Biomechanical effects of lumbar fusion surgery on adjacent segments using musculoskeletal models of the intact, degenerated and fused spine

Adjacent segment disorders are prevalent in patients following a spinal fusion surgery. Postoperative alterations in the adjacent segment biomechanics play a role in the etiology of these conditions. While experimental approaches fail to directly quantify spinal loads, previous modeling studies have numerous shortcomings when simulating the complex structures of the spine and the pre/postoperative mechanobiology of the patient. The biomechanical effects of the L4–L5 fusion surgery on muscle forces and adjacent segment kinetics (compression, shear, and moment) were investigated using a validated musculoskeletal model. The model was driven by in vivo kinematics for both preoperative (intact or severely degenerated L4–L5) and postoperative conditions while accounting for muscle atrophies. Results indicated marked changes in the kinetics of adjacent L3–L4 and L5–S1 segments (e.g., by up to 115% and 73% in shear loads and passive moments, respectively) that depended on the preoperative L4–L5 disc condition, postoperative lumbopelvic kinematics and, to a lesser extent, postoperative changes in the L4–L5 segmental lordosis and muscle injuries. Upper adjacent segment was more affected post-fusion than the lower one. While these findings identify risk factors for adjacent segment disorders, they indicate that surgical and postoperative rehabilitation interventions should focus on the preservation/restoration of patient’s normal segmental kinematics.

Biomechanical effects of lumbar fusion surgery on adjacent segments using musculoskeletal models of the intact, degenerated and fused spine Mahdi Ebrahimkhani 1 , Navid Arjmand 1* & Aboulfazl Shirazi-Adl 2 Adjacent segment disorders are prevalent in patients following a spinal fusion surgery. Postoperative alterations in the adjacent segment biomechanics play a role in the etiology of these conditions. While experimental approaches fail to directly quantify spinal loads, previous modeling studies have numerous shortcomings when simulating the complex structures of the spine and the pre/ postoperative mechanobiology of the patient. The biomechanical effects of the L4-L5 fusion surgery on muscle forces and adjacent segment kinetics (compression, shear, and moment) were investigated using a validated musculoskeletal model. The model was driven by in vivo kinematics for both preoperative (intact or severely degenerated L4-L5) and postoperative conditions while accounting for muscle atrophies. Results indicated marked changes in the kinetics of adjacent L3-L4 and L5-S1 segments (e.g., by up to 115% and 73% in shear loads and passive moments, respectively) that depended on the preoperative L4-L5 disc condition, postoperative lumbopelvic kinematics and, to a lesser extent, postoperative changes in the L4-L5 segmental lordosis and muscle injuries. Upper adjacent segment was more affected post-fusion than the lower one. While these findings identify risk factors for adjacent segment disorders, they indicate that surgical and postoperative rehabilitation interventions should focus on the preservation/restoration of patient's normal segmental kinematics.
Adjacent segment degeneration (ASDeg) and adjacent segment disease (ASDis) are commonly detected conditions, respectively without and with clinical symptoms, following spinal fusion surgeries 1 . While up to 43% of patients may develop postoperative ASDis, the prevalence of ASDeg is much greater (~ 84%) 2 . Some postoperative conditions are the likely consequences of a pre-existing or a natural progress of degeneration. In addition, the causative mechanobiology role of the altered biomechanics following a fusion surgery has been indicated. This includes postoperative alterations in the mobility of adjacent segments, disruptions in their anatomy, and iatrogenic intraoperative injuries to paraspinal muscles which altogether may change the spinal alignment and loadings thereby initiating/accelerating adjacent segment disorders (ASDs) [3][4][5][6][7] .
In vitro, in vivo and in silico biomechanical studies corroborate such postoperative alterations in the spine kinematics and kinetics [8][9][10] . Image-based in vivo studies can quantify only the postoperative alterations in vertebral kinematics/motions 11 . In vitro studies also remain limited by the assumed idealized loading/boundary conditions 12 . In silico modeling investigations, however and while simulating changes in kinematics, offer an improved insight into postoperative alterations in the kinetics of adjacent discs. Passive finite element (FE) models driven by pure moments with/without follower loads (i.e., forces that follow deformation of the spine) [13][14][15][16][17][18] or driven by image-based displacements 19 as well as musculoskeletal (MS) models with idealized passive joints [20][21][22][23][24][25] have been used to investigate the adjacent segments effects. While, force-controlled passive FE models fail to Intact MS model. Our extensively-validated nonlinear MS model 28,[44][45][46][47] evaluates forces in trunk muscles and spinal joints during in vivo activities using a kinematics-driven optimization algorithm ( Figs. 1 and 2). The  www.nature.com/scientificreports/ pelvis, T1-T12 thorax and lumbar vertebrae are rigid. T12-L1 through L5-S1 discs are simulated by 3-node nonlinear shear-deformable beams located 4 mm posterior to the disc centers to account for the shift in the disc center of rotation 32,48,49 . The trunk weight was partitioned among upper arms, forearms, hands, head, and T1-L5 segments and applied via rigid elements at their respective centers of mass 26,44,45,50 . In total, 56 trunk muscle fascicles are incorporated while considering curved lines of action of back global muscles (i.e., their wrapping around and contact forces at bony vertebrae) 46 .
For each simulated task in upright and flexed postures (see "Simulated tasks"), the flexion movement of thorax (T) and pelvis (P) are prescribed into the model based on in vivo measurements 44,45 . The lumbar flexion movement (L = T -P) is subsequently partitioned between the T12-L1 through L5-S1 segments by 11.5%, 15%, 14%, 18%, 21.5% and 20% respectively, based on in vivo studies 45,51,52 . To determine muscle forces, a multi-level optimization algorithm minimizing the sum of cubed muscle stresses is used. In this procedure, the reaction moment of each vertebra is balanced by the corresponding muscles attached to that vertebra. Calculated muscle forces are then fed back into the nonlinear MS model as external loads and muscle forces are recalculated. This iterative approach is continued till convergence is reached, i.e., almost no change in the predicted muscle forces between two successive iterations (Fig. 1). The MS model has been validated in terms of predicted muscle forces and disc compressive loads by comparisons with the measured muscle EMG and L4-L5 intradiscal pressure (IDP) values, respectively, under various tasks in upright/flexed postures in static/dynamic conditions 28,44,45,53,54 .   56,57 . Disc degeneration is modelled at the L4-L5 segment by reducing the disc height by two-third according to Pfirrmann classification 58 . Based on in vivo data collected on 44 low back pain patients, LPR decreases (the contribution of pelvis to forward trunk flexion increases) as compared to healthy individuals 59,60 . Moreover, MR images (of 259 patients) 61 indicate that the loss of motion at the degenerated (i.e., stiffened) L4-L5 segment is compensated by the hypermobility of motion segments at the thoracolumbar junction rather than the cranial or caudal adjacent segments. Therefore, in the preoperative degenerated model, LPR is dropped by 20% and the lumbar vertebral rotations are revised (compensating reduced motion at the L4-L5 by larger angles at the T12-L1 and L1-L2 levels; i.e., the contribution of L4-L5 over total lumbar rotation was reduced from 21.5% in the intact model to 8% and the contribution of T12-L1 and L1-L2 increased from 11.5% and 15% in the intact model to 19.5% and 18%, respectively) ( Fig. 4). Note that the reduced rotation at the L4-L5 segment is a direct consequence of the higher stiffness (implemented in the model by reducing the disc height) at the severely narrowed degenerated disc.
Fused postoperative MS models. Fusion simulations are solely considered for the intact MS model. This is because similar postoperative models are expected irrespective of the preoperative conditions (see "Limitations" section). Fused model is actually the intact model modified by rigidly connecting the L4 and L5 vertebrae (Fig. 2). The postoperative geometry of the spine, prescribed kinematics of vertebrae/pelvis, and physiological cross-sectional areas of injured muscles are subsequently modified based on available in vivo data as described below (also summarized in Fig. 3).   www.nature.com/scientificreports/ fusion (TLIF)] 64,65 , the shape of intervertebral cage (rectangular or wedge-shaped) 66,67 , and the surgical set-up (intraoperative patient position) 68,69 influence the postoperative posture. This adapted postoperative standing posture can be estimated through an optimization procedure 63 . Three postoperative L4-L5 lordosis angles are considered in the current MS model; normal-lordosis (no changes between pre-and postoperative L4-L5 lordoses), hyper-lordosis (increased lordosis), and hypo-lordosis (reduced lordosis) 21,70 . Postoperative changes in the lordosis of the fused segment (L4-L5) in cases of hyper-and hypo-lordoses are taken according to an in vivo study (with 42 patients) 66 . Rotations at the remaining intact segments including the sacrum are subsequently calculated via an optimization procedure that minimizes the sum of joint reaction moments and thus muscle forces in the upright standing posture under gravity loading 47,54 . These optimal rotations from the unloaded supine state to the upright standing under gravity loads are given in Supplementary Table S1 for intact, degenerated and fused models for the three normal, hyper-and hypo-lordosis configurations. Postures are depicted in Supplementary Figure S1.
Segmental kinematics in flexion tasks. Based on in vivo observations, two scenarios are adopted here; once it is assumed that the postoperative lumbopelvic rhythm (i.e., ratio of total lumbar over pelvis rotations, LPR) remains unchanged and thus the eliminated L4-L5 motion after fusion is compensated by the remaining individual lumbar segments (L1-L5) (an approach similar to that in previous MS studies 22,25 ) (Supplementary Table S2). In order to do so, the individual lumbar motion rhythms (i.e., ratio of segmental over total lumbar rotations) are modified based on upright x-ray measurements of motion rhythms in 36 patients before and after the fusion surgery 71 , i.e., the L4-L5 rotation is distributed among L1-L2 to L5-S1 by 3%, 19%, 38%, 0% and 40%, respectively (Fig. 4). According to another in vivo study on 5 patients 72 , LPR alters after the spinal fusion surgery in some patients (contribution of pelvis increases in flexion). Therefore, in the second scenario, the eliminated L4-L5 rotation after fusion is compensated by the pelvis alone, i.e., pelvis rotations were increased by 2.8°, 6°, 9.3° and 11.1° in trunk flexion angles of 20°, 40°, 60° and 80°, respectively (Supplementary Table S2). The rotations between other intact lumbar segments are again distributed according to the postoperative fused rhythms ( Fig. 4 and Supplementary Table S2). To reach the final flexed postures, the vertebral (T12 through S1) rotations for different flexed postures (Supplementary Table S2) are applied to the upright standing posture (Supplementary Table S1).
Muscle iatrogenic injuries. Iatrogenic intraoperative injuries in back muscles are simulated according to our previous measurement (on 6 patients) and modeling studies 56,57 ; the physiological cross-sectional areas (PCSAs) of multifidus (MF) and erector spinae (ES) fascicles crossing over the L4-L5 level are reduced in postoperative models by 26% and 11%, respectively. To investigate the effect of muscle damages alone on adjacent segment kinetics in flexion tasks, the largest (80°) trunk flexion task (section Simulated tasks) was re-simulated in the intact state by only implementing muscle damages (considering identical kinematics). Model outputs. Following the hypothesis that ASDs are likely a consequence of alterations in the loading patterns, the following outputs are calculated for the intact, degenerated, and fused conditions: segmental local compression/shear loads and passive joint moments at the adjacent discs mid-height planes (i.e., L3-L4 and L5-S1) as well as the muscle forces. A substantial change (assumed here to be > 25%), as compared to the preoperative intact or degenerated states, in these parameters highlights an increase in the risk to initiate/accelerate postoperative ASDs.

Results
Intact model. The intact model has been validated elsewhere 28,44,45,53,54 . Preoperative degenerated model. In the upright standing posture, predictions were found close enough to those in the intact model (changes < 10%, Table 1 and Fig. 5). In flexion tasks, the increase in the T12-L2 flexion angles (Fig. 4) and pelvic flexion (Supplementary Table S2) significantly reduced global and increased local muscle forces (Table 2). Consequently, compression forces increased by up to 21% and shear forces decreased by up to 48% at the upper adjacent segment (Table 1). All other alterations remained < 10%.
Postoperative models. Postoperative kinetics of adjacent segments altered not only with the spinal lordosis, LPR, segmental kinematics and muscle injuries but also with the level of adjacent segment, preoperative L4-L5 disc condition, and simulated task ( Table 1, Fig. 5, Supplementary Table S3). In general, alterations in LPR and muscle injuries had, respectively, the greatest and least effects on adjacent segment kinetics. Larger postoperative changes generally occurred at the upper adjacent segment and in flexion tasks especially when the postoperative LPR remained unchanged with respect to preoperative conditions (Table 1 and Fig. 5). More details on task-specific findings are provided below.
Upright standing. Compared to preoperative intact and degenerated states, the postoperative hypo-lordosis posture increased external flexion moments, local/global muscle forces ( www.nature.com/scientificreports/ (though by < 16%) at both adjacent segments (Table 1). Shear force decreased at the upper segment by up to 41% (Table 1). In contrast, the hyper-lordosis condition decreased external flexion moments, all muscle forces ( Table 2) and consequently adjacent segment compression forces by up to 19% as compared to preoperative states (Table 1). In this configuration, the shear force increased at the upper segment by up to 53%. Irrespective of the hypo or hyper-lordosis postures, passive moments at both adjacent segments and shear force at the lower segment were only slightly affected (< 10%) when compared to those in both preoperative states (Table 1). Moreover, the postoperative normal-lordosis configuration yielded results almost the same as those in the preoperative intact and degenerated states with alterations in all cases < 10% (Table 1). Muscle injuries had slight effects in local muscle forces, compression forces, shear forces and passive moments at adjacent segments (all < 10%) ( Tables 1,2).

Flexed postures.
With the fixed postoperative LPR, larger flexion angles at the adjacent segments significantly increased their passive moments by up to 73% (Table 1 and Fig. 5). This reduced local muscle forces and adjacent segment compression loads when comparing to preoperative flexed states (Tables 1, 2). At the upper adjacent segment, shear forces generally decreased with respect to the intact state but significantly increased when compared to the degenerated state (Table 1). With the altered postoperative LPR compared to the fixed LPR, adjacent segment effects were generally less pronounced with the exception of the upper adjacent segment shear loads which increased by up to 115% when compared to the degenerated state ( Table 1). The effect of muscle damages alone (with no postoperative alteration in kinematics) on adjacent segment kinetics was found < 3% in 80° trunk flexion task.

Discussion
Biomechanical effects of the L4-L5 fusion surgery on adjacent segment kinetics were investigated while considering two distinct preoperative L4-L5 disc conditions. A validated MS spine model driven by in vivo data on pre-and postoperative T12-S1 kinematics was employed while also simulating alterations in spinal lordosis and muscle injuries. In corroboration of our hypotheses, marked postoperative alterations were predicted in adjacent segment kinetics that depended on the preoperative L4-L5 disc condition, postoperative lumbopelvic kinematics and, to a lesser extent, the postoperative changes in the L4-L5 segmental lordosis and intraoperative muscle injuries. Moreover, upper (L3-L4) and lower (L5-S1) adjacent segment kinetics were affected post-fusion to different degrees.
Upper (L3-L4) adjacent segment. Alterations in the postoperative kinetics were generally more pronounced in the upper adjacent segment than the lower one (Table 1). This finding is in agreement with clinical observations showing greater prevalence of ASDs at the upper adjacent segment [74][75][76][77][78][79] . Even in the upright posture, large postoperative alterations in the L3-L4 shear load by up to 53% and 43% in the hyper-lordosis fused model with, respectively, intact and degenerated L4-L5 discs were predicted (Table 1). In agreement with other model studies 24,25 , fixed LPR postoperatively was found to generally increase passive moments and lower compression forces with respect to preoperative cases (Table 1). Moreover, L3-L4 shear load in flexed postures significantly increased by up to 84% as compared to the degenerated condition. In corroboration, in vivo animal studies and in vitro tests have shown that the shear 80,81 and hyperflexion 82 loads increase the risk of disc degeneration. It therefore appears that in these patients, the upper segment ASDs are likely associated with foregoing changes in segmental biomechanics. In patients with an altered postoperative LPR, however, these alterations were subdued and generally limited to an increase in the L3-L4 shear load by up to 115% only in those with the  www.nature.com/scientificreports/ severely degenerated preoperative state (Table 1). Finally, due to alterations in disc inclination and muscle forces ( Table 2), hyper-and hypo-lordosis configurations in flexion tasks produced the smallest and largest shear loads, respectively 21 .
Surgically altered segmental lordosis. Alterations in the L4-L5 segmental lordosis perturbed postoperative postures. For instance, the hypo-lordosis configuration caused an additional forward bending of the thorax, increased sagittal vertical axis (SVA) and retroversion of pelvis, i.e., a smaller sacral slope (Supplementary Figure S1). In contrast, the hyper-lordosis configuration resulted in the backward bending of the thorax, decreased SVA and anteversion of pelvis, i.e., a larger sacral slope (Supplementary Figure S1) thereby reducing all muscle forces (Table 2) and adjacent segment compression forces. In accordance with our findings, numerous clinical studies recommend correcting preoperative lumbopelvic abnormalities and restore the segmental lordosis during the surgery 83-86 . Traditional versus minimally invasive surgeries. The minimally invasive fusion surgery, as compared to the conventional open technique, considerably reduces the risk of iatrogenic intraoperative muscle injuries 87 and likely the ASDs 88 . Our predictions (Table 2), in agreement with a previous modeling study 57 , however, demonstrated that intraoperative muscle injuries affected primarily the load sharing among muscles with minimal direct alterations on adjacent segment kinetics. In agreement, several clinical studies have reported similar ASD prevalence among patients undergoing the minimally invasive fusion surgery or the conventional open techniques 78,89,90 . In contrast, a model study reported large changes in adjacent segment kinetics following the removal of entire muscle fascicles at treated levels 23 . In the standing posture, for instance, increases of 78% and 82% in upper adjacent segment compression and shear forces, respectively, were reported 23 .