Does primary treatment of proximal humerus fractures show favourable functional outcomes over secondary treatment with reverse shoulder arthroplasty?

Abstract

Background

This multicentre retrospective cohort study assessed whether functional outcomes after primary reverse shoulder arthroplasty (RSA) are favourable compared to secondary placement in elderly patients with displaced proximal humerus fractures (PHFs).

Methods

Fifty-three patients with primary and 32 with secondary RSA were included. Patient-reported outcome measures (PROMs) were assessed: Constant–Murley Score (CMS), Oxford Shoulder Score (OSS), Disabilities of the Arm, Shoulder and Hand (DASH) score, and Visual Analogue Scale (VAS). In addition, range of motion (ROM) was compared between groups.

Result

For PROMs, the means (SD) for primary versus secondary were 25.4 (17.7), 29.4 (19.2) for DASH; 38 (8.6), 38 (9.1) for OSS; 63 (19.8), 59 (22.0) for CMS and 2 (2.0), 3 (2.3) for VAS. For ROM, the means were the following: forward flexion 113° (33.6), 106° (34.1); abduction 103° (33.4), 96° (37.3) and external rotation 20° (19.1), 20° (17.8). There were significant differences in favour of primary treatment in forward flexion (p = 0.003, B 19.85) and abduction (p = 0.034, B 17.34).

Discussion

ROM in patients with complex displaced PHFs after primary RSA is slightly better than that after secondary treatment. Therefore, RSA could be considered primary treatment, especially when optimal ROM is of great importance to the patient.

Level of evidence

level III, retrospective comparative study treatment study

Introduction

The treatment of complex proximal humerus fractures (PHFs) remains controversial. Depending on patient and fracture characteristics and surgeon preferences, surgeons have two options: to operate or not to operate. Different surgical treatment options include open reduction and internal fixation (plate osteosynthesis), cephalo-medullary nail fixation, reverse (fracture) shoulder arthroplasty (RSA) and fracture hemiarthroplasty.^1–6

All primary head-preserving treatments, either conservative or after surgical reduction and fixation, face the risks of non-union and avascular necrosis, with an incidence of 5% and 8.5%, respectively.^7–11 In addition, poor functional outcomes and persistent pain have been reported.^7–11 In an older, low-demand patient group, the different head-preserving surgical treatments have a high failure rate, and low satisfaction has been reported, since they are prone to reconstruction collapse due to osteoporosis, rotator cuff dysfunction or poor reconstruction (e.g. persistent medial hinge displacement and/or varus deformity).^12,13 In these cases, surgical treatment with RSA is often considered.^8,14–17

For complex, displaced PHFs in the elderly, RSA is emerging as a preferred treatment option.^3,18–20 The advantages comprise the offset of the ball and socket using the RSA; through medialization of the centre of rotation, the shoulder function improves by extending the lever arm for the deltoid muscle.^21–23 Consequently, the functionality of the arm is less dependent on the rotator cuff.^23,24 However, the RSA remains the last resort and is therefore often used with precaution or as a salvage procedure.^25,26 Nonetheless, RSA as primary treatment for displaced PHFs can be favourable over secondary RSA after failed treatment. The tuberosities can be anatomically reinserted more easily in primary treatment, which renders better functional outcome as well as less risk of dislocation. ²⁷ In secondary treatment, the tuberosities are often unavailable, or the rotator cuff is atrophied and/or substantially retracted, leading to worse functional results.^28–30 Furthermore, surgery after failed conservative or surgical treatment is more demanding due to the inexistence of normal anatomical landmarks, the need to revive the edges of the fracture, and the removal of previously placed hardware. ³¹

Even though the secondary RSA is described as more challenging and is considered to have a higher number of complications, there is no sound evidence for a clinically relevant advantage of primary RSA over secondary RSA in the current literature.^32,33 This is due to the studies being small and/or having study designs that prohibit strong inferences to be made. However, those studies show a trend in which there are slightly more complications and a reduced range of motion (ROM) in the secondary RSA group. ³³

Since the RSA is becoming more popular for primary and secondary treatments, the question arises whether, for both indications, patients display the same functional outcome. We hypothesise that the primary treated patients have better functional outcomes than secondary treated patients because a salvage procedure after failed treatment is markedly more demanding than primary treatment.

Methods

In this retrospective, multicentre study (Zaandam Medical Centre and Spaarne Gasthuis, both level 2 trauma centre and tertiary referral centre for PHFs), patients treated with an RSA were included between 2016 and 2021. Patients were categorised into two groups based on the indications for RSA. The first group was composed of patients with a primary RSA, defined as surgically treated with RSA less than 3 months after the initial fracture. Inclusion criteria for this group consisted of at least two of the following indications for an RSA: age over 70 years, 3- or 4-part fracture and marked displacement, osteoporosis, head-split, or fracture dislocation. ¹⁶ Inclusion criteria for the secondary group consisted of patients with RSA after conservative treatment for at least 3 months (delayed union) and of patients with an RSA after failed surgical treatment, consisting of either plate osteosynthesis or cephalo-medullary nail fixation. Patients with a language barrier, dementia, or neurological disorders of the upper extremities were excluded from both groups.

The primary outcome measures of this study were the patient-reported outcome measures (PROMs) and ROM. For the PROMs, the following questionnaires were assessed: the Disability of the Arm, Shoulder and Hand (DASH) score, Oxford Shoulder Score (OSS), and Constant–Murley Score (CMS) absolute and adjusted to age and sex. ³⁴ Pain was assessed through the Visual Analogue Scale (VAS). The ROM of the shoulder was measured bilaterally with a protractor. One researcher assessed these outcome measures to minimise any bias. As secondary outcome measures, the complications and radiographic images were evaluated. The complications were extracted from the patient files. To assess the fracture types and complications, all radiographs at the time of injury were assessed by the researcher LSB and a trauma surgeon RJD and classified according to the Neer criteria. ³⁵ The fracture sequelae were classified based on the sequela classification by Boileau et al. ³⁶ All patients received radiographs according to the hospitals’ protocols at approximately 6 weeks and 3 months (or until the union of the tuberosities was clearly present or absent [tuberosity non-union]). As such, the last radiographs of the patients were used to assess tuberosity healing, component loosening, and scapular notching. All patients had a minimum follow-up of 1 year. Functional outcomes, in general, do not improve after 1 year, therefore justifying our cut-off point at 1 year in our opinion. ³⁷

Surgical technique

Patients were surgically treated by four (orthopaedic) surgeons with considerable expertise in shoulder injuries. Patients were positioned in a beach-chair position. Surgery was performed under general anaesthesia combined with an interscalene brachial plexus block. The Affinis Fracture Inverse© (Mathys Ltd., Bettlach, Switzerland) or the DELTA XTEND™ Reverse Shoulder System (DePuy Synthes, Warsaw, USA) was implanted through a deltopectoral or superolateral approach. Recent literature does not find significant differences between different prosthesis designs, and therefore, we decided to include both prostheses (from Mathys as well as Synthes), which allowed us to include enough patients to meet the sample size.^38,39

During the procedure, a tenotomy of the long head of the biceps was performed. The baseplate was fitted to border the inferior part of the glenoid to minimise scapular notching, with neutral or 10° of interior tilt. The tuberosities were fixated with sutures, and the supraspinatus tendon was resected in some cases and in others, if preserved, was fixated with a MaxBraid™ (Zimmer Biomet, Warsaw, USA). A few patients were treated with a Supercable (Kinamed©, Camarillo, USA), which is a synthetic cerclage wire with a metal locking mechanism, used for the adherence of the tuberosities. However, some patients experienced discomfort due to the prominent locking mechanism. Therefore, the surgeon switched to the NICELOOP™ (Wright Medical, Memphis, USA) as a cerclage ‘wire’ through the anterior and posterior rotator cuff insertions and the designated hole in the neck of the prosthesis or fixed around the stem of the prosthesis.

Statistical analyses

The sample size calculation was based on the post-operative CMS of our previous published study, with a standard deviation of 17. ⁴⁰ To detect a relevant difference between groups, we used the minimally important change of 12. ⁴¹ With an alpha of 5% and power of 80%, 32 patients were needed per group.

For statistical analyses, IBM SPSS statistics 24 was used. Patient characteristics have been described by general descriptive statistics. The data were checked, based on the histograms, for normality. To visualise the differences between both groups, boxplots were plotted. For normally distributed data, we used the one-way Student t-test, and for the non-normally distributed data, we used the non-parametric Mann–Whitney U test. For categorical variables, a chi-squared test was used. A p-value of <0.05 was considered significant.

A linear regression analysis was used to investigate confounding effects on the outcome measures. The regression model contained the PROMs and ROM as dependent variables, and Neer classification, anticoagulant use, sex and age were used as independent variables. We used a backward procedure with an α of 0.05 as a threshold for statistical significance. Beta estimates (mean differences between both groups), p-values and 95% confidence intervals are reported.

Results

Baseline results

In this study, 85 patients were included, of which 53 underwent RSA as a primary treatment and 32 patients as a secondary treatment (Figure 1). Of the 32 patients who received RSA as a secondary treatment, 17 were treated conservatively, and 15 patients in the sequela group were treated surgically. Table 1 shows the distribution of the sequela types according to the classification by Boileau et al. ³⁶ The mean age of the patients was 73.0 years (SD 9.0). The mean follow-up time was 20 months (SD 13.6). We found statistically significant differences between patients treated with RSA as a primary treatment and those who received RSA as a secondary treatment; in the primary group, there were more females, there was more anticoagulant use, and fractures had a higher Neer classification (Table 2).

Figure 1.

Flowchart inclusion of patients.

Table 1.

Sequela classification.

Type of sequela	Secondary treated patients
	(N = 32)
Type I – cephalic collapse or necrosis	11 (34.4%)
Type II – locked dislocation or fracture dislocations	5 (15.6%)
Type III – surgical neck non-unions	12 (37.5%)
Type IV – severe tuberosity malunions	4 (12.5%)

Table 2.

Baseline characteristics

Variable*	Primary treated patients	Secondary treated patients	Total	p-value
	(n = 53)	(n = 32)	(n = 85)
Sex, female	48 (90.6%)	23 (71.9%)	71 (83.5%)	0.024**
Age, years	74.4 (7.8)	70.7 (10.4)	73.0 (9.0)	0.065
Time to surgery, weeks	2.6 (2.3)	51.0 (40.6)	20.8 (34.2)	0.000**
Follow-up, months after RSA	16.5 (6.9)	25.7 (19.2)	20.0 (13.6)	0.221
ASA classification 1 2 3	6 (11.3%) 28 (52.8%) 19 (35.9%)	3 (9.4%) 20 (62.5%) 9 (28.1%)	9 (10.6%) 48 (56.5%) 28 (32.9%)	0.683
Anticoagulant use	21 (39.6%)	6 (18.8%)	27 (31.8%)	0.045**
Neer classification 1 part 2 part 3 part 4 part	0 18 (34.0%) 22 (41.5%) 13 (24.5%)	4 (12.5%) 14 (43.8%) 10 (31.3%) 4 (12.5%)	4 (4.7%) 32 (37.7%) 32 (37.7%) 17 (20.0%)	0.028**
Head-split	12 (22.6%)	3 (9.7%)	15 (17.9%)	0.134
Dislocation fracture	9 (17.0%)	5 (16.1%)	14 (16.5%)	0.919

*Continuous data are presented as mean (SD) and categoric data as number of patients (percentage of group of patients).

**Statistically significant p < 0.05.

Primary outcome measures: PROMs and ROM

With regard to the PROMs (Table 3 and Figure 2), no statistically significant differences were found, but minimal differences in favour of the primary treated group in comparison to the secondary treated group were found for the DASH scores (mean 25.40 versus 29.35, respectively) and CMS (63.0; 58.6, resp.). The linear regression analysis showed two confounding effects (sex on DASH and anticoagulant use on VAS) that were not statistically significant (B −6.400, p = 0.122; B 1.073, p = 0.241, respectively).

Table 3.

Patient-reported outcome measures.

PROMs*	Primary treated patients	Secondary treated patients	Total	p-value**
	(n = 53)	(n = 32)	(n = 85)
	Mean (SD)	Mean (SD)	Mean (SD)
VAS	2.1 (2.0)	2.5 (2.3)	2.3 (2.1)	0.489
DASH	25.4 (17.7)	29.4 (19.2)	26.9 (18.3)	0.337
OSS	37.9 (8.6)	37.8 (9.1)	37.9 (8.7)	0.957
CMS	63.0 (19.8)	58.6 (22.0)	61.4 (20.6)	0.336
CMS, adjusted (%)	75.6 (21.9)	69.7 (25.1)	73.4 (23.2)	0.255

*Data are presented as mean (standard deviation).

**Statistically significant p < 0.05.

Figure 2.

Boxplots PROMs.

The results for the ROM (Table 4 and Figure 3), the forward flexion (112.8°; 103.6°, resp.) and abduction (103.0°; 95.8°, resp.) showed minimal differences in favour of primary treatment (Table 4). The linear regression analysis showed that a combination of Neer classification and anticoagulant use had a statistically significant effect on the outcomes of forward flexion and abduction (Table 5). When corrected for this effect, results for primary treatment in forward flexion and abduction showed a statically significant and clinically relevant benefit compared to secondary treatment. ⁴²

Table 4.

Range of motion.

ROM*	Primary treated patients	Secondary treated patients	Total	p-value**
	(n = 53)	(n = 32)	(n = 85)
	Mean (SD)	Mean (SD)	Mean (SD)
Forward flexion	112.8 (33.6)	103.6 (34.1)	109.3 (33.9)	0.225
Extension	43.8 (13.5)	43.0 (12.4)	43.5 (13.1)	0.770
Abduction	103.0 (33.4)	95.8 (37.3)	100.3 (34.9)	0.360
External rotation	20.1 (19.1)	19.7 (17.8)	19.9 (18.5)	0.920

*Data are presented as mean (SD).

**Statistically significant p < 0.05.

Figure 3.

Boxplots ROM.

Table 5.

Linear regression model.

Forward flexion	Unstandardised B	p-value	95% confidence interval
Primary treatment	19.849	0.009	[5.053–34.644]
Neer classification	−49.655	0.007	[−85.061–−14.248]
Anticoagulant use	−26.541	<0.001	[−41.013–−12.068]
Abduction
Primary treatment	17.342	0.034	[1.375–33.308]
Neer classification	−47.583	0.015	[−85.791–−9.375]
Anticoagulant use	−23.053	0.004	[−38.671–−7.436]

In this study, the primary treated patients were compared to secondary treated patients with a cut-off point of 12 weeks between both groups. To avoid regression to the mean and further distinguish primary and secondary treated patients, an additional analysis was performed, in which primary treated patients who were treated up to 4 weeks of their injury till 12 weeks (the threshold for secondary treated patients) were excluded. This analysis consisted of 44 patients receiving RSA as primary treatment and 32 who received RSA as secondary treatment. The baseline characteristics showed a significant difference in Neer classification, which was higher in the primary group (p = 0.020). However, the results of this analysis did not show any statistically significant differences between the groups.

To adjust for the heterogeneity of the group (sequelae after conservative treatment as well as after surgical treatment), two sub-analyses were performed. The first analysis was a comparison between the primary group and the sequela patients who were first treated with conservative treatment. The second analysis was a comparison between the primary treated group and the sequela group of prior surgically treated patients. Both analyses did not find any statistically significant differences in the outcome measures.

Secondary outcome measures: complications and radiographic outcomes

With regard to the complication rate, no significant difference between primary and secondary treated patients was found (Table 6). However, implant infection occurred twice in secondary treated patients and not in the primary treated patients. The first patient was treated by debridement and placement of gentamicin beads. After 3 weeks, the beads were removed, and since the prosthesis was not reached, further treatment consisted of antibiotic treatment with amoxicillin (6 weeks intravenously and 6 weeks oral). The second patient had a one-step approach, with debridement and replacement of the head and glenosphere.

Table 6.

Complications.

	Complications	Primary treated patients	Secondary treated patients	Total	p-value
		(n = 53)	(n = 32)	(n = 85)
Minor
	Superficial wound infection	1	1	2
	Cuff ruptures	1		1
	Radial nerve palsy	3	2	5
	Pain supercable	2	1	3
	Impingement supercable	1		1
Total		8 (15.1%)	4 (12.5%)	12 (14.1%)
Major
	Impingement	1		1
	Loosening humeral stem, aseptic		1	1
	Dislocation	2	1	3
	Infection		2	2
	Periprosthetic fracture		1	1
Total		3 (5.7%)	5 (15.6%)	8 (9.4%)
Total		11 (20.8%)	9 (28.1%)	20 (23.5%)	0.438

Regarding the radiographic outcomes, there was no difference in the number of patients with scapular notching and no difference in tuberosity healing (Table 7).

Table 7.

Radiographic assessment.

Radiology	Primary treated patients	Secondary treated patients	Total	p-value
	(N = 52)	(N = 32)	(N = 84)
Scapular notching Grade 1 Grade 2 Grade 3 Grade 4	19 (36.5%) 2 (3.8%) 0 0	5 (15.6%) 1 (3.1%) 1 (3.1%) 0	24 (28.6%) 3 (3.6%) 1 (1.2%) 0	0.127
Tuberosity union, y	47 (90.4%)	27 (84.4%)	74 (88.1%)	0.409

From one patient, the follow-up radiographs are missing.

Primary data is available on request to the lead author.

Discussion

This study showed no statistically significant differences in PROMs and complication rates between primary and secondary treated patients with an RSA. However, clinically relevant and statistically significant differences in forward flexion (p = 0.003, B 19.85) and abduction (p = 0.034, B 17.34) were found in favour of primary treatment after correcting for the differences in baseline characteristics (a combination of Neer and anticoagulant use had a confounding effect).

Surgical considerations

Our findings can be explained by three possible reasons. Firstly, in primary treatment, the tuberosities can be reinserted more easily.^28–30 Tuberosity healing is an important part of the fracture treatment, as the ROM improves with the consolidation of the tuberosity. ⁴³ In sequela treatment, tuberosity healing can be difficult, especially for type 3 and 4 sequelae, since an osteotomy of the tuberosities and re-attachment need to be performed. ³⁶ This is supported by Seidl et al. who concluded that while acute and secondary RSA can yield successful outcomes, acute RSA results in a higher tuberosity healing rate and improved external rotation. ²⁹

Secondly, primary surgery is less cumbersome than secondary surgery. In primary treatment, the anatomical landmarks are more easily distinguished due to a lack of adhesions and scar tissue. Furthermore, the fracture edges are still viable aiding in the chance of successful tuberosity healing. Also, with every extra surgical procedure, the risk of infection increases. This is confirmed by Kristensen et al. who found a higher risk of additional surgery after failed osteosynthesis in a study that compared primary RSA with secondary RSA after failed osteosynthesis. ⁴⁴

Finally, in primary treatment, the rotator cuff is often not yet atrophied and/ or substantially retracted.^28,30 This has three main effects: (a) the RSA is less dependent on the rotator cuff; however, functional outcomes are better with a functioning rotator cuff ²³ ; (b) with an atrophied rotator cuff, the risk of dislocation of the RSA is higher compared to a functional rotator cuff and ⁴⁵ (c) when the rotator cuff is retracted, we found that the insertion of the tuberosities becomes more difficult.

Primary outcome measures: ROM and PROMs

The differences in forward flexion and abduction after correction of the baseline characteristics are statistically and clinically relevant in favour of primary treatment. Minimal important changes as reported by Simovitch et al. are 12° ± 4° for forward flexion and 7° ± 4° for abduction in comparison to our results for forward flexion (9.2°) and abduction (7.2°). ⁴² These differences are in line with literature suggesting better functional outcomes for patients who received RSA as primary treatment.^{29,33,44,46–49}

However, recently, Panagopoulos et al. found no statistically significant differences in ROM, CMS and subjective shoulder value between patients who received RSA as primary treatment and patients who received RSA as secondary treatment. Nonetheless, the time from injury to regaining pain-free good function was significantly shorter than that in the acute RSA group. Therefore, they advocate early RSA in elderly patients to provide quicker recovery. ⁵⁰

In contrast, Torchia et al. performed a review comparing primary RSA vs. secondary RSA. ⁵¹ They found no differences between both groups, and external rotation was even better in the secondary RSA group. However, by looking at the data, it seems that acute and delayed are mixed in the final publication. A small favour for primary treatment is to be expected.

Complications

Several studies which compared the outcomes of primary RSA with RSA after failed osteosynthesis suggested that secondary treatment can be awaited.^33,52,53 Shannon et al. reported a complication rate of 8% in the secondary RSA group compared to 5% in the primary RSA group, but did not find differences in functional outcomes. ⁵² Sebastia-Forcada et al. found marginally lower function scores and a higher complication rate in secondary treated patients. ³³ However, compared with their pre-operative state, they had improvement in function and pain, and the complications were manageable.

Sub-analyses: effect of choice of primary treatment on outcome of secondary RSA

In our sub-analyses, the effect of the choice of primary treatment (conservative or osteosynthesis) on the outcome of secondary treatment with RSA was assessed. No effect of the choice of primary treatment on the outcome of secondary treatment was found. For the primary choice for conservative treatment, this could be explained by the difference in fracture classification (less complex fractures), and for the primary choice for surgical treatment, by the fact that the patients in the secondary RSA group were younger, included more men and had no head-split fractures.^54,55 No correction for confounding was performed due to the smaller group sizes and since the study was not powered for these sub-analyses.

Strength and limitations

Our methodology contains several improvements compared to the currently available literature. The total number of included patients was based on sample size calculation (32 patients), enabling more firm conclusions. Furthermore, we analysed the PROMs and ROM in a standardised setting in addition to recording the complication rate and radiographic measures. Since the ROM and PROMs are used as primary outcome measures, this adds considerably to the clinical relevance of this study.

However, by including conservatively treated patients and surgically treated patients in the secondary treated group, it might have been too heterogeneous to find significant differences in all outcome measures. For further research, it would be interesting to compare the outcomes separately, to prevent any regression to the mean.

Furthermore, as a cut-off point for primary treatment and secondary treatment after conservative treatment, we used the cut-off of delayed union (3 months). Of course, this is a hypothesis-driven definition. Ideally, all patients in the primary group were treated within 1 week after their fracture. However, in practice, in consultation with patients or after secondary dislocation of the fracture parts, these fractures are treated in the following weeks. The results of our sub-analysis, in which we compared patients treated within 4 weeks after fracture with patients treated after 3 months, validated our results and showed that there is no regression to the mean.

Clinical implications

The results of this study showed that patients with complex PHFs can benefit from a primary RSA. This is especially relevant as benefits were found in the ROM, which affects the day-to-day life of patients. This would imply that with more complex fractures, more benefit is to be expected for the patient treated with RSA. In contrast, with less complex fractures, a ‘wait-and-see’ management can still be maintained. Of course, patient-specific characteristics should always be accounted for. High-demanding elderly patients (>70 years) with complex PHFs might benefit from early mobilisation, and therefore, an RSA might be indicated. While conservative treatment can be considered in low-demand, elderly patients to prevent surgical intervention. In younger, active patients without osteoporosis, plate osteosynthesis should be considered.

Conclusion

This study found no differences in PROMs for primary versus secondary treatment for PHFs with RSA. However, clinically relevant and statistically significant differences in forward flexion and abduction were found in favour of primary treatment after correcting for the differences in baseline characteristics. Therefore, we conclude that in patients with complex displaced PHFs, RSA could be considered primary treatment, especially when ROM is particularly important to the patient.