Surgical fixation of isolated greater tuberosity fractures of the humerus- systematic review and meta-analysis

Abstract

Objectives

A systematic review was undertaken based on PRISMA guidelines to study the results of surgical treatment of Greater Tuberosity fractures of the Humerus by different techniques (Open, Percutaneous or Arthroscopic) and using different fixation devices (Sutures, Screws and Plates).

Data sources

Literature searches were performed for studies in English on four online databases (MEDLINE®, Embase®, Ovid® Emcare and CINAHL®) using a pre-planned search strategy.

Study selection

Studies were screened against pre-decided inclusion and exclusion criteria. 24 articles were eligible for inclusion involving 562 patients in all.

Data extraction

A template was used for collection of relevant data regarding the study design, demographics, fracture characteristics, surgical techniques and implants used, outcomes, complications and re-operations. Methodological quality was assessed using a modified Coleman Methodology Score.

Data synthesis

Pooled analysis was conducted for (1) complications/reoperations of different devices; and (2) reported results using standard outcome measures of various techniques. Results were pooled using inverse variance method. When pooling proportions, we used arcsine transformation to account for zero events.

Conclusion

Use of screws was associated with higher rates of total complications (23.7%) as well as reoperations (13.2%), compared to sutures (14.2%; 2.6%) and plates (10.6%; 4.1%). In a subgroup analysis, use of low-profile plates (LPPs) was associated with the lowest rates of total complications (4.97%). The superiority of any surgical technique could not be established using the pooled Constant or ASES scores because there was lack of uniformity in the reporting of outcome measures. Our analysis of complications of the various fixation devices dissuades the use of screws and favours the use of low-profile plates. There is a need for prospective studies comparing arthroscopic suture fixation with open low-profile plate fixation of GT fractures.

1. Introduction

Isolated greater tuberosity (GT) fractures of the humerus account for 15–20% of all proximal humeral fractures.¹ Proximal humeral fractures mainly occur in older patients with osteoporosis and usually result from a fall from standing height.2, 3, 4 On the other hand, isolated greater tuberosity fractures can occur in a younger population with normal bone quality due to significant trauma.^5,6 A majority (85–95%) of these injuries can be treated conservatively but the remainder require surgical management.^7,8 Most authors agree that fragment displacement of more than 5 mm leads to poor outcomes and should be addressed surgically.^5,9,10 While it seems logical that a displaced GT would result in mechanical impingement, the characteristic posterior displacement of the GT fragment makes the existence of anterior acromial impingement debatable. It has been argued that the displacement adversely affects rotator cuff biomechanics and hence, shoulder function. This consideration has led Park et al. to recommend a lower displacement threshold of 3 mm for considering surgery in manual workers and athletes.¹¹

Surgical options described to treat these injuries can be broadly classified into open reduction and internal fixation, percutaneous fixation and arthroscopic fixation. In their systematic review published in 2015, Levy et al. found no difference in outcomes and complications between patients treated by open/percutaneous methods and arthroscopic methods; or between those treated using different fixation devices. However, they pooled results from only 212 patients in 13 studies, mainly due to a stringent inclusion criterion of a minimum follow-up of 2 years.¹² They also did not include any studies using plate fixation. Further, since their literature search in August 2014, a number of studies reporting on surgical treatment have been published, thereby expanding the available body of literature. Hence, we believe that this topic deserves a fresh systematic review, which we have undertaken.

The aim of the review is to report outcomes of different operative techniques (percutaneous, open and arthroscopic) and complications of different fixation devices (plates, sutures and screws) used in the surgical treatment of isolated GT fractures.

2. Methods

The study was conducted in accordance to PRISMA guidelines.¹³ A study protocol was made but not registered on a database. We did not keep any time limits and looked for all papers whose electronic prints were available for review.

Primary outcome- Comparison of pooled measures of the functional outcome in studies employing different operative techniques.

Secondary Outcome- Comparison of pooled estimates of complication rates in studies employing different fixation devices.

Study Selection criteria- These were determined before undertaking the literature search.

Inclusion criteria: Studies published in English, which.

• •Included patients aged 18 and above who underwent surgical fixation
• •Report validated outcome measures for the shoulder
• •Include 10 or more cases of isolated GT fractures (in the operative arm)

Exclusion criteria:

• •Follow up < 6 months
• •Studies with abstracts only, conference abstracts, editorial comments or expert opinion
• •Studies reporting intervention for malunion or non-union
• •Studies on open fractures
• •Studies on pathological or periprosthetic fractures
• •
  Studies with inseparable outcomes for

  • a.
    Isolated GT fractures and other Proximal humeral fractures
  • b.
    Different techniques
  • c.
    Operative and non-operative treatment

The threshold of 10 patients was applied as it was felt that meaningful conclusions could not be drawn from smaller series due to a high risk of bias.

2.1. Search strategy

We conducted a systematic search of four online databases (MEDLINE®, Embase®, Ovid® Emcare and CINAHL®) on June 03, 2020. We then independently screened the titles and abstracts against the selection criteria. Any differences were resolved mutually. The full text of selected articles was reviewed, and the bibliographies were also screened to ensure comprehensive coverage of literature. The following search strategy was used on Medline®-“(((greater tuberosity ADJ4 fracture∗).ti,ab OR (greater tuberosity ADJ4 fracture∗).ti,ab,af OR ((greater tuberosity).ti,ab AND (exp “SHOULDER FRACTURES”/) OR (shoulder fractures).ti,ab)) NOT (exp “ARTHROPLASTY, REPLACEMENT”/)) [Languages English]”

2.2. Data extraction

Details of study design, sample size, follow-up, concomitant shoulder dislocations, other associated injuries, nerve injuries, patient demographics, displacement threshold for operative intervention, any fracture characteristics which influenced technique selection, surgical details (approach, technique and fixation devices), post-operative outcomes, complications and re-operations were recorded. If authors studied results of both non-operative and operative interventions, only data from patients having operative interventions were included. All studies were included for qualitative synthesis and considered for pooling complications. We counted six complications (implant failure/suture pull out, impingement/soft tissue irritation, stiffness, re-displacement, GT resorption and post-operative nerve injury), which were the most consistently reported. We grouped them together as total complications (TCs) to facilitate the meta-analysis of proportions. For pooling outcomes, we used the two most commonly used outcomes measures among selected studies (Constant-Murley (CM) and American Shoulder and Elbow Surgeons (ASES) scores). If outcomes and complications were not reported separately for different techniques or fixation devices used, the study data was excluded from the respective calculations.

2.3. Statistical methods

The methodological quality and risk of bias in the studies was evaluated using a modified Coleman Methodology Score (MCMS) which was described by Cowan et al.¹⁴ This scoring system can be used for both randomized and non-randomized studies.¹² Its scaled potential score ranges from 0 to 100 (85–100, excellent; 70–84, good; 55–69, fair; <55, poor).

2.3.1. Pooling of outcomes

The reported Mean, Standard Deviation (SD) and Range of outcome measures was noted. Where SD was not given, it was estimated from the Range.¹⁵ Studies which did not report the SD or the Range were excluded from pooling, as the standard error (SE) could not be estimated. The weighted means (WM) were pooled for the two outcome measures in groups defined by operative technique, using the inverse variance method.

2.3.2. Pooling of proportions of TCs and re-operations

For TCs and re-operations, Freeman-Tukey double arcsine transformation for proportions was used before pooling by the inverse variance method, in order to include studies with zero events in the estimation.

A random effects model was used for pooling as most studies had a retrospective cohort design with high risk of inter and intra-study heterogeneity. The 95% Confidence Intervals (CI) and pooled means were calculated for outcome measures, proportions of TCs and re-operations. We planned to compare pooled data between groups if they had low heterogeneity (measured using Cochran's Q and I²). As this was not the case, we have presented a descriptive analysis. Online software SciStat® and StatsDirect® were used for statistical analysis.

3. Results

A total of 429 articles were obtained for title and abstract review. 25 articles from the main searches and 2 more from bibliography screening were considered for full-text review. Using our selection criteria, 24 studies were deemed suitable for synthesis. (Fig. 1). 3 of these^9,24,29 reported results from two cohorts using different techniques or fixation devices, taking the total number of cohorts to 27.

Fig. 1.

Prisma flowchart.

∗ Three studies^9,24,29 reported cohorts using different techniques/approaches †Two for inseparable results of isolated fractures of GT from other proximal humeral fractures. One for inseparable results of different fixation devices and techniques. ‡‡ One study reported both ASES and Constant scores.³¹

3.1. Description of studies included (Table 1)

Table 1.

General description of included studies.

Name of study (1st author)	Year of publication	Journal	Country of study	Study type	Number of patients	Males	Females	Age (mean, range/SD) years	Follow up (mean, range/SD) months	Imaging used	MCMS
Ma16	2016	Int Surg	China	RC	11	5	6	53.72 (38–63)	18.27 (13–26)	X ray	41
Schöffl17	2011	Arch Orthop Trauma Surg	Germany	RC	10	5	5	45.6 (29–68)	>6	NM	32
Bogdan18	2017	Arch Orthop Trauma Surg	USA	RC	10	7	3	47.1 (23–87)	8 (3–12.5)	X ray	40
Dussing19	2018	Obere Extremitat	Austria	RC	18	NM	NM	47.44 SD 17.3	60.7 SD 23.2	Xray ± CT	26
Dimakopoulos20	2006	J Orthop Trauma	Greece	RC	34	19	15	52.8 (18–84)	57.9 (24–120)	X ray ± CT	52
Chen21	2013	Orthopaedics	China	RC	19	16	3	42 (25–66)	33.2 (24–42)	Xray & CT	39
Hu22	2018	Injury	China	RC	68	48	20	38.1 (23–67)	30.5 (14–46)	X Ray	41
Ji23	2010	Arthroscopy	USA	RC	16	11	5	56.5 (33–82)	28.25 (18–42)	X Ra & CT	39
Liao24	2016	CORR	China	RC	17	13	4	50.5 (31–69)	35.2 (24–48)	Xray & CT	40
Liao24	2016	CORR	China	RC	15	10	5	45.8 (27–65)	31.7 (24–42)	Xray & CT	40
Mattyasovzsky25	2011	Acta Orthopaedica	Germany	RC	12	5	7	58.4 (31–80)	36 (9–120)	Xray ± CT	30
Flatow26	1991	JBJS	USA	RC	12	7	5	53 (34–73)	54 (24–96)	Xray ± CT	39
Elkady27	2017	IJOS	Egypt	RC	12	8	4	46 (22–63)	18 (15–30)	Xray	39
Bhatia28	2006	Injury	South Africa	RC	21	10	11	51 (17–93)	42 (12–60)	Xray	42
Xue29	2017	International Orthopaedics	China	RC	60	22	38	57 (27–87)	NM (6–24)	Xray ± CT	43
Gillespie30	2015	American Journal of Orthopaedics	USA	RC	11	4	7	60 (37–71)	21 (16–44)	NM	39
Cheng31	2019	J Chin Med Asso	Taiwan	RC	25	7	18	75 (60–88)	50.8 (22–80)	Xray	24
Platzer9	2008	J Trauma	Austria	RC	52	31	21	47 (25–66)	62 (24–132)	Xray ± CT	35
Yoon32	2017	JSES	South Korea	RC	29	19	10	56.7 (24–79)	24	Xray & 3D CT	24
Ji33	2016	Acta Orthop Trauma Surg	Germany	RC	40	20	20	55.6 (27–80)	32 (24–48)	Xray & 3D CT/MRI	34
Park34	2015	Knee Surg Sports Traumatol Arthrosc	South Korea	RC	11	2	9	64 (41–83)	26 (18–40)	Xray & CT	39
Choi35	2018	Acta Orthop Belgica	South Korea	RC	13	3	10	59.87 (40–81)	30 (12–56)	Xray & CT/MRI	28
Wang36	2012	China Med J	China	PC	23	15	8	34.8 (27–45)	20 (18–36)	Xray	44
Li37	2017	Int Orth.	China	PC	14	8	6	62.9 (49–74)	18.9 (6–30)	Xray and MRI	34
Jang38	2018	J Orth Surg	South Korea	RC	11	8	6	46.5 (23–72)	22 (17–38)	Xray & 3D CT	39

NM-Not Mentioned, sd - standard deviation, RC-retrospective cohort, PC-prospective cohort, MCMS- Modified Coleman Methodology Score.

All of the studies were retrospective cohorts, except two which were prospective.^36,37 All the studies had a mean follow-up of >6 months (range: 6–75 months). There were 562 patients in all, 53.7% being males. Mean age was 51.7 years.

3.2. Quality assessment (Table 1)

The mean MCMS was 36.79 (range 24–52). According to the grading system, all these studies were of poor quality, reflecting systematic bias as all studies are non-randomized cohort studies, with all except two^36,37 being retrospective.

3.3. Associated injuries and union time (Table 2)

Table 2.

Details of included studies.

Name of study (1st author)	Technique	Pt numbers	Fracture displacement threshold/other criteria for surgery	Implant used	surgical approach	Outcome measure	Union time -mean, range	Shoulder dislocation	BPI/NI	Effect of BPI on results	Other soft tissue injuries
Ma16	Open	11	>5 mm	F3 plate-flexible Y type-miniplates (LPP)	Lateral	CM, DASH	all united at 12 m F/U	NM	no		RCT-1
Schöffl17	Open	10	>5 mm	Calcaneal plate-cut in to 6–9 holes plate (LPP)	Lateral	CM	all united at > 6 m F/U	1	no		RCT-1, BTT -1, BL-1
Bogdan18	Open	10	>5 mm or fracture dislocation	VA LCP Mesh Plate 2.4/2.7 (LPP)	Lateral	DASH	8.5 w	9	2: both in # dislocation	poor results in both	NM
Dussing19	Percutaneous: 7; Open: 11	18	Age<65 yrs: >3 mm & age>65 yrs: >5 mm.	Screws-3mm, suture anchors, ss wires.	NM	CM, SSV, WOSII, Rowe score,	NM	all (55)	BPI-1, RN-2, AN-4.	all recovered	NM
Dimakopoulos20	Open	34	All fracture dislocations	Sutures	Lateral	CM	7-11 w	all (34)	BPI-3, AN-2, RN-2	1 stiffness	RCT-22 (SSp &/Isp avulsion-6)
Chen21	Open	19	>5 mm and fragment size not <15 mm	AO-X shaped mid foot plate (LPP)	Lateral	CM	9.4 (8–14) w	11	NM	n/a	RCT-5
Hu22	Open	68	>5 mm and Split. avulsion and depression type #s-excluded	Pre-contoured locking plate (LPP)	Lateral	CM	9.4 (8–14) w	30	BPI-5 (all # dislocation)	all had stiffness	RCT-11
Ji23	Arthroscopy	16	min-5mm, max-20mm	Suture anchors	N/A	ASES, VAS, UCLA	6-12 w	7	BPI-1 (# dislocation)	1 stiffness	BL-5, PASTA-1, CT-1, SLAP -2, pBTT-4
Liao24	Open	17	>5 mm but < 2 cm. Open If displacement >1 cm or size > 3x3 cm	PHILOS plate	deltopectoral	ASES, VAS	<12 w	NM	excluded	n/a	excluded
Liao24	Arthroscopy	15	>5 mm but < 2 cm.	Suture anchors	N/A	ASES, VAS	<12 w	NM	excluded	n/a	excluded
Flatow26	Open	12	>1 cm	Sutures	Lateral	Own criteria	NM	2	NM	n/a	NM
Elkady27	Percutaneous	12	>3–5 mm with no comminution	Screws + Washer	N/A	CM	NM	4	NM	n/a	NM
Bhatia28	Open (Arthroscopy done first)	21	>5 mm and comminution. Excluded if size < 5 mm	Suture anchors	Lateral	CM	NM	11	NM	n/a	SLAP-1, PASTA-2
Xue29	Open	43	>1cm/>45°	PHILOS plate	deltopectoral	CM	NM	excluded	NM	n/a	NM
Xue29	Open	17	>1cm/>45°	Mini plate (LPP)+ Suture	Lateral	CM	NM	excluded	NM	n/a	NM
Gillespie30	Open	11	NM	PHLP Zimmer plate + Sutures	deltopectoral	SANE, PSS	10.7 (6–21) w	NM	NM	n/a	NM
Mattyasovzsky25	Open-9, Percutaneous-3	12	<5 mm: 3, 6–10 mm: 4, >10 mm: 5	Screws: 10, Plates: 2	Lateral for ORIF	CM, DASH, ROM	NM	3	NM	n/a	NM
Cheng31	Open	25	displaced fractures	PHLP Zimmer plate	deltopectoral	CM, VAS, ASES	14 (10–18) w	excluded	NM	n/a	Excluded
Platzer9	Open	30	>5 mm	Suture: 21, TBW: 9	Lateral	CM, VSS, UCLA	NM	9	NM	n/a	RCT-16
Platzer9	Percutaneous	22	>5 mm	Screws	N/A	CM, VSS, UCLA	NM	see above	NM	n/a	NM
Yoon32	Open	29	>5 mm-split and avulsion fractures, excluded #s with severe comminution of lat wall	Screws with washer	Lateral	ASES, VAS, UCLA, SSV	<12 w	NM	NM	n/a	NM
Ji33	Arthroscopy	40	>3 mm but <2 cm with or without comminution. Excluded #s which extended beyond mid head level or surgical neck	Suture anchors	N/A	ASES, VAS, UCLA, KSS, SST	8-12 w	8	BPI-1	recovered 10 weeks	RCT-8, SLAP-7, BTT -6, pST-2.
Park34	Arthroscopic assisted plate fixation.	11	>5 mm; comminution of >2 fragments; excluded #extended up to, or beyond the surgical neck.	PHLP Zimmer plate after medial row suture anchors	MIPO-5cm incision lateral	ASES, UCLA, SST	6-12 w	2	NM	n/a	PASTA-10, pBTT-1, ST-1;
Choi35	Arthroscopy	13	NM	Suture Anchors	N/A	CM, UCLA, VAS.	11-24 w	1	NM	n/a	RCT-5, BTT-1
Wang36	Artroscopic assisted Screw fixation	23	>5 mm	Screws + Washer	N/A	CM	12 (8–24) w	NM	NM	n/a	NM
Li37	Arthroscopy	14	>3 mm: avulsion or comminution (>2 fragments). Excluded if # extended beyond surgical neck or bicipital groove	Suture anchors	N/A	ASES, UCLA, SST	<12 w	7	NM	n/a	Excluded
Jang38	Arthroscopy	11	>5 mm. Excluded if # extended to or over the surgical neck	Suture anchors	N/A	ASES, KSS,VAS, ROM.	10 (7–12) w	NM	NM	n/a	RCTs excluded. SLAP-2, BTT-4, BL-2

#- Fracture; NI: Nerve Injury; NM: Not Mentioned; N/A-Not Applicable; LPP: Low Profile Plate; Lateral approach is deltoid splitting approach; VA: variable angle; CM-Constant-Murley; SST: Simple shoulder test; SANE: Single Assessment Numerical Evaluation; SSV: Subjective shoulder value; VSS: Vienna shoulder score; UCLA: The University of California Los Angeles shoulder score; VAS: Visual Analog Scale; WOSAI: Western Ontario Shoulder instability index.

#: fractutre; mo: month; wks: weeks; yrs: years; BPI: Brachial Plexus Injury; RN: Radial Nerve; AN: Axillary Nerve; RCT: Rotator Cuff Tear; SSp: Supraspinatus; Isp: Infraspinatus; BTT: Biceps Tendon Tear; BL: Bankart's Lesion; SLAP: Superior Labrum Antero-posterior; CT: Capsular Tear; PASTA: Partial Articular Supraspinatus Avulsion.

Sixteen studies reported patients with a concomitant anterior dislocation (fracture-dislocation) in their series. Two studies reported exclusively on these patients.^19,20 In the remaining 14 studies, 33.9% of patients had fracture-dislocations.^9,17,18,21, 22, 23^,25, 26, 27, 28^,33, 34, 35^,37 Six studies reported brachial plexus injuries in 13 (5.8%) patients18, 19, 20^,22,23,33 and 9 of them had poor outcomes. Associated soft tissue lesions were reported in 12 studies,^9,16,17,20, 21, 22, 23^,28,33, 34, 35^,38 with one of them³⁸ excluding patients with rotator cuff tears. 24.8% of patients had associated rotator cuff tears.^9,16,17,20, 21, 22, 23^,28,33, 34, 35 The next most common lesions in decreasing order were Biceps tendon tears (5.5%), PASTA lesions (4.2%), SLAP tears (3.9%) and Bankart lesions (2.6%).^9,16,17,20, 21, 22, 23^,28,33, 34, 35^,38 The use of CT scan to further evaluate the injury was reported in 15 papers and was used in 63.9% patients. The corresponding figures for MRI were 3 papers and 11.9% patients (Table 1). Fifteen studies reported the union time which was less than 12 weeks.^18,20, 21, 22, 23, 24^,30, 31, 32, 33, 34, 35, 36, 37, 38

3.4. Fracture displacement/characteristics used as surgical criteria (Table 2)

2 authors fixed fractures with displacement >3 mm^33,37 and they both used pre-operative CT or MRI scan for evaluation. (Table 1). 2 authors offered fixation in fractures displaced >1 cm^26,29 5 authors did not mention any displacement criteria^{20,25,30,31,35} Remaining 15 authors accepted a 5 mm fracture displacement threshold to proceed for fixation, though one of them¹⁹ operated upon fractures displaced >3 mm if patients were younger than 65 years. For arthroscopic fixation, 2 authors^23,24 excluded fragments larger than 2 cm, and 4 excluded fractures extending beyond the surgical neck.^33,34,37,38 Interestingly, one of the latter had to convert 3 procedures from arthroscopic to open fixation due to large size of the fracture fragment.³⁸

3.5. Surgical techniques and Fixation device (Table 2)

Open techniques were most commonly used, followed by arthroscopic and percutaneous techniques. The fixation devices most commonly used were, in order, plates, sutures and screws. Plates used included Standard Locked Plates (SLPs)- pre-contoured locking plates designed for proximal humerus fractures, namely PHILOS™ (Synthes™) and PHLP™ (Zimmer™), and Low-Profile Plates (LPPs)- small or ‘mini’ locked plates that could be contoured on-table and accepted smaller screw sizes, usually designed for other anatomical locations. (For the specific names of plates used by authors, see Table 2, ‘implant used’ column). We will use these terms to describe these plates in the following sections. The commonest means used for suture fixation were suture anchors followed by trans-osseous tunnels. In the open group, all SLPs were applied via a deltopectoral approach except in 1 study³⁴ in which arthroscopy assistance was used along with small incision. The lateral deltoid split, however, was the most common approach used for open reduction overall. All the LPPs were applied via this approach. The authors who employed screws or sutures as fixation devices with open reduction, also exclusively used this approach.

3.6. Complications and re-operations (Table 3)

Table 3.

Complications and re-operations.

Main groups- Pooled analysis
Complications	Screws	Plates	Sutures/suture anchors
No. of studies	69,19,25,27,32,36	1016,17,18,21,22,24,29,30,31,34	9 (Suture anchors- 723,24,28,33,35,37,38 Sutures- 220,26)
No. of patients	114	242	176 (Suture anchors- 130, Sutures-46)
Implant failure/suture pull out	5a	1b	5
Impingement/soft tissue irritation	4	8c	4
Stiffness	14	12d	13e
Re-displacement	9	3	1
GT absorption	1	0	5f
Post op nerve injury	0	0	1g
Total Complications (TCs)	33	24	29
Pooled proportions (random effects) (95% CI)	0.2996 (29.96%) (0.1652–0.4546)	0.1057 (10.57%) (0.0359–0.2064)	0.1420 (14.20%) (0.0599–0.2516)
Q (p value)	15.0689 (P = 0.0101)	45.3083 (P < 0.0001)	27.3111 (P = 0.0006)
I2 (95% CI)	66.8% (0–84.1)	77.9% (57.1–86.3)	70.7% (28.9–83.6)
Re-operations due to reported complications$	15h	8	4
Pooled proportions (random effects) (95% CI)	0.0965 (9.65%) (0.0134–0.2430)	0.0409 (4.09%) (0.0110–0.0889)	0.0261 (2.61%) (0.0073–0.0559)
Q (p value)	23.3565 (p = 0.0003)	21.9746 (p = 0.0152)	8.6336 (p = 0.3741)
I2 (95% CI)	78.6% (40.4–88.6)	54.5% (0–75.3)	7.3% (0–57.6)
Sub-group analysis of complications within Plates group
	Standard Locked Plates	Low Profile Plates
No. of studies	524,29,30,31,34	616,17,18,21,22,29
No. of patients	107	135
Total complications (TCs)	16	7
Pooled proportions (random effects) (95% CI)	0.2097 (20.97%) (0.0368–0.4730)	0.0497 (4.97%) (0.0151–0.1025)
Q (p value)	33.1070 p < 0.0001	6.3002 p = 0.2781
I2 (95% CI)	87.9% (71.8–93.1)	20.6% (0–68.5)
Re-operations	4	4
Pooled proportions (random effects) (95% CI)	0.0432 (4.32%) (0.0033–0.1247)	0.0420 (4.2%) (0.0035–0.1202)
Q (p value)	9.3766 p = 0.0523	12.3751 p = 0.03
I2 (95% CI)	57.3% (0–82.1)	59.6% (0–81.5)

(a) 4 screws migration (19) (b) due to fall (34) (c) 6 impingements in PHILOS™ plates^24,29 and 2 soft tissue irritation in mini locked plate (mesh plate)¹⁸ (d) 6 patients had no clinical problems (30) and 5 patients had Brachial Plexus Injuries (BPI)²² (e) 2 patients had BPI^{20, 23} (f) 4 patients were asymptomatic²⁰ and 1 had an unsatisfactory result²⁸ (g) Axillary Nerve injury²⁶ (h) one author³⁶ removed screws in all 23 patients (not included in the table) as part of protocol and another³² offered screws removal to all the patients but did not mention how many underwent screw removal.

$ Reasons for reoperations:Screws group-1 for failed screw; 4 for screw migration¹⁹; 9-arthroscopy release for stiffness³²; 1-biccipital tendinitis²⁷Plates group-2 implant removal on demand¹⁷; 2 for irritation¹⁸; 2 for impingement²⁴; 1 MUA for stiffness²⁴; 1 revision for failure following fall.³⁴Sutures group- 1 capsular release for stiffness²³; 2-Bicipital tendinitis, 1-suspected Avascular Necrosis of GT-found foreign body reaction.²⁸

£ 2 plates removal for soft tissue irritation¹⁸ and 2 on patient's demand.¹⁷

#-2 removal for impingement, 1 MUA for stiffness,1 revision for implant failure.

The pooled proportions for TCs were 29.96%, 14.2%, 10.6% among studies using Screws, Sutures and Plates respectively. (Forest Plot- Fig. 2). The corresponding pooled proportions for reoperations were 9.7%, 2.6% and 4.1%. (Forest Plot- Fig. 3). There were no surgical site infections in any study except Platzer et al.⁹ who reported 2 superficial infections. However, this study could not be included for pooling proportions of TCs, as they used 2 different fixation devices, and reported their results together.⁹ In the subgroup analysis, we found a lower pooled proportion of TCs (4.97%) and re-operations (4.20%) in studies using LPPs.

Fig. 2.

Forest Plots of Complications (proportions) in studies using different implants (pooled using a Random effects Model).

Fig. 3.

Forest Plots of Reoperations (proportions) in studies using different implants (pooled using a Random effects Model).

3.7. Outcomes (Table 2, Table 4)

Table 4.

Pooled Weighted Means of Constant-Murley and ASES Scores by Surgical technique.

Technique	Open	Arthroscopy
Constant-Murley Score: Pooled Weighted Means
No. of studies	7 (16,17,20,21,22,29,31)	2(35,36)
No. of patients (Ref of papers)	227	36
Age (years)- Mean ± SD	50.8 ± 12.2	43.9 ± 8.4
Pooled WM (random effects) (95% CI)	89.14 (86.58–91.70)	82.76 (64.14–101.37)
Q (p value)	104.536 (p < 0.0001)	29.131 (p < 0.0001)
I2 (95% CI)	93.30% (89.09–95.89)	96.57% (90.70–98.73)
ASES score: Pooled weighted means
No. of studies	3 (24,31,32)	5 (23,24,33,37,38)
No of patients (Ref of papers)	71	96
Age (years)- Mean ± SD	61.7 ± 11.7	54.2 ± 12.8
Pooled WM (random effects) (95% CI)	90.09 (87.08–93.09)	92.16 (88.95–95.37)
Q (p value)	7.684 (p = 0.0215)	32.311 (p < 0.0001)
I2 (95% CI)	73.97% (13.09–92.20)	87.% (73.55–94.21)

CM score was used by 14 studies and ASES score by eight, with one author³¹ reporting both. Results from three studies which did not use any of these scoring systems could not be pooled.^8,26,30 Another study was also excluded as open fixation was performed using plate with arthroscopic assistance (combination of two techniques).³⁴

Pooled means of open and arthroscopic technique groups were 89.1 and 82.8 respectively using CM score. Corresponding figures for ASES score were 90.1 and 92.2 for both techniques in the above order (Forest Plot-Fig. 4).

Fig. 4.

Forest Plots of pooled means of Constant and ASES scores among studies grouped by surgical technique.

Among 4 studies employing percutaneous technique,^9,19,25,27 all reported CM scores. However, no study qualified for pooling as 2^19,25 clubbed outcomes together with open techniques, while in another 2^9,28 the Standard Error could not be estimated from presented data.

4. Discussion

4.1. Pooled CM and ASES

When assessing the outcomes based on surgical technique employed, pooled CM scores in the open technique group were higher than arthroscopic group, with a difference of 6.8 points. This observation was reversed when comparing the pooled ASES score, as the arthroscopic group recorded about 2 points more than open. There are three points to consider here. First, CM score is a mix of subjective and objective assessment and ASES is purely subjective score; therefore, they both should not be compared. Second, in both scenarios where scores were low, there were fewer studies in those groups-2^35,36 in arthroscopic for CM score and 3^24,31,32 in open group for ASES score (Table 4). Thirdly, the difference is too small to be of clinical relevance. For all these reasons, superiority of one technique above other couldn't be justified.

4.2. Pooled proportions of TCs and re-operations

Use of Screws was found to be associated with higher pooled proportions of total complications (29.96%) and reoperations (9.64%) than both other implant devices. Stiffness is the most common complication in all three groups with screws demonstrating the highest rate (12.28%). Plates had the lowest rate of stiffness (4.95%) and sutures came second (7.38%). A study done by Gaudelli et al. found that locking plates had better average load to failure and stiffness than suture bridge repair and tension band repair with heavy sutures for the split fractures of GT.³⁹ In another study, Lin et al. found that the suture anchor constructs were stronger than screws for the humeral GT fracture.⁴⁰ Further, Bass et al. investigated Bone Mineral Density with Dual-Energy X-ray Absorptiometry in the 35–50 age group and reported osteopenia in more than 25% of both men and women.⁴¹ The mean age of the patients in this review was 51.7 years (48 years in the screws group). Therefore, screws alone can't provide stable fixation as compared to locked plates and sutures. These considerations may explain the rate of both implant failure (4.39%) and re-displacement (7.89%), which was highest in the screw group. Locking plates had 3.3% rate for impingement/soft tissue irritation which was broadly comparable to screws (3.5%) and sutures (2.27%). (Table 4). They had less overall complication rate. However, they cannot be used for small multi-fragmentary and avulsion fractures where, we feel, sutures techniques are the mainstay.

4.3. Subgroup analysis of LPPs

Due to their low profile, these plates are not bulky and can be contoured to the GT easily. They can be applied higher on the GT and more directly onto the fractured fragment which is an advantage over the SLPs which are designed to be applied more inferiorly and can lead to impingement due to the bulkier profile if moved higher in order to achieve a similar mechanical advantage. At the same time, the LPPs maintain the utility of a fixed angle device. They behave akin to multiple screws applied with inter-connected washers which increases the pull-out strength of the construct and also provide a lateral tension-band against the pull of the rotator cuff. This may be linked to the lower incidence of both TCs (4.96%) and re-operations (4.20%) in this subgroup. There is a possibility, however, that they may have been used in a more carefully selected group of less comminuted GT fracture a source of potential bias.

4.4. Suture fixation

One argument for using sutures is that they are less prone to impingement than plates and can be employed with arthroscopic technique. Also, plates can't be practically used for small multi-fragmentary and cuff avulsion fractures where sutures are the only option. However, suture fixation by arthroscopic techniques also has limitations as large fragments and fractures extending beyond the level of the surgical neck preclude their use.^{23,24,33,34,37,38} Arthroscopic reduction is also difficult if the fragment is very displaced. In fact, some authors have not used arthroscopic means for fractures displaced more than 1 cm.²⁴ Jang et al. had to convert 3 arthroscopic surgeries to open fixation due to large size of fracture fragment.³⁸ Therefore, a pre-operative CT scan is recommended, if in doubt, to measure the size, morphology and extent of displacement of the fracture if arthroscopic fixation is being contemplated.

4.5. Fracture dislocations

5.8% of patients with fracture dislocations also had a brachial plexus injury. This association led to a poor outcome in 63.8%. Therefore, a high index of suspicion should be maintained for neurological injuries in fracture dislocations.

4.6. Surgical approach

Lateral deltoid split approach was used for all open reductions that used sutures, screws or LPPs but not for SLPs. These fractures almost invariably displace posteriorly either in inferior or superior direction.⁴² Deltopectoral is an anterior approach providing access anterior to the usual location of the displaced fragments. A lateral deltoid split approach is a recognised approach for GT fixation and offers the advantage of directly facing the fracture as well as retrieving posteriorly displaced fragments. However, the axillary nerve crosses the surgical field and a double window approach (split above and below the nerve) is required to keep the nerve safe. This likely reflects the authors’ preference in choosing the delto-pectoral approach when using the SLP.

The authors description of their method for using the low-profile plates indicates that none of them needed to dissect distal to axillary nerve.16, 17, 18^,21,22,29 There is also no need to dissect distal to the axillary nerve for sutures or screws. Therefore, this approach is ideal for using these fixation methods and provides the dual advantage of a small incision, and direct access to the fracture. There was only one study comparing LPPs to standard locked plate (PHILOS) that found LPPs to be performing better in terms of complications and outcome scores.²⁹

4.7. Limitations

There are several limitations in this review- (1) All the included studies were non-randomized cohort studies. Additionally, all were retrospective except two.^36,37 Mean MCMS is 36.79 (range 24–52). Therefore, these studies are poor on quality assessment and evidence generated from them is weak. (2) All the studies except two^9,24 are non-comparative; therefore, any comparison between pooled means would be invalid. The compared techniques in these studies were not the same so results could not be pooled (3) The included cohorts may not be similar in terms of fracture size or morphology. In practice, these factors guide the choice of implant and technique. Plates and screws are usually selected only for bigger, less comminuted fragments. On the other hand, sutures on their own or reinforced by plates^29,30,34 are accepted options for small, multi-fragmentary and avulsion fractures. However, these factors are not mentioned in nearly half the papers reviewed. Where mentioned, the criteria overlap significantly between studies. Hence, we are unable to ascertain if this has influenced the composition of the included cohorts (4) There was very high clinical (in terms of fracture size and displacement, associated fracture dislocations, types of various implants used and dissimilar post-operative rehabilitation) and statistical heterogeneity (high Q and I² > 60%) in all groups. Hence, we did not analyse the pooled outcome scores or complications between groups as such comparisons would be invalid. (5) The inclusion and exclusion criteria, and complications were not uniformly reported, which decreases the quality and reliability of assessment of their incidence. Therefore, this review reflects the need for good quality comparative studies with the use of uniform validated outcome measures and standard protocols for reporting of associated injuries and complications.

4.8. Strengths

This is the first review which had looked systematically at the outcomes and complications of operative management of isolated GT fractures. Additionally, arcsine transformation method was used in this review to calculate weights of the studies with zero complications. This increased generalizability and reduced the risk of increasing the weight of pooled proportions if these studies were to be excluded.⁴³ On the other hand, other methods that choose to exclude these studies can overestimate the effect size. Levy et al.¹² weighted the study outcome means by the number of participants in each study to pool the WMs, which is not an appropriate statistical method to combine studies. We used the inverse variance method which is a well accepted method for meta-analysis. On the other hand, Levy et al. combined complications from all the studies for each fixation device without using any method to weight the studies and did not calculate weighted proportions like us. Therefore, our review gives more confidence in the pooling of studies on this topic.

5. Conclusion

Inconsistent reporting and wide variations in the choice of outcome measures mean that there is no clear evidence to favour any of the techniques discussed on the basis of outcome measures. Our pooling of proportions of complications and re-operations according to fixation device favours the use of smaller/low profile plates and does not support the use of screws. However, a pragmatic approach is necessary as certain fracture types like small and multi-fragmentary cuff avulsions are not amenable to plating and require suture fixation methods. The evidence is weak due to high inherent bias arising form study designs which is reflected in both high statistical heterogeneity and high confidence interval of pooled proportions. We propose prospective studies for comparison between arthroscopic sutures and smaller (LPPs) locking plates with standard reporting of outcome measures, associated injuries and complications, in order to draw concrete results.

Funding sources

None.

Disclosure of funding

None.

Appendix A. Supplementary data

The following is the Supplementary data to this article: