Intubation Biomechanics: Clinical Implications of Computational Modeling of Intervertebral Motion and Spinal Cord Strain during Tracheal Intubation in an Intact Cervical Spine

Finite element modeling is a computational simulation method used to predict the behaviors of complex three-dimensional structures. The basis of finite element modeling lies in dividing a complex structure into many smaller, simpler structures called elements so that the overall structural response to loading may be mathematically calculated. The response of each element is expressed in terms of a finite number of degrees of freedom at a set of points called nodes that connect each element to other adjacent elements. To model biologic structures, finite element models require accurate three-dimensional representations of the geometry (anatomy) of all structurally distinct components such as vertebrae, intervertebral discs, ligaments, etc. These models also require knowledge of the material properties of each component (e.g., elasticity) under all conditions under which the model will be tested.¹⁵  For more than 25 yr, finite element models of the human cervical spine have been used to predict spine motions that occur under routine or hazardous (e.g., high force) conditions. In addition, these models allow for characterization of processes and dynamics occurring inside the substance of the spine or cord that are impossible to directly measure, such as mechanical stress and strain.¹⁵ 

The model used in this study is of the complete human cervical spine (from the occiput to C7) and the cervical spinal cord.¹⁶  It consists of 196,984 elements, 237,635 nodes, and has 671,997 degrees of freedom. In a previous validation study, this model closely predicted intervertebral motions (occiput–C1 through C4–C5) measured during orotracheal intubation in living patients.¹⁶  It did so with intubations accomplished with both direct laryngoscopy (Macintosh) and videolaryngoscopy.¹⁶  The results of the current study provide additional evidence of the validity of this model (see Results). Specifically, under routine (Macintosh) intubation conditions, this model predicted 1 to 2 degrees of flexion at C5–C6 and C6–C7. Flexion below C5 during Macintosh intubations has been previously reported by Turkstra et al.¹⁷  Also, under routine conditions, this model predicted the occiput translates 1.4 mm posterior to C1, with progressively less posterior translation at C1–C2 and C2–C3, changing to anterior translation at C3–C4, with 1.0 mm or less of anterior translation in the remaining caudal segments. These predicted translations are compatible with translation values calculated from the clinical intubation data reported by Mentzelopoulos et al.¹⁸  (occiput–C1 = 1.0 mm, C1–C2 = 0.6 mm, C2–C3 = 0.5 mm, C3–C4 = 0.4 mm, and C4–C5 = 0.3 mm). Therefore, this model predicts cervical spine motions measured during routine intubations reported in three different clinical studies. In addition, in our previous validation study, this model closely predicted spine motions in cadavers with type II odontoid fractures during intubations performed with both conventional and videolaryngoscopy.¹⁶ 

For additional information about this model, including development, anatomy, material properties, convergence studies, software, execution time, and code availability, see Supplemental Digital Content 1 (Finite Element Model Development, Material Properties, Calculations, and Limitations; https://links.lww.com/ALN/C740). We followed the applicable Enhancing the QUAlity and Transparency Of health Research (EQUATOR) guidelines for simulation.¹⁹ 

Although laryngoscope force is not uniformly distributed along the blade, it can be represented biomechanically as being applied at a single point.²⁰  Thus, laryngoscope blade contact force was modeled as being applied at a single point (location) having both magnitude (N) and direction (degrees), i.e., as a force vector.¹⁶  Intubation force was simulated as a force vector applied to the anterior surface of a selected cervical spine vertebral body. In a previous finite element modeling study,¹⁶  using radiographic images and simultaneous laryngoscope force distribution measurements from a previous clinical study,²⁰  the mean applied force location, magnitude, and direction for a routine intubation with a Macintosh blade were estimated to be the midpoint of the C3 vertebral body, 48.8 N, and 70 degrees from the body’s coronal plane, respectively. This specific combination of laryngoscope force location, magnitude, and direction are hereafter referred to as routine conditions, denoting force conditions occurring during a routine direct (Macintosh) laryngoscopy and intubation.

As shown in figure 1, laryngoscope force application location, magnitude, and direction were each varied over a range of values. Four laryngoscope force application locations were studied: the superior half of the anterior surface of the C2 vertebral body (C2_SUP), the inferior half of the anterior surface of the C2 body (C2_INF), the midpoint of the anterior surface of the C3 body (C3, routine location); and the midpoint of the anterior surface of the C4 body (C4). The C2_INF force application location corresponds to that observed with the Airtraq videolaryngoscope.^16,20  Four intubation force magnitudes were studied: 24.4 N, 48.8 N (routine magnitude), 73.2 N, and 97.6 N, corresponding to 50, 100, 150, and 200% of the routine force magnitude. In a study of patients who were predicted to be easy to intubate, Macintosh intubation force magnitude was 48.8 ± 15.8 N (mean ± SD), with the greatest individual patient force magnitude equal to 70.9 N.²⁰  In a different intubation study, with the utilization of manual in-line stabilization, pressures applied by a Macintosh blade were two-fold greater than without the use of manual in-line stabilization.⁴  In a previous cadaver intubation study, Macintosh intubation force magnitude was 47.1 ± 20.5 N, with the greatest individual cadaver Macintosh force magnitude equaling 93.6 N.²¹  Thus, 97.6 N (twice the routine value) appears to approximate the maximum amount of force that anesthesiologists can apply with a conventional direct (Macintosh) laryngoscope. Finally, three laryngoscope force directions were studied: 50 degrees, 70 degrees (routine direction), and 90 degrees from the body’s coronal plane. The 90-degree force direction corresponds to that observed with the Airtraq videolaryngoscope.^16,20  Thus, in total, 48 simulations were conducted, consisting of all combinations of laryngoscope force application location (n = 4), magnitude (n = 4), and direction (n = 3). The resultant spine motion and cord strain values represent quasistatic values corresponding to the maximum values occurring during tracheal intubation.

In all simulations, the occiput was allowed to rotate and translate cranially and caudally, whereas the caudal surface of the C7 vertebral body was kinematically constrained to restrict all motion (see Discussion, Limitations). In all simulations, the cervical spine was considered to start at a neutral position, with the degrees of intervertebral rotation (flexion and extension) and anterior–posterior translation (subluxation) defined as zero. Segmental intervertebral rotation and translation at each of seven intervertebral segments were calculated. Rotation was measured as the difference in rotation between reference nodes kinematically attached to adjacent vertebrae and was independent of translation. Intervertebral extension was represented by positive values and flexion by negative values. Translation was measured as the difference in anterior–posterior displacement of the centers of rotation of the two adjacent vertebrae. This method decreases the effect of intervertebral rotation on measures of translation. In a given intervertebral segment, translation of the cranial vertebral body posterior to the caudal vertebral body was defined as being posterior subluxation and is represented with positive values. Conversely, translation of the cranial vertebral body anterior to the caudal vertebral body of a segment was defined as anterior and is represented with negative values.

Because strain quantitates tissue deformation (e.g., change in length or width) as a ratio of the initial/final value of the parameter, strain is dimensionless. We utilized the logarithmic strain method to calculate strain, strain = ln(L/L₀), where L is the final length, and L₀ is the initial length. Studies show that accounting for strain in multiple simultaneous planes has a larger observed correlation with tissue injury than strain in any single plane.^13,22  Thus, we used two strain measures that each incorporate the overall three-dimensional strain field: (1) maximum principal strain (tensile strain, analogous to stretch, represented by positive values) and (2) minimum principal strain (analogous to compression, represented by negative values; see Supplemental Digital Content 1 [Finite Element Model Development, Material Properties, Calculations, and Limitations; https://links.lww.com/ALN/C740]). In animal spinal cord injury models, these two strains had the largest observed correlations with tissue injury.^13,14 

Based on a recent experimental study of cervical cord injury in nonhuman primates,¹⁴  we defined two strain values as thresholds for potential cord injury. The maximum principal strains resulting in a 50% cord injury measured histologically 14 to 17 weeks after insult were 0.26 to 0.31 for gray and white matter, respectively.¹⁴  In the same study, the 50% injury values for minimum principal strains were –0.38 to –0.42 for gray and white matter, respectively. Because 50% injury strain values are too great to use as clinical safety thresholds, we defined 50% of these values as potentially injurious, specifically 0.14 for maximum principal strain (stretch) and –0.20 for minimum principal strain (compression; see Discussion).

Because model cervical spine anatomy was derived from a single adult human subject²³  and mean material property data inputs were utilized to define the model, the model does not simulate the inherent variation across the human population. In addition, the model does not include random error from experimental measurements. The absence of these variations produces deterministic (i.e., single-valued) motion and strain values. Thus, model predictions for motion and strain are functionally equivalent to population mean values (see Discussion, Limitations). All values for cervical spine motions were rounded to a single decimal before analysis.

Figure 2 (A and B) shows the complete data set of predicted intervertebral rotations (extension and flexion) and anterior–posterior translations (subluxation) at the 7 cervical intervertebral segments, each under all (48) modeled tracheal intubation force conditions. Table 1 summarizes model maximum values for intervertebral rotation and translation and the specific force conditions that resulted in these motions.^24–30  Among the 14 combinations of motion (n = 2; rotation and translation) and intervertebral segment (n = 7), all maximum values occurred with the maximum force magnitude (97.6 N). Other intubation force characteristics (location, direction) causing maximum motions differed among motions and segments. To address our first question, table 1 also shows maximum physiologic values for intervertebral motion reported among seven clinical voluntary range of motion studies.^24–30  Maximum values for intervertebral rotation and translation predicted by the model did not meaningfully exceed physiologically normal maximum values measured during voluntary cervical flexion and extension (for additional details and discussion see Supplemental Digital Content 2 [Clinical Studies of Voluntary Cervical Intervertebral Motion; https://links.lww.com/ALN/C741]).

Table 1 also summarizes model values for rotation and translation under routine intubation conditions. Among the 7 segments, the differences between maximum and routine values for rotation and translation were 3.5 degrees or less and 1.1 mm or less, respectively. Thus, the predicted differences between maximum intervertebral motions during intubation and those occurring during a routine intubation are quantitatively small.

Although our second aim pertained only to maximum (peak) values of strain present in any portion of the cord, the model predicted cord strain to be spatially heterogenous, with peak strains present in different regions of the cord depending on the location of applied force. Figure 3 shows the distribution of spinal cord strains at each of the four force application locations. Maximum principal (tensile) strain (stretch) was very low in most of the cord. However, there were foci of increased maximum principal strain in the anterior cord. In addition, there were smaller foci of greater (peak) tensile strain in the posterior cord at C1–C2 that were present at mid-C3 when force was applied at C4. Similarly, minimum principal strain (compression) was very low in most of the cord. A focus of increased (peak) compressive strain was present in the posterior cord at C1–C2 with the two most cephalad force application locations (C2_SUP and C2_INF) and was present at mid-C3 with the two most caudal force locations (C3 and C4).

Specifically addressing our second aim, some tracheal intubation conditions did result in potentially injurious cord strains. Figure 4 shows peak maximum and minimum principal strain under all (48) intubation force conditions. Peak maximum principal strain (stretch) did not exceed the potential injury threshold (0.14) in any modeled intubation force condition (0 of 48). The peak values for maximum principal strain were insensitive to force magnitude. In contrast, peak minimum principal strain (compression) exceeded the potential injury threshold (–0.20) in 3 of 48 conditions, all with force applied at C4 with the greatest force magnitude (97.6 N). Peak values for minimum principal strain were sensitive to force magnitude; compressive strains increased markedly when force magnitude exceeded 24.4 N. Although peak compressive strains did not exceed the potential injury threshold when force was applied at the routine location (C3), compressive strains were close to the potential injury threshold with force magnitudes of 48.8 N or greater (see Discussion).

In the presence of an intact (stable) spine, are there tracheal intubation conditions in which cervical intervertebral motions exceed physiologically normal maximum values? The model predicted that the answer is no. Model predictions for intervertebral rotation (flexion/extension) and translation (subluxation) did not exceed the range of voluntary motion reported in the clinical studies. This was so even with the maximum modeled force magnitude, 97.6 N, which approximates the greatest amount of force anesthesiologists can apply with a conventional direct laryngoscope. In our models, we included two intubation force locations that may, in fact, not be clinically achievable: one very cephalad (C2_SUP) and one very caudad (C4). However, this only serves to reinforce the conclusions of this study. We modeled conditions that might truly be one in a million, and still it was practically impossible for tracheal intubation to cause an intact cervical spine to move beyond the maximum motions that occur voluntarily. This is the expected behavior of a stable cervical spine.

The second question was whether in the presence of an intact cervical spine there are tracheal intubation conditions in which potentially injurious cervical cord strains can occur? Importantly, for this second question, we obtained a different answer. The model predicted that the answer is yes, conditionally. Notably, under force conditions approximating a routine intubation using a Macintosh blade, peak strains did not exceed estimated potential cord injury thresholds for maximum and minimum principal strains. This is an expected result because if injurious cord strains occurred during routine direct laryngoscopy and intubation, intubation-related cervical cord injury would be commonplace, which it is not. In fact, even when intubation force magnitude was twice the routine value (97.6 N instead of 48.8 N), when force was applied at the routine (C3) location, compressive strain did not exceed the potential injury threshold. Again, this is consistent with clinical experience, because even with a difficult intubation, cord injury is rare. However, when maximum force was applied in a location that was more caudal than is routine (i.e., force applied at C4), predicted compressive cord strains exceeded a potentially injurious value. Admittedly, it is difficult to imagine how, with any current laryngoscope, it would be helpful or even possible to apply such high force below the level of the glottis. Thus, in patients who have an intact cervical spine, it might appear to be practically impossible for the cervical cord to experience injurious strain during tracheal intubation.

There is, however, one critically important caveat. The caveat is that model predictions of whether or not injurious cord strains occur during tracheal intubation depend entirely on the levels of cord strain that cause injury. Strain values that cause cord injury in patients are not currently known. The potential strain injury thresholds used in this study are estimates (see Discussion, Limitations, and Supplemental Content 1 [Finite Element Model Development, Material Properties, Calculations, and Limitations; https://links.lww.com/ALN/C740]). Logically, if a patient’s strain injury thresholds were less than our estimated injury thresholds (i.e., less than normal), strain-related cord injury could occur during routine intubation conditions. In other words, cord injury could occur even with normal intubation force and in the absence of pathologic cervical spine motion. In fact, there are several observations that suggest cervical cord injury does, in fact, occur in patients by this mechanism. First, in animal models, acute electrophysiologic responses (e.g., evoked potentials) serve as an indicator of neural sensitivity to acute cord strain.^9,31  Second, in a study of 38 patients who had chronic cervical spondylotic myelopathy, spinal cord evoked potential (N13) amplitudes decreased when patients were placed in 20 degrees of head/neck extension.³²  Decreased evoked potential amplitudes with extension were not associated with cervical spine stability but were associated with measures of preexisting cervical cord compression.³²  These observations suggest that patients who have spondylotic myelopathy may have less tolerance to acute increases in cord strain, even in the absence of instability. Third, in a closed claims study, 11 of 37 patients who experienced perioperative cervical cord injury did so while undergoing a noncervical spine procedure and with an apparently stable cervical spine.²  Most of these 11 patients had preoperatively unrecognized severe cervical spondylosis. Fourth, there are more than 20 case reports describing patients with severe cervical spondylosis and who, in the absence of a difficult intubation, suffered intraoperative cervical cord injury during noncervical spine surgery^33–35  (for additional references and discussion, see Supplemental Digital Content 3 [Case Reports of Perioperative Cervical Spinal Cord Injury in Patients with Cervical Spondylosis; https://links.lww.com/ALN/C742]). Accordingly, we hypothesize that patients who have severe cervical spondylosis have less tolerance to acute cord strain and consequently have greater potential to experience potentially injurious cord strain during an otherwise routine (normal force) intubation. Figure 4 shows that peak compressive strains increase markedly and approach the potential injury threshold when force magnitude exceeds 24.4 N, which is half the value applied during a routine intubation with a Macintosh blade. Thus, in patients who may have increased susceptibility to strain-related cord injury, we hypothesize that low-force laryngoscopy²⁰  may confer less risk of strain-related cervical cord injury.

The findings of this study suggest that the approach to preventing intubation-related cervical cord injury should be reconsidered. The model suggests that cervical cord injury from tracheal intubation is not directly related to the motion of the cervical spine but instead by the resultant spinal cord deformation, i.e., strain. This mechanism of injury would apply regardless of whether the cervical spine is intact (stable) or injured (unstable). We suggest that airway management of patients who have disease of the cervical spine or cord should no longer exclusively focus on minimizing cervical spine motion. Instead, an additional, more mechanistically oriented goal, should be to minimize cervical cord strain.

As previously reported, when compared to patients, the current model appears to underestimate intubation-related extension at C3–C4 and C4–C5 by 2 or 3 degrees.¹⁶  This may be caused by the imposed kinematic constraint of the C7 vertebral body. Although the difference between observed and predicted motion is small at these two segments, we cannot estimate how much this difference might affect model values for cord strains in the more caudal regions of the cervical cord. In a future version of the model, inclusion of the first thoracic (T1) vertebral segment will permit C7–T1 motion, and this may increase subaxial segmental motion.

Because model anatomy was derived from a single adult human subject and because mean material property data were utilized to define the model, the lack of geometric and material property variation produces deterministic (i.e., single-valued) motion and strain values. Accordingly, the current model does not simulate the inherent variation across the human population but instead represents an anthropometrical mean, i.e., an “average” patient. In the future, to account for variation in geometry and material properties across the general population, probabilistic methods will be dovetailed with the current model. With probabilistic analyses, model parameters (anatomy, material properties) are not defined by a single value but are sampled from a distribution that represents the population’s variation, and the model is solved many times to develop a distribution of output variables.³⁶ 

The values used to definite potential strain injury thresholds—50% of 50% cord injury values observed in nonhuman primates (see Materials and Methods)—are unavoidably speculative. In living humans, injurious cord strain values are not currently known. A recent magnetic resonance imaging study of nine healthy volunteers reported in vivo maximum and minimum principal strains in sustained extension (without pain or symptoms) were approximately 0.12 and –0.14, respectively.³⁷  Thus, potential strain injury thresholds used in our simulations (0.14 for maximum and –0.20 for minimum principal strains, respectively) were greater than cord strains that appear to be noninjurious in healthy asymptomatic patients. Thus, potential injury thresholds used in this study do not appear to be too low and, as a result, do not appear to greatly overestimate the potential for intubation-related cord injury.

The current model does not include an explicit representation of spinal cord gray and white matter. These tissues may^38,39  or may not⁴⁰  have different primary biomechanical properties, and the rostral–caudal alignment of axonal fibers in the spinal cord white matter provides a direction-specific mechanical response.³⁹  Gray matter may have lesser strain tolerances than white matter,^{10–12,14,41}  although the difference is small (10 to 20%). Thus, spinal cord strain fields^42,43  and regional (intracord) susceptibility to strain injury are certain to be more complex than are represented in the current version of our model. For additional discussion of model limitations see Supplemental Content 1 (Finite Element Model Development, Material Properties, Calculations, and Limitations; https://links.lww.com/ALN/C740).

In the presence of an intact cervical spine, computational modeling predicted that intervertebral motions during tracheal intubation did not exceed normal physiologic (voluntary) maximum values, even under high-force conditions. In contrast, under nonroutine high-force conditions, the model predicted that potentially injurious cervical cord strains could occur. In patients who have less than normal tolerance to acute cord strain (e.g., patients with cervical myelopathy), cord strains occurring during routine tracheal intubation conditions could approach potentially injurious values. In such patients, low-force laryngoscopy may reduce the risk of intubation-related (i.e., strain-related) cervical cord injury.

Supported by National Institutes of Health (Bethesda, Maryland) grant No. R01EB012048 (to Drs. Puttlitz, Hindman, and Santoni), the Department of Mechanical Engineering of Colorado State University (Fort Collins, Colorado), and the Department of Anesthesia of the University of Iowa Roy J. and Lucille A. Carver College of Medicine (Iowa City, Iowa).

Dr. Puttlitz received support from the National Institutes of Health (Bethesda, Maryland) and the Colorado Office of Economic Development and International Trade (Denver, Colorado). Dr. Gadomski and Dr. Puttlitz received external funding from Abbott (Irving, Texas); Acuitive (Allendale, New Jersey); Angiocrine (New York, New York); Anika Therapeutics, Inc. (Bedford, Massachusetts); Asahi Kasei Pharma (Tokyo, Japan); Auritec (Pasadena, California); Bioventus, Inc. (Durham, North Carolina); CGBio (Seongnam, South Korea); Chap Med, Inc. (Destin, Florida); Collagen Matrix, Inc. (Oakland, New Jersey); the U.S. Department of Defense (Washington, D.C.); DSM Biomedical (Heerlen, The Netherlands); Elute, Inc. (Salt Lake City, Utah); Evoke Medical, LLC (Lawrence, Kansas); Hyalex Ortho, Inc. (Lexington, Massachusetts); Ingeneron, Inc. (Houston, Texas); Intelligent Implants (Charlotte, North Carolina); Magic Implants (Jerusalem, Israel); Miach Orthopaedics (Westborough, Massachusetts); MiRus, LLC (Madison, Wisconsin); NeuroSigma (Los Angeles, California); Nushores Biosciences (Little Rock, Arkansas); Organogenesis (Canton, Massachusetts); Paragon 28 (Centennial, Colorado); Progenerative Medical (San Antonio, Texas); SI Tech (Fort Collins, Colorado); Smith & Nephew, Inc. (London, United Kingdom); Sparta Biopharma (Hamilton, New Jersey); University of Arkansas, Little Rock (Little Rock, Arkansas); University of Otago (Christchurch, New Zealand); Wright Medical (Franklin, Tennessee); and Zetagen Therapeutics, Inc. (Syracuse, New York). The other authors declare no competing interests.