Predicting failures of suture anchors used for rotator cuff repair: A CT-based 3-dimensional finite element analysis

Abstract

BACKGROUND: Failure of inserted anchors has been recognized as one of the major pathomechanisms of re-tearing after rotator cuff repair.

OBJECTIVE: To predict the inserted anchor failure using CT-based 3-dimensional finite element method (CT/3D-FEM).

METHODS: Among twenty patients who underwent rotator cuff repair, 5 had anchor failure (failed anchor group) and 15 had no anchor failure (stable anchor group). A 3D model of proximal humerus was developed for each patient based on the CT data. A virtual pullout testing of TWINFIX™ anchors inserted into bone at 6 different sites was performed using FEM. Then, mean failure load of 6 anchors for each patient was compared between two groups. Moreover, an optimal cut-off value of the mean failure load was determined for predicting anchor failure.

RESULTS: The mean failure load in the failed anchor group (70.3 N) was significantly lower than that in the stable anchor group (119.0 N; $p < 0.0001$ ). In our method, the optimum cut-off value of the mean failure load was 75.4 N.

CONCLUSIONS: Failure of the inserted TWINFIX™ anchor could be predicted using CT/3D-FEM. In this method, there seemed to be a high risk of anchor failure in shoulders with a mean failure load of <75.4 N.

Keywords

Rotator cuff repair suture anchor pullout strength computed tomography finite element method

1. Introduction

Suture anchors are widely used for both arthroscopic and open rotator cuff repair, which can simplify the surgical techniques needed for tendon-to-bone suturing. Surgeons typically insert several anchors into the greater and/or the lesser tuberosities depending on the shape and the size of a tear. Although many types of anchors are currently available for clinical practice [1,2], failure of inserted anchors has been still recognized as one of the major pathomechanisms of postoperative re-tearing [3–5]. Once a complication such as this occurs, both the cuff tendon and any bony defects should be repaired during revision surgery [6]. Thus, it would be beneficial for surgeons to be able to precisely estimate the risk of anchor failure preoperatively.

Recent advancements in computer aided engineering provides for making estimates of a patient’s bone strength using CT data. Bessho et al. showed that bone strength estimated by a CT-based 3-dimensional finite element method (CT/3D-FEM) was well correlated with the results of mechanical testing. They also reported that the fracture pattern observed during mechanical testing could be successfully predicted using this method [7,8]. However, there have been no studies that applied a CT/3D-FEM to assess anchor failure after rotator cuff repair.

Thus, the aim of the present study was to determine the risk of anchor failure with use of a CT/3D-FEM. In particular, we compared the predicted failure load between patients who actually had the pullout of inserted anchor during surgery and those who did not.

2. Materials and methods

2.1. Patient selection

The protocol for this study was approved by our institutional review board. A signed consent form was obtained from all rotator cuff tear patients when they were scheduled for surgical repair in our institution. Preoperative CT scanning was performed using a bone mineral reference phantom (B-MAS200; Kyoto Kagaku Co, Ltd, Kyoto, Japan) in each patient. For this study, we included patients who had small to medium-sized full-thickness rotator cuff tears (<3 cm) in order to standardize the repair tension. Patients with a partial-thickness tear were also included if the diameter of a cuff defect after conversion to a full-thickness tear was <3 cm. Consequently, 20 shoulders of 20 patients (16 full-thickness tears and 4 bursal side partial-thickness tears) were eligible for the present study. These patients included 11 males and 9 females, whose mean age was 63 years (range: 52–75 years). The repair techniques used were single-row for 2 patients, double-row for 12 patients and transosseous equivalent for 6 patients.

Among them, patients who represented anchor failure during or immediately after the surgery and those who represented no anchor failures in the follow up X-ray as well as MRI were categorized as the failed anchor group and stable anchor group, respectively.

2.2. Clinical observations

In the present series, pullout of inserted anchor was confirmed in 4 patients arthroscopically while examining the integrity of the repair in the final step of surgery. Since all of these failures were confirmed during the surgery, immediate revision was performed by open transosseous repair technique. One patient demonstrated a partial back-out of the inserted anchor in the X-ray taken immediately after surgery. Since, however, neither the further progression of back-out nor the re-tearing of cuff tendon was seen in the follow-up study, this patient was treated conservatively. These 5 patients were categorized as the failed anchor group (4 females and 1 male; mean age: 67 years). A double-row technique was used in 4 (3 lateral-row and 1 medial-row anchor failures) and a single-row technique was used in 1. On the other hand, all inserted anchors were stable and remained in place in the postoperative X-ray as well as MRI in the other 15 patients (stable anchor group: 5 females and 10 males; mean age: 62 years).

2.3. Anchor insertion model development

In this study, we used finite element method (FEM) software, Mechanical Finder (version 6.1, Extended Edition; Research Center of Computational Mechanics, Inc., Tokyo, Japan). In each patient, CT data of the affected shoulder were saved in Digital Imaging and Communications in Medicine (DICOM) format, and then imported to a workstation to develop a 3D-FE model of the lateral part of the humeral head. Geometric data of a suture anchor (TWINFIX™, 5.0 mm; Smith and Nephew Endoscopy KK, Tokyo, Japan) were also imported to the workstation to create a virtual implant. Then, virtual TWINFIX™ anchors were inserted into the bone model at 6 different sites to develop 6 anchor insertion models for each patient. The sites of anchor insertion were the medial and lateral aspects of the superior facet, the medial and lateral aspects of the middle facet, and the anterolateral and posterolateral aspects of the greater tuberosity (Fig. 1). In each model, a virtual anchor was placed perpendicularly to the bony surface and inserted 3 mm below the level of the bony surface (Fig. 2). Next a cylindrical cutting tool (diameter: 3.0 mm) was also modeled and placed around the anchor eyelets to recreate an anchor insertion hole. The bony elements inside this tool were removed from the bony surface to the screw ridge of the inserted anchor [9].

Fig. 1.

A 3D-model of the lateral part of a humeral head (lateral view). The 6 circles indicate the insertion sites used for virtual anchors. The arrow shows the bicipital groove.

Fig. 2.

Insertion of a virtual anchor into a bone model (lateral view). A virtual TWINFIX™ anchor was placed above the greater tuberosity and tilted perpendicularly against the bony surface at the insertion site (arrow: bicipital groove).

Each model was divided into 0.2 to 1.0-mm tetrahedral solid elements. The Young’s modulus of a bone element was determined from Hounsfield unit values according to previous reports by Keller and Keyak et al. [10,11]. The Poisson’s ratio was determined to be 0.4. Triangular shell elements (thickness: 0.3 mm) were modeled on the outer surface of bone to reproduce the thin cortical shell to avoid underestimating the material properties of surface elements due to partial volume effects [7]. For each triangular shell element, the Young’s modulus was set so as to be equivalent to that of the adjacent tetrahedral solid element, while the minimum Young’s modulus of the shell element was set to 10 GPa based on previous reports [12,13].

The total numbers of solid elements and shell elements were approximately 500,000 and 50,000, respectively. Regarding the inserted TWINFIX™ anchors, the material was assumed to be a single-phase linear, elastic, isotropic titanium alloy with a Young’s modulus of 110.00 GPa and Poisson’s ratio of 0.28 [9].

2.4. Virtual pullout testing

A tensile load of 400 N was applied to the anchor eyelets in 80 incremental steps. The direction of the tensile load was set along the long axis of an inserted anchor to standardize the analysis condition. The distal part of the humeral head was completely constrained in all directions so as not to move throughout the analysis. Then, we made a non-linear analysis for fracture prediction to simulate the pullout of an inserted anchor.

Location of an element failure was determined based on the maximum principal stress failure theory according to a previous study [8]. A non-linear finite element (FE) analysis was made using the Newton–Raphson method in consideration of the large displacement. The ratio of ultimate element stress to yield stress was determined to be 0.8. We hypothesized that an element crack in tension would occur when the maximum principal stress exceeded the element ultimate stress. Yield in compression was defined as occurring when the von Mises equivalent stress exceeded the element yield stress. Element failure during compression was defined as occurring when the negative value of the maximum principal strain exceeded a microstrain of 10,000 [8]. In this study, we defined the failure load as a load at which a crack would occur in at least one shell element of bone [7,11]. Then, the mean failure load of all the 6 anchors was established for each patient.

2.5. Statistical analysis

Statistical analysis was done using Graph Pad Prism software (version 5.0a, Graph Pad Software, Inc., La Jolla, CA). Mean failure loads were compared between the failed anchor group and the stable anchor group and between males and females using Mann–Whitney U-tests. Post hoc power analysis was done using GPower software (version 3.1.5, Germany) to confirm the power of our analyses ( $1 - β$ ) when comparing between the failed anchor group and stable anchor group and between males and females. To establish an optimal cut-off mean failure load value for predicting anchor pullout, we used receiver operating characteristic (ROC) curve analysis. A p-value of <0.05 was considered significant.

3. Results

The stress distribution patterns were nearly identical for all anchor insertion models. A high stress concentration was observed around an inserted anchor (Fig. 3(a), (b)). Element failure was observed around the eyelet of an inserted anchor (Fig. 4(a), (b)). The mean failure loads of the failed anchor group and the stable anchor group were 70.3 ± 25.6 N and 119.0 ± 28.3 N, respectively (Fig. 5); the failure load of those in the failed anchor group was significantly smaller than that in the stable anchor group ( $p < 0.0001$ ). Female patients had a significantly lower mean failure load (89.1 ± 31.8 N) than male patients (124.9 ± 32.6 N; $p = 0.03$ ; Fig. 6). The power analysis ( $1 - β$ ) results for these comparisons between the failed anchor and stable anchor groups and between males and females were 0.95 and 0.75, respectively.

Fig. 3.

Distribution of von Mises equivalent stress around an inserted anchor ((a) cranial view, (b) cutting surface in the coronal plane at the insertion site). A high stress concentration was observed around the inserted anchor, particularly from bony surface to subchondral area. (Colors are visible in the online version of the article; https://dx-doi-org.web.bisu.edu.cn/10.3233/BME-151535.)

Fig. 4.

Failed elements during an analysis ((a) cranial view, (b) cutting surface in the coronal plane at the insertion site). Element failures (white dots) occurred around the anchor eyelets (arrows).

Fig. 5.

Comparisons of mean failure loads between the 2 study groups. The mean failure load in the failed anchor group was significantly lower than that in the stable anchor group ( $p < 0.0001$ ).

Fig. 6.

Comparisons of mean failure loads between genders. Females had a significantly lower mean failure load than males ( $p = 0.03$ ).

Fig. 7.

ROC curve of the mean failure load for predicting anchor pullout. The point at the left upper corner represents the optimum cut-off value of the mean failure load (75.4 N). At this point, the sensitivity and specificity were 80% and 100%, respectively.

An ROC analysis showed that the optimum cut-off value of the mean failure load was 75.4 N (sensitivity: 80%; specificity: 100%; Fig. 7).

4. Discussion

Several factors can potentially contribute to successful cuff repair using suture anchors, such as local bone and tendon quality, anchor design and tear size [14]. Among these factors, we focused on the local bone quality of the greater tuberosity because rotator cuff tears are frequently seen in the middle-aged to elderly population who often suffer from osteopenia or osteoporosis [15–17]. To exclude the effects that could have been caused by other factors, we standardized the size of a tear (<3 cm) and the type of anchor used (TWINFIX™). Therefore, we believe that anchor pullout observed in the present series was mainly caused by local bone weakness around the greater tuberosity. This is the first study to show the clinical application of a CT/3D-FEM for assessing the risk of anchor failure during rotator cuff repair.

The incidence of pullout of inserted anchor in the present series (25%) was higher than that in a previous report by Benson and colleagues [3]. In their series, anchor pullout was observed in only 6 of 269 patients (2.4%). They concluded that there was a minimal risk of suture anchor pullout with small- to medium-sized tears [3]. However, they made a retrospective survey only using medical records and postoperative radiographs. Thus, the loosening or the pullout of inserted anchor might be overlooked if the patients had not presented with any symptoms. Moreover, the patients’ age and gender were quite different from ours. The average age was much younger (55 years) and there were 70% males in their study population, whereas in our study, 4 out of 5 patients (80%) in the failed anchor group were female with an average age of 67 years. This may have also affected the difference in anchor failure rate.

The relationship between bone mineral density (BMD) and the pullout strength of inserted anchors has been investigated in many studies [14,18–22]. Tingart et al. found a positive correlation between the load to failure of inserted anchors and BMD measured by quantitative computed tomography [21]. In a subsequent study they reported that there was a positive correlation between cortical BMD and the failure load for a metal anchor, whereas they found no correlation between trabecular BMD and the failure load [14]. Recently, Yackacki et al. reported that both BMD and the trabecular thickness reflected the pullout strength of inserted anchors [22]. However, because all these previous studies used human cadaver humeral bones, their results cannot be directly applied to individual patients seen in actual clinical practice. Moreover, recent studies using quantitative computed tomography showed that there was a location-dependent difference in BMD within the greater tuberosity even in the same patient [23,24]. To predict the risk of anchor pullout, these differences in pullout strength according to the location of the bone in each patient should be taken into consideration.

In comparison, with use of a CT/3D-FEM, a patient-specific analytic model can be developed that precisely reflects the local bone architecture in each patient. The Young’s modulus in each element of a patient’s bone was calculated based on its Hounsfield unit value, which was calibrated using a bone mineral reference phantom. Because the models developed in the present study reflected both the 3D-architecture and the local BMD, we assumed that CT/3D-FEM could provide reliable data for predicting the pullout risk of inserted anchors.

In the present study, our failed anchor group had a significantly lower mean anchor failure load than that of the stable anchor group. In other words, our results with CT/3D-FEM corresponded well with actual clinical observations. We concluded that anchor failure after rotator cuff repair might be predictable using CT/3D-FEM. Another interesting finding was that female patients had a lower mean failure load than that of male patients. It seemed that the local bone strength around the greater tuberosity in females might have been lower than that in males. These results were consistent with findings by Clavert et al. who described that female gender, age of >70 years, and stage 2 cuff retraction were factors associated with osteoporosis of the greater tuberosity [15]. Therefore, we assumed that female patients might have a higher risk of anchor pullout than male patients.

Our ROC analysis showed that the optimal cut-off value of the mean failure load was 75.4 N. Considering this result, it might be better that surgeons not use suture anchors if the predicted mean failure load is less than 75.4 N because there is a high risk of anchor pullout. However, it should be noted that the amount of failure load predicted by this method was not the actual pullout strength of the inserted anchor. This cut-off value should only be used for a risk assessment of anchor failure with 80% sensitivity and 100% specificity.

The results of the present study also showed the possibility of future clinical application of this method for the prediction of failure in other implants. If both the geometric data and the material properties of any implants are available, their risk of failure can be calculated using this method. Further studies would be necessary to determine the true roles of CT/3D-FEM in the prediction of implant failures.

Our study had several limitations. First, the relationship between the Hounsfield unit value and Young’s modulus was established in experiments that used human femurs [11]. Since there are no reported reference data for the human humerus, additional biomechanical testing studies using human humeral head might be needed to confirm the relationship between these two parameters in this bone. Second, the number of our patients was rather small because we only included those with cuff defect sizes of less than 3 cm to standardize the tension at the repair site. Repair tension might be higher in large or massive tears than that in small tears, which might increase the risk of pullout of inserted anchors. Third, we only used TWINFIX™ anchors for our analysis. A number of other types of anchors that are made of different materials and that have different designs are now available for clinical use. These other types of anchors might yield different results when using CT/3D-FEM. Forth, the direction of applied load was set along the long axis of inserted anchor both to simplify and to standardize the analysis conditions. This direction is known to show “a worst-case scenario” for the inserted anchors, in which the pullout strength was likely the lowest [25,26]. In the realistic condition, however, the direction of tensile load could be changed with the angle of anchor insertion as well as the arm position [9]. Further studies using other types of anchors with various loading conditions would be necessary to clarify the true roles of CT/3D-FEM in the assessment of the pull-out strength of inserted anchors.

In conclusion, TWINFIX™ anchor failure during rotator cuff repair could be predicted using CT/3D-FEM. In this method, there seems to be a high risk of failure in shoulders with a mean failure load of less than 75.4 N.

Footnotes

Acknowledgements

This study was supported by a research grant from ZENKYOREN (National Mutual Insurance Federation of Agricultural Cooperatives). The Stereolithography data for the TWINFIX™ anchors were supplied by Smith & Nephew Endoscopy KK (Tokyo, Japan) under a non-disclosure agreement.

References

[1] Barber

F.A.

and Herbert

M.A.

, Cyclic loading biomechanical analysis of the pullout strengths of rotator cuff and glenoid anchors: 2013 update, Arthroscopy 29(5) (2013), 832–844.

[2] Barber

F.A.

Herbert

M.A.

Beavis

R.C.

and Barrera Oro

, Suture anchor materials, eyelets, and designs: Update 2008, Arthroscopy 24(8) (2008), 859–867.

[3] Benson

E.C.

MacDermid

J.C.

Drosdowech

D.S.

and Athwal

G.S.

, The incidence of early metallic suture anchor pullout after arthroscopic rotator cuff repair, Arthroscopy 26(3) (2010), 310–315.

[4] Cummins

C.A.

and Murrell

G.A.

, Mode of failure for rotator cuff repair with suture anchors identified at revision surgery, J. Shoulder Elbow Surg. 12(2) (2003), 128–133.

[5] Kaar

T.K.

Schenck

R.C.

Jr. Wirth

M.A.

and Rockwood

C.A.

Jr. , Complications of metallic suture anchors in shoulder surgery: A report of 8 cases, Arthroscopy 17(1) (2001), 31–37.

[6] Park

H.B.

Keyurapan

Gill

H.S.

Selhi

H.S.

and McFarland

E.G.

, Suture anchors and tacks for shoulder surgery, Part II: The prevention and treatment of complications, Am. J. Sports Med. 34(1) (2006), 136–144.

[7] Bessho

Ohnishi

Matsuyama

Matsumoto

Imai

and Nakamura

, Prediction of strength and strain of the proximal femur by a CT-based finite element method, J. Biomech. 40(8) (2007), 1745–1753.

[8] Bessho

Ohnishi

Okazaki

Sato

Kominami

Matsunaga

et al., Prediction of the strength and fracture location of the femoral neck by CT-based finite-element method: A preliminary study on patients with hip fracture, J. Orthop. Sci. 9(6) (2004), 545–550.

[9] Sano

Takahashi

Chiba

Hatta

Yamamoto

and Itoi

, Stress distribution inside bone after suture anchor insertion: Simulation using a three-dimensional finite element method, Knee Surg. Sports Traumatol. Arthrosc. 21(8) (2013), 1777–1782.

10.

[10] Keller

T.S.

, Predicting the compressive mechanical behavior of bone, J. Biomech. 27(9) (1994), 1159–1168.

11.

[11] Keyak

J.H.

Rossi

S.A.

Jones

K.A.

and Skinner

H.B.

, Prediction of femoral fracture load using automated finite element modeling, J. Biomech. 31(2) (1998), 125–133.

12.

[12] Bayraktar

H.H.

Morgan

E.F.

Niebur

G.L.

Morris

G.E.

Wong

E.K.

and Keaveny

T.M.

, Comparison of the elastic and yield properties of human femoral trabecular and cortical bone tissue, J. Biomech. 37(1) (2004), 27–35.

13.

[13] McCalden

R.W.

McGeough

J.A.

Barker

M.B.

and Court-Brown

C.M.

, Age-related changes in the tensile properties of cortical bone. The relative importance of changes in porosity, mineralization, and microstructure, J. Bone Joint Surg. Am. 75(8) (1993), 1193–1205.

14.

[14] Tingart

M.J.

Apreleva

Lehtinen

Zurakowski

and Warner

J.J.

, Anchor design and bone mineral density affect the pull-out strength of suture anchors in rotator cuff repair: Which anchors are best to use in patients with low bone quality?, Am. J. Sports Med. 32(6) (2004), 1466–1473.

15.

[15] Clavert

Bouchaib

Sommaire

Flurin

P.H.

and Hardy Sfa

, Does bone density of the greater tuberosity change in patients over 70?, Orthop. Traumatol. Surg. Res. 100(1) (2014), 109–111.

16.

[16] Kannus

Leppala

Lehto

Sievanen

Heinonen

and Jarvinen

, A rotator cuff rupture produces permanent osteoporosis in the affected extremity, but not in those with whom shoulder function has returned to normal, J. Bone Miner. Res. 10(8) (1995), 1263–1271.

17.

[17] Yamamoto

Takagishi

Osawa

Yanagawa

Nakajima

Shitara

et al., Prevalence and risk factors of a rotator cuff tear in the general population, J. Shoulder Elbow Surg. 19(1) (2010), 116–120.

18.

[18] Goradia

V.K.

Mullen

D.J.

Boucher

H.R.

Parks

B.G.

and O’Donnell

J.B.

, Cyclic loading of rotator cuff repairs: A comparison of bioabsorbable tacks with metal suture anchors and transosseous sutures, Arthroscopy 17(4) (2001), 360–364.

19.

[19] Pietschmann

M.F.

Frohlich

Ficklscherer

Gulecyuz

M.F.

Wegener

Jansson

et al., Suture anchor fixation strength in osteopenic versus non-osteopenic bone for rotator cuff repair, Arch. Orthop. Trauma Surg. 129(3) (2009), 373–379.

20.

[20] Pietschmann

M.F.

Frohlich

Ficklscherer

Hausdorf

Utzschneider

Jansson

et al., Pullout strength of suture anchors in comparison with transosseous sutures for rotator cuff repair, Knee Surg. Sports Traumatol. Arthrosc. 16(5) (2008), 504–510.

21.

[21] Tingart

M.J.

Apreleva

Zurakowski

and Warner

J.J.

, Pullout strength of suture anchors used in rotator cuff repair, J. Bone Joint Surg. Am. 85(11) (2003), 2190–2198.

22.

[22] Yakacki

C.M.

Poukalova

Guldberg

R.E.

Lin

Saing

Gillogly

et al., The effect of the trabecular microstructure on the pullout strength of suture anchors, J. Biomech. 43(10) (2010), 1953–1959.

23.

[23] Kirchhoff

Braunstein

Milz

Sprecher

C.M.

Fischer

Tami

et al., Assessment of bone quality within the tuberosities of the osteoporotic humeral head: Relevance for anchor positioning in rotator cuff repair, Am. J. Sports Med. 38(3) (2010), 564–569.

24.

[24] Oh

J.H.

Song

B.W.

and Lee

Y.S.

, Measurement of volumetric bone mineral density in proximal humerus using quantitative computed tomography in patients with unilateral rotator cuff tear, J. Shoulder Elbow Surg. 23(7) (2014), 993–1002.

25.

[25] Serhan

Altinel

Eroglu

Verim

Demir

and Atmaca

, Suture anchor fixation strength with or without augmentation in osteopenic and severely osteoporotic bones in rotator cuff repair: A biomechanical study on polyurethane foam model, J. Orthop. Surg. Res. 9 (2014), 48.

26.

[26] Oshtory

Lindsey

D.P.

Giori

N.J.

and Mirza

F.M.

, Bioabsorbable tricalcium phosphate bone cement strengthens fixation of suture anchors, Clin. Orthop. Relat. Res. 468(12) (2010), 3406–3412.