Introduction
Lumbar fusion surgery is among the most commonly performed spinal procedures worldwide, with pedicle screw fixation forming the cornerstone of posterior instrumentation [1]. Despite considerable technical advancements, screw loosening, defined radiographically by the appearance of a radiolucent halo around the screw, remains a frequent and clinically consequential complication. Rates of 15%–30% have been reported in patients with osteoporosis, and loosening represents a leading cause of construct failure, pseudarthrosis, adjacent segment disease, and revision surgery [1,2].
Therefore, preoperative identification of patients at high risk for fixation failure is a priority. Dual-energy X-ray absorptiometry (DEXA)-based areal bone mineral density (aBMD) of the lumbar spine or proximal femur has been the standard assessment tool for decades. However, both measurement sites have well-recognized limitations in the context of degenerative lumbar disease, and the correlation each achieves with pedicle screw fixation strength remains modest [3,4]. This has prompted studies investigating computed tomography (CT)-derived Hounsfield unit (HU) measurements, particularly those targeting the planned screw trajectory, as mechanistically rational and potentially superior complementary assessment tools [5].
This review examines the growing evidence base for this emerging transition: why DEXA-based approaches are insufficient for the specific demands of pedicle screw fixation planning, what CT-derived metrics offer as complementary tools, where the trajectory-specific HU literature currently stands and what gaps remain before clinical implementation can be standardized. The framework of assessment and clinical application is summarized in Figs. 1 and 2 [6–14].
Limitations of DEXA-Based BMD in Pedicle Screw Planning
Lumbar DEXA
Lumbar DEXA measures two-dimensional aBMD of the L1–L4 vertebral bodies and surrounding structures in the posteroanterior projection. In patients with degenerative lumbar disease, which is the population most likely to undergo lumbar fusion, this measurement is subject to systematic upward bias from osteophytes, endplate sclerosis, facet joint hypertrophy, and aortic calcifications [3]. Studies have consistently demonstrated that lumbar DEXA overestimates true bone quality in this population, generating false-negative assessments of osteoporosis in a clinically significant proportion of patients, thereby masking fixation risk [3].
Femoral DEXA
Femoral neck DEXA avoids degenerative confounders of the lumbar spine and has been shown to correlate more strongly with pedicle screw insertional torque than with lumbar BMD in retrospective in vivo studies [4]. Nevertheless, correlation coefficients of approximately r=0.5 indicate that femoral BMD explains substantially less than half the variance in localized fixation strength—a finding that is mechanistically unsurprising. The fixation strength of a pedicle screw is determined by the bone quality at the screw-bone interface, not at the femoral neck, which is an anatomically remote and mechanically distinct site.
Systemic-local mismatch
The fundamental limitation of any DEXA-based metric is that it measures systemic bone density rather than local bone quality at the fixation site. Pedicle screw pullout strength and insertional torque are determined by cancellous bone density within the pedicle, cortical shell integrity, trabecular architecture, and the transition zone between the pedicle isthmus and vertebral body. These parameters vary substantially between vertebral levels, between sides at the same level, and even between adjacent regions within a single vertebral body [11,12]. No systemic measurement can evaluate this spatial heterogeneity [13,14]. A patient with osteoporotic femoral neck BMD may have a well-preserved pedicle cortical bone at the operative level and vice versa. DEXA cannot distinguish between these scenarios.
CT-Derived HU Measurements as Bone Quality Indices
Vertebral body HU
Hounsfield units derived from CT attenuation values provide a three-dimensional volumetric correlate of bone mineral content that is not susceptible to the degenerative artefacts that affect lumbar DEXA [2]. Several studies have validated CT HU thresholds for identifying osteoporosis, generally below 110–135 HU measured at L1 on routine clinical CT, and have demonstrated correlations between DEXA T-scores and vertebral fracture risk across diverse populations [2,5]. A critical operational advantage is that CT HU measurement is opportunistic; it uses preoperative CT routinely acquired for surgical planning, with no additional radiation exposure or cost to the patient pathway. Zou et al. [9] demonstrated that vertebral body HU outperformed DXA T-scores for predicting screw loosening, reinforcing the clinical relevance of CT-based assessment. It should be noted that all HU thresholds in the current literature are exploratory values derived from non-standardized CT protocols and have not been prospectively validated across centers.
Vertebral Bone Quality score
The MRI-based Vertebral Bone Quality (VBQ) score, derived as the quotient of the median signal intensity of the L1–L4 vertebrae divided by the signal intensity of cerebrospinal fluid on T1-weighted sagittal sequences, has emerged as an additional opportunistic metric [15]. It has been shown to correlate with DEXA BMD and predict screw loosening and adjacent segment fracture. Comparative studies of HU and VBQ suggest a similar predictive performance for overall screw loosening [15]; however, trajectory-specific comparisons of VBQ and trajectory-specific HU have not been reported. VBQ requires MRI acquisition, which may not be universally available or clinically indicated.
From vertebral body to trajectory-specific HU
Critical conceptual and technical advancements in CT-based bone quality assessment have led to the transition from measuring HU at the center of the vertebral body to measuring HU within a region of interest (ROI) that precisely recapitulates the planned screw trajectory [6]. This approach acknowledges that the bone encountered by the advancing screw includes not only the central cancellous vertebral body but also the pedicle cortex, pedicle isthmus cancellous bone, and transition zone to the vertebral body, which are regions that are anatomically distinct, vary independently of global BMD, and are the primary determinants of screw purchase. Li et al. [11] demonstrated that HU measured at the precise future screw insertion site was a superior predictor compared to conventional measurement locations, further validating that spatial specificity of measurement is paramount. Trajectory-specific HU aligns the bone quality measurement directly with the mechanical domain of fixation.
Evidence for Trajectory-Specific HU
Table 1 summarizes the key studies examining trajectory-specific HU as a predictor of pedicle screw insertional torque and postoperative loosening [6–13].
Establishing proof of concept: Matsukawa et al. [6] (2018)
In a study of 92 consecutive patients undergoing single-level posterior lumbar interbody fusion using the cortical bone trajectory technique, trajectory HU measured within a cylindrical ROI precisely following each screw path correlated with insertional torque at r=0.75 (p<0.001), markedly outperforming femoral (r=0.59) and lumbar (r=0.55) BMD in the same patients [6]. Multivariate logistic regression identified trajectory HU was an independent predictor of 12-month screw loosening (odds ratio [OR], 0.70; 95% confidence interval [CI], 0.56–0.84) [6]. This landmark study established a proof of concept for trajectory-specific measurements. It should be acknowledged, however, that a correlation of r=0.75 derived from n=92 patients carries a 95% CI of approximately 0.65–0.83, and the lower bound of this range is not markedly superior to the femoral DEXA correlations characterized as modest elsewhere in this literature.
Early confirmation in single-level fusion: Sakai et al. [7] (2018)
Sakai et al. [7] examined trajectory HU as a predictor of screw loosening in 52 patients who underwent single-level lumbar interbody fusion. This is one of the earliest independent confirmations that measurement along the screw path, rather than at the vertebral body center, provides clinically meaningful prediction of fixation failure. These findings corroborated the conceptual framework established by Matsukawa et al. [6] and extended it to a separate cohort and surgical context.
Three-dimensional trajectory HU: Zhao et al. [10] (2023)
Using three-dimensional HU measurement of the full screw trajectory volume in MIMICS software in 90 patients with degenerative lumbar disease, Zhao et al. [10] demonstrated that three-dimensional trajectory HU outperformed vertebral body HU and DEXA-based BMD across all receiver operating characteristic curve analyses for screw loosening prediction. Proposed exploratory threshold values of 110 HU for the screw trajectory and 160 HU for the vertebral body were identified [10]. These values were derived from a single-center, non-calibrated CT protocol and should not be applied as clinical decision points pending cross-scanner validation. The three-dimensional approach captures heterogeneity in bone quality along the full length of the trajectory, including the pedicle isthmus, which single-slice methods may sample inconsistently.
Site-specific HU at the insertion point: Li et al. [11] (2023)
Li et al. [11] demonstrated that HU measured at the precise future screw insertion site was a superior predictor of fixation outcomes compared to conventional vertebral body center measurements. This study reinforces the mechanistic rationale for trajectory-aligned measurements and provides additional evidence that the spatial specificity of HU sampling is a key determinant of predictive performance.
Extension to lumbosacral fixation: Ishikawa et al. [12] (2023)
Ishikawa et al. [12] extended the trajectory HU concept to S1 screws in lumbosacral fusion and reported an area under the receiver operating characteristic curve of 0.79 for predicting screw loosening. This demonstrates that the trajectory-specific approach is applicable beyond single-level lumbar procedures and may be relevant to more complex constructs involving the lumbosacral junction.
Single-slice trajectory HU: Yamamoto et al. [13] (2025)
A retrospective Japanese cohort study demonstrated that pedicle screw trajectory HU measured on a single axial CT slice, selected at the planned screw entry level for routine preoperative planning, was significantly lower in screws that subsequently loosened at follow-up [13]. An exploratory threshold of 123 HU was identified as a practical preoperative risk stratification tool [13]. This value was derived from a single-center, non-standardized acquisition protocol and requires prospective multicenter validation before it can be applied as a clinical decision threshold. This approach offers the advantage of technical simplicity and compatibility with the standard picture archiving and communication system (PACS) workflow.
QCT-based regional volumetric BMD: Ishikawa et al. [8] (2018)
Using quantitative CT (QCT) to measure regional volumetric BMD (vBMD) within a cylindrical volume around each screw (pedicle screw-specific volumetric BMD, PS-vBMD), Ishikawa et al. [8] demonstrated that PS-vBMD was correlated with insertional torque at r=0.61, outperforming femoral neck aBMD (r=0.53), total hip aBMD, and lumbar aBMD in the same patients. PS-vBMD was the dominant predictor of fixation strength in a multiple regression analysis [8]. Although QCT provides true vBMD (mg/cm3) rather than attenuation-based HU, the conceptual parallel—local, trajectory-aligned measurement—aligns with the broader literature.
Clinical Implications and Limitations of Current Evidence
Preoperative risk stratification and surgical planning
If trajectory-specific HU values below the exploratory thresholds of 110–123 HU, which are derived from heterogeneous, non-standardized CT acquisition protocols and have not been validated across institutions, are ultimately confirmed through prospective validation, surgeons may be able to prospectively tailor fixation strategy: planning prophylactic cement augmentation for at-risk levels, considering expandable screw designs, selecting cortical bone trajectory techniques to maximize cortical purchase, or adjusting construct length to incorporate additional fixation points at levels with adequate bone quality [6,14]. The trajectory-specific approach could theoretically provide a level-by-level and side-by-side risk map that neither DEXA nor vertebral body HU can provide.
CT scanner variability and measurement standardization
A critical limitation not fully addressed in the current literature is the dependence of HU values on CT acquisition parameters. Beam energy (kVp), reconstruction kernel, slice thickness, and patient body habitus (beam hardening artefact) all influence the measured HU for the same bone. In multi-center imaging work, uncalibrated HU values can vary by 30–40 units depending on the scanner and protocol alone. The proposed exploratory thresholds of 110–123 HU are derived from studies using heterogeneous acquisition protocols, and their inter-institutional transferability has not been validated [5]. Standardization of CT protocols or development of HU correction algorithms accounting for acquisition parameters is essential before threshold-based clinical decision-making can be universally implemented.
Methodological heterogeneity across studies
All trajectory-specific HU studies to date have a retrospective design with small to moderate sample sizes and heterogeneous measurement methodologies [6–14]. Single-slice axial, multi-planar cylindrical, and fully three-dimensional volumetric approaches have each been employed without standardized ROI dimensions, slice selection criteria, or image reconstruction requirements. No published study has prospectively compared trajectory-specific HU against concurrent femoral DEXA in the same patients under pre-specified, standardized conditions with a primary outcome of screw loosening at defined follow-up.
Statistical and methodological limitations of primary studies
The three methodological limitations of the underlying primary studies merit explicit acknowledgement. First, insertional torque, which serves as the primary correlation endpoint in most of the cited studies, is a biomechanical surrogate outcome and not a direct clinical endpoint. The relationship between insertional torque and long-term screw loosening, although mechanistically plausible, has not been definitively established in large prospective cohorts. Second, the correlation coefficients of r=0.61 to 0.75 derived from samples of 52–92 patients, have wide CIs. At n=92, the 95% CI for r=0.75 extends approximately from 0.65 to 0.83; the lower limit of this range is not markedly different from that of the femoral DEXA correlation values that the literature characterizes as modest. Therefore, point estimates from small studies may overestimate the true effect size. Third, all key trajectory-specific HU studies were conducted in East Asian populations, and the generalizability of both the correlation strengths and the proposed thresholds to other demographic groups, with different body habitus, bone microarchitecture and CT acquisition conventions, has not been assessed.
Software integration and clinical scalability
Manual trajectory HU measurement using standard PACS tools is technically feasible but time-consuming, limiting uptake in high-volume clinical practice. Integration into preoperative surgical navigation platforms and robotic surgery systems as automated bone quality heat maps along planned trajectories is in active development at multiple centers. Artificial intelligence-assisted segmentation of screw trajectories on preoperative CT with automated HU extraction and risk flagging is the key pathway to clinical scalability and reproducibility across operators and institutions.
Future Directions
The following research priorities can be elucidated from the current evidence base. Prospective validation of trajectory HU thresholds in multi-center cohorts with standardized CT acquisition protocols and pre-specified follow-up intervals is the single most important next step [6–13]. Without prospective data, the current exploratory retrospective threshold values of 110–123 HU cannot be applied to preoperative decision-making with the confidence required for guideline incorporation.
Standardization of measurement methodology, particularly the choice between single-slice, multi-planar cylindrical, and three-dimensional volumetric approaches, requires head-to-head comparison in the same patient cohorts, with reproducibility analysis across operators and institutions [5,10]. The inter- and intra-rater reliability of manual trajectory HU measurement and its dependence on imaging parameters have not been systematically characterized.
The development of composite risk scores incorporating trajectory HU with clinical variables (age, body mass index, fusion level count, sagittal alignment parameters and pharmacological osteoporosis treatment status) may improve predictive accuracy beyond HU alone. Machine learning approaches applied to preoperative CT data, extracting radiomics features beyond the mean HU, represent an emerging avenue.
Validation in high-risk subgroups, such as those with adult spinal deformity, revision surgery, sacropelvic fixation, and long-term corticosteroids or immunosuppressants, is required before broad clinical extrapolation from the predominantly single-level posterior lumbar interbody fusion/transforaminal lumbar interbody fusion populations studied to date [8]. This includes validation across diverse demographic groups beyond the East Asian populations represented in the current literature.
Conclusions
A convergent body of retrospective evidence supports the mechanistic and empirical superiority of trajectory-specific CT-derived HU over systemic DEXA-based BMD for predicting pedicle screw insertional torque and postoperative loosening in degenerative lumbar disease [6–14]. The approach is mechanistically rational; it identifies the bone the screw actually engages, which is clinically feasible using routine preoperative CT without additional patient burden. This strategy is supported by multiple independent studies employing three distinct measurement methodologies and extending across a range of fusion types, including lumbosacral constructs.
The important limitations of the current evidence, including its exclusively retrospective design, small sample sizes predominantly from East Asian populations, heterogeneous measurement protocols, and the unresolved influence of CT acquisition variability on proposed HU thresholds, must be addressed before this approach can be incorporated into operative planning guidelines [5]. Prospective validation, protocol standardization, cross-scanner calibration and integration into automated surgical navigation platforms are the priority steps for clinical translation.
Trajectory-specific CT-derived HU assessment represents a promising complementary tool to the existing DEXA-based evaluation for pedicle screw fixation planning. Evidence supports a transition toward this localized, mechanistically coupled approach; however, this transition should be understood as an emerging framework to be validated and standardized rather than an established replacement for systemic bone density assessment. The field’s priority must be to fill the critical gaps with respect to prospective validation, protocol standardization, and cross-scanner calibration, as this will determine whether these exploratory thresholds can be translated into reliable clinical decision tools.
PDF Links
PubReader
XML Download
Full text via DOI
Download Citation
