Associations between radiographic parameters in the reverse shoulder arthroplasty and patient outcomes

Abstract

Background

Various radiological parameters have been measured in the Reverse Shoulder Arthroplasty (RSA) and correlated to patient outcomes, to determine best baseplate position. Results remain unclear with respect to certain parameters such as inferior baseplate tilt. We have investigated our series of patients to clarify the relationship between radiological parameters and patient outcomes.

Methods

We conducted a UK based bi-centre retrospective review of 156 prostheses. Critical shoulder angle (CSA), RSA angle (RSAA), Overhang and Deltoid Lever Arm (DLA) were measured on preoperative and postoperative radiographs. Range of motion and Oxford Shoulder Score (OSS) (range 1–8 years) were obtained. We assessed for scapular notching at minimum 1 year follow up (n = 138).

Results

Decreased postoperative CSA and increased DLA were associated with higher OSSs (P = 0.001 and 0.019). Increased overhang and DLA were associated with increased flexion (P = 0.033 and 0.024 respectively). Multivariate analysis showed that CSA and DLA affected notching rate (P = 0.002 and 0.007).

Conclusions

Baseplate tilt in relation to the acromion (CSA) and DLA are the most predictive parameters for notching and OSS. We recommend considering a maximum CSA of 26 degrees to decrease notching rate and improve OSS. We recommend considering an Overhang of at least 6 mm to improve FF.

Introduction

The Reverse Shoulder Arthroplasty (RSA) is now a widely used prosthesis design for rotator cuff deficient patients who require a shoulder replacement. Grammont was aiming to increase the Deltoid Lever Arm (DLA) by medialising the Centre of Rotation (COR) of the shoulder to strengthen the deltoid and compensate for the deficient rotator cuff. ¹ The resulting increase in the DLA increases the abduction moment of the middle deltoid and recruits in addition to the middle deltoid some anterior and posterior deltoid fibres. ² Distalization of the humerus re-tensions the deltoid which adds to strength. ² Scapular notching was a frequent occurrence with earlier RSA designs (e.g. the Delta III prosthesis), due to impingement of the humeral components against the glenoid neck. ³ To decrease scapular notching lateralised designs emerged which lateralise at the glenoid-bone interphase, baseplate, glenosphere and/or the humeral side (by decreasing the Neck Shaft Angle (NSA) or the Polyethylene Opening Angle (POA, also known as insert-NSA) from 155 to 150, 147,145 or 135). ⁴ Lateralised glenoid designs and a lower POA/NSA were associated with a lower rate of scapular notching.^5,6

Numerous radiological parameters were investigated as potential factors affecting the rate of notching including overhang and inferior baseplate tilt (measured by the Critical Shoulder Angle (CSA) or the more recently described Reverse Shoulder Arthroplasty Angle (RSAA). ^{7 )}

Regarding baseplate tilt, there have been conflicting findings. Falaise et al. suggested implanting the glenoid in inferior tilt to decrease the rate of notching as they found that scapular notching was associated with a decreased inferior tilt of the baseplate. ⁸ In contrast, Edwards et al. ⁹ found no influence of tilt on notching while Rhee et al. ¹⁰ and Duethman et al. ⁵ found increased inferior tilt to be associated with scapular notching. The findings with respect to overhang in the literature are widely consistent that increased overhang decreases the rate of scapular notching.^{3,5,6,8,10,11}

Regarding biomechanical outcomes, lateralised designed showed better abduction, less force to initiate elevation ¹² and a better external rotation by re-tensioning any remaining intact cuff ¹³ in biomechanical studies.

In recent in-vivo studies, numerous radiological parameters were investigated as potential factors influencing Range of Motion (ROM) postoperatively including CSA, Acromio-humeral Index (AHI), Acromial Index (AI), DLA and overhang, the findings being sometimes conflicting. Roberson et al. found that a lower CSA was associated with higher Forward Flexion (FF) postoperatively in-vivo ¹⁴ while Lädermann et al. found no effect of CSA on ROM in their 3D computer simulation study ¹⁵ but found that lateralisation of COR increased ROM. Haidamous et al found that overhang is associated with a better FF postoperatively while humeral lateralisation or distalization were not associated with a better ROM. ¹⁶ However, Rhee et al. found no effect of overhang on ROM but a positive effect on Patient Reported Outcome Measures (PROMs). ¹⁰ Rhee et al. also found that medialising the COR decreases External Rotation (ER), echoing previous studies. ¹⁰

The optimum position of the glenosphere to decrease scapular notching is inferior on the glenoid with some overhang. However, it is less clear whether the glenoid should be implanted with more inferior or more neutral tilt based on some new studies which call into question previous wisdom of increasing inferior baseplate tilt to decrease notching. Glenoid component position and its effect on ROM in-vivo is scarcely investigated in the literature.

The challenges in using postoperative radiographs to measure parameters accurately and reliably should be noted. An Anteroposterior view radiograph (AP) is required with minimal rotation to accurately measure angles; Suter et al. showed that only 5 degrees of anteversion can affect the measured CSA by as much as 2 degrees. ¹⁷ Werner et al. concluded that Acromio-humeral distance is not a reliable measurement of distalization in AP radiographs. ¹⁸

Therefore, our primary objective was to resolve controversies in the literature particularly around the role of inferior baseplate tilt and whether radiological parameters are associated with ROM in-vivo. We considered both the glenoid and the humeral side preoperatively and postoperatively investigating a broad range of radiological parameters. We only included cases where the radiograph showed a standard AP view to allow reliable measurement.

Patients and methods

The Royal Cornwall Hospital's (RCH) research and development department waived the ethics requirement assessing the project as a service evaluation rather than research. Patients were not subject to additional investigations or clinical assessments outside of their routine care. A retrospective review of RSAs at the RCH and the North Devon District Hospital (NDDH) over the period 2010 to 2020, identified 156 cases meeting both these inclusion criteria:

1. A suitable standard postoperative AP radiograph (with insignificant rotation as assessed by the primary author) to allow reliable measurement of radiological parameters.

2. Presence of a minimum 1-year Oxford Shoulder Score (OSS) or ROM assessment

Indications for surgery, procedure details and implants used

The indications for RSA were rotator cuff arthropathy in 57 patients, osteoarthritis with a cuff tear or a weak cuff in 43 patients, irreparable/massive tear or failed repair in 12 patients, trauma in 9 patients (acute fracture of the proximal humerus), old fracture sequelae (osteoarthritis/cuff arthropathy) in 28 patients (1 patient had an old screw fixation 22 years ago of the humeral head) and 7 revisions for periprosthetic fracture or complications of Copeland resurfacing hemiarthroplasty/ anatomical shoulder replacement.

156 prostheses were implanted in 150 patients across the two centres (66 Left, 90 Right, 94 females, 56 males). The procedures involved six primary surgeons in the two centres, and all were performed in Beach-Chair position via a deltopectoral approach. The humerus was implanted in 20 degrees of retroversion. Proximal humerus fracture patients received cemented stems with the rest receiving uncemented stems and 3 receiving stemless implants.

Prostheses implanted were Zimmer Trabecular (75 implants, 36 mm glenosphere, 40 mm glenosphere in 1 patient), Biomet Comprehensive (14 implants, 36 mm glenosphere in all patients), Equinox (56 implants, 38 mm glenosphere in all) and Aequalis (11 implants, 36 mm glenosphere), Table 1.

Table 1.

Published features of implants used⁴.

Implant	Number of implants	Eccentric positioning	Inlay/ onlay	NSA	POA	Glenosphere laterization	Glenoid LO	Humeral LO	Global LO	Notching cases
Biomet comprehensive	14	Yes	Onlay	135	147	5.2 mm	14.8	15	29.8	2
Exactech, Equinox	56	Yes	Onlay	132.5	145	3.3 mm	12.9	13.5	26.4	9
Zimmer trabecular	75	No	Inlay	150	150	1.9 mm	11.5	7.5	19	6
Tornier, Aequalis	11	No	Inlay	155	155	−2 mm	7.6	8	15.6	4

NSA: neck shaft angle; POA: polyethylene opening angle; LO: Lateral offset.

Range of motion and oxford shoulder score assessments

The final ROM assessment available (N = 138) occurred on average at 2.4 years (1 to 8 years). In total 18 clinicians documented the ROM, 76 of which were documented by one specialist physiotherapist

Letters with information about the study with an opportunity to be excluded were sent to all alive patients for postoperative OSS unless already obtained in clinic at a year postoperatively. Some patients who did not reply were followed up with a phone call. 3 patients who have not initially returned their questionnaires preferred to be taken through the questionnaire on the phone by the primary author rather than fill another posted questionnaire. 94 filled questionnaires were returned by patients. Average time from surgery to OSS was 4.5 years and a minimum of a year (N = 105). Partially filled questionnaires were excluded. Several patients commented that they were scoring low due to a recent injury, prosthetic infection, surgery, or conditions such as, c-spine pain, rheumatoid arthritis, stroke, and Parkinson's. These scores were also excluded (16 questionnaires excluded in total). The remainder of the patients were deceased (13) or lost to follow up due to address change or did not respond and were subsequently uncontactable (15 patients). They were included in the review as a minimum 1 year ROM assessment was already documented in an earlier postoperative clinic appointment. Postoperative ROM was obtained from the clinic letters including FF, ER, and Internal Rotation (IR).

Radiographic measures

The radiological parameters were measured as shown in Figure 1, 2 and 3 by the primary author to eliminate inter-rater variability. The CSA was measured as per the study by Roberson et al. ¹⁴ with a line drawn across the glenoid baseplate and a line from the distal end point of the baseplate and tip of the acromion. The RSAA was measured as the angle between a line across the baseplate and a line across the sclerotic line in the supraspinatus fossa. ⁷ Humeral lateralisation (Lateral offset (LO)) was measured as the horizontal distance from the superior point of the glenoid to the most lateral extent of the greater tuberosity. A ratio value was calculated of LO against the distance from the glenoid to the most lateral extent of the acromion (Acromion Index, AI). Distalization or AHI was measured as the vertical distance between the acromion and the humerus. DLA was measured by the horizontal distance from the COR of the shoulder to a line drawn from the most lateral extent of the acromion to the deltoid tuberosity. Postoperative magnification factor was calculated by dividing the measured diameter of the glenosphere over its known size. Measured postoperative overhang, DLA and AHI were divided by the magnification factor to obtain adjusted values. Minimum 1 year postoperative radiograph were available for 138 patients to assess for notching (range: 1 to 7 years, average 2.5 years).

Figure 1.

Preoperative parameters. (a) CSA measurement, a non-rotated standard AP radiograph of the glenosphere is required (b) RSAA measurement, a line through the sclerotic line of the supraspinatus fossa and another from the inferior to the superior point of the glenoid, if the angle is >90, then the RSAA is the negative value of the difference (e.g. 100 degrees means the RSAA is −10 to suggest superior tilt by 10 degrees). (c) DLA, length of the perpendicular line from the COR to a line drawn from the most lateral point of acromion to the deltoid tuberosity. (d) AHI, the vertical distance between two horizontal lines drawn from the top of the GT and the under surface of the acromion.

CSA: critical shoulder angle; RSAA: revere shoulder arthroplasty angle; DLA: deltoid lever arm; AHI: acromio-humeral index; COR: centre of rotation; GT: greater tuberosity.

Figure 2.

Postoperative parameters in a Zimmer trabecular and a Biomet comprehensive prosthesis (a) CSA measurement, a non-rotated standard AP radiograph of the glenosphere is required (b) RSAA measurement, a line through the sclerotic line of the supraspinatus fossa and another on the base of the base plate, in this case the RSAA is −5.7 degrees, which is the amount of superior tilt of the plate relative to a perpendicular line to the sclerotic line of the supraspinatus fossa (c) For illustration DLA is being measured here from a postoperative radiograph of a Biomet comprehensive. (d) LO, horizontal distance from the superior edge of the glenoid to the lateral aspect of the GT marked by a vertical perpendicular line. For ratio with Acromion, another line is drawn from the glenoid superior edge to the most lateral point of the acromion and then the LO is divided by this distance to give a LO/acromion ratio, AI.

CSA: critical shoulder angle; LO: lateral offset or humeral laterization; DLA: deltoid lever arm; RSAA: reverse shoulder arthroplasty angle; AI: acromial index.

Figure 3.

(a, b and c) OH, the distance from the inferior edge of the scapular neck and the inferior edge of the glenospherein a Biomet, Zimmer and Equinox implant respectively. The distance between one line parallel to post of baseplate from inferior edge of scapular neck and another also parallel to post from inferior edge of prosthesis. (d) AHI: The vertical distance from the under-surface of the acromion to the humerus demonstrated here in an Equinox implant. (e and f) Measurement of CSA and RSAA in an Equinox implant. Since the baseplate is not a parallel, the base of the baseplate close to the prosthesis rather than the glenoid was used to measure the angle.

OH: overhang; AHI: acromio-humeral Index; CSA: critical shoulder angle; RSAA: reverse shoulder arthroplasty angle.

Statistical analysis

All statistics were performed using IBM SPSS. Simple linear regression analysis was used to detect correlations between scale values. One-tailed t-test was used to compare means for parametric series. Significant P values were taken as less than 0.05. Receiver Operator Characteristic (ROC) curve and multivariate analysis were performed to isolate radiological measures significantly predictive of notching.

Results

Demographics, change in range of motion and radiological parameters

Table 1 summarises the frequencies of each implant and their published features. IR was found to be recorded in clinical letters by referencing to anatomical landmarks (E.g. buttock, sacrum, and lumbar spine level). These were converted to numerical equivalents to allow data analysis, Table 2. Table 3 summarises the demographics of patients from each centre and the average value of preoperative radiological parameters from each. Table 4 summarises the average change postoperatively in range of motion and radiological parameters from preoperative values. There were significant correlations between all preoperative and postoperative radiological parameters.

Table 2.

Internal rotation estimation.

Quoted internal rotation in the notes	Numerical equivalent
0	−1
‘GT’	0
‘Lateral thigh’	0
‘Side’	0
‘Buttock’	1
‘Sacrum’	2
‘Belt line’	3
‘L5’	4
‘L4/L5’	4
‘Lower lumbar spine’	4
‘L4’	5
‘Lumbar’	6
‘Lumbar spine’	6
‘L3’	6
‘Mid lumbar spine’	6
‘L2’	7
‘L1’	8
‘T11’	9
‘T12’	9
‘T/L junction’	9
‘T10’	10
‘T9’	11
‘T7 or T8’	12
‘Bra’	13

Table 3.

Demographics by hospital summarising age, gender, preoperative radiological parameters, and postoperative outcomes.

Hospital:	NDDH (39 Equinox)	RCH (75 Zimmer, 14 Biomet, 17 Equinox, 11 Aequalis)	Significance of difference
Age	75.8	74.76	P = 0.407
Preop CSA	40.9 N = 34	35.4 N = 107	P < 0.001
Preop RSAA	−17.5 N = 34	−17.6 N = 107	P = 0.975
Preop LO	5.0 N = 33	4.7 N = 96	P =  0.012
Preop AHI	0.64 N = 27	0.82 N = 86	P = 0.069
Preop DLA	1.7 N = 31	1.9 N = 94	P = 0.123
Postop CSA	37.4 N = 35	32.7 N = 117	P = 0.003
Postop RSAA	−13.5 N = 35	−8.4 N = 117	P = 0.006
Postop LO	4.6 N = 34	3.9 N = 114	P < 0.001
Postop AHI	2.9 N = 34	2.7 N = 113	P = 0.278
Postop DLA	3.9 N = 34	4.0 N = 116	P = 0.242
Postop OH	3.6 N = 34	4.0 N = 112	P = 0.368
Postop OSS	37.9 N = 30	43.2 N = 72	P = 0.002
Postop FF	121.6 N = 25	129.3 N = 105	P = 0.237
Postop ER	29.1 N = 17	19.4 N = 94	P = 0.024
Postop IR	5.9 N = 17	5.3 N = 97	P = 0.546

CSA: critical shoulder angle; RSAA: reverse shoulder arthroplasty angle; LO: lateral offset; AHI: acromio-humeral index; DLA: deltoid lever arm; OH: overhang; FF: forward flexion; ER: external rotation; IR: internal rotation; AHI, DLA and LO are in centimetres. OH is in millimetres. Preop/Postop: preoperative/postoperative; NDDH: North Devon District Hospital; RCH: Royal Cornwall Hospital.

Table 4.

Change in ROM and radiological parameters with correlations between preoperative and postoperative parameters.

Variable	Preoperative	Postoperative	Change + 95% confidence interval	Significance of change	Correlation between preop and postop values
FF	75.73	127.61	59 (67.9 to 50.2)	<0.001	P = 0.915 N = 72
ER at side	10.2	19	9.6 (15.7 to 3.5)	0.03	P = 0.211 N = 59
IR	3	4.5	1.4 (2.8 to 0.1)	0.066	p = 0.541 N = 59
CSA	36.9	34.8	2.7 (1.3 to 4.2)	<0.001	P < 0.001 N = 140
RSAA	−17.2	−11.2	7.6 (9.6 to 5.6)	<0.001	P = 0.012 N = 140
AHI	0.8	3.0	2.2 (2.3 to 2.0)	<0.001	P = 0.006 N = 112
DLA	1.8	4.5	2.5 (2.6 to 2.4)	<0.001	P < 0.001 N = 124
LO	4.9	4.6	0.29 (0.1 to 0.4)	<0.001	P < 0.001 N = 125

CSA: critical shoulder angle; RSAA: reverse shoulder arthroplasty angle; LO: lateral offset; AHI: acromio-humeral index; DLA: deltoid lever arm; FF: forward flexion; ER: external rotation; IR: internal rotation; AHI, DLA and LO are in centimetres. Preop/postop: preoperative/postoperative.

Preoperative parameters and postoperative outcomes per implant type

Preoperative and postoperative variables were compared between implant types. Table 5 summarises the average values per implant. Analysis of Variance (ANOVA) done in SPSS, showed that preop CSA, preop LO, postop CSA, Postop RSAA, OH, OSS, Postop FF and Postop ER were significantly different between implants (Table 6). Post-hoc Tukey HSD and Games-Howell showed maintained significance for all the mentioned parameters except OH. The significant combinations on post-hoc are shown in Table 7.

Table 5.

Averages for radiological parameters and postoperative outcomes within each implant followed by the number of data values for each parameter/ outcome per implant (N).

Implant:	Zimmer (N)	Aequalis (N)	Equinox (N)	Biomet (N)
Age	75.1	77.01	74.7	73.72
Gender
Preop CSA	34.7 (69)	35.8 (11)	40.4 (48)	34.9 (13)
Preop RSAA	−16.8 (69)	−17.6 (11)	−18.7 (48)	−17 (13)
Preop LO	4.6 (60)	5.1 (9)	4.97 (48)	4.7 (13)
Preop AHI	0.82 (56)	1 (6)	0.71 (40)	0.7(11)
Preop DLA	1.97 (59)	1.78 (9)	1.795 (45)	1.79 (12)
Postop CSA	34.9 (75)	32.2 (11)	34.2 (52)	27.5 (14)
Postop RSAA	−11.6 (75)	−6.1 (11)	−9.8 (52)	−0.44 (14)
Postop LO	4.12 (75)	3.93 (9)	4.89 (51)	4.93 (13)
Postop AHI	2.9(74)	3.23 (9)	3.11(51)	3.01 (13)
Postop DLA	3.99 (75)	4.35 (10)	3.9 (51)	4.17 (14)
Postop OH	4.44 (74)	2.67 (8)	3.47 (50)	3.82 (14)
Postop OSS	44.2 (51)	35.7(3)	39.8 (46)	40.6 (5)
Postop FF	135 (68)	124.4 (9)	121 (39)	114 (14)
Postop ER	21.4 (60)	4.44 (9)	24.7 (30)	21.25 (12)
Postop Abd	102.5 (12)	90 (1)	100.2 (22)	75 (2)
Postop IR	5.53 (64)	3.78 (9)	5.47 (30)	5.27 (11)

CSA: critical shoulder angle; RSAA: reverse shoulder arthroplasty angle; LO: lateral offset; AHI: acromio-humeral index; DLA: deltoid lever arm; OH: overhang; FF: forward flexion; ER: external rotation; IR: internal rotation; AI: acromial index; OSS: oxford shoulder score; Preop/Postop: preoperative/postoperative.

Table 6.

Pairwise comparison of each radiological parameter/outcome measure, P values shown for each comparison. Significant values are marked with an asterisk.

Implant:	Zimmer Vs Equinox	Biomet Vs Aequalis	Aequalis Vs Equinox	Zimmer Vs Biomet	Zimmer Vs Aequalis	Biomet Vs Equinox
Age	0.369	0.108	0.158	0.445	0.186	0.596
Preop CSA*	0.000*	0.757	0.065	0.938	0.665	0.012*
Preop RSAA	0.296	0.895	0.763	0.952	0.797	0.608
Preop LO*	0.002*	0.211	0.565	0.406	0.029*	0.208
Preop AHI	0.236	0.244	0.175	0.411	0.350	0.962
Preop DLA*	0.119	0.944	0.912	0.366	0.376	0.991
Postop CSA*	0.655	0.130	0.529	0.000*	0.255	0.018*
Postop RSAA*	0.309	0.024*	0.264	0.000*	0.050	0.002*
Postop LO*	0.000*	0.007*	0.000*	0.000*	0.040*	0.289
Postop AHI	0.438	0.676	0.691	0.557	0.969	0.938
Postop DLA*	0.400	0.503	0.027*	0.283	0.060	0.134
Postop OH*	0.021*	0.211	0.441	0.262	0.018*	0.645
Postop OSS*	0.006*	0.027*	0.010*	0.202	0.000*	0.831
Postop FF*	0.017	0.374	0.758	0.011	0.281	0.481
Postop ER*	0.340	0.052	0.004*	0.972	0.006*	0.481
Postop Abd	0.809	0.667	0.740	0.056	0.484	0.258
Postop IR	0.941	0.400	0.231	0.847	0.246	0.872

CSA: critical shoulder angle; RSAA: reverse shoulder arthroplasty angle; LO: lateral offset; AHI: acromio-humeral index; DLA: deltoid lever arm; OH: overhang; FF: forward flexion; ER: external rotation; IR: internal rotation; AI: acromial index; OSS: oxford shoulder score; Preop/Postop: preoperative/postoperative.

Table 7.

Post hoc tests after ANOVA, Tukey and games howell (GH) post hoc tests were done, significant findings are summarised here. Notably, Preop CSA, Preop LO, Postop CSA, Postop RSA, Postop FF and ER have shown maintained significance difference between implants on Post hoc. The significance detected for Preoperative CSA and LO between Equinox and the Zimmer or Aequalis, is due to implantation of these implants in different patient populations with different preoperative radiological parameters (Table 3).

Dependent variable				Mean difference (I-J)	Std. error	Sig.	95% Confidence interval
							Lower bound	Upper bound
Preop CSA	Tukey	Equinox	Zimmer	5.682*	1.395	0.000	2.05	9.31
	GH	Equinox	Zimmer	5.682*	1.374	0.000	2.10	9.27
		Equinox	Aequalis	5.505*	1.943	0.045	0.10	10.91
Preop LO	Tukey	Equinox	Zimmer	.380616*	0.118212	0.009	0.07280	0.68843
	GH	Equinox	Zimmer	.380616*	0.116204	0.008	0.07709	0.68414
Postop CSA	Tukey	Zimmer	Biomet	7.3364*	2.3636	0.012	1.195	13.478
		Equinox	Biomet	6.6703*	2.4444	0.036	0.319	13.022
	GH	Zimmer	Biomet	7.3364*	1.7945	0.003	2.320	12.353
		Equinox	Biomet	6.6703*	2.0973	0.016	1.005	12.336
Postop RSAA	Tukey	Biomet	Zimmer	11.1451*	2.6399	0.000	4.286	18.005
		Biomet	Equinox	9.3860*	2.7302	0.004	2.292	16.480
	GH	Biomet	Zimmer	11.1451*	1.6507	0.000	6.669	15.621
		Biomet	Equinox	9.3860*	1.9597	0.000	4.168	14.604
Postop FF	GH	Zimmer	Biomet	20.788*	7.402	0.049	0.09	41.48
Postop ER	Tukey	Zimmer	Aequalis	16.972*	5.639	0.017	2.25	31.69
		Equinox	Aequalis	20.222*	5.996	0.006	4.57	35.87
Postop OSS	Tukey	Zimmer	Equinox	4.370	1.611	0.039	0.16	8.58
	GH	Zimmer	Equinox	4.370	1.578	0.035	0.22	8.52

CSA: critical shoulder angle; RSAA: reverse shoulder arthroplasty angle; LO: lateral offset; OH: overhang; FF: forward flexion; ER: external rotation; Preop/Postop: preoperative/postoperative.

The Zimmer implant had a significantly higher forward flexion compared to the Biomet and Equinox (Power 0.857 and 0.751 respectively). The Aequalis implant had a significantly lower External rotation compared to the Zimmer and Equinox (Power 0.586 and 0.704 respectively). The significance detected for Preoperative CSA and LO between Equinox and the Zimmer or Aequalis, is due to implantation of these implants in different patient populations with different preoperative radiological parameters (Table 3). There was a significant difference in Postoperative OSS on Post Hoc between the Zimmer and the Equinox (4.4 points).

Correlation between radiological parameters and postoperative outcomes

Table 8 summarises the correlations between radiological parameters, ROM, and OSS. Notably, FF was found to have a significant positive correlation with overhang (P = 0.033), Figure 4. ER had an inverse correlation with DLA (P = 0.024), Figure 5. OSS was inversely correlated with preoperative and postoperative CSA (P = 0.001, 0.003 respectively).

Table 8.

Correlations with significant P values are marked with asterisks. Inverse correlations are indicated with a minus (−). When a correlation is significant N is included.

Parameter	FF N	ER N	IR N	Postop OSS	Notching
Preop CSA	0.097	0.927	0.168	<0.001 95	0.090
Postop CSA	0.715	0.369	0.114	0.003 105	0.005* 137
Preop RSAA	0.8	0.593	0.373	0.431	0.341
Postop RSAA	0.03*(−)125	0.551	0.733	0.127	0.737
OH	0.033* 119	0.193	0.594	0.341	0.188
Preop DLA	0.593	0.687	0.035* 89	0.835	0.032*(−) 114
Postop DLA	0.797	0.024 *(−) 107	0.537	0.925	0.185
Preop LO	0.957	0.007*(−) 87	0.480	0.104	0.2
Postop LO	0.037*(−)122	0.299	0.415	0.013*(−) 102	0.525
Preop AHI	0.077	0.135	0.026*(−) 80	0.143	0.663
Postop AHI	0.881	0.893	0.837	0.481	0.892
Preop AI	0.585	0.219	0.045 (−) 88	0.623	0.811
Postop AI	0.251	0.966	0.502	0.514	0.536
Notching	0.029 (−) 120	0.424	0.607	0.008(−) 94
Postop OSS	<0.001 82	0.001 65	0.018 70		0.008*(−) 94

CSA: critical shoulder angle; RSAA: reverse shoulder arthroplasty angle; LO: lateral offset; AHI: acromio-humeral index; DLA: deltoid lever arm; OH: overhang; FF: forward flexion; ER: external rotation; IR: internal rotation; AI: acromial index; OSS: oxford shoulder score; GLO: published values for implant global lateral offset; Preop/Postop: preoperativepostoperative.

Figure 4.

(A) simple regression tests demonstrating significant effect of postoperative OH with FF. (b) Postoperative external rotation correlations with postoperative deltoid lever arm.

OH: overhang; FF: forward flexion; DLA adjusted: deltoid lever arm adjusted for magnification.

Figure 5.

Effect of implant type (global lateral offset) (a) and measured LO (b) on FF.

LO adjusted: lateral offset adjusted for magnification; postop FF: postoperative forward flexion.

To eliminate the effect of implant, the correlations were checked within each of the Zimmer and the Equinox implant groups separately as they were the most frequently implanted .

Tables 9 and 10 Show the correlations described in Table 8 tested again within the Zimmer and the Equinox respectively.

Table 9.

Zimmer implant. Correlations with significant P values are marked with asterisks. Inverse correlations are indicated with a minus (−). When a correlation is significant N is included.

Parameter	FF N	ER N	IR N	Postop OSS	Notching
Preop CSA	0.480	0.733	0.175	0.012*(−)46	0.788
Postop CSA	0.359	0.787	0.295	0.118	0.261
Preop RSAA	0.231	0.099	0.849	0.151	0.612
Postop RSAA	0.512	0.487	0.824	0.165	0.676
OH	0.063	0.661	0.872	0.824	0.942
Preop DLA	0.282	0.854	0.006* 47	0.702	0.029*(−)55
Postop DLA	0.021* 67	0.277	0.672	0.106	0.326
Preop LO	0.303	0.058	0.320	0.887	0.357
Postop LO	0.088	0.046*(−) 59	0.797	0.298	0.127
Preop AHI	0.110	0.078	0.137	0.006*35	0.965
Postop AHI	0.551	0.442	0.886	0.870	0.774
Preop AI	0.321	0.240	0.023*(−) 48	0.458	0.902
Postop AI	0.380	0.483	0.975	0.595	0.481
Notching	0.016*(−) 65	0.136	0.042*(−) 61	0.881
Postop OSS	0.006* 46	0.110	0.113		0.881

CSA: critical shoulder angle; RSAA: reverse shoulder arthroplasty angle; LO: lateral offset; AHI: acromio-humeral Index; DLA: deltoid lever arm; OH: overhang; FF: forward flexion; ER: external rotation; IR: internal rotation; AI: acromial index; OSS: oxford shoulder score; GLO: published values for implant global lateral offset; Preop/Postop: preoperative/postoperative.

Table 10.

Equinox implant. Correlations with significant P values are marked with asterisks. Inverse correlations are indicated with a minus (−). When a correlation is significant N is included.

Parameter	FF N	ER N	IR N	Postop OSS	Notching
Preop CSA	0.053	0.838	0.127	0.128	0.416
Postop CSA	0.999	0.290	0.249	0.023* (−)45	0.021*44
Preop RSAA	0.609	0.246	0.012* 25	0.628	0.321
Postop RSAA	0.361	0.371	0.798	0.150	0.397
OH	0.693	0.830	0.659	0.112	0.705
Preop DLA	0.490	0.353	0.354	0.611	0.439
Postop DLA	0.625	0.714	0.817	0.392	0.129
Preop LO	0.758	0.035*(−) 24	0.296	0.672	0.856
Postop LO	0.761	0.660	0.505	0.524	0.063
Preop AHI	0.927	0.810	0.409	0.921	0.948
Postop AHI	0.541	0.011* (−) 26	0.777	0.898	0.641
Preop AI	0.267	0.472	0.202	0.783	0.874
Postop AI	0.764	0.984	0.801	0.619	0.009*(−)44
Notching	0.461	0.849	0.094	0.018(-)*38
Postop OSS	0.002* 28	0.290	0.224		0.018*(−)38

CSA: critical shoulder angle; RSAA: reverse shoulder arthroplasty angle; LO: lateral offset; AHI: acromio-humeral index; DLA: deltoid lever arm; OH: overhang; FF: forward flexion; ER: external rotation; IR: internal rotation; AI: acromial index; OSS: oxford shoulder score; GLO: published values for implant global lateral offset; Preop/Postop: preoperative/postoperative.

The correlations in Table 8 were detected across all implants. Consequently, some of these correlations may have been caused by the type of implant. For instance, Postop RSA had a negative correlation overall with FF (Table 8). This is likely been caused by implant type as lateralised implants (which have lower FF, Figure 5), have higher RSA (Figure 6). Expectedly, the correlation was lost when analysed within the Zimmer and Equinox (Tables 9 and 10). Among the significantly correlated parameters in Table 8, the following parameters had significant correlations with outcomes within either the Zimmer or Equinox: Preop CSA, Postop CSA, preop LO, postop LO, preop AHI and postop DLA (Tables 9 and 10). These radiological parameters can therefore affect outcomes independent from implant type. Notably Postop CSA was correlated with OSS (P value = 0.003 overall and 0.023 within Equinox). The power for the correlation between OSS and CSA was >0.8 overall. Postoperative DLA measured values had no significant correlation with OSS and/or notching overall. Corrected values were obtained by dividing by the glenoid size as patient size is likely to affect how much DLA is significant. The corrected values showed positive correlation with OSS and notching overall (P = 0.031 (N = 100) and 0.027 (N = 113)). Corrected DLA values were also correlated with OSS in the Zimmer implant group (P = 0.044, N = 50).

Figure 6.

(a) Variation of measured RSAA with different implant types as shown by the different global lateral offsets. (b) CSA correlation with OSSs.

CSA: critical shoulder angle; OSSs: oxford shoulder score; Postop: postoperative.

For OH, although no significant correlation was detected within either Zimmer or Equinox (Only across all implant types); an effect independent from implant was suspected (Note a weak positive correlation between OH and FF in Zimmer 0.063). According to Table 5, Equinox, Biomet and Aequalis were on average implanted in less overhang than Zimmer. All three implants had less FF compared to the Zimmer. Figure 5 shows that FF is negatively correlated with lateralisation. Therefore, the observed correlation between OH and FF may have been caused by the lateralised implants (Biomet and Equinox) which on average have less OH than the Zimmer. To establish if this is the case; patients were divided into those with FF >120 and those ≤120 and an analysis of averages done per implant type (Table 11).

Table 11.

Average overhang in each FF group (FF > 120 and FF ≤ 120) for each implant and overall. A power calculation is shown for the difference in means in overhang between the two groups across all implants.

Implant	Postop FF > 120	Postop FF ≤ 120	P value of difference (power)
	Average OH (N)
Zimmer	4.47 (47)	3.97 (20)
Aequalis	4.43 (3)	1.29 (3)
Equinox	4.7 (15)	3.44 (18)
Biomet	4.23 (5)	3.6 (9)
Total	4.5 (70)	3.54 (50)	0.006 (0.810)

OH: overhang; FF: forward flexion; Preop/Postop: preoperative/postoperative.

Overall, Patients with FF at >120 had an average OH of 4.5 mm, those with FF ≤120 had an average OH of 3.54 mm. The means were significantly different with sufficient power (P = 0.006 and Power = 0.810), Table 11. Within each implant group, the mean for OH was consistently less in those with FF ≤120 compared with those >120 FF, Table 11. Therefore, OH as a radiological parameter does affect FF independently of implant type. The significance was lost within implant groups likely due to insufficient numbers (Note P value narrowly missed significance at 0.063 for OH versus FF within Zimmer implants, Table 9). An ROC analysis suggests a cut off value of 6 mm overhang to improve FF.

Correlations between radiological parameters and notching

Notching was found in 21 out of 138 suitable cases (minimum year follow up radiograph), grade 4 (2 Aequalis), grade 2 (1 Biomet Comprehensive, 1 Aequalis, 1 Equinox and 1 Zimmer Trabecular) and grade 1 (5 Zimmer Trabecular, 8 Equinox, 1 Aequalis and 1 Biomet Comprehensive), Table 1.

Multivariate analysis which included CSA, RSAA, NSA, DLA and OH showed CSA to be significantly correlated with Notching (P = 0.006) followed by postop DLA (P = 0.033) with the rest of the parameters having no significant effects. CSA's individual correlation with Notching was significant at P = 0.005 and a power of 0.816. Postop DLA when corrected for glenoid size, was correlated with notching (p = 0.007).

The same multivariate analysis was performed within the Equinox implant, CSA, DLA and RSA were correlated with notching (P = 0.001, 0.001 and 0.008 respectively, inverse correlation in case of DLA and RSA) with no significant effect from OH. The multivariate analysis when performed in the Zimmer showed preoperative DLA only as significant (P = 0.013, inverse correlation). Aequalis showed no significant correaltions. In Biomet, overhang was significantly correlated with Notching (P = 0.006, inverse correlation) with no significance from any other parameter.

ROC curve analysis identified increased CSA as the most significant predictive factor for notching with an Area Under Curve (AUC) of 0.726, followed by preoperative/postoperative DLA then Preoperative CSA Figure 7. OH, RSAA and NSA, had poor to negligible effects on the rate of notching overall, according to ROC.

Figure 7.

ROC curve analysis postop CSA (graph b) as most predictive for notching followed by preop and postop DLA (graph a)). CSA is shown in a separate graph as larger values are associated with notching while the opposite holds true for RSAA, DLA and OH. © AUC for each parameter in order of influence.

Adj OH: overhang adjusted for magnification; LO adjusted: lateral offset adjusted for magnification; RSAA: reverse shoulder arthroplasty angle; CSA: critical shoulder angle; NSA: neck shaft angle; DLA adjusted: deltoid lever arm adjusted for mangification; ROC: receiver operator characteristic; Postop/Preop: postoperative/preoperative.

ROC analysis suggests aiming for a CSA of no more than 26 degrees to decrease the rate of notching. and improve OSS to values over 35/48 (sensitivity and specificity >0.8). This corresponds to a superior tilt of no more than 2 degrees relative to the sclerotic fossa (or RSAA = −2 degrees).

Mean overhang in those who Notched was 0.71 mm less than those who did not (the means being not significantly different between those who notched and those who did not). ROC analysis suggested a less significant effect from OH compared to CSA and DLA.

Discussion

The study confirms the RSA as an effective intervention for most patients with rotator cuff issues and osteoarthritis. 32 patients achieved OSS of 48/48 with 78 patients achieving scores of at least 40, about 30% of the OSSs were less than 40. 50% of patients achieved FF of at least 130 degrees postoperatively. The Zimmer implant showed a positive correlation between postoperative DLA and Postoperative FF. Across all implants, DLA (Corrected for glenoid size) was correlated with OSS, and this was also the case within the Zimmer implant. This illustrates the biomechanical principle behind the RSA; centre of rotation medialisation to increase deltoid moment arm which offloads the rotator cuff.

Implant design and radiological parameters correlation with postoperative range of motion

The more recent RSA designs lateralise the centre of rotation to decrease scapular notching and increase external rotation. However, there is still controversy around the benefits of lateralisation in the RSA particularly in decreasing Notching. There are also concerns that departing from Grammont's principle by lateralising the centre of rotation can lead to early failure.

The Aequalis implant (a relatively medialised implant) had a significantly lower External rotation compared to the Zimmer and Equinox with the significance maintained on Post-hoc tests. Biomet, Equinox, and Zimmer did not differ significantly in their external rotation. Mean difference in ER was insignificant between Zimmer and Equinox, the two most frequently implanted prostheses (N = 60 and 30 for Postoperative ER) at only 3.3 degrees (higher in Equinox). The Zimmer and Equinox only have a 7.4 mm difference in Global Lateral offset and a 1.4 mm difference in Glenoid lateral offset. The Aequalis and Equinox have a 10.8 and 5.3 difference in Global and Glenoid lateral offset respectively. Based on this, the lack of a significant difference between Equinox and Zimmer may be due to their close lateral offsets. Based on the standard deviations obtained for postoperative ER in Equinox and Zimmer; to obtain sufficient power, a sample size of 728 implants is required. A much smaller sample size of 50 is required to show a powered ER difference between Aequalis and Equinox based on the values obtained (N = 39). Therefore, any advantage in ER in the Equinox relative to the Zimmer; can be difficult to demonstrate due to their close offsets. Verdano et al. showed that ER improved with the arm in abduction but not in adduction in lateralised designs while Greiner et al. showed that ER only improved in patients with intact Teres minor. ¹³ Valenti et al. showed ER in adduction improved by 15 degrees versus 30 degrees in abduction. ¹⁹ The improvements in ER appear to be due to increases in cuff tension. ¹² Patients with Massive Cuff tear were included, which may have confounded the results. The known negative effect a standard RSA has on ER has been confirmed, by demonstrating an inverse relationship between DLA (which would increase with medialisation) and ER. This agrees with the findings by Frankle et al. Rhee et al. and Cuff et al.^10,20,21

FF was found to be inversely correlated with LO and significantly decreased in lateralised implants. The Zimmer had a higher FF compared to the Equinox and Biomet (P = 0.017 and 0.011, power = 0.857 and 0.751 respectively). These findings agree with Verdano et al. ²² and Rhee et al. ¹⁰ Overhang was significantly associated with better FF echoing findings by Haidamous et al. ¹⁶ but disagreeing with those by Rhee et al. ¹⁰ For OH when the correlation was re-checked within each implant, the significance was lost, likely due to insufficient numbers (In the Zimmer P value narrowly missed significance at 0.063 for correlation between OH and FF). However, in all implant groups, those with FF ≤120 had a consistently lower average OH compared with those with FF >120. Overall, the power of the difference in OH between patients with FF ≤120 and >120, was 0.810 with a P of 0.006.

Radiological parameters correlation with notching and OSS

21 cases of notching (out of 138 cases) were detected at an average 2.5 year follow-up radiograph (1–7 years). Scapular notching was associated with a higher CSA. This association was more significant than associations between notching and each of NSA, OH and RSAA. Lateralised implants did not lower notching rate in contrast to previous studies^19,23,24 but in agreement with Rhee et al. ¹⁰ The findings by Katz et al. showed that while lateralised designs decreased the rate of notching, they did not eliminate it altogether with a rate of 29% (19). The study by Verdano et al. showed that low grade notching was in fact higher in the lateralised group. ²²

Inferior baseplate tilt has been controversial in its role in reducing notching with studies advocating it in reducing notching such as the one by Huri et al. (65 prostheses) ²⁴ which had a 2 year minimum follow up. Edwards (42 prostheses) and Kempton et al. (65 prostheses) suggested it does not affect the rate of notching, at one year follow up.^5,9,25 Duethman et al. had a five year follow up and found that notching increased with inferior baseplate tilt (147 prostheses). ⁵ Duethman reviewed both lateralised and non-lateralised implants while Kempton et al. only reviewed Delta prostheses. Huri showed that inferior tilt is relevant in decreasing notching in lateralised prostheses (They reviewed 18 Grammont type prostheses and 45 Lateralised designs). The benefits of inferior tilt also extend to increased stability and decreased impingement.^26,27 As an example, for further controversy, two biomechanical studies have come to opposite conclusions regarding whether to implant in inferior tilt.^28,29 Gutierrez et al. suggested that the effect of inferior baseplate tilt may be different in lateralised versus inferior offset prostheses. ³⁰ In their computer model, forces were most evenly distributed in lateralised glenospheres when implanted in inferior tilt, while inferior offset glenospheres had most even force when implanted in neutral. Our findings support this; in the lateralised Equinox multivariate analysis showed that decreased CSA improved OSSs and inferior baseplate tilt had a protective effect against notching. This was not the case in the Zimmer implant or the Aequalis. This may help to resolve the controversy showing that inferior baseplate tilt is at the very least more beneficial in lateralised implants than it is in medialised implants. Two recent studies showed that a small amount of baseplate superior tilt is tolerated with no significant effect on Scapular notching or PROMs.^31,32

To eliminate the effect of implant on Notching or postoperative outcomes the analysis was repeated within the Zimmer and Equinox implant groups as they were most frequently implanted (N = 75 and 56 respectively). Postoperative CSA was significantly correlated with Notching and OSS in the Equinox implant and when all implants were pooled. Overall, the power was >0.8 for OSS and notching respectively. Preoperative CSA was correlated with OSS in the Zimmer (P = 0.01)

RSAA was significantly higher in the Biomet Comprehensive. However, this did not translate to improved OSSs in Biomet implants. This may be due to a compensatory negative affect on OSSs through a decreased flexion or through increased stresses at the glenoid-bone interphase of lateralised implants as shown by Yang et al. and Roche et al.^33,34 Zimmer and Equinox had significantly different OSSs (Higher in the Zimmer by 4 points) with the significance maintained on Post hoc analysis. Increased preoperative AHI was associated with increased postoperative OSS. Preoperative AHI, being a marker of the degree of rotator cuff disease, humeral head superior migration and posterior subluxation (due to Infraspinatus tear) ³⁵ was shown here and in previous studies to affect postoperative outcomes and notching.^36,37

Aetiology of notching

Preoperative CSA and DLA were inversely correlated with postoperative OSS and Notching respectively. Postoperative CSA and Postoperative DLA were also inversely correlated with notching. Postoperative CSA and DLA were correlated with preoperative values explaining how patient preoperative shoulder morphology may have significant effects on their postoperative outcomes in RSA.

How preoperative or postoperative DLA can affect notching rate is not immediately obvious, however, it is known that a higher DLA (medialised Centre of rotation) decreases joint reaction forces, shear forces and micromotion. The mechanical hypothesis of impingement of the humeral component against the scapular neck is the most widely accepted for scapular notching pathophysiology.^38,39 Some authors have explained that notching can occur or be exacerbated by polyethylene wear from micromotion and subsequent reactional osteolysis.^3,38,40 Osteolysis better explains why the Notch is often more extensive than what would be expected from impingement alone. ³⁸ A lower DLA will result in less compressive forces and more shear, ⁴¹ leading to a less stable prosthesis with more micromotion, ⁴¹ wear and resultant osteolytic notching. Preoperative DLA was inversely correlated with Notching. Postoperative DLA raw values were not significantly correlated with notching, however, when corrected for the patient's size (by dividing the measured value over glenoid size), an inverse correlation was detected at P = 0.017 (as measured DLAs will vary in their effects from patient to patient depending on the patients’ sizes). Postoperative DLA divided by glenoid size was also correlated with OSS across all implants and within the Zimmer implant (P = 0.031 and 0.044 respectively).

A superiorly tilted prosthesis can cause conversion of compressive forces to shear forces and resultant notching.^29,30 Since, the deltoid muscle attaches to the acromion the amount of shear or compression generated will be influenced by the acromion's morphology. The CSA is measured to the tip of the acromion. This may explain why the CSA showed more influence on OSS and notching than the RSAA which is not measured from the acromion tip. The effect of CSA on OSS and notching were significant overall with power >0.8 but also demonstrated within the Equinox. It is noteworthy that postoperative AI (measured by dividing LO over Acromion distance from the glenoid's superior tip) was inversely correlated with notching in Equinox. This suggests that patients with a higher acromion index are more likely to develop notching.

Overall, these results demonstrate how acromion morphology can potentially influence the amount of shear generated by the deltoid. Too much shear in patients with superiorly tilted baseplates, or those with a decreased DLA can cause instability leading to notching.

Clinical implications of notching

OSSs were positively correlated with FF, ER, IR, Abduction and inversely correlated with Notching. Notching was inversely correlated with FF and IR. These findings suggest that notching can be a marker of poor postoperative outcomes. This is in contrast to earlier findings by Levigne, Werner and Boileau^37,38,42 but in agreement with the more recent findings by Sirveaux, Simovitch and Jang.^3,43,44

Conclusion

Limitations of our study include its relatively moderate sample size of 156 implants, being retrospective and the lack of preoperative OSSs for comparison. However, compared to the available studies by Huri, Edwards, Kempton and Duethman et al.^5,6,9,24 we have provided significant data on notching in 70 Zimmer implants and 45 Equinox implants with an average 2.5 year follow up (minimum a year). We have not assessed the influence of surgeon which may have confounded the results. We have not investigated complications postoperatively. Our study provides in-vivo evidence of a benefit of decreased CSA (or more inferior baseplate tilt) as shown by higher postoperative OSSs. The role of the acromion and deltoid in influencing OSS and Notching was demonstrated.

 The range of RSAA values measured (N = 152, −36 to 10, Mean = −9.6, Standard deviation 9.6); suggest a tendency for the baseplate to be implanted in superior tilt relative to the supraspinatus fossa sclerotic line. Future studies can consider the role of robotic assistance, intraoperative navigation, preoperative planning, and mixed reality in controlling baseplate tilt intraoperatively.

CSA was significantly correlated with Notching and OSS overall, but also within the Equinox implant group separately (power > 0.8 overall). ROC analysis specifies a maximum CSA of 26 degrees (or RSAA of no less than −2 degrees) to avoid notching and improve OSSs to values over 35/48.

Increased overhang increased FF with a power of >0.8. To our knowledge, there is only one other recent study showing a positive association between overhang and FF. ¹⁶ ROC analysis specifies a minimum of 6 mm overhang when implanting the baseplate to improve FF to values over 120 degrees (power > 0.8) with a sensitivity and specificity >0.8.

In conclusion, baseplate tilt in relation to the acromion (CSA) and DLA are more predictive of notching and OSS than baseplate tilt (RSAA) and OH. This suggests a role of the Acromion and Deltoid muscle in the pathophysiology of notching especially in patients with superiorly tilted prosthesis and/or a low DLA.