to measure proximal humerus BMD using DXA. Oh et al. reported obtaining consistent data by measuring in the 30° Nshoulder external rotation (ER) position, thereby preventing overlap of the bicipital groove [4]. However, the Regions of Interest (ROIs) in this study (Greater Tuberosity, head center, and surgical neck) were less segmented, considering the complex anatomy of the humeral head. Consequently, the differences in BMD among the regions were less pronounced than in previous studies [13, 14, 15]. Therefore, there is a necessity to re‐establish the measurement position and ROIs to avoid overlapping and obtain more accurate BMD for each region.

This study aims to address these limitations by focusing on the standardization and clinical utility of proximal humeral DXA measurements. Based on results obtained using this method, we intend to clarify the pattern of change in proximal humerus BMD in patients with rotator cuff disease and identify the factors that influence it. We also plan to set an appropriate cut‐off value for proximal humerus BMD by analyzing its correlation with lumbar spine and hip BMD. Our hypotheses are that: (1) The BMD values for each region measured by our standardized protocol will show consistency with existing data utilizing various imaging modalities; (2) Factors indicating disease progression, such as fatty infiltration (FI), will be correlated with a decrease in proximal humerus BMD; (3) LS and hip BMD will have a positive correlation with proximal humerus BMD, and we can set the cut‐off value for proximal humerus BMD using the already standardized cut‐off values for LS and hip BMD.

2. Materials and Methods

2.1. Patient Enrollment

The hospital's Institutional Review Board approved this retrospective cohort study. All patients provide informed consent before surgery. The inclusion criteria were a full‐thickness rotator cuff tear treated with arthroscopic rotator cuff repair and preoperative BMD measurements by DXA of the bilateral shoulders, hips, and lumbar spine. From February 2012 to May 2014, 296 patients with a full‐thickness rotator cuff tear underwent arthroscopic repair. The 33 patients were excluded: 20 with bilateral symptomatic image‐proven rotator cuff tears, six with an isolated subscapularis tear, and seven treated with revisional repair. Thirty‐nine patients who did not have bilateral shoulder BMD measured before surgery were additionally excluded. Therefore, 224 patients were included in the study.

2.2. BMD Measurement

The BMD of the bilateral shoulders, hips, and lumbar spine was measured preoperatively using DXA (Lunar Prodigy, GE Medical Systems, Madison, WI). Given that no standardized protocol for measuring BMD in the shoulder region has been established to date, we developed a new protocol to minimize overlapping among microstructures and isolate them. We improvised a positioning device composed of a pillow and a handle (Figure 1). The pillow is made of a firm mattress with a 30° inclination, considering scapular anteversion, to make the glenoid surface vertical to the plane of the examining floor, preventing overlapping of the glenoid and humeral head [10]. The handle is made of plastic that can be inserted under the pillow and positioned to fix the shoulder at 30° of internal rotation and 15° of abduction, with the elbow flexed to 90° in neutral position. Thus, the greater tuberosity (GT) is placed in the lateral 1/3 of the humeral head, and the bicipital groove is put forward in axial view so that the greater and lesser tuberosities can be easily visualized on the scan without overlapping. In defining ROIs, the humeral head was divided vertically into three columns of equal width (Figure 2). From the most inferior margin of the articular surface to the most superior margin of the humeral head, four horizontal lines were drawn to make three horizontal rows with an even height. Square regions measuring 1 × 1 cm were drawn at the center of the middle row and defined as ROIs. In three different ROIs of the humeral head, the most lateral square was considered as the greater tuberosity (R₃). The middle square was considered as the lesser tuberosity (R₄), the most medial square was arbitrarily named as the medial aspect of the humeral head (R₂), and the average value of these three ROIs was regarded as the average humeral head (R₂₃₄). The surgical neck of the humerus (R5) and glenoid (R1) were measured as the control. Hip and lumbar spine BMD were measured and recorded according to a conventional standardized protocol [16].

Figure 1.

Patient position during proximal humerus dual‐energy X‐ray absorptiometry. The positioning device is composed of a pillow on the back of the patient and a handle to grab. The pillow was made with a 30° inclination, and the handle makes the proximal humerus 30° internal rotation to make the bicipital groove midway between the greater tuberosity and the lesser tuberosity.

Figure 2.

Region of interest (ROI) of proximal humerus dual‐energy x‐ray absorptiometry. The humeral head was divided into quarters with equal height, and two vertical lines were drawn to divide the head into three columns with equal width. 1 × 1 cm squares drawn in the middle row were designated as ROIs. In three different ROIs of the humeral head, the most lateral square was considered as the greater tuberosity (R3). The middle square was considered the lesser tuberosity (R4), and the most medial square was arbitrarily named the medial aspect of the humeral head. (R2) The surgical neck of the humerus, which took account of two squares, (1 × 2 cm, R5) and glenoid (1 × 1 cm‐sized square in the lower row, R1), was measured as the control.

2.3. Clinical Assessments

All 224 patients completed a standardized preoperative outcome assessment questionnaire. Clinical outcomes were assessed with respect to: (1) pain, (2) range of motion (ROM), (3) strength, (4) functional scores. A visual analog scale (VAS; 10 cm marked from “no pain” to “unbearable pain”) was used to evaluate pain at rest, during motion, and at night. Mean pain scores were calculated and compared. In addition, the worst pain was also recorded. ROM was measured using a goniometer in active forward flexion, abduction, external rotation at the side, and internal rotation. Internal rotation was measured at vertebral levels, which were translated into numbers from 1 (buttocks) to 17 (T2). The strength of muscles, including supraspinatus (forward elevation in the scapular plane), infraspinatus (external rotation with the arm at the side), and subscapularis (internal rotation with the arm at the side), was measured using a handheld electronic scale (CHS, CAS, Yangju, Korea). The functional scoring systems used were: the American Shoulder and Elbow Surgeons (ASES), the Constant score, the University of California, Los Angeles (UCLA) score, the Simple Shoulder Test (SST), and the Shoulder Pain and Disability Index (SPADI).

2.4. Preoperative MRI Assessments

MRI was performed using a 3.0‐T scanner (Achieva 3.0‐T, Philips Healthcare, Eindhoven, Netherlands) with a dedicated shoulder coil. Assessment was made using imaging processing software (Marosis M‐View 5.4; INFINITT Healthcare, Seoul, Korea). Fatty infiltration (FI) was assessed using the Goutallier grading system as modified by Fuchs et al. [17, 18]. A global fatty degeneration index (GFDI) was calculated as the mean Goutallier grade of the supraspinatus, infraspinatus, and subscapularis muscles. Muscle atrophy (MA) was assessed using the tangent sign (TS), occupation ratio (OR) of the supraspinatus and infraspinatus [19]. Two fellowship‐trained orthopedic surgeons, unaware of the MRI findings, separately measured all variables as previously described. Discrepancies were resolved by consensus.

2.5. Statistical Analysis

All statistical analyses were carried out using IBM SPSS 29.0 software, and a p‐value of < 0.05 was considered as the level of statistical significance. To evaluate differences in BMD across the five regions of interest (ROIs) within a single shoulder, repeated‐measures ANOVA was performed, followed by Bonferroni‐corrected post hoc comparisons. To compare BMD between bilateral proximal humeri, a paired t‐test was used. Stepwise multiple linear regression was performed after univariate analysis to identify parameters associated with differences in bilateral shoulder BMD. For univariate analysis, a simple linear regression model was used. Pearson correlation coefficients were calculated to assess the correlation between the BMDs of the proximal humerus and hip and lumbar spine. A simple linear regression model was used to describe the relationship between the ST‐score of the femoral neck and the BMD of the proximal humerus.

3. Results

Baseline characteristics of the enrolled patients are summarized in Table 1.

Table 1.

Patient characteristics.

	Mean ± SD or No.
	(n = 224)
Demographic parameters
Age (years)	62.1 ± 7.7
Sex (female:male)	137:87
BMI (kg/m2)	24.8 ± 4.3
Operative parameters
Tear size, AP (mm)	27.4 ± 14.7
Tear size, ML (mm)	17.9 ± 10.3
Radiologic parameters
GFDI	1.6 ± 0.7
Tangent sign (1:2:3)	134:64:21
Clinical parameters
Dominant side	175:49
(dominant:non‐dominant)
Symptom duration (months)	16.1 ± 24.4
Osteoprosis parameters
Osteoporosis (normal:osteopenia:osteoporosis)	77:104:43
BMD at femur (g/cm2)	0.84 ± 0.15
T‐score at femur	−0.51 ± 1.04
BMD at lumbar spine (g/cm2)	1.04 ± 0.20
T‐score at lumbar spine	−0.96 ± 1.58

Note: Bone status was classified based on the lowest T‐score measured at either the lumbar spine or proximal femur. Normal bone density was defined as a T‐score > −1.0, osteopenia as a T‐score between −1.0 and −2.5, and osteoporosis as a T‐score ≤ −2.5.

Abbreviations: AP, anteroposterior; BMD, bone mineral density; BMI, body mass index; GFDI, global fatty degeneration index; ML, mediolateral; SD, standard deviation.

3.1. Regional and Side‐to‐Side Differences in Proximal Humerus BMD

There were significant differences in BMD values among the five ROIs in both the symptomatic and asymptomatic shoulders. In both groups, Bonferroni‐adjusted post hoc comparisons demonstrated a significant difference between R₃ and R₄ (symptomatic shoulder: p = 0.03; asymptomatic shoulder: p = 0.007), whereas all other pairwise comparisons showed p‐values < 0.001. The BMD value was lowest in R₃ and highest in R₂. Although every BMD value of the affected humerus was lower than that of the asymptomatic side (1.02%–5.44%), only R₂ and average value of humeral head (R₂₃₄) showed statistically significant reductions compared with the asymptomatic side (p < 0.001, 0.011, respectively) (Table 2).

Table 2.

Bilateral shoulder bone mineral density measurement.

	BMD (g/cm2) (mean ± SD)		BMD diff. (g/cm2) (Asx. − Aff.)	p‐value	Difference (%)(Diff./Asx.)
	Asymptomatic side	Affected side
R1 (glenoid)	0.678 ± 0.226	0.692 ± 0.201	−0.014	0.350	−1.999
R2 (medial humeral head)	0.806 ± 0.192	0.763 ± 0.206	0.044	< 0.001*	5.439
R3 (greater tuberosity)	0.504 ± 0.153	0.499 ± 0.135	0.005	0.559	1.015
R4 (lesser tuberosity)	0.535 ± 0.170	0.524 ± 0.156	0.011	0.248	2.116
R5 (surgical neck of humerus)	0.590 ± 0.189	0.577 ± 0.183	0.013	0.156	2.197
R234 (average value of humeral head)	0.615 ± 0.153	0.595 ± 0.148	0.019	0.011*	3.089
Rmin (minimum value of humerus)	0.504 ± 0.153	0.500 ± 0.134	0.005	0.297	1.015

Abbreviations: Aff., affected side; Asx., asymptomatic side; Diff., difference; BMD, bone mineral density; SD, standard deviation.

*p < 0.05 was defined as statistically significant.

3.2. Multivariate Analysis of Factors Affecting Side‐to‐Side BMD Differences

Stepwise multivariate linear regression identified distinct predictors of side‐to‐side BMD differences across ROIs, including rotator cuff fatty infiltration, muscle strength imbalance, arm dominance, sex, and pain level. Notably, supraspinatus and infraspinatus fatty infiltration emerged as consistent predictors in multiple regions, while functional variables such as forward elevation and internal rotation strength differences were associated with regional BMD asymmetry. Detailed regression results for each ROI (R₁–R₅) are presented in (Table 3).

Table 3.

Multivariate stepwise linear regression for regional bone mineral density differences.

Region	Relevant factors	Equation	R 2 adj.	p‐value
R1 (Glenoid)	Dominancy, FIIS	D1 = 0.011 − 0.112 Dom + 0.042 FIIS	0.063	< 0.001*
R2 (Medial humeral head)	Internal rotation power difference, Dominancy, FISS	D2 = −0.107 + 0.008 pwrIR + 0.076 Dom + 0.028 FISS	0.105	< 0.001*
R3 (Greater tuberosity)	Forward elevation power difference, FISS	D3 = −0.061 + 0.005 pwrFE + 0.021 FISS	0.040	0.005*
R4 (Lesser tuberosity)	Internal rotation power difference, Sex, FISS	D4 = −0.118 + 0.012 pwrIR + 0.026 FISS + 0.041 sex	0.128	< 0.001*
R5 (Surgical neck of humerus)	Resting VAS score	D5 = 0.041 − 0.008 resting VAS	0.015	0.040*

Abbreviations: BMD, bone mineral density; Dom, dominancy; FIIS, fatty infiltration of infraspinatus muscle; FISS, fatty infiltration of supraspinatus muscle; pwrFE, forward elevation power difference; pwrIR, internal rotation power difference; VAS, visual analogue scale score.

^* p < 0.05 was defined as statistically significant.

3.3. Subgroup Analysis According to the Degree of Fatty Infiltration of the Supraspinatus

Stepwise multivariate linear regression demonstrated that fatty infiltration of the supraspinatus muscle was a common factor influencing BMD differences. Accordingly, we performed a subgroup analysis based on the degree of supraspinatus fatty infiltration. Subgroup analysis revealed distinct patterns of BMD asymmetry between shoulders, depending on the degree of FI of the supraspinatus.

In patients with mild to moderate FI (stages 0–2, n = 166), the affected shoulders showed significantly lower BMD than the contralateral side at the glenoid (R₁, p = 0.026) and medial humeral head (R₂, p = 0.003), while no significant differences were observed in the greater or lesser tuberosities (R₃–R₄) or at the surgical neck (R₅).

In contrast, patients with advanced FI (Stages 3–4, n = 52) exhibited more extensive and pronounced reductions in the affected shoulders, with significantly lower BMD across all regions analyzed, including the glenoid (R₁, p = 0.023), medial humeral head (R₂, p < 0.001), greater and lesser tuberosities (R₃–R₄, p = 0.005 and p = 0.002, respectively), and the surgical neck (R₅, p = 0.010) (Table 4).

Table 4.

Subgroup analysis according to fatty infiltration of supraspinatus muscle.

Group	ROI	BMD in affected shoulder(g/cm2) (mean ± SD)	BMD in asymptomatic shoulder (g/cm2) (mean ± SD)	p‐value
FISS = 0, 1, 2	R1	0.70 ± 0.21	0.67 ± 0.22	0.026*
(n = 166)	R2	0.77 ± 0.20	0.80 ± 0.20	0.003*
	R3	0.50 ± 0.14	0.50 ± 0.15	0.713
	R4	0.53 ± 0.16	0.53 ± 0.18	0.944
	R5	0.58 ± 0.18	0.58 ± 0.19	0.945
FISS = 3, 4	R1	0.64 ± 0.14	0.71 ± 0.25	0.023*
(n = 52)	R2	0.74 ± 0.21	0.82 ± 0.19	< 0.001*
	R3	0.49 ± 0.13	0.54 ± 0.14	0.005*
	R4	0.51 ± 0.14	0.56 ± 0.13	0.002*
	R5	0.56 ± 0.18	0.61 ± 0.16	0.010*

Abbreviations: BMD, bone mineral density; FISS, fatty infiltration of supraspinatus muscle; ROI, region of interest.

^* p < 0.05 was defined as statistically significant.

3.4. Relationship Between Hip/Lumbar Spine BMD and Proximal Humerus BMD: A Linear Regression‐Based Prediction Model

Femoral neck and lumbar spine BMD showed a moderate positive correlation with the average humeral head BMD (Table 5). Among the linear regression models tested, the model using femoral neck BMD and average humeral head BMD had the highest R [2] (0.488) and the lowest root mean square error (RMSE) (0.107), indicating it was the best‐performing model (Table 6). Based on these findings, a mathematical equation was obtained to predict the proximal humeral BMD of the affected side by using the T‐score of the femoral neck.

Table 5.

Association of shoulder and lumber/femur bone mineral density and T‐score.

	Affected shoulder (R234)		Asymptomatic shoulder (C234)
	PCC	p‐value	PCC	p‐value
BMD in femur neck (g/cm2)	0.687	< 0.001	0.629	< 0.001
T‐score in fermur neck	0.659	< 0.001	0.622	< 0.001
BMD in lumbar spine (g/cm2)	0.616	< 0.001	0.591	< 0.001
T‐score in lumbar spine	0.603	< 0.001	0.579	< 0.001

Abbreviatiosn: BMD, bone mineral density; C234, average humeral head BMD of contralateral shoulder; PCC, Pearson correlation coefficient; R234, average humeral head BMD of affected shoulder.

Table 6.

Humeral bone mineral density prediction based on lumbar and femoral T‐scores.

Model	Independent variable	Outcome	R 2	CVRMSE
1	FN	R234	0.488	0.107
2	FN	R2	0.398	0.160
3	FN	R4	0.438	0.117
4	Avg of FN and L14 T‐score	R234	0.475	0.109
5	Avg of FN and L14 T‐score	R2	0.407	0.159
6	Avg of FN and L14 T‐score	R4	0.426	0.119
7	Min of FN and L14 T‐score	R234	0.416	0.114
8	Min of FN and L14 T‐score	R2	0.364	0.163
9	Min of FN and L14 T‐score	R4	0.366	0.124

Abbreviations: Avg, average; BMD, bone mineral density; FN, T‐score of femoral neck; L14, L1–4 lumbar vertebrae; Min, minimum; RMSE, root mean squared error.

$$\mathrm{Proximal}\mathrm{humeral}\mathrm{BMD}\mathrm{of}\mathrm{affected}\mathrm{side}=0.669+0.088\times \mathrm{T}‐\mathrm{score}\mathrm{of}\mathrm{femoral}\mathrm{neck}(R^2=0.472,F=168.7,P<0.001).$$

In the same way, an equation between the proximal humeral BMD of the asymptomatic side and the T‐score of the femoral neck was obtained:

$$\mathrm{Proximal}\mathrm{humeral}\mathrm{BMD}\mathrm{of}\mathrm{asymptomatic}\mathrm{side}=0.685+0.083\times \mathrm{T}‐\mathrm{score}\mathrm{of}\mathrm{femoral}\mathrm{neck}(R^2=0.395,F=138.8,p<0.001).$$

Considering −2.5 of the T‐score of the femoral neck is the cut‐off value to diagnose osteoporosis, according to the equation above, the corresponding BMD of the unaffected side humerus is 0.478 g/cm², and substitution of this value in the equation of proximal humeral BMD of the affected side shows the T‐score of the femoral neck as −2.17 (Figure 3).

Figure 3.

Equations between average humeral head BMD and femoral neck T‐score. A diagnostic criterion of average humeral head BMD ≤ 0.478 for osteoporosis was determined based on the estimated regression line. When the femoral neck T‐score corresponding to this BMD threshold was calculated on the symptomatic side, the resulting value was −2.17. When humeral head BMD is not available, a femoral neck T‐score of ≤ −2.17 can be used as a surrogate threshold, providing a sensitivity of 0.40 and a specificity of 0.93. BMD, bone mineral density; C234, average humeral head BMD of the contralateral side; FN_T, femoral neck T‐score; R234, average humeral head BMD of the affected side.

4. Discussion

According to previous studies based on three‐dimensional imaging and cadaveric analyses, proximal humerus BMD tends to be higher on the medial side near the joint and lower on the lateral side, including the greater tuberosity [13, 14, 15, 20]. This pattern is consistent with the findings of our study. By analyzing factors associated with BMD differences, we confirmed that the progression of fatty infiltration contributes to greater BMD discrepancies. Finally, we identified a moderate positive correlation between lumbar spine and hip BMD and proximal humerus BMD. We established a cutoff value for proximal humerus BMD using a linear regression‐based modeling approach. Thus, we were able to validate all three of our initial hypotheses.

Studies on femoral and lumbar DXA measurements demonstrate that proper patient positioning is critical for obtaining accurate BMD values. During femoral BMD assessment, the leg is internally rotated to prevent shortening or overlapping of the femoral neck. Similarly, flexion of the hips and knees is used to reduce lumbar lordosis during lumbar spine measurements. In addition, overlapping of the ischium on the femur can introduce measurement errors, and appropriate adjustments are made to avoid such artifacts [9, 12, 21]. By applying these principles to the proximal humerus, we established a standardized measurement position and defined our regions of interest. Prior attempts to develop a proximal humerus DXA protocol have been made. Oh et al. and Lee et al. performed measurements at 30° of external rotation, but their assessments were limited to the greater tuberosity and the middle column, without evaluating the medial column, representing the periarticular region [2, 4]. However, our findings indicate that, in patients with rotator cuff disease, the medial column demonstrates the earliest and most pronounced decline in BMD. Using a measurement technique that eliminates glenoid overlap, we accurately evaluated this area and confirmed that the medial column consistently shows the highest BMD, consistent with previous studies [13, 14, 15].

In our analysis, significant side‐to‐side differences were observed in the medial column. In contrast, the greater tuberosity (GT) and lesser tuberosity (LT) did not show meaningful differences in the overall cohort. However, subgroup analysis based on supraspinatus fatty infiltration (FI 0–2 vs. FI 3–4) revealed that patients in the FI 3–4 group exhibited significant reductions in BMD across all regions, including the GT and LT. These findings suggest that in the early stage of rotator cuff disease, BMD loss is most prominent around the medial periarticular surface. In contrast, BMD reduction in the GT and LT becomes more apparent as the disease progresses. This pattern may reflect the sequence of biomechanical changes: early alterations in joint reaction forces predominantly affect the medial column. In contrast, mechanical traction changes at the tendon insertion sites manifest later in the disease course. In the multivariate analysis, muscle strength was correlated with the degree of BMD difference, and the associations between GT BMD changes and external rotation power, as well as between LT BMD changes and internal rotation power, further support this relationship. The observation that R₂ demonstrates greater BMD reduction when the dominant arm is involved also indicates that changes in BMD are closely related to alterations in the functional status of the muscles.

It is often difficult to directly measure humeral BMD, and the absence of an established threshold value presents an additional challenge. To address these issues, it is necessary to examine the relationship between humeral BMD and lumbar or femoral BMD. We found that femoral BMD showed a stronger correlation than lumbar BMD, and the model incorporating femoral neck BMD and average humeral head BMD provided the best fit. A humeral head average BMD of 0.478 (g/cm²) can be used as a cutoff value, which corresponds, when back‐calculated, to a femoral neck T‐score of –2.17. When defining proximal humerus osteoporosis as a BMD of 0.478 (g/cm²) or below and using a femur neck T‐score of –2.17 or below as the diagnostic criterion, the sensitivity and specificity calculated from our dataset were 0.40 and 0.93, respectively. (Figure 3) Therefore, in situations where humerus DXA cannot be performed, Femoral neck DXA may be useful as a screening tool for proximal humeral local osteoporosis; however, it cannot be used as a definitive diagnostic method. If the femoral neck T‐score is higher than –2.17, local osteoporosis can be reasonably excluded. However, a femoral neck T‐score below −2.17 does not necessarily indicate the presence of local osteoporosis. Considering that an apparent reduction in humeral head BMD becomes evident only when rotator cuff tears progress to FI grades 3–4, it is likely that early‐stage tears may not yet demonstrate local osteoporosis. In summary, local osteoporosis can be diagnosed using humerus DXA when the BMD is below the cutoff value of 0.478 (g/cm²). If humerus DXA cannot be performed, a femur neck T‐score lower than –2.17 combined with advanced rotator cuff degeneration (fatty infiltration grades 3–4) may indicate a higher likelihood of local osteoporosis. These criteria can be applied to guide postoperative medication decisions and preoperative surgical planning.

A notable limitation of this study is that, in patients with a unilateral rotator cuff tear, the contralateral asymptomatic shoulder often harbors an unrecognized tear [22]. Despite this, the fact that significant bilateral differences were still observed suggests that rotator cuff pathology meaningfully contributes to local osteoporosis, and the true magnitude of this effect may, in fact, be greater than what is demonstrated in our study. Future investigations could focus specifically on patients with confirmed imaging of both shoulders to more accurately isolate the effect of rotator cuff disease on proximal humerus BMD. Another limitation relates to the establishment of a diagnostic cutoff. Standard DXA thresholds rely on T‐scores derived from healthy young adult reference populations [12, 16]. Because such normative data for the proximal humerus are currently unavailable, we estimated the cutoff using regression modeling based on femoral neck BMD. Given the moderate correlation between femur neck and humeral head BMD (r = 0.6–0.7), the cutoff derived in this manner likely contains some degree of error. Nonetheless, considering the practical difficulty of obtaining humoral BMD data from a large, healthy young population, our approach represents a reasonable methodological solution. Future work incorporating larger datasets may allow refinement and validation of a more precise proximal humerus–specific thresholds [11].

5. Conclusion

Using a newly developed proximal humerus DXA protocol, we obtained valid measurements. In the early stages of rotator cuff disease (FI 0–2), BMD reduction was primarily observed in the medial aspect of the humeral head, adjacent to the joint. As fatty infiltration progressed to Goutallier grades 3–4, BMD loss became evident across the entire humeral head. Femur neck T‐scores demonstrated a moderate positive correlation with humeral head BMD, and regression‐based modeling allowed us to define an appropriate diagnostic cutoff. A humeral head average BMD of ≤ 0.478 g/cm², or a femur neck T‐score of ≤ –2.17 in patients with FI grade ≥ 3, indicates a high likelihood of local osteoporosis.

Author Contributions

Jae Woo Cho: contributed to data curation, investigation and writing of the paper. Chris Hyunchul Jo: contributed to methodology, conceptualization and supervision of the paper. Dong Min Kim: contributed to data collection and analysis. Sohee Oh: contributed to statistical analysis of data.

Ethics Statement

Seoul Metropolitan Governance—Seoul National University Boramae Medical Center Institutional Review Board approved this study (IRB No. 20‐2019‐60).

Conflicts of Interest

The authors declare no conflicts of interest.