Biomechanical testing was performed on nine fresh-frozen, cadaveric humeri obtained from eight donors (Science Care, Aurora, Co, USA) with a median age of fifty-nine years (range: 40 - 81). The median body mass index was 28.7 kg/m² (range: 17 - 29.5). Specimens were without a history of humeral fracture, tumor, infection or instrumentation.

The physiologic rotator cuff forces that are present even when wearing a sling were simulated in order to accurately assess the interfragmentary stability afforded by the two forms of internal fixation. We constructed a testing model that included the insertions of the rotator cuff muscles which were loaded similarly to their native force vectors, under previously described physiologic loads [28] [29] [30] .

Proximal humeri were dissected to maintain the infraspinatus (ISP), supraspinatus (SSP) and subscapularis (SSC) tendons and a three-part fracture (Neer Classification) [31] of the proximal humerus was created. An osteotomy of the GT was performed 5 mm posterior to the bicipital groove and completed with the saw blade exiting at the perimeter of the SSP, ISP and teres minor tendon insertions [32] [33] . Care was taken to preserve the insertions in all specimens. We specifically intended to evaluate the integrity of fixation in a comminuted GT fracture model. To simulate this comminution in a standardized method a second osteotomy of the GT fragment was performed at its midpoint with an osteotome in the horizontal plane. The surgical neck fracture was then created with an oscillating saw perpendicular to the humeral shaft. Specimens were frozen at 0˚C for storage and each was thawed twelve hours prior to testing. Fractures were reduced and fixed with a Synthes Proximal Humerus Internal Locking System (PHILOS®) (Synthes, Inc., West Chester, PA). The device was implanted on the lateral aspect of the proximal humerus 0.5 cm posterior to the lateral edge of the bicipital groove. Six locked screws were placed proximally and three non-locking screws were placed into the shaft. Internal fixation of the GT fragment was achieved in each specimen.

The myotendinous junctions of the SSC, SSP and ISP were sewn with heavy, nonabsorbable suture in Krackow fashion. Size 2 metal rings (Essentials-South Bend, Northbrook, Illinois) were sutured to each of the tendons three centimeters proximal to the plate in the twelve and three o’clock positions for left sided constructs and twelve and nine o’clock positions for right sided plates. Two rings were placed into the suture holes of the plate in the twelve and three o’clock positions for left-sided plates and the twelve and nine o’clock positions for right-sided plates (Figure 1(a) and Figure 1(b)). Plastic tie straps were used rather than suture to eliminate bias that would be introduced if the sutures were hand tied. The tie straps were uniformly tensioned to 30 N as determined by a tensionmeter [24] .

All humeri were cut nineteen centimeters distal to the humeral head. The distal aspect of the specimens was placed in a polyvinyl chloride (PVC) cylinder and then into a 120 cm × 76 cm aluminum container. Vertical alignment of the shaft relative to the base was determined by centering the distal humeral shaft in the PVC cylinder with a micrometer. The humeral shaft/PVC construct was then secured to the aluminum container with screws and polymethylmethacrylate resin. Constructs were bolted to the base of the testing system in neutral rotation with the humeral head in contact with a rigidly fixed, total shoulder glenoid component (Figure 2).

Testing was performed using a servoelectric materials testing machine (Test Resources, Model 800 L, Shakopee, MN) and a 3D motion capture system (Optotrak Certus™, Northern Digital, Inc., Waterloo, Ontario, Canada). The humeral shaft/PVC construct was rigidly attached to a torsion actuator and under load control of a 2.2 kN axial load cell (Test Resources) attached to the ISP tendon (Figure 3). Individual, static forces applied to the remaining rotator cuff muscles were based upon previously published values observed in vivo [28] [29] [30] . The SSC was statically tensioned using a pulley and 6.77 kg of weight to simulate a force of 66.3 N in line with the muscle’s vector of action [28] [29] . The SSP was statically tensioned using a pulley and 2.61 kg of weight to simulate a force of 25.6 N in line with the muscle’s vector of pull. The torsion motor applied 2000 cycles of sinusoidal force to the ISP tendon, ranging from 5 to 66.3 N (applying a moment of 0.1 to 1.25 N-m) [30] , at 1 Hz simulating contraction of both the

subscapularis and infraspinatus tendons. The moment applied to the testing construct was calculated by multiplying the force applied to the ISP tendon by the torsion motor and the distance of the ISP insertion from the central axis of the humeral shaft. Optoelectronic LED marker triads (Northern Digital, Inc.) were rigidly attached to the humeral head, GT, and humeral shaft to quantify the relative rotations between the fracture fragments during cyclic, torsional loading of the repaired constructs. For the current study, interfragmentary stability was defined as the measured relative rotation between the GT and humeral head fragments and was calculated from the slope of the linear portion of the moment/angular displacement curve using the following equation:

where K is the torsional stiffness (N-m/deg); M is the change in moment applied to the GT fragment (N-m); and θ is the relative change in angular displacement of the GT fragment relative to the humeral head (˚) about a coordinate axis that was defined with the Optotrak system to be parallel to the central axis of the humeral shaft. Interfragmentary stability at time zero, and after 500, 1000, 1500 and 2000 cycles of loading was calculated for comparison between fixation groups. At each of these pre-defined intervals, the construct was loaded 5 times and the average K was derived and used for statistical comparison.

To mitigate the potential bias caused by testing sequence, specimens were randomly assigned to (1) locked plate (LP) alone testing first (Figure 1(a)) or (2) locked plate with suture augmentation (LPSA) testing first (Figure 1(b)). The two fixation configurations were tested with each specimen serving as their own control.

Due to the failure of 2 of the specimens during testing, statistical analysis was performed with n = 7. Torsional stiffness data was initially screened for normality using the Shapiro-Wilk statistical test. As it was determined that all data sets satisfied the normality assumption, torsional stiffness data within each construct type were compared with a one-way repeated measure analysis of variance (ANOVA) with post hoc multiple comparison of groups using Bonferroni correction to determine the effects of cycle number (0, 500, 1000, 1500 and 2000) on construct rigidity. Differences in the measured torsional stiffness between the two forms of repair after 2000 cycles of loading were compared with a paired t-test. All analyses were performed using SPSS v.20 (IBM, Chicago, IL) at a significance level of 0.05.