Functional outcomes and complications after locked plating of fracture-dislocations of the proximal humerus and the significance of the intertuberosity periosteal bridge

Abstract

Background

We aimed to evaluate the functional outcomes and complications following locked plating of three- or four-part proximal humerus fracture-dislocations (PHFDs).

Methods

We evaluated the functional outcomes and range of motion (ROM) deficit of 26 shoulders with PHFD (AO/OTA 11B1.1(5a), 11B1.1(5b), 11B1.2(5b), 11C3.1(5a), 11C3.1(5b), 11C3.2(5a)), in 25 patients after locking plate fixation. Three patients with avascular necrosis (AVN) of humeral head were followed for one year, and the remaining 22 patients were followed for 24–89 months.

Results

At the most recent follow-up, the shoulder subjective value was 77 (60–85), the normalized Constant score was 70 (53–83), the visual analogue scale pain scores were 1 (0–2), the elevation ROM deficit was 35° (13–41), the external rotation ROM deficit was 20° (10–33), and internal rotation ROM (number of vertebrae) deficit was 4 (1–6). AVN was noted in six (23%) shoulders; the shoulders with AVN had worse outcomes than those without AVN. Absent capsular attachments (p < .001), an opened intertuberosity periosteum, and a four-part fracture configuration (p = .004) were significantly associated with AVN.

Conclusion

To conclude, functional outcomes after locked plating of fracture-dislocations were acceptable in the majority of the patients. However, complications such as AVN may occur in 23% of patients, leading to poor outcomes.

Level of evidence

Level 4; case series.

Introduction

Proximal humerus fracture-dislocation (PHFD) is an uncommon fracture associated with high postfixation complications. ¹ A dislocated head in the setting of a three-fourth-part fracture is assumed to have lost a significant part of its vascularity,^2,3 but may still be salvaged in many situations.^4–6 Concomitant fracture of the tuberosities and soft tissue disruption due to the dislocation may increase the complexity of the surgery, making functional outcomes after internal fixation unpredictable. Thus, the reported incidence of avascular necrosis (AVN) after fixation of PHFD has varied widely (20–82%). ⁵ Additionally, there are contrasting reports on whether posttraumatic AVN does^1,6,7 or does not^4,8 influence the eventual functional outcomes. Although replacement may be considered in patients over 65 years of age, internal fixation offers the advantages of restoring native anatomy. It may be preferred in active patients and those of middle age.

Because PHFDs are uncommonly encountered, most studies have typically pooled cases from multiple surgeons (4–11 number of surgeons)^4,9,10; hence, a technique variation or a surgeon factor ¹¹ itself may influence the eventual outcome and development of complications. Moreover, it is unclear which potential technical or injury-related factors are associated with the development of AVN and poor outcomes. Additionally, patient selection most suitable for internal fixation, based on technical factors, needs further study. Therefore, we aimed to evaluate the functional outcomes and complications after locked plating of three- and four-part PHFD.

Methods

The retrospective study was conducted after approval from the hospital's ethics committee. All patients were operated on by the senior author, a fellowship-trained shoulder surgeon. The study included patients who were operated between December 2016 and December 2022. The data were drawn from a prospectively maintained database of patients who underwent surgeries for proximal humerus fractures.

Patients

We included patients with (1) a fracture of the humeral neck with a dislocated humeral head, (2) with or without a head-split component, and (3) with a displaced fracture of either the greater tuberosity (GT) or the lesser tuberosity (LT) or both, (4) aged 18–80 years who underwent locked plating. We excluded two-part fractures, fractures with an intact humeral neck, patients with brachial plexus injury, and patients who presented ≥ three months postinjury. During the study period, the indications for internal fixation included all patients with two-, three-, or four-part fracture dislocations. Replacement was performed for one patient with a fracture dislocation who had inadequate bone stock in the humeral head and with severe comminution of the tuberosities. As our unit is a referral unit, in some situations, patients referred from remote locations could be operated upon only after a few weeks.

During the study period, 30 shoulders with PHFD in 29 patients underwent locked plating. Four patients were lost to follow-up, and 26 shoulders in 25 patients were included. Twenty-two patients had a minimum two-year follow-up. Three patients, who developed AVN of the humeral head at one-year, declined further follow-up or underwent revision surgery. Their data at one-year follow-up were included in the study. The average follow-up duration of n = 25 patients was 35 ± 19.6 months (range 12–89 months).

Surgical technique

Locking plate fixation via the deltopectoral approach was performed for all shoulders. The incidence of AVN in fracture dislocation has been reported to be around 20–80%, with a high incidence of screw penetration in the joint due to collapse related to AVN. We found no clinical evidence on the minimum number of screws to be used, nor any clinical evidence in support of subchondral-located screws. Therefore, to avoid future screw penetration due to collapse related to AVN in some cases with high AVN risk related to the time lag in surgery^12,13 and/or four-part fracture configuration ⁶ (Figures 1–3), we: (1) attempted to keep the screw length shorter than the subchondral location via the sounding technique. ¹⁴ In this technique, only the proximal cortex was drilled, and the standard depth gauge with a blunt end was advanced in the humeral head until it reached a hard stop, and (2) used at least five sutures through the cuff tendon junction and tied them to the plate. We used between three and seven screws in the humeral head, depending on the bone purchase and the stability achieved, to create a final stable construct. In all cases, stability was checked intraoperatively by taking the shoulder through the full rotational range of motion (ROM). Bone grafts or bone graft substitutes were not used during internal fixation of the fractures.

Figure 1.

(a) Preoperative anteroposterior radiograph of the shoulder showing three-part anterior fracture-dislocation in a 38-year-old man, who presented six days after injury. (b) Postoperative anteroposterior shoulder radiograph showing acceptable reduction, four screws proximally and two screws distally. (c) Anteroposterior shoulder radiograph done two years postoperatively showing bony healing. (d) Clinical follow-up at two-year follow-up showing no elevation deficit compared to the opposite side. We used four screws in the humeral head apart from the four sutures through the cuff tendons to give us a final stable construct. There was a high risk of avascular necrosis (AVN) because he was operated on after six days; hence, fewer screws and shorter than subchondral-located screws were used to avoid any future screw penetration in case of an AVN+ collapse. However, the fracture healed without further collapse, loss of fixation, or AVN. Because there is no clinical evidence on the minimum number of screws and the length of the screws to be used in the humeral head segment, we use between three and seven screws in the humeral head to give us a stable construct.

Figure 3.

(a) Preoperative anteroposterior radiograph of the shoulder showing four-part anterior fracture-dislocation of the proximal humerus in a 72-year-old woman, operated after a delay of 14 days. (b) Postoperative anteroposterior shoulder radiograph showing acceptable reduction, with three screws proximally and two screws distally. (c) Anteroposterior shoulder radiograph at five years postoperatively showing bony union and hardware removed on patient request (d). Clinical follow-up at five years of follow-up showed no abduction deficit compared to the opposite side. We used three screws in the humeral head, in addition to the five sutures through the cuff tendons, to achieve a final stable construct. There was a high risk of AVN because she was 72 years old with a four-part fracture and operated after a delay of 14 days; hence, fewer screws and shorter than subchondral-located screws were used to avoid any future screw penetration in case of AVN and collapse. The fracture healed without further collapse, loss of fixation, or AVN.

Note: AVN: avascular necrosis.

Figure 2.

(a) Preoperative anteroposterior radiograph of the shoulder showing six-week-old, four-part anterior fracture-dislocation of proximal humerus in a 24-year-old man. (b) Axial computed tomography section showing the dislocated head fragment lying anterior to the glenoid rim. (c) Immediate postoperative anteroposterior shoulder radiograph showing acceptable reduction, four screws proximally and two screws distally. (d) Anteroposterior shoulder radiograph taken two-year postoperatively showing bony union. In this patient, we used four screws in the humeral head apart from the four sutures through the cuff tendons to give us a final stable construct. There was a high risk of AVN because he had a four-part fracture and was operated on after a delay of six weeks; hence, fewer screws were used to avoid any future screw penetration in case of AVN and collapse. Because there is no clinical evidence on the minimum number of screws and the optimal length of screws to be used in the humeral head segment, we used four screws to achieve a stable construct. The fracture healed without further collapse, loss of fixation, or AVN.

Note: AVN: avascular necrosis.

Intraoperative absence of capsular attachments (if >2 cm) as an intraoperative variable was noted by the primary surgeon.

Reduction of the dislocated humeral head

In cases of anteriorly dislocated heads, we aimed to relocate the head indirectly through the rotator interval or via a split in the middle of the subscapularis without opening the GT–LT (intertuberosity) periosteum bridge. However, in two anteriorly dislocated cases, the head had escaped anteroinferior to the subscapularis and had lost all capsular attachments. In the first three cases of posteriorly dislocated heads, the heads were directly relocated by opening the intertuberosity periosteum and osteotomizing the intertuberosity bony bridge. In the latter four shoulders, we indirectly reduced the head via the rotator interval (Figure 4). The humeral head was retrieved in the surgeon's hand and relocated back in two shoulders (one anterior and one posterior). Arthroplasty was not performed in them because of their young age.

Figure 4.

(a) Preoperative anteroposterior radiograph of the shoulder with a posterior fracture-dislocation of the proximal humerus in a 48-year-old woman in whom the dislocated head was reduced via Cobb elevator through the rotator interval. (b) Postoperative anteroposterior shoulder radiograph showing acceptable reduction, six screws proximally and three screws distally. (c) Anteroposterior shoulder radiograph taken two-year postoperatively showing bony healing. (d) Clinical follow-up at two-year follow-up showing no elevation deficit compared to the opposite side.

Outcome evaluation

Clinical evaluation

An independent research assistant performed the postoperative functional outcome evaluation (Table 1) and ROM measurements at the most recent follow-up. ROM evaluations included elevation (elev) and external rotation (ER) using a long-arm goniometer (standard error of measurement (SEM) = 5°) ¹⁵ and internal rotation (IR) with the extended thumb reaching the highest vertebrae at the back) deficits (compared to opposite normal limb in unilateral affection).

Table 1.

Postoperative clinical and radiological variables that were recorded.

Clinical outcomes

Radiological variables recorded in postoperative X-rays

Outcomes scores 1. 1. Age and gender normalized &Constant score (nConstant) 16 2. Shoulder subjective value 3. Residual pain visual analog scale scores Range of motion (ROM) 1. *Elevation ROM deficit (compared to opposite normal) 2. *External rotation (ER) ROM with elbow adducted (compared to opposite normal) deficit 3. Internal rotation ROM (vertebrae level) deficit (compared to opposite normal)-thumb reaching the highest vertebrae at the back

1. Fracture reduction quality assessed as per Schnetzke et al. 17 The anteroposterior X-rays were evaluated for: a. Greater tuberosity (GT) cranialization (< or >5 mm) b. Head shaft displacement (< or >5 mm) c. The neck-shaft angle: Achievement of the NSA between 110° and 150° was considered acceptable Acceptable and satisfactory reduction: If all three parameters were achieved • cranialization of GT <5mm • head-shaft displacement <5mm • NSA between 110° and 150° Unsatisfactory/unacceptable reduction: If any of the above parameters not achieved 2. Medial calcar hinge restoration (< or >5 mm) 3. Concentric articular reduction of the humeral head 4. Number of screws in the head fragment 5. Screw placement subchondral (yes or no) 6. Calcar screw placement (yes or no)

Note: ^&Abduction was measured only for the purpose of recording the Constant score.

The ROM variables separately recorded were Elevation, External Rotation, and Internal Rotation.

*An independent research assistant performed postoperative outcome evaluation and range of motion measurements using a long-arm goniometer.

mm- millimeter.

Radiological evaluation

The preoperative and postoperative evaluations of the biplanar X-rays were performed by two independent orthopedic surgeons who were blinded to the postoperative functional outcomes and who did not participate in the operations.

Preoperative evaluation

The preoperative biplanar X-rays and computed tomography (CT) scans were evaluated for Neer classification with a “part” being labeled when there was >5 mm of displacement, AO/OTA classification, presence of a head-split component (>25% of articular involvement ¹⁸ ), disruption of the medial hinge, ² the presence/absence of 8-mm medial spike and the presence of surgical or anatomical neck fracture. Furthermore, the concurrence of the two authors was done before classifying them as head-split fractures or three-fourth-part fractures. The preoperative anteroposterior X-rays were also evaluated for the proportion of the dislocated humeral head below the inferior glenoid margin or medial to the coracoid base in nonhead-split anteriorly dislocated shoulders (Figure 5).

Figure 5.

Anteroposterior radiographs of anteriorly dislocated humeral heads and the method of measuring the percentage of humeral heads below the inferior glenoid margin or medial to the coracoid base. (a) Dashed black vertical line: A vertical line (reference line) is drawn vertically downwards from the coracoid base. The horizontal blue solid line represents the horizontal diameter of the humeral head. The humeral head is 100% medial to the coracoid base in this X-ray. (b) Dashed black horizontal line: A horizontal dashed (reference) line is drawn tangentially to the inferior glenoid margin. The yellow vertical solid line represents the vertical diameter of the humeral head. The head is 100% displaced below the glenoid in this X-ray. (c) The humeral head is lateral to the coracoid base (dashed black vertical line), and >50% of the humeral head is inferior to the glenoid inferior margin (Dashed black horizontal line). Images A and B represent instances where the humeral head has escaped the capsular and subscapularis envelope. Of the analysis done in 17 anteriorly dislocated in which the head was not split, we found that the head was 100% displaced below the glenoid in only one shoulder, <50% displaced below the glenoid in six shoulders and >50% but <100% displaced below the glenoid in 10 shoulders; additionally the head was 100% displaced medial to the coracoid base in only one shoulder and lateral to glenoid in the rest 16 shoulders.

Postoperative evaluation

The anteroposterior and “Y” view postoperative x-rays (within the first week) were evaluated for quality of fracture reduction (as per Schnetzke et al. ¹⁷ ) and other radiographic variables, as noted in Table 1.

Complications noted at the most recent follow-up were penetration of the screws in the joint, varus collapse of the fracture (>10°), need for plate removal due to pain in activities of daily living, nonunion (absence of bony bridge across fracture site), and AVN of the humeral head. AVN was classified as partial if there were subchondral sclerosis and cysts, but maintained sphericity, and as complete if there was a head collapse and loss of sphericity.

Poor outcomes

Poor outcomes of surgical fixation were classified in any one of the following situations: AVN or nonunion, varus collapse leading to revision surgery, shoulder subjective value (SSV) or normalized Constant score (nConstant) < 50 (as per Schnetzke et al. ¹⁷ ), or if the hardware was removed due to shoulder pain in activities of daily living.

Primary and secondary outcome variables

The primary outcome variables were the nConstant (normalized Constant score), ¹⁶ the SSV, the visual analogue scale (VAS) pain scores, and the ROM deficits. The secondary variables were the proportion of patients with complications and AVN. The tertiary variables were the bivariate comparisons with the chosen variables.

Statistical analysis

All data were checked for normality; normal data were evaluated with a t-test, and nonnormal data with the Mann–Whitney U test. Normal data were represented by mean ± standard deviation (SD), and nonnormal data were represented by median ± interquartile range. Fourteen patient-related or injury-related factors (independent variable), identified as clinically relevant in the literature, were examined in a bivariate analysis for their potential association with poor outcomes or AVN (dependent variable). The methodology for selecting the factors is outlined in Supplemental Table S1. The Benjamini-Hochberg correction was applied in the multiple bivariate analysis to control the false discovery rate. For all other calculations, the p-value was set at .05.

Results

Baseline characteristics

The patients were aged 49 ± 13 (Table 2), with a majority of men population (n = 16 (64%)). We included 19 anterior fracture-dislocations and seven posterior fracture-dislocations. The fracture configuration included 18 three-part and eight four-part fractures. Head split fractures comprised 15% (four of 26) of the fractures. Four patients presented after a delay of ≥ 21 days.

Table 2.

Demographic and baseline characteristics of patient with proximal humerus fracture-dislocation).

Variable	Value
Age at surgery (years), mean (SD)	49 (13)
Male/female, n (%)	16 (64)/9 (36)
Side affected: right/left, n (%)	15 (58)/11 (42)
Duration between injury and Surgery, n (%)
≤5 days	13 (50)
6–20 days	9 (35)
>21 days	4 (15)
Fracture pattern
3-part	18 (69)
4-part	8 (31)
Pre-op medial hinge intact, n (%)	7 (27)
Presence of metaphyseal spike, n (%)	10 (38)
Head-split fractures, n (%)	4 (15)
Anatomical neck, n (%)/surgical neck, n (%)	13 (50)/13 (50)
Absent capsular attachment, n (%)	5 (19)
Anterior dislocation/posterior dislocation	19 (73)/7 (27)
AO/OTA#	11B1.1(5a), n = 10
11B1.1	11B1.1(5b), n = 1
1B1.2	11B1.2(5b), n = 2
11C3.1	11C3.1(5a), n = 6
	11C3.1(5b), n = 4
11C3.2	11C3.2(5a), n = 3

Note: SD: standard deviation; # AO/OTA classification system 2018 modification.

Functional outcomes

At the most recent follow-up, the SSV was 77 (60–85), the nConstant was 70 (53–83), the VAS pain score was 1 (0–2), the elevation ROM deficit was 35°(13°−41°), the ER ROM deficit was 20°(10°−33°), and IR ROM (number of vertebrae) deficit was (4 (1–6)) (Table 3).

Table 3.

Clinical outcomes of patients with proximal humerus fracture-dislocation at their most recent follow-up.

Variable	Value
Normalized Constant Score (nConstant), median (IQR)	70 (53–83)
Shoulder Subjective Value, median (IQR)	77 (60–85)
Pain Visual Analog Scale score, median (IQR)	1 (0–2)
Deficit in elevation (°), median (IQR)	35 (13–41)
Elevation of opposite normal side, median (IQR)	150 (149–153)
Deficit in external rotation (ER) (°), median (IQR)	20 (10–33)
ER of opposite normal side, median (IQR)	60 (53–62)
Deficit in IR (difference in the number of vertebrae), median (IQR)	4 (1–6)
IR of opposite normal side (Thoracic vertebrae level), median (IQR)	T7 (T7-T10)
Follow-up duration (months), mean (SD)	35 (19.6)

Note: SD: standard deviation; IQR: Interquartile Range; ER: external rotation with elbow adducted; IR: internal rotation with the extended thumb reaching the highest vertebrae at the back.

Radiographic outcomes

The postoperative medial hinge integrity was present in 12 (46%) shoulders; the postoperative reduction was acceptable in 11 (42%) shoulders; and intraoperative absent capsular attachment was noted in five shoulders (Table 4).

Table 4.

Complications, avascular necrosis, and poor outcomes in patients with proximal humerus fracture-dislocation after locked plating.

Variable	Value
AVN + nonunion, n (%)	6 (23) + 1 (4) AVN's included: 3 anterior PHFD (2 complete AVN and 1 partial AVN), 3 posterior PHFD (all 3 complete AVN)
Poor outcomes, n (%)	9 (35) Including: 6 AVN heads (4 had nConstant < 50) 1 (nonunion) 1 (hardware removal due to pain) 1 with poor score (nConsant < 50)
Metal hardware removal (without arthrolysis), n (%)	3 (11.5) 1 patient for pain in daily activities and in 2 patients as per their request 1 p
Revision surgery, n (%)	2 (7.7) 1. 1 patient with AVN revised to arthroplasty 2. 1 nonunion revised to revision reconstruction
Postoperative medial hinge intact, n (%)	12 (46)
Calcar support screw, n (%)	12 (46)
Acceptable reduction, n (%)	11 (42)
Number of screws in the head a
<5 screws, n (%)	19 (73)
≥5 Screws, n (%)	7 (27)
Screw penetration at most recent follow-up, n (%)	9 (35) b (6 had AVN humeral head, 1 had head-split fracture, and 2 others)

Note: AVN: avascular necrosis.

^a We used at least three sutures through the rotator cuff tendons (anchored to the plate) and three to seven screws in the humeral head. The number of screws was decided by the fixation stability achieved intraoperatively. Because of the younger age in some patients, three to four screws in the head were considered sufficient for adequate stability.

^b Of the total n = 9 shoulders that had screw penetration at the final follow-up, n = 3 had five screws in the humeral head, n = 3 had three screws in the head, n = 2 had six screws in the head and n = 1 had four screws in the head. The proportion of shoulders with screw penetration tended to be higher (5/7 (71%)) in the ≥5 screw group than in the <5 screw group [4/19(21%)].

AVN and nonunion

Radiographic evidence for AVN was present in six (23%) shoulders (Tables 4 and 5). One patient had a nonunion of the surgical neck. AVN was noted in three anteriorly dislocated heads (complete AVN in two and partial AVN in one) and three posteriorly dislocated heads (all with complete AVN). The two anteriorly dislocated heads that developed complete AVN had escaped the subscapularis envelope and were noted to be either 100% inferior to the glenoid (n = 1) or 100% medial to the coracoid base (n = 1) in the preoperative anteroposterior X-rays (Figure 5). Of the two cases in which the humeral head was retrieved in the surgeon's hand, complete resorption of the head was noted at 5.5-year follow-up in one case (Figure 6) and complete AVN in the second. Overall, the shoulders with AVN (n = 6) had significantly worse nConstant (40 ± 15.5) than those without AVN (74 ± 15, p < .0001).

Table 5.

Details of six shoulders that had avascular necrosis (AVN) of the humeral head after locked plating.

Serial number	Age (years)	Sex	Duration between injury and surgery (days)	Fracture type (three-part or four-part, and head-split)	Location of fracture-(anatomical/surgical neck)	Dislocation direction (anterior/posterior)	Metaphyseal spike (present/absent)	AVN (complete/partial)	Capsular opening/detachment	Complication	Revision surgery	Normalized Constant score	SSV
1	62	M	4	4	Anatomical	Posterior	Present	Complete	Yes	Screw penetration	None	28.3	20
2	62	M	8	4	Anatomical	Posterior	Present	Complete	Yes	Screw penetration	None	40.2	25
3	69	F	3	4	Anatomical	Anterior	Absent	Complete	Yes	Screw penetration	None	20.5	30
4	45	M	30	4	Anatomical	Posterior	Absent	Complete	Yes	Screw penetration + total head resorption	None	58.5	60
5	55	F	3	4	Anatomical	Anterior	Absent	Complete	Yes	Screw penetration	Replacement	35.7	30
6	42	F	21	3	Anatomical	Anterior	Present	Partial	No	Screw penetration	None	58.1	80

Note: M: male; F: female; SSV: shoulder subjective value; N/A: not applicable.

Figure 6.

Follow-up images at 5.5-year of a patient in whom the posteriorly dislocated head was retrieved in the surgeon's hand and put back in the glenoid (a) the anteroposterior radiograph shows complete humeral head resorption (b) clinical photos (hardware removal not yet done) show an active elevation of 130°, though the patient complained of mild pain in the shoulder.

Poor outcomes, screw penetrations, and revisions

Poor outcomes were noted in nine (35%) shoulders (Table 4). Among the nine shoulders with poor outcomes, six had AVN of the humeral head (four of them had nConstant < 50), one had a nonunion of the neck, one had hardware removal due to pain in activities of daily living, and one had a poor score (nConstant < 50).

Screw penetration was noted at the final follow-up in nine (35%) shoulders (Table 4). Of the nine shoulders with screw penetration, six of them had concomitant AVN of the humeral head, one had a head-split fracture, and there were two others. Additionally, of the total n = 9 shoulders with screw penetration in the joint, n = 3 had five screw-fixations in the humeral head, n = 3 had three screw-fixations in the humeral head, n = 2 had six screw-fixations in the humeral head, and n = 1 had four screw-fixations. Thus, 71% (five of seven) of shoulders in the ≥5 screw group had screw penetration, and 21% (four of 19) in the <5 screw group had screw penetration.

Metal hardware was removed in 3 patients, and revision surgery for failures was performed for two patients.

Bivariate analysis

In bivariate analysis, the absence of capsular attachment (p = .002) was significantly associated with poor outcomes (Supplemental Table S2).

Additionally, absent capsular attachment (p < .001), an opened intertuberosity periosteum (p < .001), and a four-part fracture configuration (p = .004) were significantly associated with AVN (Supplemental Table S3). However, the time lag between injury and surgery (either greater than five days or greater than 21 days) was not associated with poor outcomes or AVN. Additionally, absent preoperative or postoperative medial hinge integrity was not associated with poor outcomes.

Discussion

Our study found that although locked plating of proximal humerus fracture dislocations resulted in acceptable functional outcomes and pain scores in most patients, the proportion of patients with poor outcomes was high. Thus, our results indicate that fracture-dislocations are complex and terrible injuries. However, the findings of our study may help decide indications for replacement in some instances. Some of the fracture dislocations, such as the ones in which the humeral head has completely escaped the capsular envelope and/or lost all capsular attachments, akin to Robinson's Type 2 injury ¹⁹ should be counseled regarding the high incidence of AVN and poor outcomes, with a possible need for a replacement to be considered as a part of shared decision-making. These fractures have been identified in our study in the preoperative X-rays as being either 100% inferior to the inferior glenoid margin or 100% medial to the coracoid base. We also recommend against putting the humeral head back in the joint where it was retrieved in the surgeon's hand because of the observed complete AVN in one shoulder at one year and total head resorption in the second shoulder at 5.5-year follow-up; a replacement may be an option in such instances. ¹

The patients in our cohort were younger (50 years), consistent with published studies.^5,19,20 The Constant scores in our study (70) are approximately similar to the average values reported in a systematic review on PHFD (73). ⁵ Although 50% of our patients presented after a delay of six days or more, this delay was not associated with poor outcomes or AVN, consistent with the conclusion of a recently reported study on proximal humerus fracture fixation. ²¹ Therefore, patients presenting more than five days after the injury may not be strictly contraindicated for internal fixation.

The AVN rate after fixation of PHFD in our study (23%) is approximately similar to the pooled AVN rate (20%) reported in a systematic review by Miltenberg et al. ⁵ Additionally, we found that the shoulders with AVN had worse outcomes than those without AVN.

As per earlier studies^6,17 and our bivariate analysis, a four-part configuration and absent capsular attachments may be associated with AVN and poor outcomes. However, the integrity of the intertuberosity periosteum may also be an important factor that has never been investigated. Opening the periosteal bridge for reduction has reportedly led to AVN in 82% of the PHFD cases. ²² To keep the surgery least invasive to the periosteal vasculature, ²³ we reduced the head through the rotator interval or the subscapularis-split but without disturbing the intertuberosity periosteum in 17 anteriorly dislocated heads, and 16 of them did not develop AVN, with only one that had partial AVN. However, in the initial three posterior PHFD, we opened the intertuberosity bridge and mobilized the GT and the LT to reduce the dislocated head. We found that the capsular attachments were absent in these cases, and all these heads eventually developed AVN. However, it may also be argued that some of the four-part fracture-dislocations may need an opening of the intertuberosity bridge to visualize and reduce the head in the glenoid cavity and to achieve an anatomic reduction. According to the published literature, many surgeons prefer to directly visualize and relocate the dislocated heads by opening the intertuberosity periosteal bridge.^{6,19,22,24,25} Others, however, have preserved the soft tissue attachment and reduced the head indirectly.^4,26 After the study concluded, the authors now prefer to perform a primary replacement if the head has escaped the subscapularis, as evident on preoperative radiographs, or if it cannot be reduced without opening the periosteum bridge. An acceptable reduction was found in 42% of shoulders in our study and has been similarly reported using the mentioned criteria in earlier studies on fracture-dislocations and complex fractures.^13,17

The anastomotic branches of the posterior circumflex artery (PCA) form an arterial ring on the capsule before supplying the humeral head. ²⁷ The anterior circumflex artery (ACA) supplies the head through the ascending anterolateral branch. ²⁸ Therefore, maintaining the GT–LT bridge and the capsular attachments to the head should be prioritized ²⁹ to preserve the ACA, the PCA, and their anastomotic channels in case of possible interruption to either artery. Additionally, preserving the periosteum and its vasculature may be crucial for maintaining tuberosity perfusion and promoting healing. ²³

Some authors have argued that the dislocation component does not necessarily influence AVN rates.^2,20 However, concerns regarding the increased incidence of AVN and poor outcomes after PHFD fixation have been highlighted in several studies.^{1,4–6,9,10,13} Although few studies have reported no influence of AVN on functional scores,^4,8 several other studies have reported that functional outcomes are poor in cases with AVN.^1,6,7

There are several limitations in our study. Most patients were followed for two years or more; however, some fractures may develop AVN on longer follow-ups. Additionally, as the study is retrospective, it is susceptible to selection bias. Our study also did not include a control group; therefore, we cannot say whether replacement may have been a better option for some of the older patients. Lastly, due to the retrospective design, our study could only investigate the association between factors such as capsular detachment and the occurrence of AVN, rather than establishing a cause-and-effect relationship. Our strength lies in the fact that one surgeon operated on all patients, thereby minimizing the surgeon factor, and independent researchers conducted the clinical and radiological evaluations in a blinded manner.

Conclusion

To conclude, functional outcomes and pain scores, after locked plating of fracture-dislocations, were acceptable in most patients. However, AVN may develop in approximately 23% of the shoulders and may lead to poor outcomes. Factors such as absent capsular attachments and opening of the intertuberosity periosteum may be associated with the occurrence of AVN.