Comparative analysis of accuracy and safety in guided versus freehand lateral‐to‐medial humeral transcondylar drilling

Abstract

Objective

To compare the performance of three guided humeral transcondylar drilling techniques and the freehand drilling technique based on anatomical landmarks.

Study design

Randomized block design ex vivo study.

Sample population

Canine cadavers (n = 48; 96 elbows).

Methods

Elbows were scanned using computed tomography (CT) to manufacture patient‐specific three‐dimensional (3D)‐printed drill guides (3D‐PDG) and were randomly assigned to one of four groups: universal aiming device (UAD), fluoroscopy‐assisted drilling (FAD), 3D‐PDG, and freehand drilling (FHD) performed by two board‐certified surgeons. All elbows were CT scanned. Drilled tunnels (DT) were compared with an optimal drilling axis (ODA). Outcome measures included drilling time, DT entry and exit point translations, angulation, distance to the articular surface, and joint infringement.

Results

Drilling technique influenced both entry and exit point translation. Freehand drilling was the fastest and safest method (0% joint infringement). The UAD exhibited the highest incidence of joint infringement (26.1%), followed by 3D‐PDG (21.7%) and FAD (4.2%). Most joint infringements were caused by unacceptable (>10°) craniocaudal or proximodistal DT angulation.

Conclusion

When performed by experienced surgeons, FHD is accurate and safe, and may outperform guided techniques. Laterally applied 3D‐PDG necessitates specific design modifications to improve stability, as well as surgeon training and feedback. The UAD technique may be adversely affected by the lack of a surgical assistant and by limb position. Fluoroscopy‐assisted drilling is accurate, but it results in a longer procedure time and requires equipment that may not be readily accessible.

Clinical significance

Experienced surgeons can reliably perform FHD utilizing anatomical landmarks with satisfactory accuracy for humeral condylar drilling.

Abbreviations

3D‐PDG

three‐dimensional–printed drill guide (Also appears as “3‐D PDG” and “3D‐printed guide”)

ACVS

American College of Veterinary Surgeons

AMI

area moment of inertia

CT

computed tomography

CVETS

Columbia veterinary emergency trauma and specialty

DT

drilled tunnel

ERL

epicondylar reference line

FAD

fluoroscopy‐assisted drilling

FHD

free‐hand drilling

HCD

humeral condylar diameter

HIF

humeral intracondylar fissure

IQR

interquartile range

K‐wire

kirschner wire

MinTAD

minimum tunnel–articular distance

MPR

multiplanar reconstruction

ODA

optimal drilling axis

SD

standard deviation

SPSS

statistical package for the social sciences

TCD

transcondylar drilling

TCS

transcondylar screw

UAD

universal aiming device

1. INTRODUCTION

Humeral condylar fractures account for 41% of all humeral fractures seen in veterinary practice and are classified as lateral, medial, and Y fractures. Lateral condylar fractures are the most common type, accounting for 56% to 67% of fractures. ¹ , ² Humeral condylar fractures commonly occur in certain breeds, including spaniels and the French bulldog. They are particularly common in skeletally immature dogs and those with humeral intracondylar fissures (HIF), often following low‐energy trauma. ¹ , ² , ³

The use of a transcondylar screw (TCS) is an integral part of rigid transcondylar fracture fixation, providing interfragmentary stability while counteracting the cyclic forces generated during weight bearing. For maximum rigidity, the TCS should have a high area moment of inertia (AMI) relative to the diameter of the isthmus of the humeral condyle. ⁴ If interfragmentary compression is desired, the TCS must also be applied perpendicularly to the fracture line. ⁵ If this orientation is not achieved, the shearing forces generated during compression will render the reestablishment of joint surface congruity nearly impossible. The challenge of placing a maximally sized TCS perpendicularly through a narrow bone isthmus underscores the need for specific implant positioning and an accurate transcondylar drilling (TCD) method. Barnes et al. established guidelines based on anatomical landmarks for locating entry and exit points of drill tunnels for TCS insertion. ⁴ They reported that the lateral‐to‐medial approach had a substantially larger tolerance in the frontal plane. In their study, the lateral proximodistal tolerance angle averaged 26.9° (±4.7°) whereas the medial‐to‐lateral tolerance angle averaged only 16.2° (±3.5°). This means the lateral‐to‐medial approach offered a tolerance that was about 10.7° greater than that attained when drilling medial‐to‐lateral. ⁴

Various guided techniques have been proposed to improve the accuracy of TCD over FHD based on palpable anatomical features, including three‐dimensional (3D) printed drill guides (3D‐PDG) and universal aiming devices (UAD). As far as we are aware, studies directly comparing the performance of 3D‐PDG and UAD to fluoroscopy‐assisted drilling (FAD) and freehand drilling (FHD) techniques have not been conducted.

The objectives of this study were twofold. First, to compare the accuracy and safety of four TCD methods—UAD, FAD, 3D‐PDG, and FHD—where accuracy is defined as the method's ability to replicate the optimal drilling axis (ODA), and safety is defined as its ability to avoid joint space infringement. Second, the study aimed to evaluate the TCD time using these four methods. We hypothesized that the accuracy and safety of 3D‐PDG would be similar to FAD and UAD—that is, less than 10° relative to the ODA, ⁴ and that all three methods would be superior to FHD; that none of the tunnels drilled with 3D‐PDG, FAD, or UAD would infringe on the joint space or supratrochlear foramen; that all tunnels drilled with 3D‐PDG would be within 5° of the ODA, and that FHD would be the most time‐efficient method.

2. MATERIALS AND METHODS

2.1. Study design

A randomized matched‐pair block design ex vivo study was designed to compare the four techniques.

2.2. Subjects

Cadaver forelimbs (n = 96) were harvested from predominantly Staffordshire terrier mixed‐breed canine cadavers with body weights ranging from 15 to 25 kg, including both sexes, and estimated ages ranging from 8 months to 8 years. The cadavers were donated by a local animal control agency (Lexington County Animal Services, Lexington, South Carolina). The animals were euthanized by the animal control agency for reasons unrelated to the study, following guidelines established by the American Veterinary Medical Association. The limbs were harvested in pairs over 6 months, catalogued, scanned using computed tomography (CT) to rule out congenital, developmental, degenerative, and traumatic elbow abnormalities, and stored at −18°C. Computed tomography scans were performed with a 64‐slice CT scanner (GE Optima 660, General Electric, Boston, MA), using a helical protocol with 0.625 mm slice thickness, 0.625 mm distance between slices, and 120 kVp. Images were evaluated using a bone algorithm, with a window width of 1500 HU (2500 kg/m³) and a window level of 300 HU (1300 mg/m³). Using a computerized randomizer (Google Sheets randomizer), limbs were assigned equally to each of two surgeons, board certified by the American College of Veterinary Surgeons (ACVS) with at least 10 years of experience, and the four TCD methods. The experiment sequence was similarly randomized to minimize the effect of consecutive learning. For the 3D‐PDG group, a patient‐specific drill guide was designed and manufactured by a veterinary 3D‐printing company (Cartesian Medical, Culver City, California) with a footprint designed to fit snugly against bone without interposing soft tissue that could impede proper application. The guide also included a K‐wire channel to improve stability during drilling. The two surgeons recruited for the study did not have prior experience using the 3D‐printed drill guides.

2.3. Procedure

Before the experiment, the limbs were thawed for 48 h, and the hair surrounding the elbow was clipped circumferentially. The limbs were vertically suspended in extension using a hanging leg technique and a grooming table arm (Duckygoo Adjustable Grooming Arm, Amazon.com, Seattle, Washington) as support. The scapula was stabilized using a table vise (IQ Vise, Amazon.com) Sea with the lateral aspect of the leg directed toward the surgeon. A standardized limited lateral approach to the distal humerus was performed by the principal investigator (RR) on all limbs. A skin incision measuring approximately 8 cm was performed over the lateral humeral epicondyle. The antebrachial fascia was incised along the cranial border of the lateral head of the triceps muscle, exposing the brachialis muscle and radial nerve, which were cranially retracted. The origin of the extensor carpi radialis and anconeus muscles was elevated from the epicondylar crest and humeral condyle to allow precise contact and a secure fit between the 3D‐PDG and the bone. To allow standardization, the surgeons were not permitted to modify or extend the lateral surgical approach.

During the experiment, the surgeons were provided with a surgical drill (Aesculap Acculan 4 Mini Drill, B. Braun Vet Care GmbH, Germany), a disposable #10 scalpel (Shield Line, Bayonne, New Jersey), and a medium Gelpi retractor (Jorgensen Laboratories LLC, Loveland, Colorado). A printed image from a reference paper illustrating the ODA based on anatomical features was also available for reference. ⁴ For the 3D‐PDG group, a 3D‐printed humeral bone model was available to demonstrate the proper fit of the drill guide based on the precise placement of the guide pods and the position of the stabilizing K‐wire (Figure 1). The surgeons were provided with 1.5 mm K‐wires (Movora NA, Saint Augustine, Florida) and 2 mm quick‐coupling drill bits (Movora NA). Neither surgeon had prior contact with or experience using 3D‐printed guides, so the principal investigator offered each surgeon brief and similar instructions on the 3D‐PDG placement technique but did not assist during the experiment. Specifically, the surgeons were shown where the guide pods were intended to contact the cranial and caudal aspects of the lateral epicondylar crest and the expected position of the drilling channel relative to the lateral humeral condyle. For the UAD group, surgeons were provided with a standard Universal Aiming Device (IMEX Veterinary Inc., Longview, Texas) and 2.5 mm quick‐coupling drill bits (Movora NA). The surgeons were permitted to perform a 1 cm stab incision over the medial humeral epicondyle to allow more secure medial bone placement of the UAD when deemed necessary for optimal application; however, no assistance was offered during guide placement or drilling (Figure 2). For the FAD group, a fluoroscopy device (GE OEC 9800, GE Healthcare, Chicago, Illinois) was positioned horizontally across from the surgeon to obtain orthogonal views of the limb (Figure 3). The surgeons were provided with 0.045″ K‐wires (Movora NA) and 2.5 mm cannulated drill bits (Arthrex, Naples, Florida), and were allowed to obtain as many fluoroscopy images and redirect the K‐wire as many times as necessary to avoid joint infringement and closely match the ODA before overdrilling. For the FHD group, surgeons were provided with a 2.5 mm quick‐coupling drill bit (Movora NA) and a 2.5 mm Universal Drill Guide (Movora NA). All procedures were timed, starting when the surgeon first touched the cadaver limb and ending when the transcortex was breached. Drill bits were replaced every six limbs for all four techniques to minimize the effect of dulling.

FIGURE 1.

Original three‐dimensional patient‐specific humeral drilling guide showing the position of the drilling channel, stabilizing pods, and K‐wire.

FIGURE 2.

Universal aiming device application and single‐hand stabilization.

FIGURE 3.

Cadaveric limb setup for fluoroscopy‐assisted drilling transcondylar humeral drilling.

Following the experiment, CT scans of all limbs were repeated. Using three‐dimensional and multiplanar (MPR) reconstructions on an open‐source 3D imaging software (3D Slicer, https://www.slicer.org), the location and orientation of the drilled tunnels were compared with an optimal drilling axis (ODA), established based on a previous publication using anatomical landmarks. ⁴ A sagittal projection was used to measure the humeral condylar diameter (HCD) at the level of the condylar isthmus and the minimal distance between the drilled tunnel (DT) and the humeral articular surface (MinTAD) (Figure 4). Next, an epicondylar reference line (ERL) was drawn between the most prominent aspects of the medial and lateral epicondyles. The caudal border of the medial supracondylar crest was used as a reference for the frontal plane, and the transverse plane was established perpendicular to the frontal plane and parallel to the ERL. Python code was written to establish the ODA using reference points automatically marked at 0.3 times and 0.2 times the HCD medially and 0.3 times the HCD cranial and distal laterally. ⁶ Finally, two points in the visual center of the drilled tunnel at its most lateral and medial extent were marked. A line connecting these two points established the drilled tunnel (DT) (Figures 5 and 6).

FIGURE 4.

Sagittal plane view of the distal humerus at the level of the condylar isthmus, showing the cranial and caudal articular surfaces and the center of the drilled tunnel. These landmarks are used to determine humeral condylar diameter and the minimum distance between the drilled tunnel and the closest humeral articular surface.

FIGURE 5.

Multiplanar reconstruction and three‐dimensional (3D) volume rendering showing the optimal drilling axis (green line) and drilled tunnel (red line) relative to the medial epicondyle. The purple dot represents the center of the optimal drilling axis.

FIGURE 6.

Three‐dimensional volume rendering showing the relationship between the epicondylar reference line (blue line), optimal drilling axis (green line), and drilled tunnel (red line).

2.4. Variables and measurements

Using the custom Python code, the following measurements and result calculations were made on the 3D reconstructions following the guidelines proposed by Pardo et al.: ⁷

• The minimum tunnel‐articular distance (MinTAD) was measured in the sagittal plane, representing the distance between the center of the DT and the closest humeral articular surface.

• The lateral and medial tunnel entry point translation was measured in the sagittal plane relative to the medial and lateral origins of the ODA.

• Tunnel angulation (TA) was measured within the frontal and transverse planes, expressed in degrees relative to the ODA.

Joint infringement following hypothetical screw placement was defined as having a minimum distance of less than 1.75 mm between the center of the drilled transcondylar tunnel and the nearest articular surface for the 3.5 mm screw, and less than 2.25 mm for the 4.5 mm screw.

Transcondylar screws without joint infringement were categorized as optimally placed if they were within 5° of planned craniocaudal and proximodistal angulation; they were categorized as suboptimal if they were within 5–10° of planned angulation; and they were categorized as unacceptable if they were greater than 10° from planned angulation.

2.5. Data analysis

A sample size calculation was performed using the mean and standard deviation of maximum angular tunnel deviation data published in three studies evaluating placement using 3D‐PDG and UAD. ⁸ , ⁹ For an alpha of 0.05 and a power of 0.80, a total of 4 groups of 14 limbs each, or 56 limbs, would be required for the study. We obtained an additional 40 limbs for our research to increase statistical power and to account for any potential attrition due to unsuitability, osteoarthritis, or technical errors during the experiment.

2.6. Statistical methods

Mean and standard deviation were used to describe the angular screw deviation and the deviation of the entry and exit points from the proposed locations. The direction of deviation was described using the mode.

Surgical techniques (FAD, FHD, 3D‐PDG, UAD), the surgeon, limb laterality, and their interactions were included as fixed effects in a series of linear mixed models for each dependent variable (humeral condylar entry point translation, exit point translation, craniocaudal angular deviation, proximodistal angular deviation, and surgical time). Dog was included as a random effect. Limb order, humeral condylar diameter, and MinTAD were included as unstructured covariates. The restricted maximum likelihood method was used to evaluate model fit. Statistically nonsignificant fixed effects were removed from the final models. Residuals were assessed through a histogram and were found to be normally distributed.

The odds of experiencing joint infringement after tunnel drilling for each surgical technique were compared using the χ² test. The odds of having joint infringement with each surgical technique following the hypothetical placement of 3.5 and 4.5 mm cortical screws (actual screw placement was not performed in this study) were also calculated and compared using the χ² test. Joint infringement following hypothetical screw placement was defined as having a minimum distance of less than 1.75 mm between the center of the drilled transcondylar tunnel and the nearest articular surface for the 3.5 mm screw, and less than 2.25 mm for the 4.5 mm screw.

Transcondylar screws without joint infringement were categorized as described under section 2.4. A Kruskal–Wallis test was utilized to compare the likelihood of these surgical outcomes among groups. All analyses were conducted using IBM SPSS Statistics version 29 (IBM, Armonk, New York), with significance set at α = .05.

3. RESULTS

Two limbs were excluded from the analysis due to the inadvertent loss of study data, leaving 94 limbs for the final data analysis. In bivariable analyses, MinTAD was associated with surgical technique (p < .01) (Table 1), so was tested as a confounder in each multivariable model. Humeral condylar diameter, limb order, and laterality were not associated with surgical technique.

TABLE 1.

Mean (SD) of surgical time (s), angular and positional error (mm), MinTAD (mm), and humeral condylar diameter (mm). Median (IQR) of limb order by surgical technique.

	FAD (n = 24)	FHD (n = 25)	3D PDG (n = 23)	UAD (n = 22)	p a
Surgical time	160.0 (53.0)	43.6 (14.5)	81.3 (27.4)	73.7 (27.5)	<.01
Humeral condylar entry point translation	2.8 (0.8)	2.9 (1.1)	3.3 (1.1)	5.3 (4.0)	<.01
Humeral condylar exit point translation	4.0 (2.0)	4.1 (2.1)	6.3 (3.9)	4.8 (2.5)	.16
Craniocaudal angular deviation	7.2 (4.9)	13.2 (33.1)	6.2 (5.2)	6.1 (5.8)	.73
Proximodistal angular deviation	3.4 (2.9)	10.1 (34.9)	5.3 (4.4)	3.2 (2.4)	.24
MinTAD	4.8 (1.5)	4.7 (1.4)	3.0 (1.7)	3.1 (2.0)	<.01
Humeral condylar diameter	13.6 (2.0)	13.7 (1.5)	13.6 (1.7)	14.0 (1.5)	.87
Limb order	6.5 (6)	7.0 (6.0)	6.0 (6.0)	6.0 (6.0)	.9

^a Statistical significance was assessed using the Kruskal–Wallis test when data were non‐normally distributed; otherwise, one‐way ANOVA was applied.

3.1. Surgical time

There was a difference in surgical time among the four techniques used in bivariable (p < .01) (Table 1) and multivariable (p < .01) (Table 2) analyses. Freehand drilling took the least surgical time: 43.6 ± 14.6 s (mean ± standard deviation)—less than all other methods (p < .05). The next fastest was the UAD (73.7 ± 27.5 s) and then the 3D‐PDG (81.3 ± 27.4 s). Surgical time was longest with FAD (160.0 ± 53.0 s). In the multivariable model adjusting for multiple comparisons, all methods were different from each other (p < .01) except for 3D‐PDG (p = .83) and UAD (p = .76). MinTAD was not significantly associated with surgical time, so it was removed from the model.

TABLE 2.

Linear mixed models predicting surgical time and positional and angular error by surgical technique.

	Surgical time a
	B	SE	p value
Surgical technique			<.01
Fluoroscopy‐assisted drilling	reference category	‐	‐
Freehand drilling	−116.6	9.4	<.01
Three‐dimensional printed guide	−78.9	9.6	<.01
Universal aiming device	−86.4	9.6	<.01
Intercept	160	7.1	<.01
	Humeral condylar entry point translation b
Surgical technique			<.01
Fluoroscopy‐assisted drilling	reference category	‐	‐
Freehand drilling	−0.01	0.05	.83
Three‐dimensional printed guide	−0.02	0.05	.76
Universal aiming device	0.17	0.05	<.01
MinTAD	−0.04	0.01	<.01
Intercept	0.61	0.07	<.01
	Humeral condylar exit point translation c
Surgical technique			<.01
Fluoroscopy‐assisted drilling	reference category	‐	‐
Freehand drilling	−0.6	1.53	.7
Three‐dimensional printed guide	3.39	1.34	.01
Universal aiming device	−0.64	1.27	.62
MinTAD	−1	0.22	<.01
Surgical technique × MinTAD			<.01
Fluoroscopy‐assisted drilling	reference category	‐	‐
Freehand drilling	0.15	0.3	.61
Three‐dimensional printed guide	−0.84	0.3	.01
Universal aiming device	−0.01	0.28	.97
Intercept	8.53	1.14	<.01
	Craniocaudal angular deviation
Surgical technique			.42
Fluoroscopy‐assisted drilling	reference category	‐	‐
Freehand drilling	6.03	4.94	.22
Three‐dimensional printed guide	−1.01	5.05	.84
Universal aiming device	−1.07	5.1	.83
Intercept	7.21	3.53	.04
	Proximodistal angular deviation
Surgical technique			.5
Fluoroscopy‐assisted drilling	reference category	‐	‐
Freehand drilling	6.72	5.11	.19
Three‐dimensional printed guide	1.97	5.22	.71
Universal aiming device	−0.14	5.28	.98
Intercept	3.38	3.65	.35

^a Random intercept variance for surgeon (SE): 9.78 (32.88).

^b Random intercept variance for dog (SE): 0.01 (0.01); values were log₁₀‐transformed to meet normality assumptions.

^c Random intercept variance for dog (SE): 1.03 (0.49).

3.2. Entry and exit point translation

Surgical technique impacted entry point translation in both bivariable (Table 1) and multivariable (Table 2) analyses. Fluoroscopy‐assisted drilling had the least amount of entry point translation (2.8 ± 0.8 mm), similar to that of FHD (2.9 ± 1.1 mm) and 3D‐PDG (3.3 ± 1.1 mm). After adjusting for multiple comparisons in the multivariable model, FAD, FHD, and 3D‐PDG had lower levels of entry point translation when compared with UAD (5.3 ± 4.0 mm). The most common entry point translation in each group was craniodistal (Table 3). MinTAD was associated with entry point translation (p < .01).

TABLE 3.

N (%) of joint infringements, direction of entry and exit point deviations, laterality, and screw placement outcome by surgical technique.

	FAD (n = 24)	FHD (n = 25)	3D PDG (n = 23)	UAD (n = 22)	p a
Joint infringement prior to placement	1 (4.2)	0 (0.0)	5 (21.7)	6 (27.3)	<.01
Joint infringement 3.5 mm	1 (4.2)	0 (0.0)	7 (30.4)	6 (27.3)	<.01
Joint infringement 4.5 mm	1 (4.2)	0 (0.0)	8 (53.3)	6 (27.3)	<.01
Direction of entry point deviation
Caudal	0 (0.0)	0 (0.0)	1 (4.4)	0 (0.0)	<.01
Caudoproximal	1 (4.2)	0 (0.0)	4 (17.4)	1 (4.6)	‐
Caudodistal	1 (4.2)	0 (0.0)	6 (26.1)	1 (4.6)	‐
Cranial	2 (8.3)	3 (12.0)	0 (0.0)	0 (0.0)	‐
Cranioproximal	7 (29.2)	6 (24.0)	0 (0.0)	3 (13.6)	‐
Craniodistal	11 (45.8)	15 (60.0)	10 (43.5)	17 (77.3)	‐
Distal	2 (8.3)	0 (0.0)	2 (8.7)	0 (0.0)	‐
Proximal	0 (0.0)	1 (4.0)	0 (0.0)	0 (0.0)	‐
Direction of exit point deviation
Caudal	2 (8.3)	2 (8.0)	1 (4.4)	1 (4.6)	.13
Caudoproximal	3 (12.5)	6 (24.0)	5 (21.7)	1 (4.6)	‐
Caudodistal	8 (33.3)	5 (20.0)	5 (21.7)	5 (22.7)	‐
Cranial	0 (0.0)	2 (8.0)	0 (0.0)	1 (4.6)	‐
Cranioproximal	3 (12.5)	6 (24.0)	1 (4.4)	2 (9.1)	‐
Craniodistal	7 (29.2)	2 (8.0)	11 (47.8)	10 (45.5)	‐
Distal	1 (4.2)	2 (8.0)	0 (0.0)	2 (9.1)	‐
Proximal	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	‐
Laterality
Right	14 (58.3)	13 (52.0)	8 (34.8)	11 (50.0)	.42
Left	10 (41.7)	12 (48.0)	15 (65.2)	11 (50.0)	‐
Screw placement outcome b
Optimal	6 (26.1)	11 (44.0)	7 (46.7)	9 (56.3)	.26
Suboptimal	11 (47.8)	5 (20.0)	6 (40.0)	4 (25.0)	‐
Unacceptable	6 (26.1)	9 (36.0)	2 (13.3)	3 (18.8)	‐

Abbreviations: 3D‐PDG, three‐dimentional printed drill guide; FAD, fluoroscopy‐assisted drilling; FHD, free‐hand drilling; UAD, universal aiming device.

^a Statistical significance was assessed using the χ² test or Fisher's exact (if any expected cell counts were <5).

^b Calculated among limbs without joint infringement (n = 79).

Surgical technique also impacted exit point translation. In bivariable analyses (Table 1), FAD had the lowest exit point translation (4.0 ± 2.0 mm), followed by FHD (4.1 ± 2.1 mm), UAD (4.8 ± 2.5 mm), and 3D‐PDG (6.3 ± 3.9 mm). However, the multivariable models indicated that exit point translation by surgical technique depended on MinTAD, as indicated by the significant interaction term in the multivariable model (Table 3). Pairwise comparisons of slopes indicated that only 3D‐PDG differed from the others (p < .01). Specifically, the effect of MinTAD on exit translation was substantially steeper for 3D‐PDG than all other techniques. When the tunnel is positioned further from the joint surface (larger MinTAD), exit translation improves much more dramatically with 3D‐PDG, whereas the other methods show little to no change. Groups differed in the location of their exit point. The most common FAD exit point deviation was caudodistal, whereas FHD was bimodal, with exit points most commonly either cranioproximal or caudoproximal. Three‐dimensional printed guides and UAD most commonly had exit point deviations that were craniodistal (Table 3).

3.3. Angular deviation

Surgical technique did not have an impact on craniocaudal angular deviation or proximodistal angular deviation of tunnel placement (Tables 1 and 2). Standard deviations were notably higher for craniocaudal angular deviation and proximodistal angular deviation in the FHD group than for the other groups (Table 1).

3.4. Joint infringement, drilled tunnel and hypothetical screw placement

Joint infringement before hypothetical screw placement occurred in 12 of the 94 limbs (12.8%), with a statistical difference in joint infringement observed among groups (p < .01) (Table 2). The lowest level of joint infringement was found in the FHD group (0/25, 0%), followed by the FAD group (1/24, 4.2%), the 3D‐PDG group (5/23, 21.7%), and the UAD group (6/22, 27.3%).

Hypothetical placement of a 3.5 mm screw would lead to joint infringement in two additional limbs (14/94, 14.9%) with a statistical difference among groups (p < .01) (Table 3). Hypothetical 4.5 mm screw placement would have resulted in joint infringement in one additional limb (15/94, 16.0%) also with a statistical difference in joint infringement among groups (p < .01) (Table 3).

Among drilled tunnels without joint infringement, the surgical procedure would not have affected the likelihood of achieving optimal, suboptimal, or unacceptable hypothetical screw placement (Table 3). About half of all joint infringements (8/15, 53.3%) would have resulted from unacceptable (>10°) craniocaudal and/or proximodistal angulation in comparison with the ODA.

4. DISCUSSION

This study compared the accuracy, safety, and efficiency of four TCD methods—FHD, 3D‐PDG, UAD, and FAD—in canine cadaver limbs. The first hypothesis, which stated that FHD would have inferior accuracy and safety compared to the other three techniques, was rejected. Instead, we found that FHD was associated with the lowest incidence of joint space infringement (0%) among the tested methods, lower than the high incidence of joint infringement observed with UAD (p < .01) (27.3%). These findings contrast with most of the available literature comparing FHD with guided drilling methods. In a study by Faux et al. using a synthetic bone model, UAD and FAD were significantly more accurate than FHD. ¹⁰ In their study, UAD was superior to FAD, a finding that also contrasted with ours. However, the authors acknowledged that in clinical cases, UAD would be more prone to slippage, making FAD a potentially more accurate choice. Driver et al. compared the accuracy of FAD to FHD in cadaveric canine vertebrae, concluding that optimal drill hole placement was significantly more likely with fluoroscopy. ¹¹ This aligns with Brioschi et al., who concluded that a surgeon's ability to drill accurately by hand, defined in their paper as within 4° of perpendicular to the drilling surface, was limited and further impaired by drilling angles less than 60° from perpendicular. ¹² Freehand drilling was associated with greater craniocaudal and proximodistal deviations in our study compared to the other three techniques. The choice of technique did not affect tunnel angular deviation.

We chose to group limb outcomes separately based on whether DTs, regardless of their angulation, encroached on the joint space. This approach was meant to enhance the clinical relevance of our findings, as joint infringement is considered an indication for immediate surgical revision. ⁴ , ⁷ None of the surgical techniques proved superior to the others for optimal DT placement and orientation.

Surgeons must use their judgment to decide whether suboptimal TCS placement that does not infringe the joint space requires immediate revision. Previous studies have indicated that TCSs applied at angles greater than 10° relative to the epicondylar line were associated with a 4.8 times greater odds of persistent intracondylar fracture gap on postoperative craniocaudal radiographs, a risk factor for implant failure and osteoarthritis. ¹³ In this study, we used DTs as surrogates for TCSs to minimize the effect of metal implant artifacts on our ability to accurately measure DTs on postexperiment CT images. Considering recommendations to apply a TCS measuring 30% to 50% of the humeral condyle diameter, surgical revision of a suboptimally applied screw that is less than 10° relative to the ODA requires a challenging decision from the surgeon. ¹⁴ The availability of condylar bone stock, the chances of creating partially overlapping bone tunnels that would reduce screw thread purchase, and the effect of prolonged surgical procedures on postsurgical complications must be considered carefully in the clinical setting. ¹³

Our second hypothesis was that DTs created using the 3D‐PDG, FAD, or UAD would not invade the joint space or supratrochlear foramen. To our surprise, DTs created with 3D‐PDG and UAD presented a relatively high rate of joint infringement (22% and 26%, respectively), leading us to reject this hypothesis. The rate of joint infringement with either device in our study was higher than previously reported in the literature. Easter et al. reported a 0% rate of joint infringement using a 3D‐printed guide of a different design, whereas Neal et al. reported a 10% rate of joint infringement with a similar UAD. ⁸ , ⁹ The reasons for the high rate of joint infringement with the UAD cannot be inferred directly from the study data. Neal et al. reported an eightfold increase in the risk of joint involvement with the UAD compared to fluoroscopy. ⁹ In our research, UAD was associated with approximately 6 times the likelihood (26%) of joint infringement compared with FAD (4%). Although our comparative findings between UAD and FAD agree with Neal et al., the lack of a surgical assistant and possible surgeon bias should be noted. The UAD used in this study was made from stainless steel, making it a sturdy device with minimal tendency to “flex” during application. However, it was also relatively heavy and challenging to keep in place without assistance, which may have led to slight movement of the device at the onset of drilling. Given the lack of perpendicularity between the drill bit's orientation and the condyle's lateral aspect, even minor motions could cause loss of intended alignment, resulting in suboptimal or unacceptable drilling trajectories.

Finally, both surgeons reported difficulty stabilizing the UAD tip against the medial epicondyle, an issue also reported by Kershaw et al. in a recent publication. ¹⁵ The authors recognize that drilling a small indentation on the medial epicondyle at the planned tunnel exit site could have improved the accuracy of the UAD method. Our findings concerning the UAD should therefore be interpreted with caution, considering previous studies indicating the usefulness and accuracy of this device. ⁷ , ⁹ , ¹⁶

Our third hypothesis, that all DTs created with the 3D‐PDG would stay within 5° of the ODA, was also rejected. Only 39% met this criterion, and nearly a third were unacceptable. Additionally, 30% of the limbs using the 3D‐PDG technique would have theoretically shown joint infringement with a 3.5 mm screw, which is much higher than the rates with FAD (4%) and FHD (0%). These findings contrast with those of Easter et al., who reported a mean maximum screw angulation of 5.2° (2.1°) using a medially applied 3D‐printed guide with features different from ours. ⁸ The relatively low accuracy in our study with the 3D‐PDG may be due to factors related to the surgeons and the device.

On the surgeons' side, unfamiliarity with 3D‐PDG, insufficient training before the experiment, and a lack of feedback may have negatively affected the device's drilling accuracy. Surgical skills are acquired through deliberate practice, which is defined as repetitive practice of a focused skill with specific feedback, resulting in improved performance of that skill. ¹⁷ , ¹⁸ Structured feedback, in particular, is crucial for honing surgical skills. ¹⁹ Although both surgeons received similar brief instructions on how to use the 3D‐PDG at the surgical site and had access to a 3D‐printed bone model featuring the guide, more comprehensive instructions and opportunities to ask questions and practice the skill could have improved the results. Moreover, while drilling became progressively faster when the surgeons used the 3D‐PDG during the experiment, their overall performance did not improve.

Regarding the 3D‐PDG device, the authors identify several potential shortcomings and highlight opportunities for further improvement. The polymer construction and minimalistic design of the 3D‐PDG may have affected the results negatively. Although polymer drilling guides are sufficiently rigid for most procedures, slight flex and movement can still occur during drilling, even with a stabilizing K‐wire. This can be further supported by the finding that the greater the distance from the hard subchondral bone plate (larger MinTAD), the lower the exit translation with the 3D‐PDG, indicating that this device is particularly vulnerable to drill bit deflection.

In a recent study by Kershaw et al., the authors reported good accuracy using a 3D‐printed guide applied to the medial aspect of the humeral condyle. ¹⁵ Their guide construction and application differed from ours in several ways, and resembled the design reported by Easter et al. ⁸ First, their guides were printed as a simple, “solid” construct, whereas ours followed a skeletonized design. The authors acknowledge that a denser design is likely less prone to deflection during application and drilling and would consider this an important modification of our guide for future research. Second, our guide incorporated bone‐contact “pods” to minimize the need for soft‐tissue dissection, whereas Kershaw and Easter's guide utilized a larger footprint. While the goal of minimizing soft‐tissue trauma remains unquestionable, guide‐to‐bone contact has been recognized as critical to the accuracy and stability of 3D‐printed guides in several studies. ²⁰ , ²¹ , ²² Hall et al. also observed that the presence of interposing soft tissues in the guide footprint region could have influenced guide performance, a conclusion shared by Violini et al. ²⁰ , ²¹ In a future guide design, the authors would remove the contact pods in favor of a broad contact area between the guide, the epicondyle, and the epicondylar crest. The addition of a second K‐wire channel cranial to the drilling channel could further improve guide stability by allowing divergent wires to be placed. More comprehensive soft tissue elevation and a larger surgical approach could also have positively contributed to the performance of the 3D‐guide in our study. Last, and perhaps most important, the choice to design and apply a 3D‐printed guide laterally rather than medially likely influenced guide performance to a large extent. The medial aspect of the humeral condyle is relatively broad and displays a prominent epicondyle but the lateral aspect is more rounded and sharply tapers proximally into the epicondylar crest. These features make the medial aspect of the humeral condyle a superior surface for 3D‐guide anchoring, despite the reportedly narrower safe‐drilling corridor. ⁴ The authors recommend taking these factors into consideration for future studies.

The fourth hypothesis was supported: FHD was the fastest method. No additional instrumentation was required to drill a tunnel using this technique and both surgeons had extensive experience with this drilling modality, which likely contributed to efficiency. These results may reflect expertise and preference but they may be clinically relevant. Time efficiency in the operating room is desirable, as prolonged surgical time has been associated with a higher rate of surgical site infections. ¹³ Equipment and supplies, such as fluoroscopy and surgical guides, may not be available in all circumstances. Condylar fractures are frequently challenging and associated with prolonged general anesthetic events, which may also be associated with increased patient morbidity. ²³

We acknowledge several limitations of this study beyond the 3D‐guide modifications previously discussed. Estimating the risk of joint space infringement based on DT angulation and MinTAD‐to‐screw radius, rather than observing screw positions directly through postexperiment imaging, may impact the validity of the results. Although the same methodology has been used by Kershaw et al., ¹⁵ factors such as manual overdrilling with 3.2 mm drill bit for a 4.5 mm screw and variations in angulation during screw insertion can cause the final screw path to deviate from the intended trajectory in a clinical case. ⁸ Using 2.0 mm drill bits for the 3D‐PDG and 2.5 mm drill bits for the other three techniques, although necessary for logistical reasons, may have also introduced additional variability by increasing deflection with the smaller drill bit.

The custom Python code developed to extract imaging data from both MPR and 3D renderings was tested for repeatability during the pilot phase. However, it has not been validated against any commercially available software because, to the best of the authors' knowledge, such software does not currently exist. The anatomical landmarks for establishing the epicondylar reference line, from which the ODA was based, were identified by a single observer (the primary author) on a virtual three‐dimensional reconstruction. Although the landmarks were identified strictly following the method described by Barnes et al., the observer's familiarity with anatomical landmarks and individual variations could have affected the accuracy of the results. ⁴

This study was conducted using limbs harvested from medium to large‐breed dogs, primarily Staffordshire terrier mixed breeds, which are not representative of the population typically affected by condylar fractures and humeral intracondylar fissures. This choice was made due to the challenges associated with obtaining limbs from commonly affected breeds for a study with sufficient statistical power. The visual and tactile recognition of anatomical landmarks may be less accurate in breeds such as French bulldogs, which are more commonly affected. Consequently, the performance of the methods tested in this study may vary in these breeds. ² , ⁹ , ²⁴

All of the TCD methods were evaluated using nonfractured limbs. The performance of these methods in fractured humeri can pose substantial challenges even after fracture reduction due to the presence of reduction instruments and implants, which could affect the performance of the drilling devices tested in this experiment. As previously suggested, the lack of an assistant during the procedures might have negatively impacted the performance of aiming devices like 3D‐PDG and UAD, favoring the less cumbersome FHD method for solo surgeons. All limbs were tested in a hanging limb setup to standardize the experiment; however, this differs from the way in which the enrolled surgeons typically perform transcondylar drilling. Both surgeons favor a free limb position, with the drill oriented vertically rather than horizontally, which raises questions about the effect of limb position on the performance of each drilling method. Finally, each surgeon tested all limbs in a single‐day session. The fatigue associated with this approach may have negatively impacted the surgeons' performance compared with what might be expected in a clinical setting.

In conclusion, FHD was accurate, safe, and not inferior to guided drilling techniques when performed by experienced surgeons. Fluoroscopy‐assisted drilling was accurate, but it required specialized equipment that is not always accessible. The 3D‐PDG used in this study, applied laterally to the humeral condyle, was insufficiently stable and this approach therefore requires design modifications, extensive training, feedback, and practice to improve its performance. The availability of an assistant and limb position may influence the effectiveness of UAD.

FUNDING INFORMATION

This study was supported by the Veterinary Orthopedic Society Wade O. Brinker Research Award and Columbia Veterinary Emergency Trauma and Specialty.

CONFLICT OF INTEREST

The authors declare no conflict of interest.