Preoperative bone mineral density quantitatively assessed by Hounsfield units is associated with failed back surgery syndrome after lumbar fusion surgery: a retrospective study

Article information

Asian Spine J. 2025;.asj.2025.0129

Publication date (electronic) : 2025 September 2

doi : https://doi.org/10.31616/asj.2025.0129

Longlong Qiu^1,*, Haocheng Xu^2,*, Liming Yu^1,*, Xiaojie Chen¹, Junwei Qu¹, Xinlei Xia², Chaojun Zheng², Qiwang Chen¹

¹Department of Orthopaedics, Ningbo Ninth Hospital, Ningbo, China

²Department of Orthopaedics, Huashan Hospital, Fudan University, Shanghai, China

Corresponding author: Qiwang Chen, Department of Orthopaedics, Ningbo Ninth Hospital, No. 68 Xiangbei Road, Jiangbei District, Ningbo 315020, China, Tel: +86-0574-55683345, Fax: +86-0574-87672961, E-mail: 36982901@qq.com

Co-corresponding author: Chaojun Zheng, Department of Orthopaedics, Huashan Hospital, Fudan University, 12 Mid-Wulumuqi Road, Shanghai 200040, China, Tel: +86-021-5288-7136, Fax: +86-021-6248-9191, E-mail: Countd2388@163.com; Cjzheng17@fudan.edu.cn

*These authors contributed equally to this work and should be considered co-first authors.

Received 2025 March 5; Revised 2025 May 3; Accepted 2025 May 15.

Abstract

Introduction

Surgical intervention remains a standard and effective approach for managing degenerative lumbar diseases (DLDs) in patients who do not respond to conservative treatment strategies [1]. Although many prior studies have documented the excellent clinical outcomes of lumbosacral spinal surgeries, serious complications can occur [2]. Among these complications, failed back surgery syndrome (FBSS) is a term that describes an unsatisfactory outcome following anatomically and radiologically successful lumbosacral spinal surgery, characterized by persistent, worsening, or recurrent low back pain (LBP), with or without radiating pain to the lower limbs [3,4]. Significantly, 10%–40% of people receiving lumbosacral spinal surgery experience this debilitating consequence [3,4]. Given the serious negative impact of FBSS on patients’ psychological and physiological statuses, attempts to mitigate the risk of this complication are crucial.

Although the precise pathophysiological mechanism of FBSS remains unclear, increasing evidence has revealed a strong correlation between postoperative spinal instability and FBSS [3,5]. This spinal instability affects not only the surgical segments but also the adjacent segments [3]. Numerous previous studies have demonstrated that osteoporosis is a significant risk factor for postoperative spinal instability [6,7]. Furthermore, osteoporosis is commonly associated with persistent LBP [8]. Thus, decreased bone mineral density (BMD) may play a role in the etiology of FBSS. However, few studies have investigated the correlation between BMD and FBSS.

Computerized tomography (CT)-derived Hounsfield unit (HU) values have been extensively utilized in the evaluation of bone quality in patients with spinal disorders [9,10]. A growing body of studies have demonstrated a strong relationship between CT-derived HU values and the BMD assessments provided by dual-energy X-ray absorptiometry (DEXA) [9,10]. Several previous studies have further demonstrated that compared with DEXA, CT-derived HU values yield significantly more accurate BMD assessments of the lumbar spine and demonstrate a superior diagnostic accuracy for osteoporosis, especially in patients undergoing spine surgery [11,12].

Thus, this study aimed to assess BMD, as quantified by preoperative CT-derived HU values, in patients with DLDs and to explore the relationship between BMD and FBSS following lumbar fusion surgery, thereby providing a theoretical basis for its clinical prevention.

Materials and Methods

This retrospective study was performed in accordance with the Declaration of Helsinki, and all study protocols were approved by the Human Ethics Committee (2024LIK02).

Data measurements

CT-derived HU values

In accordance with standard procedures outlined in previous studies [9,10], all patients underwent preoperative quantitative CT imaging (Lightspeed 64-VCT; General Electric, Boston, MA, USA). Two-dimensional reconstructions were obtained in the sagittal plane using the CT parameters, which included a slice thickness of 1.25 mm with a 0.625-mm interval. The Picture Archiving and Communication Systems imaging software (Tomorrow Medical Network Technology Co. Ltd., Ningbo, China) was used to calculate the HU values. The largest possible elliptical region of interest was placed within the medullary space of the vertebral body to minimize beam hardening and volume averaging artifacts from adjacent cortical bone (Fig. 1). Regions of interest were measured on axial images at the mid-vertebral level from L1 to L4, and the mean HU value for each patient was derived from the average of these four different measurements (Fig. 1). The measurements were performed by two independent clinicians who were blinded to the patients’ information, and the final values were obtained by averaging their results.

Fig. 1

Computed tomography scans illustrating the method of determining the Hounsfield unit (HU) value with use of an elliptical region of interest area. (A) The left image shows the axial planes of interest area on a sagittal slice of a computed tomography scan of L1–L4 lumbar vertebral body. (B–E) The right images show the elliptical region of interest area on the axial slice of L1–L4 lumbar vertebral body, along with the HU values generated by the imaging software program.

T scores

For DEXA measurements, the average T scores from the L1 to L4 vertebrae were determined using a Lunar Prodigy Advance densitometer (General Electric).

Demographic and other medical data collection

The following perioperative data were extracted from the patients’ medical records: age, sex, body mass index, primary diagnosis, preoperative disease duration, number of surgical segments, operative time, and intraoperative bleeding volume.

All patients completed an NRS assessment to evaluate LBP at rest and during movement, conducted preoperatively, 1–3 days postoperatively, and at 1-year follow-up. In addition, the Oswestry Disability Index (ODI) was administered to these patients both before and 1-year following the procedure. Prior to the procedure, all patients filled out a number of questionnaires, such as the Pain Catastrophizing Scale (PCS), Beck Anxiety Inventory (BAI), and Beck Depression Index (BDI). Furthermore, all patients with postoperative chronic LBP underwent further imaging assessments—including X-ray, CT, and magnetic resonance imaging (MRI)—at the 1-year postoperative follow-up. Dynamic lateral lumbar X-ray detected neither obvious activities in the surgical segments nor implant loosening/displacement, and the latter was further verified on CT scans (e.g., no obvious bone resorption around the pedicle screws and intervertebral fusion device with or without sagittal CT scans demonstrating the formation of bone connections in the surgical segments). Furthermore, lumbar MRI detection was used to confirm adequate decompression of both the spinal canal and intervertebral foramina and to exclude adjacent segment spinal stenosis or disc herniation.

Results

This study included 115 patients (Fig. 2), and their clinical characteristics are summarized in Table 1. Compared to preoperative outcomes, the 115 patients who underwent lumbar fusion surgery demonstrated significantly reduced NRS scores both at rest (1.1±1.5 vs. 0.5±1.1, p<0.05) and during movement (4.3±2.1 vs. 1.3±1.5, p<0.05), as well as decreased ODI scores (51.0%±25.8% vs. 20.2%±23.4%, p<0.05) at the 1-year postoperative follow-up.

Fig. 2

Design of patient selection. DLDs, degenerative lumbar diseases; TLIF, transforaminal lumbar interbody fusion; FBSS, failed back surgery syndrome.

Table 1

Medical characteristic between the patients with and without FBSS

At the 1-year postoperative follow-up, 14 patients (14/115, 12.2%) were detected to have developed FBSS. Among the 14 patients with FBSS, nine patients (9/14, 64.3%) exhibited increased 1-year postoperative LBP at rest or with movement compared to their preoperative assessments. Three of them also experienced lower limb pain; however, only one of them reported lower limb pain NRS ratings of more than 3. Furthermore, the other five patients (9/14, 35.7%) still exhibited NRS scores of LBP greater than 3 at the 1-year postoperative assessments, which was similar to the preoperative LBP intensities, and only one of these five patients also experienced mild persistent lower limb pain (NRS <3). Importantly, 1-year postoperative imaging—including dynamic X-ray, computed tomography (CT), and MRI—along with electrophysiological assessments, confirmed complete decompression of both the spinal canal and intervertebral foramina. No internal implant failure, adjacent segment spinal stenosis or disc herniation, or progressive lumbosacral nerve injuries were observed in any of the FBSS patients included in this study.

Among these 14 patients with FBSS, the 1-year postoperative NRS score at rest was correlated with both the preoperative PCS score (r=0.58, p<0.05) and the average NRS score 1–3 days postoperatively (r=0.58, p<0.05) (Fig. 3), and the 1-year postoperative NRS score with movements was associated with the T score (r=−0.58, p<0.05) and the HU value (r=−0.64, p<0.05) (Fig. 3). In addition, the 1-year postoperative ODI scores were correlated with HU values (r=−0.66, p<0.05) (Fig. 3).

Fig. 3

(A) Postoperative 1-year Numeric Rating Pain Scale (NRS) scores of low back pain (LBP) at rest are positively associated with preoperative Pain Catastrophizing Scale (PCS) scores in failed back surgery syndrome (FBSS) patients. (B) Postoperative 1-year NRS scores of LBP at rest are positively associated with average NRS scores of LBP 1–3 days after operation in FBSS patients. (C) Postoperative 1-year NRS scores of LBP with movements are negatively associated with preoperative T scores in FBSS patients. (D) Postoperative 1-year NRS scores of LBP with movements are negatively associated with preoperative computed tomography (CT)-derived Hounsfield unit (HU) values in FBSS patients. (E) Postoperative 1-year Oswestry Disability Index (ODI) scores are negatively associated with preoperative CT-derived HU values in FBSS patients.

Comparison between patients with and without FBSS

Preoperatively, the patients with FBSS exhibited higher NRS scores at rest, higher PCS scores, and higher BAI scores, as well as lower T scores and HU values, than those without FBSS (Table 2). Furthermore, the FBSS patients presented with comparatively greater intraoperative hemorrhage and extended operative times, as well as higher average NRS scores 1–3 days after operation, than did those without FBSS (Tables 1, 2). Patients with FBSS had significantly higher NRS ratings at rest and with movements, as well as higher ODIs, than patients without FBSS at the 1-year preoperative evaluation (Table 2).

Table 2

Preoperative and postoperative assessments between patients with and without FBSS

Perioperative predictors of FBSS after lumbar fusion surgery

Multivariate logistic regression was used to examine all preoperative covariates with significant differences (p<0.05) between patients with and without FBSS. These included the NRS scores at rest, PCS scores, BAI scores, T scores, and HU values, as well as the average NRS scores 1–3 days after operation, intraoperative bleeding, and the operative time. The findings demonstrated that the preoperative NRS score at rest, preoperative BAI score, preoperative PCS score, and preoperative HU value were independently associated with FBSS (Table 3). ROC curve analysis revealed that the CVs for the prediction of FBSS were 0.5 for the preoperative NRS score at rest, 3.5 for the preoperative BAI score, 15.5 for the preoperative PCS score, and 106.0 for the preoperative HU value (p<0.05) (Fig. 4).

Table 3

Multivariate regression analysis of significant predicting factors of FBSS

Fig. 4

Receiver operating characteristic (ROC) curve analysis revealed the cutoff value of preoperative low back pain intensity at rest (A), preoperative Beck Anxiety Inventory scores (B), preoperative Pain Catastrophizing Scale scores (C), and preoperative computed tomography-derived Hounsfield unit values (D), for predicting failed back surgery syndrome after lumbar fusion. CV, cutoff value; AUC, area under the ROC curve.

Discussion

The notable incidence of chronic LBP after lumbar fusion surgery (12.2%, 14/115 patients) in this study highlights that more than one in 10 patients who underwent lumbar fusion developed FBSS. Additionally, during the 1-year postoperative follow-up, all patients in the FBSS group had evident pain-related impairment, and around two-thirds of the patients required ongoing pain medication. Therefore, clinicians should be mindful of this serious complication and exercise caution when considering using lumbar fusion surgery for treating patients with DLDs.

The study results demonstrated that low preoperative BMD was linked to chronic postoperative LBP. Many previous studies have demonstrated that osteoporosis is a significant risk factor for both pedicle screw loosening and fusion device displacement/settlement [15,16], and these conditions have been shown to be strongly linked to FBSS [3,5]. However, the 1-year postoperative imaging assessments in this study contradicted this possibility. There is growing evidence that patients with osteoporosis are more prone to experience fusion-induced adjacent segment degeneration [6]. Although no patients with FBSS in this study presented with spinal stenosis or disc herniation of adjacent segments, Sun et al. [6] demonstrated that low bone mass may negatively impact the functional integrity of the endplate vascular microstructure and disturb matrix metabolism in the nucleus pulposus and annulus fibrosus, which may stimulate the sinovertebral nerve and subsequently lead to chronic postoperative LBP.

According to previous studies [17,18], the co-occurrence of osteoporosis and sarcopenia is prevalent in middle-aged and older individuals, and this syndrome is termed “osteosarcopenia.” Although the exact pathophysiological process of osteosarcopenia remains unclear, a recent study revealed that osteocalcin released from the bone plays a critical role in the communication between the muscle and bone [19], and the Wnt/β-catenin signaling pathway has been reported to mediate bone-muscle crosstalk by regulating osteoblastic activity [20]. Therefore, low BMD influences the interaction between myokines and osteokines, which may not only induce a loss of muscle mass and strength but also impact muscle function, thereby causing poor recovery of soft tissues at surgical sites. Numerous earlier investigations have demonstrated the crucial role of the lumbar muscles in maintaining spinal stability [21,22]; thus, inadequate posterior lumbar muscle recovery after operation may exacerbate the instability of the surgical and/or adjacent segments. Importantly, poor recovery following operation may impair blood circulation within the muscles at the surgical site, which could facilitate the local buildup of inflammatory mediators during physical activities [23]. This local buildup stimulates the peripheral terminals of primary sensory neurons and induces LBP [23]. These presumptions were further confirmed by the study’s findings that BMD measures were linked to 1-year postoperative LBP with movements but not with LBP at rest. Thus, osteosarcopenia may be another potential reason for the correlation between low preoperative BMD and FBSS.

Another key finding of this study is that CT-derived HU values, but not T values, were independently associated with FBSS, despite both being feasible methods for quantifying BMD. According to previous studies [9,11], the trabecular bone is more affected by osteoporosis than is the cortical bone. Therefore, lumbar spine images obtained by CT allow for the selective direct measurement of HU values of the trabecular bone [9,10]. Thus, this approach can circumvent the influence of the cortical bone and osteophyte growth, providing a more accurate reflection of the vertebral quality. In contrast, the T scores measured by DEXA reflect an average BMD that includes both cortical and trabecular bone [11]. Therefore, their measurements may be prone to overestimation owing to variations in the bone size and the presence of surrounding bone hypertrophy. Although DEXA-derived T scores have demonstrated relatively high efficacy in predicting thoracolumbar fragility fractures, it is important to acknowledge the potential influence of factors such as spinal degeneration, osteophyte formation, and osteosclerosis on BMD, as emphasized in numerous previous studies [11]. Equally important, preoperative CTs are routinely performed for many DLD patients before they undergo lumbar spinal surgery. Therefore, utilizing HU values can help avoid additional costs and radiation exposure from DEXA.

Similar to previous studies [24,25], this study discovered a link between preoperative LBP at rest and FBSS. In contrast to pain with movement, pain at rest has been linked to central sensitization (CS). Therefore, preoperative central hypersensitivity induced by impaired endogenous pain regulation may be another potential risk factor for FBSS, as suggested by many previous studies [26]. Importantly, this hypothesis may explain the predictive value of the preoperative BAI and PCS scores for FBSS. According to previous studies [26,27], neuroplastic changes originating from nociceptive pathways may spread to diverse brain regions and be attributed to the negative psychosocial symptoms (e.g., pain catastrophizing thoughts, anxiety, and depression). These psychosocial symptoms can also enhance forebrain activity, thereby increasing CS and subsequently causing FBSS [28].

Because the mechanism of FBSS is uncertain, the results of this study should be regarded cautiously. The present study demonstrated that low BMD, as quantified by CT-derived HU values, may be linked to FBSS. However, approximately one-fifth of the patients with FBSS presented HU values greater than 106, suggesting that there are other pathogenic mechanisms related to FBSS. Another clinical limitation of this study is that other measurements, such as the cross-sectional area and fatty infiltration of the posterior lumbar muscles, were not analyzed. However, these measurements cannot fully reflect sarcopenia. Furthermore, this study was a retrospective data review from a single center, which may have caused selection bias during patient enrollment and the small sample size. Thus, more significant results (e.g., different pathogenic mechanisms and their roles in FBSS) may be achieved in a future prospective study with more measurements and a larger number of cases.

Conclusions

The findings of this study support the existence of low BMD in patients with FBSS. Importantly, low preoperative BMD may be involved in the pathophysiological process of FBSS and contribute to LBP intensity and pain-related disability in patients with FBSS. Preoperative assessments of CT-derived HU values may provide additional information for identifying patients who may be at risk for developing FBSS, which may be beneficial for the preoperative stratification of patients and the identification of appropriate treatments to prevent this complication.

Article information Continued